2018-03-22 01:09:35 +08:00
=============================
2017-06-07 04:31:59 +08:00
User Guide for AMDGPU Backend
=============================
.. contents ::
:local:
2014-11-14 22:08:00 +08:00
2020-04-14 12:55:43 +08:00
.. toctree ::
:hidden:
AMDGPU/AMDGPUAsmGFX7
AMDGPU/AMDGPUAsmGFX8
AMDGPU/AMDGPUAsmGFX9
AMDGPU/AMDGPUAsmGFX900
AMDGPU/AMDGPUAsmGFX904
AMDGPU/AMDGPUAsmGFX906
AMDGPU/AMDGPUAsmGFX908
AMDGPU/AMDGPUAsmGFX10
AMDGPU/AMDGPUAsmGFX1011
AMDGPUModifierSyntax
AMDGPUOperandSyntax
AMDGPUInstructionSyntax
AMDGPUInstructionNotation
2020-07-04 06:31:53 +08:00
AMDGPUDwarfExtensionsForHeterogeneousDebugging
2020-04-14 12:55:43 +08:00
2014-11-14 22:08:00 +08:00
Introduction
============
2017-06-07 04:31:59 +08:00
The AMDGPU backend provides ISA code generation for AMD GPUs, starting with the
R600 family up until the current GCN families. It lives in the
2019-12-12 13:47:56 +08:00
`` llvm/lib/Target/AMDGPU `` directory.
2014-11-14 22:08:00 +08:00
2017-06-07 04:31:59 +08:00
LLVM
====
2014-11-14 22:08:00 +08:00
2017-06-07 04:31:59 +08:00
.. _amdgpu-target-triples:
Target Triples
--------------
2020-12-01 12:12:04 +08:00
Use the Clang option `` -target <Architecture>-<Vendor>-<OS>-<Environment> ``
to specify the target triple:
2017-06-07 04:31:59 +08:00
2017-11-10 09:00:54 +08:00
.. table :: AMDGPU Architectures
:name: amdgpu-architecture-table
============ ==============================================================
Architecture Description
============ ==============================================================
`` r600 `` AMD GPUs HD2XXX-HD6XXX for graphics and compute shaders.
`` amdgcn `` AMD GPUs GCN GFX6 onwards for graphics and compute shaders.
============ ==============================================================
.. table :: AMDGPU Vendors
:name: amdgpu-vendor-table
============ ==============================================================
Vendor Description
============ ==============================================================
`` amd `` Can be used for all AMD GPU usage.
`` mesa3d `` Can be used if the OS is `` mesa3d `` .
============ ==============================================================
.. table :: AMDGPU Operating Systems
2020-12-01 12:12:04 +08:00
:name: amdgpu-os
2017-11-10 09:00:54 +08:00
============== ============================================================
OS Description
============== ============================================================
*<empty>* Defaults to the *unknown* OS.
`` amdhsa `` Compute kernels executed on HSA [HSA]_ compatible runtimes
2020-12-01 12:12:04 +08:00
such as:
2020-12-10 11:38:32 +08:00
- AMD's ROCm™ runtime [AMD-ROCm]_ using the *rocm-amdhsa*
loader on Linux. See *AMD ROCm Platform Release Notes*
[AMD-ROCm-Release-Notes]_ for supported hardware and
software.
- AMD's PAL runtime using the *pal-amdhsa* loader on
Windows.
2020-12-01 12:12:04 +08:00
`` amdpal `` Graphic shaders and compute kernels executed on AMD's PAL
2020-12-10 11:38:32 +08:00
runtime using the *pal-amdpal* loader on Windows and Linux
Pro.
`` mesa3d `` Graphic shaders and compute kernels executed on AMD's Mesa
3D runtime using the *mesa-mesa3d* loader on Linux.
2017-11-10 09:00:54 +08:00
============== ============================================================
.. table :: AMDGPU Environments
:name: amdgpu-environment-table
============ ==============================================================
Environment Description
============ ==============================================================
2018-03-24 02:45:18 +08:00
*<empty>* Default.
2017-11-10 09:00:54 +08:00
============ ==============================================================
2017-06-07 04:31:59 +08:00
.. _amdgpu-processors:
Processors
----------
2020-12-01 12:12:04 +08:00
Use the Clang options `` -mcpu=<target-id> `` or `` --offload-arch=<target-id> `` to
specify the AMDGPU processor together with optional target features. See
:ref: `amdgpu-target-id` and :ref: `amdgpu-target-features` for AMD GPU target
specific information.
2017-06-07 04:31:59 +08:00
2020-12-24 17:44:01 +08:00
Every processor supports every OS ABI (see :ref: `amdgpu-os` ) with the following exceptions:
* `` amdhsa `` is not supported in `` r600 `` architecture (see :ref: `amdgpu-architecture-table` ).
2017-06-07 04:31:59 +08:00
.. table :: AMDGPU Processors
2017-11-10 09:00:54 +08:00
:name: amdgpu-processor-table
2017-06-07 04:31:59 +08:00
2020-12-17 10:45:47 +08:00
=========== =============== ============ ===== ================= =============== =============== ======================
Processor Alternative Target dGPU/ Target Target OS Support Example
Processor Triple APU Features Properties *(see* Products
Architecture Supported `amdgpu-os`_
*and
corresponding
runtime release
notes for
current
information and
level of
support)*
=========== =============== ============ ===== ================= =============== =============== ======================
2017-10-19 01:59:20 +08:00
**Radeon HD 2000/3000 Series (R600)** [AMD-RADEON-HD-2000-3000]_
2020-12-17 10:45:47 +08:00
-----------------------------------------------------------------------------------------------------------------------
2020-12-01 12:12:04 +08:00
`` r600 `` `` r600 `` dGPU - Does not
support
generic
address
space
`` r630 `` `` r600 `` dGPU - Does not
support
generic
address
space
`` rs880 `` `` r600 `` dGPU - Does not
support
generic
address
space
`` rv670 `` `` r600 `` dGPU - Does not
support
generic
address
space
2017-10-19 01:59:20 +08:00
**Radeon HD 4000 Series (R700)** [AMD-RADEON-HD-4000]_
2020-12-17 10:45:47 +08:00
-----------------------------------------------------------------------------------------------------------------------
2020-12-01 12:12:04 +08:00
`` rv710 `` `` r600 `` dGPU - Does not
support
generic
address
space
`` rv730 `` `` r600 `` dGPU - Does not
support
generic
address
space
`` rv770 `` `` r600 `` dGPU - Does not
support
generic
address
space
2017-10-19 01:59:20 +08:00
**Radeon HD 5000 Series (Evergreen)** [AMD-RADEON-HD-5000]_
2020-12-17 10:45:47 +08:00
-----------------------------------------------------------------------------------------------------------------------
2020-12-01 12:12:04 +08:00
`` cedar `` `` r600 `` dGPU - Does not
support
generic
address
space
`` cypress `` `` r600 `` dGPU - Does not
support
generic
address
space
`` juniper `` `` r600 `` dGPU - Does not
support
generic
address
space
`` redwood `` `` r600 `` dGPU - Does not
support
generic
address
space
`` sumo `` `` r600 `` dGPU - Does not
support
generic
address
space
2017-10-19 01:59:20 +08:00
**Radeon HD 6000 Series (Northern Islands)** [AMD-RADEON-HD-6000]_
2020-12-17 10:45:47 +08:00
-----------------------------------------------------------------------------------------------------------------------
2020-12-01 12:12:04 +08:00
`` barts `` `` r600 `` dGPU - Does not
support
generic
address
space
`` caicos `` `` r600 `` dGPU - Does not
support
generic
address
space
`` cayman `` `` r600 `` dGPU - Does not
support
generic
address
space
`` turks `` `` r600 `` dGPU - Does not
support
generic
address
space
2017-10-19 01:59:20 +08:00
**GCN GFX6 (Southern Islands (SI))** [AMD-GCN-GFX6]_
2020-12-17 10:45:47 +08:00
-----------------------------------------------------------------------------------------------------------------------
`` gfx600 `` - `` tahiti `` `` amdgcn `` dGPU - Does not - *pal-amdpal*
2020-12-01 12:12:04 +08:00
support
generic
address
space
2020-12-17 10:45:47 +08:00
`` gfx601 `` - `` pitcairn `` `` amdgcn `` dGPU - Does not - *pal-amdpal*
2020-12-01 12:12:04 +08:00
- `` verde `` support
generic
address
space
2020-12-17 10:45:47 +08:00
`` gfx602 `` - `` hainan `` `` amdgcn `` dGPU - Does not - *pal-amdpal*
2020-12-01 12:12:04 +08:00
- `` oland `` support
generic
address
space
2017-10-19 01:59:20 +08:00
**GCN GFX7 (Sea Islands (CI))** [AMD-GCN-GFX7]_
2020-12-17 10:45:47 +08:00
-----------------------------------------------------------------------------------------------------------------------
`` gfx700 `` - `` kaveri `` `` amdgcn `` APU - Offset - *rocm-amdhsa* - A6-7000
flat - *pal-amdhsa* - A6 Pro-7050B
scratch - *pal-amdpal* - A8-7100
- A8 Pro-7150B
- A10-7300
- A10 Pro-7350B
- FX-7500
- A8-7200P
- A10-7400P
- FX-7600P
`` gfx701 `` - `` hawaii `` `` amdgcn `` dGPU - Offset - *rocm-amdhsa* - FirePro W8100
flat - *pal-amdhsa* - FirePro W9100
scratch - *pal-amdpal* - FirePro S9150
- FirePro S9170
`` gfx702 `` `` amdgcn `` dGPU - Offset - *rocm-amdhsa* - Radeon R9 290
flat - *pal-amdhsa* - Radeon R9 290x
scratch - *pal-amdpal* - Radeon R390
- Radeon R390x
`` gfx703 `` - `` kabini `` `` amdgcn `` APU - Offset - *pal-amdhsa* - E1-2100
- `` mullins `` flat - *pal-amdpal* - E1-2200
scratch - E1-2500
- E2-3000
- E2-3800
- A4-5000
- A4-5100
- A6-5200
- A4 Pro-3340B
`` gfx704 `` - `` bonaire `` `` amdgcn `` dGPU - Offset - *pal-amdhsa* - Radeon HD 7790
flat - *pal-amdpal* - Radeon HD 8770
scratch - R7 260
- R7 260X
`` gfx705 `` `` amdgcn `` APU - Offset - *pal-amdhsa* *TBA*
flat - *pal-amdpal*
scratch .. TODO::
Add product
names.
2020-10-22 15:15:41 +08:00
2017-10-19 01:59:20 +08:00
**GCN GFX8 (Volcanic Islands (VI))** [AMD-GCN-GFX8]_
2020-12-17 10:45:47 +08:00
-----------------------------------------------------------------------------------------------------------------------
`` gfx801 `` - `` carrizo `` `` amdgcn `` APU - xnack - Offset - *rocm-amdhsa* - A6-8500P
flat - *pal-amdhsa* - Pro A6-8500B
scratch - *pal-amdpal* - A8-8600P
- Pro A8-8600B
- FX-8800P
- Pro A12-8800B
- A10-8700P
- Pro A10-8700B
- A10-8780P
- A10-9600P
- A10-9630P
- A12-9700P
- A12-9730P
- FX-9800P
- FX-9830P
- E2-9010
- A6-9210
- A9-9410
`` gfx802 `` - `` iceland `` `` amdgcn `` dGPU - Offset - *rocm-amdhsa* - Radeon R9 285
- `` tonga `` flat - *pal-amdhsa* - Radeon R9 380
scratch - *pal-amdpal* - Radeon R9 385
`` gfx803 `` - `` fiji `` `` amdgcn `` dGPU - *rocm-amdhsa* - Radeon R9 Nano
- *pal-amdhsa* - Radeon R9 Fury
- *pal-amdpal* - Radeon R9 FuryX
- Radeon Pro Duo
- FirePro S9300x2
- Radeon Instinct MI8
\ - `` polaris10 `` `` amdgcn `` dGPU - Offset - *rocm-amdhsa* - Radeon RX 470
flat - *pal-amdhsa* - Radeon RX 480
scratch - *pal-amdpal* - Radeon Instinct MI6
\ - `` polaris11 `` `` amdgcn `` dGPU - Offset - *rocm-amdhsa* - Radeon RX 460
flat - *pal-amdhsa*
scratch - *pal-amdpal*
`` gfx805 `` - `` tongapro `` `` amdgcn `` dGPU - Offset - *rocm-amdhsa* - FirePro S7150
flat - *pal-amdhsa* - FirePro S7100
scratch - *pal-amdpal* - FirePro W7100
- Mobile FirePro
M7170
`` gfx810 `` - `` stoney `` `` amdgcn `` APU - xnack - Offset - *rocm-amdhsa* *TBA*
flat - *pal-amdhsa*
scratch - *pal-amdpal* .. TODO::
Add product
names.
2020-10-22 15:15:41 +08:00
2020-12-13 09:06:18 +08:00
**GCN GFX9 (Vega)** [AMD-GCN-GFX9]_
2020-12-17 10:45:47 +08:00
-----------------------------------------------------------------------------------------------------------------------
`` gfx900 `` `` amdgcn `` dGPU - xnack - Absolute - *rocm-amdhsa* - Radeon Vega
flat - *pal-amdhsa* Frontier Edition
scratch - *pal-amdpal* - Radeon RX Vega 56
- Radeon RX Vega 64
- Radeon RX Vega 64
Liquid
- Radeon Instinct MI25
`` gfx902 `` `` amdgcn `` APU - xnack - Absolute - *rocm-amdhsa* - Ryzen 3 2200G
flat - *pal-amdhsa* - Ryzen 5 2400G
scratch - *pal-amdpal*
`` gfx904 `` `` amdgcn `` dGPU - xnack - *rocm-amdhsa* *TBA*
- *pal-amdhsa*
- *pal-amdpal* .. TODO::
Add product
names.
`` gfx906 `` `` amdgcn `` dGPU - sramecc - Absolute - *rocm-amdhsa* - Radeon Instinct MI50
- xnack flat - *pal-amdhsa* - Radeon Instinct MI60
scratch - *pal-amdpal* - Radeon VII
- Radeon Pro VII
2021-03-26 10:01:18 +08:00
`` gfx908 `` `` amdgcn `` dGPU - sramecc - *rocm-amdhsa* - Radeon Instinct MI100 Accelerator
2020-12-17 10:45:47 +08:00
- xnack - Absolute
2021-03-26 10:01:18 +08:00
flat
2020-12-17 10:45:47 +08:00
scratch
`` gfx909 `` `` amdgcn `` APU - xnack - Absolute - *pal-amdpal* *TBA*
flat
scratch .. TODO::
Add product
names.
2021-02-18 05:37:46 +08:00
`` gfx90a `` `` amdgcn `` dGPU - sramecc - Absolute - *rocm-amdhsa* *TBA*
- tgsplit flat
- xnack scratch .. TODO::
- Packed
work-item Add product
IDs names.
2020-12-17 10:45:47 +08:00
`` gfx90c `` `` amdgcn `` APU - xnack - Absolute - *pal-amdpal* - Ryzen 7 4700G
flat - Ryzen 7 4700GE
scratch - Ryzen 5 4600G
- Ryzen 5 4600GE
- Ryzen 3 4300G
- Ryzen 3 4300GE
- Ryzen Pro 4000G
- Ryzen 7 Pro 4700G
- Ryzen 7 Pro 4750GE
- Ryzen 5 Pro 4650G
- Ryzen 5 Pro 4650GE
- Ryzen 3 Pro 4350G
- Ryzen 3 Pro 4350GE
2020-10-07 01:23:59 +08:00
2020-12-13 09:06:18 +08:00
**GCN GFX10 (RDNA 1)** [AMD-GCN-GFX10-RDNA1]_
2020-12-17 10:45:47 +08:00
-----------------------------------------------------------------------------------------------------------------------
`` gfx1010 `` `` amdgcn `` dGPU - cumode - Absolute - *rocm-amdhsa* - Radeon RX 5700
- wavefrontsize64 flat - *pal-amdhsa* - Radeon RX 5700 XT
- xnack scratch - *pal-amdpal* - Radeon Pro 5600 XT
- Radeon Pro 5600M
2021-03-26 10:01:18 +08:00
`` gfx1011 `` `` amdgcn `` dGPU - cumode - *rocm-amdhsa* - Radeon Pro V520
2020-12-17 10:45:47 +08:00
- wavefrontsize64 - Absolute - *pal-amdhsa*
- xnack flat - *pal-amdpal*
2021-03-26 10:01:18 +08:00
scratch
2020-12-17 10:45:47 +08:00
`` gfx1012 `` `` amdgcn `` dGPU - cumode - Absolute - *rocm-amdhsa* - Radeon RX 5500
- wavefrontsize64 flat - *pal-amdhsa* - Radeon RX 5500 XT
- xnack scratch - *pal-amdpal*
2020-12-13 09:06:18 +08:00
**GCN GFX10 (RDNA 2)** [AMD-GCN-GFX10-RDNA2]_
2020-12-17 10:45:47 +08:00
-----------------------------------------------------------------------------------------------------------------------
2021-03-26 10:01:18 +08:00
`` gfx1030 `` `` amdgcn `` dGPU - cumode - Absolute - *rocm-amdhsa* - Radeon RX 6800
- wavefrontsize64 flat - *pal-amdhsa* - Radeon RX 6800 XT
scratch - *pal-amdpal* - Radeon RX 6900 XT
`` gfx1031 `` `` amdgcn `` dGPU - cumode - Absolute - *rocm-amdhsa* - Radeon RX 6700 XT
2020-12-17 10:45:47 +08:00
- wavefrontsize64 flat - *pal-amdhsa*
2021-03-26 10:01:18 +08:00
scratch - *pal-amdpal*
2020-12-17 10:45:47 +08:00
`` gfx1032 `` `` amdgcn `` dGPU - cumode - Absolute - *rocm-amdhsa* *TBA*
- wavefrontsize64 flat - *pal-amdhsa*
scratch - *pal-amdpal* .. TODO::
2020-10-22 15:15:41 +08:00
2020-12-17 10:45:47 +08:00
Add product
names.
2020-10-22 15:15:41 +08:00
2020-12-17 10:45:47 +08:00
`` gfx1033 `` `` amdgcn `` APU - cumode - Absolute - *pal-amdpal* *TBA*
- wavefrontsize64 flat
scratch .. TODO::
2020-12-01 12:12:04 +08:00
2020-12-17 10:45:47 +08:00
Add product
names.
2020-12-01 12:12:04 +08:00
2020-12-17 10:45:47 +08:00
=========== =============== ============ ===== ================= =============== =============== ======================
2017-11-10 09:00:54 +08:00
.. _amdgpu-target-features:
Target Features
---------------
Target features control how code is generated to support certain
2017-12-11 23:35:27 +08:00
processor specific features. Not all target features are supported by
all processors. The runtime must ensure that the features supported by
the device used to execute the code match the features enabled when
generating the code. A mismatch of features may result in incorrect
execution, or a reduction in performance.
2020-12-01 12:12:04 +08:00
The target features supported by each processor is listed in
2017-12-11 23:35:27 +08:00
:ref: `amdgpu-processor-table` .
2017-11-10 09:00:54 +08:00
2020-12-01 12:12:04 +08:00
Target features are controlled by exactly one of the following Clang
options:
`` -mcpu=<target-id> `` or `` --offload-arch=<target-id> ``
The `` -mcpu `` and `` --offload-arch `` can specify the target feature as
optional components of the target ID. If omitted, the target feature has the
`` any `` value. See :ref: `amdgpu-target-id` .
`` -m[no-]<target-feature> ``
Target features not specified by the target ID are specified using a
separate option. These target features can have an `` on `` or `` off ``
value. `` on `` is specified by omitting the `` no- `` prefix, and
`` off `` is specified by including the `` no- `` prefix. The default
if not specified is `` off `` .
2017-11-10 09:00:54 +08:00
For example:
2020-12-01 12:12:04 +08:00
`` -mcpu=gfx908:xnack+ ``
2017-12-11 23:35:27 +08:00
Enable the `` xnack `` feature.
2020-12-01 12:12:04 +08:00
`` -mcpu=gfx908:xnack- ``
2017-12-11 23:35:27 +08:00
Disable the `` xnack `` feature.
2020-12-01 12:12:04 +08:00
`` -mcumode ``
Enable the `` cumode `` feature.
`` -mno-cumode ``
Disable the `` cumode `` feature.
2017-11-10 09:00:54 +08:00
.. table :: AMDGPU Target Features
2020-12-01 12:12:04 +08:00
:name: amdgpu-target-features-table
=============== ============================ ==================================================
Target Feature Clang Option to Control Description
Name
=============== ============================ ==================================================
cumode - `` -m[no-]cumode `` Control the wavefront execution mode used
when generating code for kernels. When disabled
native WGP wavefront execution mode is used,
when enabled CU wavefront execution mode is used
(see :ref: `amdgpu-amdhsa-memory-model` ).
sramecc - `` -mcpu `` If specified, generate code that can only be
- `` --offload-arch `` loaded and executed in a process that has a
matching setting for SRAMECC.
2021-02-27 08:39:29 +08:00
If not specified for code object V2 to V3, generate
code that can be loaded and executed in a process
with SRAMECC enabled.
If not specified for code object V4, generate
code that can be loaded and executed in a process
with either setting of SRAMECC.
2020-12-01 12:12:04 +08:00
2021-02-18 05:37:46 +08:00
tgsplit `` -m[no-]tgsplit `` Enable/disable generating code that assumes
work-groups are launched in threadgroup split mode.
When enabled the waves of a work-group may be
launched in different CUs.
2020-12-01 12:12:04 +08:00
wavefrontsize64 - `` -m[no-]wavefrontsize64 `` Control the wavefront size used when
generating code for kernels. When disabled
native wavefront size 32 is used, when enabled
wavefront size 64 is used.
xnack - `` -mcpu `` If specified, generate code that can only be
- `` --offload-arch `` loaded and executed in a process that has a
matching setting for XNACK replay.
2021-02-27 08:39:29 +08:00
If not specified for code object V2 to V3, generate
code that can be loaded and executed in a process
with XNACK replay enabled.
If not specified for code object V4, generate
code that can be loaded and executed in a process
with either setting of XNACK replay.
XNACK replay can be used for demand paging and
page migration. If enabled in the device, then if
a page fault occurs the code may execute
incorrectly unless generated with XNACK replay
enabled, or generated for code object V4 without
specifying XNACK replay. Executing code that was
generated with XNACK replay enabled, or generated
for code object V4 without specifying XNACK replay,
on a device that does not have XNACK replay
enabled will execute correctly but may be less
performant than code generated for XNACK replay
disabled.
2020-12-01 12:12:04 +08:00
=============== ============================ ==================================================
.. _amdgpu-target-id:
Target ID
---------
AMDGPU supports target IDs. See `Clang Offload Bundler
2020-12-02 11:04:28 +08:00
<https://clang.llvm.org/docs/ClangOffloadBundler.html> `_ for a general
description. The AMDGPU target specific information is:
2020-12-01 12:12:04 +08:00
**processor**
2021-02-28 02:09:23 +08:00
Is an AMDGPU processor or alternative processor name specified in
2020-12-01 12:12:04 +08:00
:ref: `amdgpu-processor-table` . The non-canonical form target ID allows both
the primary processor and alternative processor names. The canonical form
target ID only allow the primary processor name.
**target-feature**
Is a target feature name specified in :ref: `amdgpu-target-features-table` that
is supported by the processor. The target features supported by each processor
2021-01-20 11:38:55 +08:00
is specified in :ref: `amdgpu-processor-table` . Those that can be specified in
2020-12-01 12:12:04 +08:00
a target ID are marked as being controlled by `` -mcpu `` and
`` --offload-arch `` . Each target feature must appear at most once in a target
ID. The non-canonical form target ID allows the target features to be
specified in any order. The canonical form target ID requires the target
features to be specified in alphabetic order.
2021-01-20 11:38:55 +08:00
.. _amdgpu-target-id-v2-v3:
Code Object V2 to V3 Target ID
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
The target ID syntax for code object V2 to V3 is the same as defined in `Clang
Offload Bundler <https://clang.llvm.org/docs/ClangOffloadBundler.html>`_ except
when used in the :ref: `amdgpu-assembler-directive-amdgcn-target` assembler
directive and the bundle entry ID. In those cases it has the following BNF
syntax:
.. code ::
<target-id> ::== <processor> ( "+" <target-feature> )*
Where a target feature is omitted if *Off* and present if *On* or *Any* .
.. note ::
The code object V2 to V3 cannot represent *Any* and treats it the same as
*On* .
2020-12-01 12:12:04 +08:00
.. _amdgpu-embedding-bundled-objects:
Embedding Bundled Code Objects
------------------------------
AMDGPU supports the HIP and OpenMP languages that perform code object embedding
as described in `Clang Offload Bundler
2020-12-02 11:04:28 +08:00
<https://clang.llvm.org/docs/ClangOffloadBundler.html> `_.
2017-06-07 04:31:59 +08:00
2021-01-20 11:38:55 +08:00
.. note ::
The target ID syntax used for code object V2 to V3 for a bundle entry ID
differs from that used elsewhere. See :ref: `amdgpu-target-id-v2-v3` .
2017-06-07 04:31:59 +08:00
.. _amdgpu-address-spaces:
2016-04-06 09:29:19 +08:00
Address Spaces
--------------
2019-12-12 13:47:56 +08:00
The AMDGPU architecture supports a number of memory address spaces. The address
space names use the OpenCL standard names, with some additions.
2020-02-20 15:08:59 +08:00
The AMDGPU address spaces correspond to target architecture specific LLVM
address space numbers used in LLVM IR.
2019-12-12 13:47:56 +08:00
The AMDGPU address spaces are described in
:ref: `amdgpu-address-spaces-table` . Only 64-bit process address spaces are
supported for the `` amdgcn `` target.
.. table :: AMDGPU Address Spaces
:name: amdgpu-address-spaces-table
================================= =============== =========== ================ ======= ============================
.. 64-Bit Process Address Space
--------------------------------- --------------- ----------- ---------------- ------------------------------------
Address Space Name LLVM IR Address HSA Segment Hardware Address NULL Value
Space Number Name Name Size
================================= =============== =========== ================ ======= ============================
Generic 0 flat flat 64 0x0000000000000000
Global 1 global global 64 0x0000000000000000
Region 2 N/A GDS 32 *not implemented for AMDHSA*
Local 3 group LDS 32 0xFFFFFFFF
Constant 4 constant *same as global* 64 0x0000000000000000
2020-06-19 05:24:12 +08:00
Private 5 private scratch 32 0xFFFFFFFF
Constant 32-bit 6 *TODO* 0x00000000
2019-12-12 13:47:56 +08:00
Buffer Fat Pointer (experimental) 7 *TODO*
================================= =============== =========== ================ ======= ============================
**Generic**
2020-12-01 12:12:04 +08:00
The generic address space is supported unless the *Target Properties* column
of :ref: `amdgpu-processor-table` specifies *Does not support generic address
space*.
The generic address space uses the hardware flat address support for two fixed
ranges of virtual addresses (the private and local apertures), that are
outside the range of addressable global memory, to map from a flat address to
a private or local address. This uses FLAT instructions that can take a flat
address and access global, private (scratch), and group (LDS) memory depending
on if the address is within one of the aperture ranges.
Flat access to scratch requires hardware aperture setup and setup in the
kernel prologue (see :ref: `amdgpu-amdhsa-kernel-prolog-flat-scratch` ). Flat
access to LDS requires hardware aperture setup and M0 (GFX7-GFX8) register
setup (see :ref: `amdgpu-amdhsa-kernel-prolog-m0` ).
2019-12-12 13:47:56 +08:00
To convert between a private or group address space address (termed a segment
address) and a flat address the base address of the corresponding aperture
can be used. For GFX7-GFX8 these are available in the
:ref: `amdgpu-amdhsa-hsa-aql-queue` the address of which can be obtained with
Queue Ptr SGPR (see :ref: `amdgpu-amdhsa-initial-kernel-execution-state` ). For
GFX9-GFX10 the aperture base addresses are directly available as inline
constant registers `` SRC_SHARED_BASE/LIMIT `` and `` SRC_PRIVATE_BASE/LIMIT `` .
In 64-bit address mode the aperture sizes are 2^32 bytes and the base is
aligned to 2^32 which makes it easier to convert from flat to segment or
segment to flat.
A global address space address has the same value when used as a flat address
so no conversion is needed.
**Global and Constant**
The global and constant address spaces both use global virtual addresses,
which are the same virtual address space used by the CPU. However, some
virtual addresses may only be accessible to the CPU, some only accessible
by the GPU, and some by both.
Using the constant address space indicates that the data will not change
during the execution of the kernel. This allows scalar read instructions to
2020-11-07 19:47:57 +08:00
be used. As the constant address space could only be modified on the host
side, a generic pointer loaded from the constant address space is safe to be
assumed as a global pointer since only the device global memory is visible
and managed on the host side. The vector and scalar L1 caches are invalidated
of volatile data before each kernel dispatch execution to allow constant
memory to change values between kernel dispatches.
2019-12-12 13:47:56 +08:00
**Region**
The region address space uses the hardware Global Data Store (GDS). All
wavefronts executing on the same device will access the same memory for any
given region address. However, the same region address accessed by wavefronts
executing on different devices will access different memory. It is higher
performance than global memory. It is allocated by the runtime. The data
store (DS) instructions can be used to access it.
**Local**
The local address space uses the hardware Local Data Store (LDS) which is
automatically allocated when the hardware creates the wavefronts of a
work-group, and freed when all the wavefronts of a work-group have
terminated. All wavefronts belonging to the same work-group will access the
same memory for any given local address. However, the same local address
accessed by wavefronts belonging to different work-groups will access
different memory. It is higher performance than global memory. The data store
(DS) instructions can be used to access it.
**Private**
The private address space uses the hardware scratch memory support which
2020-02-20 15:08:59 +08:00
automatically allocates memory when it creates a wavefront and frees it when
2019-12-12 13:47:56 +08:00
a wavefronts terminates. The memory accessed by a lane of a wavefront for any
given private address will be different to the memory accessed by another lane
of the same or different wavefront for the same private address.
If a kernel dispatch uses scratch, then the hardware allocates memory from a
pool of backing memory allocated by the runtime for each wavefront. The lanes
of the wavefront access this using dword (4 byte) interleaving. The mapping
used from private address to backing memory address is:
`` wavefront-scratch-base +
((private-address / 4) * wavefront-size * 4) +
(wavefront-lane-id * 4) + (private-address % 4)``
If each lane of a wavefront accesses the same private address, the
interleaving results in adjacent dwords being accessed and hence requires
fewer cache lines to be fetched.
There are different ways that the wavefront scratch base address is
determined by a wavefront (see
:ref: `amdgpu-amdhsa-initial-kernel-execution-state` ).
Scratch memory can be accessed in an interleaved manner using buffer
instructions with the scratch buffer descriptor and per wavefront scratch
offset, by the scratch instructions, or by flat instructions. Multi-dword
access is not supported except by flat and scratch instructions in
GFX9-GFX10.
**Constant 32-bit**
*TODO*
**Buffer Fat Pointer**
The buffer fat pointer is an experimental address space that is currently
unsupported in the backend. It exposes a non-integral pointer that is in
the future intended to support the modelling of 128-bit buffer descriptors
plus a 32-bit offset into the buffer (in total encapsulating a 160-bit
*pointer* ), allowing normal LLVM load/store/atomic operations to be used to
model the buffer descriptors used heavily in graphics workloads targeting
the backend.
2017-06-07 04:31:59 +08:00
.. _amdgpu-memory-scopes:
Memory Scopes
-------------
This section provides LLVM memory synchronization scopes supported by the AMDGPU
backend memory model when the target triple OS is `` amdhsa `` (see
:ref: `amdgpu-amdhsa-memory-model` and :ref: `amdgpu-target-triples` ).
The memory model supported is based on the HSA memory model [HSA]_ which is
based in turn on HRF-indirect with scope inclusion [HRF]_ . The happens-before
2020-02-20 15:08:59 +08:00
relation is transitive over the synchronizes-with relation independent of scope
2019-12-12 13:47:56 +08:00
and synchronizes-with allows the memory scope instances to be inclusive (see
2017-11-10 09:00:54 +08:00
table :ref: `amdgpu-amdhsa-llvm-sync-scopes-table` ).
2017-06-07 04:31:59 +08:00
This is different to the OpenCL [OpenCL]_ memory model which does not have scope
inclusion and requires the memory scopes to exactly match. However, this
is conservatively correct for OpenCL.
2017-11-10 09:00:54 +08:00
.. table :: AMDHSA LLVM Sync Scopes
:name: amdgpu-amdhsa-llvm-sync-scopes-table
2017-06-07 04:31:59 +08:00
2019-03-26 04:50:21 +08:00
======================= ===================================================
LLVM Sync Scope Description
======================= ===================================================
*none* The default: `` system `` .
Synchronizes with, and participates in modification
and seq_cst total orderings with, other operations
(except image operations) for all address spaces
(except private, or generic that accesses private)
provided the other operation's sync scope is:
- `` system `` .
- `` agent `` and executed by a thread on the same
agent.
- `` workgroup `` and executed by a thread in the
[AMDGPU] AMDGPUUsage define call convention ABI
Reviewers: scott.linder, arsenm, b-sumner
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, dstuttard, tpr, kerbowa, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D74861
2020-02-13 14:19:25 +08:00
same work-group.
2019-03-26 04:50:21 +08:00
- `` wavefront `` and executed by a thread in the
same wavefront.
`` agent `` Synchronizes with, and participates in modification
and seq_cst total orderings with, other operations
(except image operations) for all address spaces
(except private, or generic that accesses private)
provided the other operation's sync scope is:
- `` system `` or `` agent `` and executed by a thread
on the same agent.
- `` workgroup `` and executed by a thread in the
[AMDGPU] AMDGPUUsage define call convention ABI
Reviewers: scott.linder, arsenm, b-sumner
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, dstuttard, tpr, kerbowa, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D74861
2020-02-13 14:19:25 +08:00
same work-group.
2019-03-26 04:50:21 +08:00
- `` wavefront `` and executed by a thread in the
same wavefront.
`` workgroup `` Synchronizes with, and participates in modification
and seq_cst total orderings with, other operations
(except image operations) for all address spaces
(except private, or generic that accesses private)
provided the other operation's sync scope is:
- `` system `` , `` agent `` or `` workgroup `` and
[AMDGPU] AMDGPUUsage define call convention ABI
Reviewers: scott.linder, arsenm, b-sumner
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, dstuttard, tpr, kerbowa, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D74861
2020-02-13 14:19:25 +08:00
executed by a thread in the same work-group.
2019-03-26 04:50:21 +08:00
- `` wavefront `` and executed by a thread in the
same wavefront.
`` wavefront `` Synchronizes with, and participates in modification
and seq_cst total orderings with, other operations
(except image operations) for all address spaces
(except private, or generic that accesses private)
provided the other operation's sync scope is:
- `` system `` , `` agent `` , `` workgroup `` or
`` wavefront `` and executed by a thread in the
same wavefront.
2020-02-20 15:08:59 +08:00
`` singlethread `` Only synchronizes with and participates in
2019-03-26 04:50:21 +08:00
modification and seq_cst total orderings with,
other operations (except image operations) running
in the same thread for all address spaces (for
example, in signal handlers).
`` one-as `` Same as `` system `` but only synchronizes with other
operations within the same address space.
`` agent-one-as `` Same as `` agent `` but only synchronizes with other
operations within the same address space.
`` workgroup-one-as `` Same as `` workgroup `` but only synchronizes with
other operations within the same address space.
`` wavefront-one-as `` Same as `` wavefront `` but only synchronizes with
other operations within the same address space.
`` singlethread-one-as `` Same as `` singlethread `` but only synchronizes with
other operations within the same address space.
======================= ===================================================
2017-06-07 04:31:59 +08:00
2020-02-20 15:08:59 +08:00
LLVM IR Intrinsics
------------------
2017-06-07 04:31:59 +08:00
2018-06-15 00:40:10 +08:00
The AMDGPU backend implements the following LLVM IR intrinsics.
2017-06-07 04:31:59 +08:00
*This section is WIP.*
2019-12-12 13:47:56 +08:00
.. TODO ::
List AMDGPU intrinsics.
2017-06-07 04:31:59 +08:00
2020-02-20 15:08:59 +08:00
LLVM IR Attributes
------------------
2018-06-15 00:40:10 +08:00
The AMDGPU backend supports the following LLVM IR attributes.
.. table :: AMDGPU LLVM IR Attributes
:name: amdgpu-llvm-ir-attributes-table
======================================= ==========================================================
LLVM Attribute Description
======================================= ==========================================================
"amdgpu-flat-work-group-size"="min,max" Specify the minimum and maximum flat work group sizes that
will be specified when the kernel is dispatched. Generated
by the `` amdgpu_flat_work_group_size `` CLANG attribute [CLANG-ATTR]_ .
"amdgpu-implicitarg-num-bytes"="n" Number of kernel argument bytes to add to the kernel
argument block size for the implicit arguments. This
varies by OS and language (for OpenCL see
:ref: `opencl-kernel-implicit-arguments-appended-for-amdhsa-os-table` ).
"amdgpu-num-sgpr"="n" Specifies the number of SGPRs to use. Generated by
the `` amdgpu_num_sgpr `` CLANG attribute [CLANG-ATTR]_ .
"amdgpu-num-vgpr"="n" Specifies the number of VGPRs to use. Generated by the
`` amdgpu_num_vgpr `` CLANG attribute [CLANG-ATTR]_ .
"amdgpu-waves-per-eu"="m,n" Specify the minimum and maximum number of waves per
execution unit. Generated by the `` amdgpu_waves_per_eu ``
CLANG attribute [CLANG-ATTR]_ .
2019-04-06 02:06:42 +08:00
"amdgpu-ieee" true/false. Specify whether the function expects the IEEE field of the
mode register to be set on entry. Overrides the default for
the calling convention.
"amdgpu-dx10-clamp" true/false. Specify whether the function expects the DX10_CLAMP field of
the mode register to be set on entry. Overrides the default
for the calling convention.
2018-06-15 00:40:10 +08:00
======================================= ==========================================================
2020-02-20 15:08:59 +08:00
.. _amdgpu-elf-code-object:
ELF Code Object
===============
2017-06-07 04:31:59 +08:00
The AMDGPU backend generates a standard ELF [ELF]_ relocatable code object that
can be linked by `` lld `` to produce a standard ELF shared code object which can
be loaded and executed on an AMDGPU target.
2020-02-20 15:08:59 +08:00
.. _amdgpu-elf-header:
2017-06-07 04:31:59 +08:00
Header
------
The AMDGPU backend uses the following ELF header:
.. table :: AMDGPU ELF Header
:name: amdgpu-elf-header-table
2017-10-04 04:54:07 +08:00
========================== ===============================
2017-06-07 04:31:59 +08:00
Field Value
2017-10-04 04:54:07 +08:00
========================== ===============================
2017-06-07 04:31:59 +08:00
`` e_ident[EI_CLASS] `` `` ELFCLASS64 ``
`` e_ident[EI_DATA] `` `` ELFDATA2LSB ``
2017-11-10 09:00:54 +08:00
`` e_ident[EI_OSABI] `` - `` ELFOSABI_NONE ``
- `` ELFOSABI_AMDGPU_HSA ``
- `` ELFOSABI_AMDGPU_PAL ``
- `` ELFOSABI_AMDGPU_MESA3D ``
2020-12-01 12:12:04 +08:00
`` e_ident[EI_ABIVERSION] `` - `` ELFABIVERSION_AMDGPU_HSA_V2 ``
- `` ELFABIVERSION_AMDGPU_HSA_V3 ``
- `` ELFABIVERSION_AMDGPU_HSA_V4 ``
2017-11-10 09:00:54 +08:00
- `` ELFABIVERSION_AMDGPU_PAL ``
- `` ELFABIVERSION_AMDGPU_MESA3D ``
`` e_type `` - `` ET_REL ``
- `` ET_DYN ``
2017-06-07 04:31:59 +08:00
`` e_machine `` `` EM_AMDGPU ``
`` e_entry `` 0
2020-12-01 12:12:04 +08:00
`` e_flags `` See :ref: `amdgpu-elf-header-e_flags-v2-table` ,
:ref: `amdgpu-elf-header-e_flags-table-v3` ,
and :ref: `amdgpu-elf-header-e_flags-table-v4`
2017-10-04 04:54:07 +08:00
========================== ===============================
2017-06-07 04:31:59 +08:00
..
.. table :: AMDGPU ELF Header Enumeration Values
:name: amdgpu-elf-header-enumeration-values-table
2017-10-04 05:14:14 +08:00
=============================== =====
Name Value
=============================== =====
`` EM_AMDGPU `` 224
2017-11-10 09:00:54 +08:00
`` ELFOSABI_NONE `` 0
2017-10-04 05:14:14 +08:00
`` ELFOSABI_AMDGPU_HSA `` 64
`` ELFOSABI_AMDGPU_PAL `` 65
`` ELFOSABI_AMDGPU_MESA3D `` 66
2020-12-01 12:12:04 +08:00
`` ELFABIVERSION_AMDGPU_HSA_V2 `` 0
`` ELFABIVERSION_AMDGPU_HSA_V3 `` 1
`` ELFABIVERSION_AMDGPU_HSA_V4 `` 2
2017-10-04 05:14:14 +08:00
`` ELFABIVERSION_AMDGPU_PAL `` 0
`` ELFABIVERSION_AMDGPU_MESA3D `` 0
=============================== =====
2017-06-07 04:31:59 +08:00
`` e_ident[EI_CLASS] ``
2017-11-10 09:00:54 +08:00
The ELF class is:
* `` ELFCLASS32 `` for `` r600 `` architecture.
2019-12-12 13:47:56 +08:00
* `` ELFCLASS64 `` for `` amdgcn `` architecture which only supports 64-bit
process address space applications.
2017-06-07 04:31:59 +08:00
`` e_ident[EI_DATA] ``
2017-11-10 09:00:54 +08:00
All AMDGPU targets use `` ELFDATA2LSB `` for little-endian byte ordering.
2017-06-07 04:31:59 +08:00
`` e_ident[EI_OSABI] ``
2020-02-20 15:08:59 +08:00
One of the following AMDGPU target architecture specific OS ABIs
2020-12-01 12:12:04 +08:00
(see :ref: `amdgpu-os` ):
2017-10-04 04:54:07 +08:00
2017-11-10 09:00:54 +08:00
* `` ELFOSABI_NONE `` for *unknown* OS.
2017-10-04 04:54:07 +08:00
2017-11-10 09:00:54 +08:00
* `` ELFOSABI_AMDGPU_HSA `` for `` amdhsa `` OS.
2017-06-07 04:31:59 +08:00
2017-11-10 09:00:54 +08:00
* `` ELFOSABI_AMDGPU_PAL `` for `` amdpal `` OS.
* `` ELFOSABI_AMDGPU_MESA3D `` for `` mesa3D `` OS.
2017-10-04 05:14:14 +08:00
2017-06-07 04:31:59 +08:00
`` e_ident[EI_ABIVERSION] ``
2020-02-20 15:08:59 +08:00
The ABI version of the AMDGPU target architecture specific OS ABI to which the code
2017-10-04 04:54:07 +08:00
object conforms:
2020-12-01 12:12:04 +08:00
* `` ELFABIVERSION_AMDGPU_HSA_V2 `` is used to specify the version of AMD HSA
runtime ABI for code object V2. Specify using the Clang option
`` -mcode-object-version=2 `` .
* `` ELFABIVERSION_AMDGPU_HSA_V3 `` is used to specify the version of AMD HSA
runtime ABI for code object V3. Specify using the Clang option
2020-12-18 00:09:32 +08:00
`` -mcode-object-version=3 `` . This is the default code object
version if not specified.
2020-12-01 12:12:04 +08:00
* `` ELFABIVERSION_AMDGPU_HSA_V4 `` is used to specify the version of AMD HSA
runtime ABI for code object V4. Specify using the Clang option
2020-12-18 00:09:32 +08:00
`` -mcode-object-version=4 `` .
2017-10-04 04:54:07 +08:00
* `` ELFABIVERSION_AMDGPU_PAL `` is used to specify the version of AMD PAL
runtime ABI.
2017-06-07 04:31:59 +08:00
2017-10-04 05:14:14 +08:00
* `` ELFABIVERSION_AMDGPU_MESA3D `` is used to specify the version of AMD MESA
2017-11-10 09:00:54 +08:00
3D runtime ABI.
2017-10-04 05:14:14 +08:00
2017-06-07 04:31:59 +08:00
`` e_type ``
Can be one of the following values:
`` ET_REL ``
2019-12-12 13:47:56 +08:00
The type produced by the AMDGPU backend compiler as it is relocatable code
2017-06-07 04:31:59 +08:00
object.
`` ET_DYN ``
The type produced by the linker as it is a shared code object.
The AMD HSA runtime loader requires a `` ET_DYN `` code object.
`` e_machine ``
2017-11-10 09:00:54 +08:00
The value `` EM_AMDGPU `` is used for the machine for all processors supported
by the `` r600 `` and `` amdgcn `` architectures (see
:ref: `amdgpu-processor-table` ). The specific processor is specified in the
2020-12-01 12:12:04 +08:00
`` NT_AMD_HSA_ISA_VERSION `` note record for code object V2 (see
:ref: `amdgpu-note-records-v2` ) and in the `` EF_AMDGPU_MACH `` bit field of the
`` e_flags `` for code object V3 to V4 (see
:ref: `amdgpu-elf-header-e_flags-table-v3` and
:ref: `amdgpu-elf-header-e_flags-table-v4` ).
2017-06-07 04:31:59 +08:00
`` e_entry ``
The entry point is 0 as the entry points for individual kernels must be
selected in order to invoke them through AQL packets.
`` e_flags ``
2017-11-10 09:00:54 +08:00
The AMDGPU backend uses the following ELF header flags:
2020-12-01 12:12:04 +08:00
.. table :: AMDGPU ELF Header `` e_flags `` for Code Object V2
:name: amdgpu-elf-header-e_flags-v2-table
===================================== ===== =============================
Name Value Description
===================================== ===== =============================
`` EF_AMDGPU_FEATURE_XNACK_V2 `` 0x01 Indicates if the `` xnack ``
target feature is
enabled for all code
contained in the code object.
If the processor
does not support the
`` xnack `` target
feature then must
be 0.
See
:ref: `amdgpu-target-features` .
`` EF_AMDGPU_FEATURE_TRAP_HANDLER_V2 `` 0x02 Indicates if the trap
handler is enabled for all
code contained in the code
object. If the processor
does not support a trap
handler then must be 0.
See
:ref: `amdgpu-target-features` .
===================================== ===== =============================
.. table :: AMDGPU ELF Header `` e_flags `` for Code Object V3
:name: amdgpu-elf-header-e_flags-table-v3
================================= ===== =============================
Name Value Description
================================= ===== =============================
`` EF_AMDGPU_MACH `` 0x0ff AMDGPU processor selection
mask for
`` EF_AMDGPU_MACH_xxx `` values
defined in
:ref: `amdgpu-ef-amdgpu-mach-table` .
`` EF_AMDGPU_FEATURE_XNACK_V3 `` 0x100 Indicates if the `` xnack ``
target feature is
enabled for all code
contained in the code object.
If the processor
does not support the
`` xnack `` target
feature then must
be 0.
See
:ref: `amdgpu-target-features` .
`` EF_AMDGPU_FEATURE_SRAMECC_V3 `` 0x200 Indicates if the `` sramecc ``
target feature is
enabled for all code
contained in the code object.
If the processor
does not support the
`` sramecc `` target
feature then must
be 0.
See
:ref: `amdgpu-target-features` .
================================= ===== =============================
.. table :: AMDGPU ELF Header `` e_flags `` for Code Object V4
:name: amdgpu-elf-header-e_flags-table-v4
============================================ ===== ===================================
Name Value Description
============================================ ===== ===================================
`` EF_AMDGPU_MACH `` 0x0ff AMDGPU processor selection
mask for
`` EF_AMDGPU_MACH_xxx `` values
defined in
:ref: `amdgpu-ef-amdgpu-mach-table` .
`` EF_AMDGPU_FEATURE_XNACK_V4 `` 0x300 XNACK selection mask for
`` EF_AMDGPU_FEATURE_XNACK_*_V4 ``
values.
`` EF_AMDGPU_FEATURE_XNACK_UNSUPPORTED_V4 `` 0x000 XNACK unsuppored.
`` EF_AMDGPU_FEATURE_XNACK_ANY_V4 `` 0x100 XNACK can have any value.
`` EF_AMDGPU_FEATURE_XNACK_OFF_V4 `` 0x200 XNACK disabled.
`` EF_AMDGPU_FEATURE_XNACK_ON_V4 `` 0x300 XNACK enabled.
`` EF_AMDGPU_FEATURE_SRAMECC_V4 `` 0xc00 SRAMECC selection mask for
`` EF_AMDGPU_FEATURE_SRAMECC_*_V4 ``
values.
`` EF_AMDGPU_FEATURE_SRAMECC_UNSUPPORTED_V4 `` 0x000 SRAMECC unsuppored.
`` EF_AMDGPU_FEATURE_SRAMECC_ANY_V4 `` 0x400 SRAMECC can have any value.
`` EF_AMDGPU_FEATURE_SRAMECC_OFF_V4 `` 0x800 SRAMECC disabled,
`` EF_AMDGPU_FEATURE_SRAMECC_ON_V4 `` 0xc00 SRAMECC enabled.
============================================ ===== ===================================
2017-11-10 09:00:54 +08:00
.. table :: AMDGPU `` EF_AMDGPU_MACH `` Values
:name: amdgpu-ef-amdgpu-mach-table
2020-10-22 15:15:41 +08:00
==================================== ========== =============================
Name Value Description (see
:ref: `amdgpu-processor-table` )
==================================== ========== =============================
`` EF_AMDGPU_MACH_NONE `` 0x000 *not specified*
`` EF_AMDGPU_MACH_R600_R600 `` 0x001 `` r600 ``
`` EF_AMDGPU_MACH_R600_R630 `` 0x002 `` r630 ``
`` EF_AMDGPU_MACH_R600_RS880 `` 0x003 `` rs880 ``
`` EF_AMDGPU_MACH_R600_RV670 `` 0x004 `` rv670 ``
`` EF_AMDGPU_MACH_R600_RV710 `` 0x005 `` rv710 ``
`` EF_AMDGPU_MACH_R600_RV730 `` 0x006 `` rv730 ``
`` EF_AMDGPU_MACH_R600_RV770 `` 0x007 `` rv770 ``
`` EF_AMDGPU_MACH_R600_CEDAR `` 0x008 `` cedar ``
`` EF_AMDGPU_MACH_R600_CYPRESS `` 0x009 `` cypress ``
`` EF_AMDGPU_MACH_R600_JUNIPER `` 0x00a `` juniper ``
`` EF_AMDGPU_MACH_R600_REDWOOD `` 0x00b `` redwood ``
`` EF_AMDGPU_MACH_R600_SUMO `` 0x00c `` sumo ``
`` EF_AMDGPU_MACH_R600_BARTS `` 0x00d `` barts ``
`` EF_AMDGPU_MACH_R600_CAICOS `` 0x00e `` caicos ``
`` EF_AMDGPU_MACH_R600_CAYMAN `` 0x00f `` cayman ``
`` EF_AMDGPU_MACH_R600_TURKS `` 0x010 `` turks ``
*reserved* 0x011 - Reserved for `` r600 ``
0x01f architecture processors.
`` EF_AMDGPU_MACH_AMDGCN_GFX600 `` 0x020 `` gfx600 ``
`` EF_AMDGPU_MACH_AMDGCN_GFX601 `` 0x021 `` gfx601 ``
`` EF_AMDGPU_MACH_AMDGCN_GFX700 `` 0x022 `` gfx700 ``
`` EF_AMDGPU_MACH_AMDGCN_GFX701 `` 0x023 `` gfx701 ``
`` EF_AMDGPU_MACH_AMDGCN_GFX702 `` 0x024 `` gfx702 ``
`` EF_AMDGPU_MACH_AMDGCN_GFX703 `` 0x025 `` gfx703 ``
`` EF_AMDGPU_MACH_AMDGCN_GFX704 `` 0x026 `` gfx704 ``
*reserved* 0x027 Reserved.
`` EF_AMDGPU_MACH_AMDGCN_GFX801 `` 0x028 `` gfx801 ``
`` EF_AMDGPU_MACH_AMDGCN_GFX802 `` 0x029 `` gfx802 ``
`` EF_AMDGPU_MACH_AMDGCN_GFX803 `` 0x02a `` gfx803 ``
`` EF_AMDGPU_MACH_AMDGCN_GFX810 `` 0x02b `` gfx810 ``
`` EF_AMDGPU_MACH_AMDGCN_GFX900 `` 0x02c `` gfx900 ``
`` EF_AMDGPU_MACH_AMDGCN_GFX902 `` 0x02d `` gfx902 ``
`` EF_AMDGPU_MACH_AMDGCN_GFX904 `` 0x02e `` gfx904 ``
`` EF_AMDGPU_MACH_AMDGCN_GFX906 `` 0x02f `` gfx906 ``
`` EF_AMDGPU_MACH_AMDGCN_GFX908 `` 0x030 `` gfx908 ``
`` EF_AMDGPU_MACH_AMDGCN_GFX909 `` 0x031 `` gfx909 ``
2020-10-07 01:23:59 +08:00
`` EF_AMDGPU_MACH_AMDGCN_GFX90C `` 0x032 `` gfx90c ``
2020-10-22 15:15:41 +08:00
`` EF_AMDGPU_MACH_AMDGCN_GFX1010 `` 0x033 `` gfx1010 ``
`` EF_AMDGPU_MACH_AMDGCN_GFX1011 `` 0x034 `` gfx1011 ``
`` EF_AMDGPU_MACH_AMDGCN_GFX1012 `` 0x035 `` gfx1012 ``
`` EF_AMDGPU_MACH_AMDGCN_GFX1030 `` 0x036 `` gfx1030 ``
`` EF_AMDGPU_MACH_AMDGCN_GFX1031 `` 0x037 `` gfx1031 ``
`` EF_AMDGPU_MACH_AMDGCN_GFX1032 `` 0x038 `` gfx1032 ``
2020-10-30 16:21:12 +08:00
`` EF_AMDGPU_MACH_AMDGCN_GFX1033 `` 0x039 `` gfx1033 ``
2020-10-22 15:15:41 +08:00
`` EF_AMDGPU_MACH_AMDGCN_GFX602 `` 0x03a `` gfx602 ``
`` EF_AMDGPU_MACH_AMDGCN_GFX705 `` 0x03b `` gfx705 ``
`` EF_AMDGPU_MACH_AMDGCN_GFX805 `` 0x03c `` gfx805 ``
2021-02-19 09:44:54 +08:00
*reserved* 0x03d Reserved.
*reserved* 0x03e Reserved.
`` EF_AMDGPU_MACH_AMDGCN_GFX90A `` 0x03f `` gfx90a ``
2021-03-23 06:28:11 +08:00
*reserved* 0x040 Reserved.
2021-03-23 23:27:51 +08:00
*reserved* 0x041 Reserved.
2020-10-22 15:15:41 +08:00
==================================== ========== =============================
2017-06-07 04:31:59 +08:00
Sections
--------
An AMDGPU target ELF code object has the standard ELF sections which include:
.. table :: AMDGPU ELF Sections
:name: amdgpu-elf-sections-table
================== ================ =================================
Name Type Attributes
================== ================ =================================
`` .bss `` `` SHT_NOBITS `` `` SHF_ALLOC `` + `` SHF_WRITE ``
`` .data `` `` SHT_PROGBITS `` `` SHF_ALLOC `` + `` SHF_WRITE ``
`` .debug_ ` ` \ *\** ` ` SHT_PROGBITS `` *none*
`` .dynamic `` `` SHT_DYNAMIC `` `` SHF_ALLOC ``
`` .dynstr `` `` SHT_PROGBITS `` `` SHF_ALLOC ``
`` .dynsym `` `` SHT_PROGBITS `` `` SHF_ALLOC ``
`` .got `` `` SHT_PROGBITS `` `` SHF_ALLOC `` + `` SHF_WRITE ``
`` .hash `` `` SHT_HASH `` `` SHF_ALLOC ``
`` .note `` `` SHT_NOTE `` *none*
`` .rela ` ` \ *name* ` ` SHT_RELA `` *none*
`` .rela.dyn `` `` SHT_RELA `` *none*
`` .rodata `` `` SHT_PROGBITS `` `` SHF_ALLOC ``
`` .shstrtab `` `` SHT_STRTAB `` *none*
`` .strtab `` `` SHT_STRTAB `` *none*
`` .symtab `` `` SHT_SYMTAB `` *none*
`` .text `` `` SHT_PROGBITS `` `` SHF_ALLOC `` + `` SHF_EXECINSTR ``
================== ================ =================================
These sections have their standard meanings (see [ELF]_ ) and are only generated
if needed.
`` .debug ` ` \ *\**
2020-02-20 15:08:59 +08:00
The standard DWARF sections. See :ref: `amdgpu-dwarf-debug-information` for
2020-04-14 12:55:43 +08:00
information on the DWARF produced by the AMDGPU backend.
2020-02-20 15:08:59 +08:00
2020-04-14 12:55:43 +08:00
`` .dynamic `` , `` .dynstr `` , `` .dynsym `` , `` .hash ``
The standard sections used by a dynamic loader.
2020-02-20 15:08:59 +08:00
2020-04-14 12:55:43 +08:00
`` .note ``
See :ref: `amdgpu-note-records` for the note records supported by the AMDGPU
backend.
2020-02-20 15:08:59 +08:00
2020-04-14 12:55:43 +08:00
`` .rela ` ` \ *name*, `` .rela.dyn``
For relocatable code objects, *name* is the name of the section that the
relocation records apply. For example, `` .rela.text `` is the section name for
relocation records associated with the `` .text `` section.
2020-02-20 15:08:59 +08:00
2020-04-14 12:55:43 +08:00
For linked shared code objects, `` .rela.dyn `` contains all the relocation
records from each of the relocatable code object's `` .rela ` ` \ *name* sections.
2020-02-20 15:08:59 +08:00
2020-04-14 12:55:43 +08:00
See :ref: `amdgpu-relocation-records` for the relocation records supported by
the AMDGPU backend.
2020-02-20 15:08:59 +08:00
2020-04-14 12:55:43 +08:00
`` .text ``
The executable machine code for the kernels and functions they call. Generated
as position independent code. See :ref: `amdgpu-code-conventions` for
information on conventions used in the isa generation.
2020-02-20 15:08:59 +08:00
2020-04-14 12:55:43 +08:00
.. _amdgpu-note-records:
2020-02-20 15:08:59 +08:00
2020-04-14 12:55:43 +08:00
Note Records
------------
2020-02-20 15:08:59 +08:00
2020-04-14 12:55:43 +08:00
The AMDGPU backend code object contains ELF note records in the `` .note ``
section. The set of generated notes and their semantics depend on the code
object version; see :ref: `amdgpu-note-records-v2` and
2020-12-01 12:12:04 +08:00
:ref: `amdgpu-note-records-v3-v4` .
2020-02-20 15:08:59 +08:00
2020-04-14 12:55:43 +08:00
As required by `` ELFCLASS32 `` and `` ELFCLASS64 `` , minimal zero-byte padding
must be generated after the `` name `` field to ensure the `` desc `` field is 4
byte aligned. In addition, minimal zero-byte padding must be generated to
ensure the `` desc `` field size is a multiple of 4 bytes. The `` sh_addralign ``
field of the `` .note `` section must be at least 4 to indicate at least 8 byte
alignment.
2020-02-20 15:08:59 +08:00
2020-04-14 12:55:43 +08:00
.. _amdgpu-note-records-v2:
2020-02-20 15:08:59 +08:00
2020-12-01 12:12:04 +08:00
Code Object V2 Note Records
~~~~~~~~~~~~~~~~~~~~~~~~~~~
2020-02-20 15:08:59 +08:00
2020-12-01 12:12:04 +08:00
.. warning ::
Code object V2 is not the default code object version emitted by
this version of LLVM.
2020-02-20 15:08:59 +08:00
2020-04-14 12:55:43 +08:00
The AMDGPU backend code object uses the following ELF note record in the
2020-12-01 12:12:04 +08:00
`` .note `` section when compiling for code object V2.
The note record vendor field is "AMD".
2020-02-20 15:08:59 +08:00
2020-04-14 12:55:43 +08:00
Additional note records may be present, but any which are not documented here
are deprecated and should not be used.
2020-02-20 15:08:59 +08:00
2020-04-14 12:55:43 +08:00
.. table :: AMDGPU Code Object V2 ELF Note Records
2020-12-01 12:12:04 +08:00
:name: amdgpu-elf-note-records-v2-table
===== ===================================== ======================================
Name Type Description
===== ===================================== ======================================
"AMD" `` NT_AMD_HSA_CODE_OBJECT_VERSION `` Code object version.
"AMD" `` NT_AMD_HSA_HSAIL `` HSAIL properties generated by the HSAIL
Finalizer and not the LLVM compiler.
"AMD" `` NT_AMD_HSA_ISA_VERSION `` Target ISA version.
"AMD" `` NT_AMD_HSA_METADATA `` Metadata null terminated string in
YAML [YAML]_ textual format.
"AMD" `` NT_AMD_HSA_ISA_NAME `` Target ISA name.
===== ===================================== ======================================
2020-02-20 15:08:59 +08:00
2020-04-14 12:55:43 +08:00
..
2020-02-20 15:08:59 +08:00
2020-04-14 12:55:43 +08:00
.. table :: AMDGPU Code Object V2 ELF Note Record Enumeration Values
2020-12-01 12:12:04 +08:00
:name: amdgpu-elf-note-record-enumeration-values-v2-table
===================================== =====
Name Value
===================================== =====
`` NT_AMD_HSA_CODE_OBJECT_VERSION `` 1
`` NT_AMD_HSA_HSAIL `` 2
`` NT_AMD_HSA_ISA_VERSION `` 3
*reserved* 4-9
`` NT_AMD_HSA_METADATA `` 10
`` NT_AMD_HSA_ISA_NAME `` 11
===================================== =====
`` NT_AMD_HSA_CODE_OBJECT_VERSION ``
Specifies the code object version number. The description field has the
following layout:
.. code ::
struct amdgpu_hsa_note_code_object_version_s {
uint32_t major_version;
uint32_t minor_version;
};
The `` major_version `` has a value less than or equal to 2.
`` NT_AMD_HSA_HSAIL ``
Specifies the HSAIL properties used by the HSAIL Finalizer. The description
field has the following layout:
.. code ::
struct amdgpu_hsa_note_hsail_s {
uint32_t hsail_major_version;
uint32_t hsail_minor_version;
uint8_t profile;
uint8_t machine_model;
uint8_t default_float_round;
};
`` NT_AMD_HSA_ISA_VERSION ``
Specifies the target ISA version. The description field has the following layout:
.. code ::
struct amdgpu_hsa_note_isa_s {
uint16_t vendor_name_size;
uint16_t architecture_name_size;
uint32_t major;
uint32_t minor;
uint32_t stepping;
char vendor_and_architecture_name[1];
};
`` vendor_name_size `` and `` architecture_name_size `` are the length of the
vendor and architecture names respectively, including the NUL character.
`` vendor_and_architecture_name `` contains the NUL terminates string for the
vendor, immediately followed by the NUL terminated string for the
architecture.
This note record is used by the HSA runtime loader.
Code object V2 only supports a limited number of processors and has fixed
settings for target features. See
:ref: `amdgpu-elf-note-record-supported_processors-v2-table` for a list of
processors and the corresponding target ID. In the table the note record ISA
name is a concatenation of the vendor name, architecture name, major, minor,
and stepping separated by a ":".
The target ID column shows the processor name and fixed target features used
by the LLVM compiler. The LLVM compiler does not generate a
`` NT_AMD_HSA_HSAIL `` note record.
A code object generated by the Finalizer also uses code object V2 and always
generates a `` NT_AMD_HSA_HSAIL `` note record. The processor name and
`` sramecc `` target feature is as shown in
:ref: `amdgpu-elf-note-record-supported_processors-v2-table` but the `` xnack ``
target feature is specified by the `` EF_AMDGPU_FEATURE_XNACK_V2 `` `` e_flags ``
bit.
`` NT_AMD_HSA_ISA_NAME ``
Specifies the target ISA name as a non-NUL terminated string.
This note record is not used by the HSA runtime loader.
See the `` NT_AMD_HSA_ISA_VERSION `` note record description of the code object
V2's limited support of processors and fixed settings for target features.
See :ref: `amdgpu-elf-note-record-supported_processors-v2-table` for a mapping
from the string to the corresponding target ID. If the `` xnack `` target
feature is supported and enabled, the string produced by the LLVM compiler
will may have a `` +xnack `` appended. The Finlizer did not do the appending and
instead used the `` EF_AMDGPU_FEATURE_XNACK_V2 `` `` e_flags `` bit.
`` NT_AMD_HSA_METADATA ``
2020-04-14 12:55:43 +08:00
Specifies extensible metadata associated with the code objects executed on HSA
2020-12-01 12:12:04 +08:00
[HSA]_ compatible runtimes (see :ref: `amdgpu-os` ). It is required when the
target triple OS is `` amdhsa `` (see :ref: `amdgpu-target-triples` ). See
:ref: `amdgpu-amdhsa-code-object-metadata-v2` for the syntax of the code object
metadata string.
.. table :: AMDGPU Code Object V2 Supported Processors and Fixed Target Feature Settings
:name: amdgpu-elf-note-record-supported_processors-v2-table
2021-04-09 04:42:05 +08:00
===================== ==========================
Note Record ISA Name Target ID
===================== ==========================
`` AMD:AMDGPU:6:0:0 `` `` gfx600 ``
`` AMD:AMDGPU:6:0:1 `` `` gfx601 ``
`` AMD:AMDGPU:6:0:2 `` `` gfx602 ``
`` AMD:AMDGPU:7:0:0 `` `` gfx700 ``
`` AMD:AMDGPU:7:0:1 `` `` gfx701 ``
`` AMD:AMDGPU:7:0:2 `` `` gfx702 ``
`` AMD:AMDGPU:7:0:3 `` `` gfx703 ``
`` AMD:AMDGPU:7:0:4 `` `` gfx704 ``
`` AMD:AMDGPU:7:0:5 `` `` gfx705 ``
`` AMD:AMDGPU:8:0:0 `` `` gfx802 ``
`` AMD:AMDGPU:8:0:1 `` `` gfx801:xnack+ ``
`` AMD:AMDGPU:8:0:2 `` `` gfx802 ``
`` AMD:AMDGPU:8:0:3 `` `` gfx803 ``
`` AMD:AMDGPU:8:0:4 `` `` gfx803 ``
`` AMD:AMDGPU:8:0:5 `` `` gfx805 ``
`` AMD:AMDGPU:8:1:0 `` `` gfx810:xnack+ ``
`` AMD:AMDGPU:9:0:0 `` `` gfx900:xnack- ``
`` AMD:AMDGPU:9:0:1 `` `` gfx900:xnack+ ``
`` AMD:AMDGPU:9:0:2 `` `` gfx902:xnack- ``
`` AMD:AMDGPU:9:0:3 `` `` gfx902:xnack+ ``
`` AMD:AMDGPU:9:0:4 `` `` gfx904:xnack- ``
`` AMD:AMDGPU:9:0:5 `` `` gfx904:xnack+ ``
`` AMD:AMDGPU:9:0:6 `` `` gfx906:sramecc-:xnack- ``
`` AMD:AMDGPU:9:0:7 `` `` gfx906:sramecc-:xnack+ ``
`` AMD:AMDGPU:9:0:12 `` `` gfx90c:xnack- ``
===================== ==========================
2020-12-01 12:12:04 +08:00
.. _amdgpu-note-records-v3-v4:
Code Object V3 to V4 Note Records
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2020-02-20 15:08:59 +08:00
2020-04-14 12:55:43 +08:00
The AMDGPU backend code object uses the following ELF note record in the
2020-12-01 12:12:04 +08:00
`` .note `` section when compiling for code object V3 to V4.
The note record vendor field is "AMDGPU".
2020-02-20 15:08:59 +08:00
2020-04-14 12:55:43 +08:00
Additional note records may be present, but any which are not documented here
are deprecated and should not be used.
2020-02-20 15:08:59 +08:00
2020-12-01 12:12:04 +08:00
.. table :: AMDGPU Code Object V3 to V4 ELF Note Records
:name: amdgpu-elf-note-records-table-v3-v4
2020-02-20 15:08:59 +08:00
2020-04-14 12:55:43 +08:00
======== ============================== ======================================
Name Type Description
======== ============================== ======================================
"AMDGPU" `` NT_AMDGPU_METADATA `` Metadata in Message Pack [MsgPack]_
binary format.
======== ============================== ======================================
2020-02-20 15:08:59 +08:00
2020-04-14 12:55:43 +08:00
..
2020-02-20 15:08:59 +08:00
2020-12-01 12:12:04 +08:00
.. table :: AMDGPU Code Object V3 to V4 ELF Note Record Enumeration Values
:name: amdgpu-elf-note-record-enumeration-values-table-v3-v4
2020-02-20 15:08:59 +08:00
2020-04-14 12:55:43 +08:00
============================== =====
Name Value
============================== =====
*reserved* 0-31
`` NT_AMDGPU_METADATA `` 32
============================== =====
2020-02-20 15:08:59 +08:00
2020-04-14 12:55:43 +08:00
`` NT_AMDGPU_METADATA ``
2020-12-01 12:12:04 +08:00
Specifies extensible metadata associated with an AMDGPU code object. It is
encoded as a map in the Message Pack [MsgPack]_ binary data format. See
:ref: `amdgpu-amdhsa-code-object-metadata-v3` and
:ref: `amdgpu-amdhsa-code-object-metadata-v4` for the map keys defined for the
`` amdhsa `` OS.
2020-02-20 15:08:59 +08:00
2020-04-14 12:55:43 +08:00
.. _amdgpu-symbols:
2020-02-20 15:08:59 +08:00
2020-04-14 12:55:43 +08:00
Symbols
-------
2020-02-20 15:08:59 +08:00
2020-04-14 12:55:43 +08:00
Symbols include the following:
2020-02-20 15:08:59 +08:00
2020-04-14 12:55:43 +08:00
.. table :: AMDGPU ELF Symbols
:name: amdgpu-elf-symbols-table
2020-02-20 15:08:59 +08:00
2020-04-14 12:55:43 +08:00
===================== ================== ================ ==================
Name Type Section Description
===================== ================== ================ ==================
*link-name* `` STT_OBJECT `` - `` .data `` Global variable
- `` .rodata ``
- `` .bss ``
*link-name* \ `` .kd `` `` STT_OBJECT `` - `` .rodata `` Kernel descriptor
*link-name* `` STT_FUNC `` - `` .text `` Kernel entry point
*link-name* `` STT_OBJECT `` - SHN_AMDGPU_LDS Global variable in LDS
===================== ================== ================ ==================
2020-02-20 15:08:59 +08:00
2020-04-14 12:55:43 +08:00
Global variable
Global variables both used and defined by the compilation unit.
2020-02-20 15:08:59 +08:00
2020-04-14 12:55:43 +08:00
If the symbol is defined in the compilation unit then it is allocated in the
appropriate section according to if it has initialized data or is readonly.
2020-02-20 15:08:59 +08:00
2020-04-14 12:55:43 +08:00
If the symbol is external then its section is `` STN_UNDEF `` and the loader
will resolve relocations using the definition provided by another code object
or explicitly defined by the runtime.
2020-02-20 15:08:59 +08:00
2020-04-14 12:55:43 +08:00
If the symbol resides in local/group memory (LDS) then its section is the
special processor specific section name `` SHN_AMDGPU_LDS `` , and the
`` st_value `` field describes alignment requirements as it does for common
symbols.
2020-02-20 15:08:59 +08:00
2020-04-14 12:55:43 +08:00
.. TODO ::
2020-02-20 15:08:59 +08:00
2020-04-14 12:55:43 +08:00
Add description of linked shared object symbols. Seems undefined symbols
are marked as STT_NOTYPE.
2020-02-20 15:08:59 +08:00
2020-04-14 12:55:43 +08:00
Kernel descriptor
Every HSA kernel has an associated kernel descriptor. It is the address of the
kernel descriptor that is used in the AQL dispatch packet used to invoke the
kernel, not the kernel entry point. The layout of the HSA kernel descriptor is
defined in :ref: `amdgpu-amdhsa-kernel-descriptor` .
2020-02-20 15:08:59 +08:00
2020-04-14 12:55:43 +08:00
Kernel entry point
Every HSA kernel also has a symbol for its machine code entry point.
2020-02-20 15:08:59 +08:00
2020-04-14 12:55:43 +08:00
.. _amdgpu-relocation-records:
2020-02-20 15:08:59 +08:00
2020-04-14 12:55:43 +08:00
Relocation Records
------------------
2020-02-20 15:08:59 +08:00
2020-04-14 12:55:43 +08:00
AMDGPU backend generates `` Elf64_Rela `` relocation records. Supported
relocatable fields are:
2020-02-20 15:08:59 +08:00
2020-04-14 12:55:43 +08:00
`` word32 ``
This specifies a 32-bit field occupying 4 bytes with arbitrary byte
alignment. These values use the same byte order as other word values in the
AMDGPU architecture.
2020-02-20 15:08:59 +08:00
2020-04-14 12:55:43 +08:00
`` word64 ``
This specifies a 64-bit field occupying 8 bytes with arbitrary byte
alignment. These values use the same byte order as other word values in the
AMDGPU architecture.
2020-02-20 15:08:59 +08:00
2020-04-14 12:55:43 +08:00
Following notations are used for specifying relocation calculations:
2020-02-20 15:08:59 +08:00
2020-04-14 12:55:43 +08:00
**A**
Represents the addend used to compute the value of the relocatable field.
2020-02-20 15:08:59 +08:00
2020-04-14 12:55:43 +08:00
**G**
Represents the offset into the global offset table at which the relocation
entry's symbol will reside during execution.
2020-02-20 15:08:59 +08:00
2020-04-14 12:55:43 +08:00
**GOT**
Represents the address of the global offset table.
2020-02-20 15:08:59 +08:00
2020-04-14 12:55:43 +08:00
**P**
Represents the place (section offset for `` et_rel `` or address for `` et_dyn `` )
of the storage unit being relocated (computed using `` r_offset `` ).
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
2020-04-14 12:55:43 +08:00
**S**
Represents the value of the symbol whose index resides in the relocation
entry. Relocations not using this must specify a symbol index of
`` STN_UNDEF `` .
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
2020-04-14 12:55:43 +08:00
**B**
Represents the base address of a loaded executable or shared object which is
the difference between the ELF address and the actual load address.
Relocations using this are only valid in executable or shared objects.
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
2020-04-14 12:55:43 +08:00
The following relocation types are supported:
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
2020-04-14 12:55:43 +08:00
.. table :: AMDGPU ELF Relocation Records
:name: amdgpu-elf-relocation-records-table
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
2020-04-14 12:55:43 +08:00
========================== ======= ===== ========== ==============================
Relocation Type Kind Value Field Calculation
========================== ======= ===== ========== ==============================
`` R_AMDGPU_NONE `` 0 *none* *none*
`` R_AMDGPU_ABS32_LO `` Static, 1 `` word32 `` (S + A) & 0xFFFFFFFF
Dynamic
`` R_AMDGPU_ABS32_HI `` Static, 2 `` word32 `` (S + A) >> 32
Dynamic
`` R_AMDGPU_ABS64 `` Static, 3 `` word64 `` S + A
Dynamic
`` R_AMDGPU_REL32 `` Static 4 `` word32 `` S + A - P
`` R_AMDGPU_REL64 `` Static 5 `` word64 `` S + A - P
`` R_AMDGPU_ABS32 `` Static, 6 `` word32 `` S + A
Dynamic
`` R_AMDGPU_GOTPCREL `` Static 7 `` word32 `` G + GOT + A - P
`` R_AMDGPU_GOTPCREL32_LO `` Static 8 `` word32 `` (G + GOT + A - P) & 0xFFFFFFFF
`` R_AMDGPU_GOTPCREL32_HI `` Static 9 `` word32 `` (G + GOT + A - P) >> 32
`` R_AMDGPU_REL32_LO `` Static 10 `` word32 `` (S + A - P) & 0xFFFFFFFF
`` R_AMDGPU_REL32_HI `` Static 11 `` word32 `` (S + A - P) >> 32
*reserved* 12
`` R_AMDGPU_RELATIVE64 `` Dynamic 13 `` word64 `` B + A
========================== ======= ===== ========== ==============================
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
2020-04-14 12:55:43 +08:00
`` R_AMDGPU_ABS32_LO `` and `` R_AMDGPU_ABS32_HI `` are only supported by
the `` mesa3d `` OS, which does not support `` R_AMDGPU_ABS64 `` .
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
2020-04-14 12:55:43 +08:00
There is no current OS loader support for 32-bit programs and so
`` R_AMDGPU_ABS32 `` is not used.
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
2020-05-30 07:03:50 +08:00
.. _amdgpu-loaded-code-object-path-uniform-resource-identifier:
Loaded Code Object Path Uniform Resource Identifier (URI)
---------------------------------------------------------
The AMD GPU code object loader represents the path of the ELF shared object from
which the code object was loaded as a textual Unifom Resource Identifier (URI).
Note that the code object is the in memory loaded relocated form of the ELF
shared object. Multiple code objects may be loaded at different memory
addresses in the same process from the same ELF shared object.
The loaded code object path URI syntax is defined by the following BNF syntax:
.. code ::
code_object_uri ::== file_uri | memory_uri
file_uri ::== "file://" file_path [ range_specifier ]
memory_uri ::== "memory://" process_id range_specifier
range_specifier ::== [ "#" | "?" ] "offset=" number "&" "size=" number
file_path ::== URI_ENCODED_OS_FILE_PATH
process_id ::== DECIMAL_NUMBER
number ::== HEX_NUMBER | DECIMAL_NUMBER | OCTAL_NUMBER
**number**
Is a C integral literal where hexadecimal values are prefixed by "0x" or "0X",
and octal values by "0".
**file_path**
Is the file's path specified as a URI encoded UTF-8 string. In URI encoding,
every character that is not in the regular expression `` [a-zA-Z0-9/_.~-] `` is
2020-08-10 10:29:38 +08:00
encoded as two uppercase hexadecimal digits proceeded by "%". Directories in
2020-05-30 07:03:50 +08:00
the path are separated by "/".
**offset**
Is a 0-based byte offset to the start of the code object. For a file URI, it
is from the start of the file specified by the `` file_path `` , and if omitted
defaults to 0. For a memory URI, it is the memory address and is required.
**size**
Is the number of bytes in the code object. For a file URI, if omitted it
defaults to the size of the file. It is required for a memory URI.
**process_id**
Is the identity of the process owning the memory. For Linux it is the C
unsigned integral decimal literal for the process ID (PID).
For example:
.. code ::
file:///dir1/dir2/file1
file:///dir3/dir4/file2#offset=0x2000&size=3000
memory://1234#offset=0x20000&size=3000
2020-02-20 15:08:59 +08:00
.. _amdgpu-dwarf-debug-information:
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
2020-02-20 15:08:59 +08:00
DWARF Debug Information
=======================
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
2020-02-20 15:08:59 +08:00
.. warning ::
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
2020-07-04 06:31:53 +08:00
This section describes **provisional support** for AMDGPU DWARF [DWARF]_ that
is not currently fully implemented and is subject to change.
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
2020-02-20 15:08:59 +08:00
AMDGPU generates DWARF [DWARF]_ debugging information ELF sections (see
:ref: `amdgpu-elf-code-object` ) which contain information that maps the code
object executable code and data to the source language constructs. It can be
2020-04-14 12:55:43 +08:00
used by tools such as debuggers and profilers. It uses features defined in
2020-07-04 06:31:53 +08:00
:doc: `AMDGPUDwarfExtensionsForHeterogeneousDebugging` that are made available in
2020-04-14 12:55:43 +08:00
DWARF Version 4 and DWARF Version 5 as an LLVM vendor extension.
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
2020-02-20 15:08:59 +08:00
This section defines the AMDGPU target architecture specific DWARF mappings.
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
2020-02-20 15:08:59 +08:00
.. _amdgpu-dwarf-register-identifier:
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
2020-02-20 15:08:59 +08:00
Register Identifier
-------------------
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
2020-02-20 15:08:59 +08:00
This section defines the AMDGPU target architecture register numbers used in
DWARF operation expressions (see DWARF Version 5 section 2.5 and
:ref: `amdgpu-dwarf-operation-expressions` ) and Call Frame Information
instructions (see DWARF Version 5 section 6.4 and
:ref: `amdgpu-dwarf-call-frame-information` ).
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
2020-02-20 15:08:59 +08:00
A single code object can contain code for kernels that have different wavefront
sizes. The vector registers and some scalar registers are based on the wavefront
size. AMDGPU defines distinct DWARF registers for each wavefront size. This
simplifies the consumer of the DWARF so that each register has a fixed size,
rather than being dynamic according to the wavefront size mode. Similarly,
distinct DWARF registers are defined for those registers that vary in size
according to the process address size. This allows a consumer to treat a
specific AMDGPU processor as a single architecture regardless of how it is
configured at run time. The compiler explicitly specifies the DWARF registers
that match the mode in which the code it is generating will be executed.
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
2020-02-20 15:08:59 +08:00
DWARF registers are encoded as numbers, which are mapped to architecture
registers. The mapping for AMDGPU is defined in
:ref: `amdgpu-dwarf-register-mapping-table` . All AMDGPU targets use the same
mapping.
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
2020-02-20 15:08:59 +08:00
.. table :: AMDGPU DWARF Register Mapping
:name: amdgpu-dwarf-register-mapping-table
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
2020-02-20 15:08:59 +08:00
============== ================= ======== ==================================
DWARF Register AMDGPU Register Bit Size Description
============== ================= ======== ==================================
0 PC_32 32 Program Counter (PC) when
executing in a 32-bit process
address space. Used in the CFI to
describe the PC of the calling
frame.
1 EXEC_MASK_32 32 Execution Mask Register when
executing in wavefront 32 mode.
[AMDGPU] Define DWARF encoding for condition code registers
Summary:
- Define DWARF register numbers for vector and scalar condition codes.
- Document intended purpose of reserved DWARF register numbers.
Reviewers: yaxunl, kzhuravl, arsenm, rampitec, b-sumner
Subscribers: jvesely, wdng, nhaehnle, aprantl, dstuttard, tpr, kerbowa, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D82519
2020-06-24 16:11:01 +08:00
2-15 *Reserved* *Reserved for highly accessed
registers using DWARF shortcut.*
2020-02-20 15:08:59 +08:00
16 PC_64 64 Program Counter (PC) when
executing in a 64-bit process
address space. Used in the CFI to
describe the PC of the calling
frame.
17 EXEC_MASK_64 64 Execution Mask Register when
executing in wavefront 64 mode.
[AMDGPU] Define DWARF encoding for condition code registers
Summary:
- Define DWARF register numbers for vector and scalar condition codes.
- Document intended purpose of reserved DWARF register numbers.
Reviewers: yaxunl, kzhuravl, arsenm, rampitec, b-sumner
Subscribers: jvesely, wdng, nhaehnle, aprantl, dstuttard, tpr, kerbowa, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D82519
2020-06-24 16:11:01 +08:00
18-31 *Reserved* *Reserved for highly accessed
registers using DWARF shortcut.*
2020-02-20 15:08:59 +08:00
32-95 SGPR0-SGPR63 32 Scalar General Purpose
Registers.
[AMDGPU] Define DWARF encoding for condition code registers
Summary:
- Define DWARF register numbers for vector and scalar condition codes.
- Document intended purpose of reserved DWARF register numbers.
Reviewers: yaxunl, kzhuravl, arsenm, rampitec, b-sumner
Subscribers: jvesely, wdng, nhaehnle, aprantl, dstuttard, tpr, kerbowa, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D82519
2020-06-24 16:11:01 +08:00
96-127 *Reserved* *Reserved for frequently accessed
registers using DWARF 1-byte ULEB.*
2020-08-20 09:09:12 +08:00
128 STATUS 32 Status Register.
[AMDGPU] Define DWARF encoding for condition code registers
Summary:
- Define DWARF register numbers for vector and scalar condition codes.
- Document intended purpose of reserved DWARF register numbers.
Reviewers: yaxunl, kzhuravl, arsenm, rampitec, b-sumner
Subscribers: jvesely, wdng, nhaehnle, aprantl, dstuttard, tpr, kerbowa, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D82519
2020-06-24 16:11:01 +08:00
129-511 *Reserved* *Reserved for future Scalar
Architectural Registers.*
512 VCC_32 32 Vector Condition Code Register
when executing in wavefront 32
mode.
513-1023 *Reserved* *Reserved for future Vector
Architectural Registers when
executing in wavefront 32 mode.*
2020-08-20 09:09:12 +08:00
768 VCC_64 64 Vector Condition Code Register
[AMDGPU] Define DWARF encoding for condition code registers
Summary:
- Define DWARF register numbers for vector and scalar condition codes.
- Document intended purpose of reserved DWARF register numbers.
Reviewers: yaxunl, kzhuravl, arsenm, rampitec, b-sumner
Subscribers: jvesely, wdng, nhaehnle, aprantl, dstuttard, tpr, kerbowa, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D82519
2020-06-24 16:11:01 +08:00
when executing in wavefront 64
mode.
769-1023 *Reserved* *Reserved for future Vector
Architectural Registers when
executing in wavefront 64 mode.*
1024-1087 *Reserved* *Reserved for padding.*
1088-1129 SGPR64-SGPR105 32 Scalar General Purpose Registers.
1130-1535 *Reserved* *Reserved for future Scalar
General Purpose Registers.*
2020-02-20 15:08:59 +08:00
1536-1791 VGPR0-VGPR255 32*32 Vector General Purpose Registers
when executing in wavefront 32
mode.
[AMDGPU] Define DWARF encoding for condition code registers
Summary:
- Define DWARF register numbers for vector and scalar condition codes.
- Document intended purpose of reserved DWARF register numbers.
Reviewers: yaxunl, kzhuravl, arsenm, rampitec, b-sumner
Subscribers: jvesely, wdng, nhaehnle, aprantl, dstuttard, tpr, kerbowa, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D82519
2020-06-24 16:11:01 +08:00
1792-2047 *Reserved* *Reserved for future Vector
General Purpose Registers when
executing in wavefront 32 mode.*
2020-02-20 15:08:59 +08:00
2048-2303 AGPR0-AGPR255 32*32 Vector Accumulation Registers
when executing in wavefront 32
[AMDGPU] Define DWARF encoding for condition code registers
Summary:
- Define DWARF register numbers for vector and scalar condition codes.
- Document intended purpose of reserved DWARF register numbers.
Reviewers: yaxunl, kzhuravl, arsenm, rampitec, b-sumner
Subscribers: jvesely, wdng, nhaehnle, aprantl, dstuttard, tpr, kerbowa, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D82519
2020-06-24 16:11:01 +08:00
mode.
2304-2559 *Reserved* *Reserved for future Vector
Accumulation Registers when
executing in wavefront 32 mode.*
2020-02-20 15:08:59 +08:00
2560-2815 VGPR0-VGPR255 64*32 Vector General Purpose Registers
when executing in wavefront 64
mode.
[AMDGPU] Define DWARF encoding for condition code registers
Summary:
- Define DWARF register numbers for vector and scalar condition codes.
- Document intended purpose of reserved DWARF register numbers.
Reviewers: yaxunl, kzhuravl, arsenm, rampitec, b-sumner
Subscribers: jvesely, wdng, nhaehnle, aprantl, dstuttard, tpr, kerbowa, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D82519
2020-06-24 16:11:01 +08:00
2816-3071 *Reserved* *Reserved for future Vector
General Purpose Registers when
executing in wavefront 64 mode.*
2020-02-20 15:08:59 +08:00
3072-3327 AGPR0-AGPR255 64*32 Vector Accumulation Registers
when executing in wavefront 64
mode.
[AMDGPU] Define DWARF encoding for condition code registers
Summary:
- Define DWARF register numbers for vector and scalar condition codes.
- Document intended purpose of reserved DWARF register numbers.
Reviewers: yaxunl, kzhuravl, arsenm, rampitec, b-sumner
Subscribers: jvesely, wdng, nhaehnle, aprantl, dstuttard, tpr, kerbowa, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D82519
2020-06-24 16:11:01 +08:00
3328-3583 *Reserved* *Reserved for future Vector
Accumulation Registers when
executing in wavefront 64 mode.*
2020-02-20 15:08:59 +08:00
============== ================= ======== ==================================
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
2020-02-20 15:08:59 +08:00
The vector registers are represented as the full size for the wavefront. They
are organized as consecutive dwords (32-bits), one per lane, with the dword at
the least significant bit position corresponding to lane 0 and so forth. DWARF
location expressions involving the `` DW_OP_LLVM_offset `` and
`` DW_OP_LLVM_push_lane `` operations are used to select the part of the vector
register corresponding to the lane that is executing the current thread of
execution in languages that are implemented using a SIMD or SIMT execution
model.
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
2020-02-20 15:08:59 +08:00
If the wavefront size is 32 lanes then the wavefront 32 mode register
definitions are used. If the wavefront size is 64 lanes then the wavefront 64
mode register definitions are used. Some AMDGPU targets support executing in
both wavefront 32 and wavefront 64 mode. The register definitions corresponding
to the wavefront mode of the generated code will be used.
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
2020-02-20 15:08:59 +08:00
If code is generated to execute in a 32-bit process address space, then the
32-bit process address space register definitions are used. If code is generated
to execute in a 64-bit process address space, then the 64-bit process address
space register definitions are used. The `` amdgcn `` target only supports the
64-bit process address space.
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
2020-02-20 15:08:59 +08:00
.. _amdgpu-dwarf-address-class-identifier:
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
2020-02-20 15:08:59 +08:00
Address Class Identifier
------------------------
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
2020-02-20 15:08:59 +08:00
The DWARF address class represents the source language memory space. See DWARF
2020-07-04 06:31:53 +08:00
Version 5 section 2.12 which is updated by the *DWARF Extensions For
Heterogeneous Debugging* section :ref: `amdgpu-dwarf-segment_addresses` .
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
2020-02-20 15:08:59 +08:00
The DWARF address class mapping used for AMDGPU is defined in
:ref: `amdgpu-dwarf-address-class-mapping-table` .
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
2020-02-20 15:08:59 +08:00
.. table :: AMDGPU DWARF Address Class Mapping
:name: amdgpu-dwarf-address-class-mapping-table
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
2020-02-20 15:08:59 +08:00
========================= ====== =================
DWARF AMDGPU
-------------------------------- -----------------
Address Class Name Value Address Space
========================= ====== =================
`` DW_ADDR_none `` 0x0000 Generic (Flat)
`` DW_ADDR_LLVM_global `` 0x0001 Global
`` DW_ADDR_LLVM_constant `` 0x0002 Global
`` DW_ADDR_LLVM_group `` 0x0003 Local (group/LDS)
`` DW_ADDR_LLVM_private `` 0x0004 Private (Scratch)
`` DW_ADDR_AMDGPU_region `` 0x8000 Region (GDS)
========================= ====== =================
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
2020-07-04 06:31:53 +08:00
The DWARF address class values defined in the *DWARF Extensions For
Heterogeneous Debugging* section :ref: `amdgpu-dwarf-segment_addresses` are used.
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
2020-02-20 15:08:59 +08:00
In addition, `` DW_ADDR_AMDGPU_region `` is encoded as a vendor extension. This is
available for use for the AMD extension for access to the hardware GDS memory
which is scratchpad memory allocated per device.
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
2020-02-20 15:08:59 +08:00
For AMDGPU if no `` DW_AT_address_class `` attribute is present, then the default
address class of `` DW_ADDR_none `` is used.
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
2020-02-20 15:08:59 +08:00
See :ref: `amdgpu-dwarf-address-space-identifier` for information on the AMDGPU
mapping of DWARF address classes to DWARF address spaces, including address size
and NULL value.
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
2020-02-20 15:08:59 +08:00
.. _amdgpu-dwarf-address-space-identifier:
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
2020-02-20 15:08:59 +08:00
Address Space Identifier
------------------------
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
2020-02-20 15:08:59 +08:00
DWARF address spaces correspond to target architecture specific linear
2020-07-04 06:31:53 +08:00
addressable memory areas. See DWARF Version 5 section 2.12 and *DWARF Extensions
For Heterogeneous Debugging* section :ref: `amdgpu-dwarf-segment_addresses` .
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
2020-02-20 15:08:59 +08:00
The DWARF address space mapping used for AMDGPU is defined in
:ref: `amdgpu-dwarf-address-space-mapping-table` .
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
2020-02-20 15:08:59 +08:00
.. table :: AMDGPU DWARF Address Space Mapping
:name: amdgpu-dwarf-address-space-mapping-table
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
2020-02-20 15:08:59 +08:00
======================================= ===== ======= ======== ================= =======================
DWARF AMDGPU Notes
--------------------------------------- ----- ---------------- ----------------- -----------------------
Address Space Name Value Address Bit Size Address Space
--------------------------------------- ----- ------- -------- ----------------- -----------------------
.. 64-bit 32-bit
process process
address address
space space
======================================= ===== ======= ======== ================= =======================
2020-10-29 06:20:51 +08:00
`` DW_ASPACE_none `` 0x00 64 32 Global *default address space*
`` DW_ASPACE_AMDGPU_generic `` 0x01 64 32 Generic (Flat)
`` DW_ASPACE_AMDGPU_region `` 0x02 32 32 Region (GDS)
`` DW_ASPACE_AMDGPU_local `` 0x03 32 32 Local (group/LDS)
2020-02-20 15:08:59 +08:00
*Reserved* 0x04
2020-10-29 06:20:51 +08:00
`` DW_ASPACE_AMDGPU_private_lane `` 0x05 32 32 Private (Scratch) *focused lane*
`` DW_ASPACE_AMDGPU_private_wave `` 0x06 32 32 Private (Scratch) *unswizzled wavefront*
2020-02-20 15:08:59 +08:00
======================================= ===== ======= ======== ================= =======================
See :ref: `amdgpu-address-spaces` for information on the AMDGPU address spaces
including address size and NULL value.
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
2020-02-20 15:08:59 +08:00
The `` DW_ASPACE_none `` address space is the default target architecture address
space used in DWARF operations that do not specify an address space. It
therefore has to map to the global address space so that the `` DW_OP_addr* `` and
related operations can refer to addresses in the program code.
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
2020-02-20 15:08:59 +08:00
The `` DW_ASPACE_AMDGPU_generic `` address space allows location expressions to
specify the flat address space. If the address corresponds to an address in the
local address space, then it corresponds to the wavefront that is executing the
focused thread of execution. If the address corresponds to an address in the
private address space, then it corresponds to the lane that is executing the
focused thread of execution for languages that are implemented using a SIMD or
SIMT execution model.
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
2020-02-20 15:08:59 +08:00
.. note ::
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
2020-02-20 15:08:59 +08:00
CUDA-like languages such as HIP that do not have address spaces in the
language type system, but do allow variables to be allocated in different
address spaces, need to explicitly specify the `` DW_ASPACE_AMDGPU_generic ``
address space in the DWARF expression operations as the default address space
is the global address space.
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
2020-02-20 15:08:59 +08:00
The `` DW_ASPACE_AMDGPU_local `` address space allows location expressions to
specify the local address space corresponding to the wavefront that is executing
the focused thread of execution.
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
2020-02-20 15:08:59 +08:00
The `` DW_ASPACE_AMDGPU_private_lane `` address space allows location expressions
to specify the private address space corresponding to the lane that is executing
the focused thread of execution for languages that are implemented using a SIMD
or SIMT execution model.
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
2020-02-20 15:08:59 +08:00
The `` DW_ASPACE_AMDGPU_private_wave `` address space allows location expressions
to specify the unswizzled private address space corresponding to the wavefront
that is executing the focused thread of execution. The wavefront view of private
memory is the per wavefront unswizzled backing memory layout defined in
:ref: `amdgpu-address-spaces` , such that address 0 corresponds to the first
location for the backing memory of the wavefront (namely the address is not
offset by `` wavefront-scratch-base `` ). The following formula can be used to
convert from a `` DW_ASPACE_AMDGPU_private_lane `` address to a
`` DW_ASPACE_AMDGPU_private_wave `` address:
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
2020-02-20 15:08:59 +08:00
::
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
2020-02-20 15:08:59 +08:00
private-address-wavefront =
((private-address-lane / 4) * wavefront-size * 4) +
(wavefront-lane-id * 4) + (private-address-lane % 4)
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
2020-02-20 15:08:59 +08:00
If the `` DW_ASPACE_AMDGPU_private_lane `` address is dword aligned, and the start
of the dwords for each lane starting with lane 0 is required, then this
simplifies to:
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
2020-02-20 15:08:59 +08:00
::
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
2020-02-20 15:08:59 +08:00
private-address-wavefront =
private-address-lane * wavefront-size
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
2020-02-20 15:08:59 +08:00
A compiler can use the `` DW_ASPACE_AMDGPU_private_wave `` address space to read a
complete spilled vector register back into a complete vector register in the
CFI. The frame pointer can be a private lane address which is dword aligned,
which can be shifted to multiply by the wavefront size, and then used to form a
private wavefront address that gives a location for a contiguous set of dwords,
one per lane, where the vector register dwords are spilled. The compiler knows
the wavefront size since it generates the code. Note that the type of the
address may have to be converted as the size of a
`` DW_ASPACE_AMDGPU_private_lane `` address may be smaller than the size of a
`` DW_ASPACE_AMDGPU_private_wave `` address.
.. _amdgpu-dwarf-lane-identifier:
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
2020-02-20 15:08:59 +08:00
Lane identifier
---------------
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
2020-02-20 15:08:59 +08:00
DWARF lane identifies specify a target architecture lane position for hardware
that executes in a SIMD or SIMT manner, and on which a source language maps its
threads of execution onto those lanes. The DWARF lane identifier is pushed by
the `` DW_OP_LLVM_push_lane `` DWARF expression operation. See DWARF Version 5
2020-07-04 06:31:53 +08:00
section 2.5 which is updated by *DWARF Extensions For Heterogeneous Debugging*
section :ref: `amdgpu-dwarf-operation-expressions` .
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
2020-02-20 15:08:59 +08:00
For AMDGPU, the lane identifier corresponds to the hardware lane ID of a
wavefront. It is numbered from 0 to the wavefront size minus 1.
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
2020-02-20 15:08:59 +08:00
Operation Expressions
---------------------
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
2020-02-20 15:08:59 +08:00
DWARF expressions are used to compute program values and the locations of
program objects. See DWARF Version 5 section 2.5 and
:ref: `amdgpu-dwarf-operation-expressions` .
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
2020-02-20 15:08:59 +08:00
DWARF location descriptions describe how to access storage which includes memory
and registers. When accessing storage on AMDGPU, bytes are ordered with least
significant bytes first, and bits are ordered within bytes with least
significant bits first.
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
2020-02-20 15:08:59 +08:00
For AMDGPU CFI expressions, `` DW_OP_LLVM_select_bit_piece `` is used to describe
unwinding vector registers that are spilled under the execution mask to memory:
the zero-single location description is the vector register, and the one-single
location description is the spilled memory location description. The
`` DW_OP_LLVM_form_aspace_address `` is used to specify the address space of the
memory location description.
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
2020-02-20 15:08:59 +08:00
In AMDGPU expressions, `` DW_OP_LLVM_select_bit_piece `` is used by the
`` DW_AT_LLVM_lane_pc `` attribute expression where divergent control flow is
controlled by the execution mask. An undefined location description together
with `` DW_OP_LLVM_extend `` is used to indicate the lane was not active on entry
to the subprogram. See :ref: `amdgpu-dwarf-dw-at-llvm-lane-pc` for an example.
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
2020-02-20 15:08:59 +08:00
Debugger Information Entry Attributes
-------------------------------------
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
2020-02-20 15:08:59 +08:00
This section describes how certain debugger information entry attributes are
used by AMDGPU. See the sections in DWARF Version 5 section 2 which are updated
2020-07-04 06:31:53 +08:00
by *DWARF Extensions For Heterogeneous Debugging* section
:ref: `amdgpu-dwarf-debugging-information-entry-attributes` .
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
2020-02-20 15:08:59 +08:00
.. _amdgpu-dwarf-dw-at-llvm-lane-pc:
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
2020-02-20 15:08:59 +08:00
`` DW_AT_LLVM_lane_pc ``
~~~~~~~~~~~~~~~~~~~~~~
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
2020-02-20 15:08:59 +08:00
For AMDGPU, the `` DW_AT_LLVM_lane_pc `` attribute is used to specify the program
location of the separate lanes of a SIMT thread.
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
If the lane is an active lane then this will be the same as the current program
location.
If the lane is inactive, but was active on entry to the subprogram, then this is
the program location in the subprogram at which execution of the lane is
conceptual positioned.
If the lane was not active on entry to the subprogram, then this will be the
undefined location. A client debugger can check if the lane is part of a valid
work-group by checking that the lane is in the range of the associated
2020-02-20 15:08:59 +08:00
work-group within the grid, accounting for partial work-groups. If it is not,
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
then the debugger can omit any information for the lane. Otherwise, the debugger
may repeatedly unwind the stack and inspect the `` DW_AT_LLVM_lane_pc `` of the
calling subprogram until it finds a non-undefined location. Conceptually the
lane only has the call frames that it has a non-undefined
`` DW_AT_LLVM_lane_pc `` .
2020-02-20 15:08:59 +08:00
The following example illustrates how the AMDGPU backend can generate a DWARF
location list expression for the nested `` IF/THEN/ELSE `` structures of the
following subprogram pseudo code for a target with 64 lanes per wavefront.
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
.. code ::
:number-lines:
SUBPROGRAM X
BEGIN
a;
IF (c1) THEN
b;
IF (c2) THEN
c;
ELSE
d;
ENDIF
e;
ELSE
f;
ENDIF
g;
END
The AMDGPU backend may generate the following pseudo LLVM MIR to manipulate the
2020-02-20 15:08:59 +08:00
execution mask (`` EXEC `` ) to linearize the control flow. The condition is
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
evaluated to make a mask of the lanes for which the condition evaluates to true.
First the `` THEN `` region is executed by setting the `` EXEC `` mask to the
logical `` AND `` of the current `` EXEC `` mask with the condition mask. Then the
`` ELSE `` region is executed by negating the `` EXEC `` mask and logical `` AND `` of
the saved `` EXEC `` mask at the start of the region. After the `` IF/THEN/ELSE ``
region the `` EXEC `` mask is restored to the value it had at the beginning of the
region. This is shown below. Other approaches are possible, but the basic
concept is the same.
.. code ::
:number-lines:
$lex_start:
a;
%1 = EXEC
%2 = c1
$lex_1_start:
EXEC = %1 & %2
$if_1_then:
b;
%3 = EXEC
%4 = c2
$lex_1_1_start:
EXEC = %3 & %4
$lex_1_1_then:
c;
EXEC = ~EXEC & %3
$lex_1_1_else:
d;
EXEC = %3
$lex_1_1_end:
e;
EXEC = ~EXEC & %1
$lex_1_else:
f;
EXEC = %1
$lex_1_end:
g;
$lex_end:
2020-02-20 15:08:59 +08:00
To create the DWARF location list expression that defines the location
description of a vector of lane program locations, the LLVM MIR `` DBG_VALUE ``
pseudo instruction can be used to annotate the linearized control flow. This can
be done by defining an artificial variable for the lane PC. The DWARF location
list expression created for it is used as the value of the
`` DW_AT_LLVM_lane_pc `` attribute on the subprogram's debugger information entry.
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
A DWARF procedure is defined for each well nested structured control flow region
which provides the conceptual lane program location for a lane if it is not
2020-02-20 15:08:59 +08:00
active (namely it is divergent). The DWARF operation expression for each region
conceptually inherits the value of the immediately enclosing region and modifies
it according to the semantics of the region.
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
For an `` IF/THEN/ELSE `` region the divergent program location is at the start of
the region for the `` THEN `` region since it is executed first. For the `` ELSE ``
region the divergent program location is at the end of the `` IF/THEN/ELSE ``
region since the `` THEN `` region has completed.
The lane PC artificial variable is assigned at each region transition. It uses
the immediately enclosing region's DWARF procedure to compute the program
location for each lane assuming they are divergent, and then modifies the result
by inserting the current program location for each lane that the `` EXEC `` mask
indicates is active.
By having separate DWARF procedures for each region, they can be reused to
2020-02-20 15:08:59 +08:00
define the value for any nested region. This reduces the total size of the DWARF
operation expressions.
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
The following provides an example using pseudo LLVM MIR.
.. code ::
:number-lines:
$lex_start:
DEFINE_DWARF %__uint_64 = DW_TAG_base_type[
DW_AT_name = "__uint64";
DW_AT_byte_size = 8;
DW_AT_encoding = DW_ATE_unsigned;
];
DEFINE_DWARF %__active_lane_pc = DW_TAG_dwarf_procedure[
DW_AT_name = "__active_lane_pc";
DW_AT_location = [
DW_OP_regx PC;
DW_OP_LLVM_extend 64, 64;
DW_OP_regval_type EXEC, %uint_64;
DW_OP_LLVM_select_bit_piece 64, 64;
];
];
DEFINE_DWARF %__divergent_lane_pc = DW_TAG_dwarf_procedure[
DW_AT_name = "__divergent_lane_pc";
DW_AT_location = [
DW_OP_LLVM_undefined;
DW_OP_LLVM_extend 64, 64;
];
];
DBG_VALUE $noreg, $noreg, %DW_AT_LLVM_lane_pc, DIExpression[
DW_OP_call_ref %__divergent_lane_pc;
DW_OP_call_ref %__active_lane_pc;
];
a;
%1 = EXEC;
DBG_VALUE %1, $noreg, %__lex_1_save_exec;
%2 = c1;
$lex_1_start:
EXEC = %1 & %2;
$lex_1_then:
DEFINE_DWARF %__divergent_lane_pc_1_then = DW_TAG_dwarf_procedure[
DW_AT_name = "__divergent_lane_pc_1_then";
DW_AT_location = DIExpression[
DW_OP_call_ref %__divergent_lane_pc;
2020-07-01 16:17:28 +08:00
DW_OP_addrx &lex_1_start;
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
DW_OP_stack_value;
DW_OP_LLVM_extend 64, 64;
DW_OP_call_ref %__lex_1_save_exec;
DW_OP_deref_type 64, %__uint_64;
DW_OP_LLVM_select_bit_piece 64, 64;
];
];
DBG_VALUE $noreg, $noreg, %DW_AT_LLVM_lane_pc, DIExpression[
DW_OP_call_ref %__divergent_lane_pc_1_then;
DW_OP_call_ref %__active_lane_pc;
];
b;
%3 = EXEC;
DBG_VALUE %3, %__lex_1_1_save_exec;
%4 = c2;
$lex_1_1_start:
EXEC = %3 & %4;
$lex_1_1_then:
DEFINE_DWARF %__divergent_lane_pc_1_1_then = DW_TAG_dwarf_procedure[
DW_AT_name = "__divergent_lane_pc_1_1_then";
DW_AT_location = DIExpression[
DW_OP_call_ref %__divergent_lane_pc_1_then;
2020-07-01 16:17:28 +08:00
DW_OP_addrx &lex_1_1_start;
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
DW_OP_stack_value;
DW_OP_LLVM_extend 64, 64;
DW_OP_call_ref %__lex_1_1_save_exec;
DW_OP_deref_type 64, %__uint_64;
DW_OP_LLVM_select_bit_piece 64, 64;
];
];
DBG_VALUE $noreg, $noreg, %DW_AT_LLVM_lane_pc, DIExpression[
DW_OP_call_ref %__divergent_lane_pc_1_1_then;
DW_OP_call_ref %__active_lane_pc;
];
c;
EXEC = ~EXEC & %3;
$lex_1_1_else:
DEFINE_DWARF %__divergent_lane_pc_1_1_else = DW_TAG_dwarf_procedure[
DW_AT_name = "__divergent_lane_pc_1_1_else";
DW_AT_location = DIExpression[
DW_OP_call_ref %__divergent_lane_pc_1_then;
2020-07-01 16:17:28 +08:00
DW_OP_addrx &lex_1_1_end;
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
DW_OP_stack_value;
DW_OP_LLVM_extend 64, 64;
DW_OP_call_ref %__lex_1_1_save_exec;
DW_OP_deref_type 64, %__uint_64;
DW_OP_LLVM_select_bit_piece 64, 64;
];
];
DBG_VALUE $noreg, $noreg, %DW_AT_LLVM_lane_pc, DIExpression[
DW_OP_call_ref %__divergent_lane_pc_1_1_else;
DW_OP_call_ref %__active_lane_pc;
];
d;
EXEC = %3;
$lex_1_1_end:
DBG_VALUE $noreg, $noreg, %DW_AT_LLVM_lane_pc, DIExpression[
DW_OP_call_ref %__divergent_lane_pc;
DW_OP_call_ref %__active_lane_pc;
];
e;
EXEC = ~EXEC & %1;
$lex_1_else:
DEFINE_DWARF %__divergent_lane_pc_1_else = DW_TAG_dwarf_procedure[
DW_AT_name = "__divergent_lane_pc_1_else";
DW_AT_location = DIExpression[
DW_OP_call_ref %__divergent_lane_pc;
2020-07-01 16:17:28 +08:00
DW_OP_addrx &lex_1_end;
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
DW_OP_stack_value;
DW_OP_LLVM_extend 64, 64;
DW_OP_call_ref %__lex_1_save_exec;
DW_OP_deref_type 64, %__uint_64;
DW_OP_LLVM_select_bit_piece 64, 64;
];
];
DBG_VALUE $noreg, $noreg, %DW_AT_LLVM_lane_pc, DIExpression[
DW_OP_call_ref %__divergent_lane_pc_1_else;
DW_OP_call_ref %__active_lane_pc;
];
f;
EXEC = %1;
$lex_1_end:
DBG_VALUE $noreg, $noreg, %DW_AT_LLVM_lane_pc DIExpression[
DW_OP_call_ref %__divergent_lane_pc;
DW_OP_call_ref %__active_lane_pc;
];
g;
$lex_end:
The DWARF procedure `` %__active_lane_pc `` is used to update the lane pc elements
2020-02-20 15:08:59 +08:00
that are active, with the current program location.
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
Artificial variables %__lex_1_save_exec and %__lex_1_1_save_exec are created for
the execution masks saved on entry to a region. Using the `` DBG_VALUE `` pseudo
2020-02-20 15:08:59 +08:00
instruction, location list entries will be created that describe where the
artificial variables are allocated at any given program location. The compiler
may allocate them to registers or spill them to memory.
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
2020-02-20 15:08:59 +08:00
The DWARF procedures for each region use the values of the saved execution mask
artificial variables to only update the lanes that are active on entry to the
region. All other lanes retain the value of the enclosing region where they were
last active. If they were not active on entry to the subprogram, then will have
the undefined location description.
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
Other structured control flow regions can be handled similarly. For example,
loops would set the divergent program location for the region at the end of the
loop. Any lanes active will be in the loop, and any lanes not active must have
exited the loop.
An `` IF/THEN/ELSEIF/ELSEIF/... `` region can be treated as a nest of
`` IF/THEN/ELSE `` regions.
The DWARF procedures can use the active lane artificial variable described in
:ref: `amdgpu-dwarf-amdgpu-dw-at-llvm-active-lane` rather than the actual
2020-02-20 15:08:59 +08:00
`` EXEC `` mask in order to support whole or quad wavefront mode.
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
.. _amdgpu-dwarf-amdgpu-dw-at-llvm-active-lane:
2020-02-20 15:08:59 +08:00
`` DW_AT_LLVM_active_lane ``
~~~~~~~~~~~~~~~~~~~~~~~~~~
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
2020-02-20 15:08:59 +08:00
The `` DW_AT_LLVM_active_lane `` attribute on a subprogram debugger information
entry is used to specify the lanes that are conceptually active for a SIMT
thread.
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
2020-02-20 15:08:59 +08:00
The execution mask may be modified to implement whole or quad wavefront mode
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
operations. For example, all lanes may need to temporarily be made active to
2020-02-20 15:08:59 +08:00
execute a whole wavefront operation. Such regions would save the `` EXEC `` mask,
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
update it to enable the necessary lanes, perform the operations, and then
2020-02-20 15:08:59 +08:00
restore the `` EXEC `` mask from the saved value. While executing the whole
wavefront region, the conceptual execution mask is the saved value, not the
`` EXEC `` value.
[AMDGPU] Update AMDGPUUsage with DWARF proposal
Summary:
- Add AMDGPU DWARF proposal.
- Add references for gfx10 ISA and SemVer.
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, aprantl, dstuttard, tpr, jfb, dmgreen, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D70523
2020-02-19 12:02:02 +08:00
This is handled by defining an artificial variable for the active lane mask. The
active lane mask artificial variable would be the actual `` EXEC `` mask for
normal regions, and the saved execution mask for regions where the mask is
2020-02-20 15:08:59 +08:00
temporarily updated. The location list expression created for this artificial
variable is used to define the value of the `` DW_AT_LLVM_active_lane ``
attribute.
2017-08-16 04:47:41 +08:00
2020-02-20 15:08:59 +08:00
`` DW_AT_LLVM_augmentation ``
~~~~~~~~~~~~~~~~~~~~~~~~~~~
2017-08-16 04:47:41 +08:00
2020-02-20 15:08:59 +08:00
For AMDGPU, the `` DW_AT_LLVM_augmentation `` attribute of a compilation unit
debugger information entry has the following value for the augmentation string:
2017-08-16 04:47:41 +08:00
2020-02-20 15:08:59 +08:00
::
2018-02-24 07:01:06 +08:00
2020-02-20 15:08:59 +08:00
[amdgpu:v0.0]
The "vX.Y" specifies the major X and minor Y version number of the AMDGPU
extensions used in the DWARF of the compilation unit. The version number
conforms to [SEMVER]_ .
Call Frame Information
----------------------
DWARF Call Frame Information (CFI) describes how a consumer can virtually
*unwind* call frames in a running process or core dump. See DWARF Version 5
section 6.4 and :ref: `amdgpu-dwarf-call-frame-information` .
For AMDGPU, the Common Information Entry (CIE) fields have the following values:
1. `` augmentation `` string contains the following null-terminated UTF-8 string:
::
[amd:v0.0]
The `` vX.Y `` specifies the major X and minor Y version number of the AMDGPU
extensions used in this CIE or to the FDEs that use it. The version number
conforms to [SEMVER]_ .
2. `` address_size `` for the `` Global `` address space is defined in
:ref: `amdgpu-dwarf-address-space-identifier` .
3. `` segment_selector_size `` is 0 as AMDGPU does not use a segment selector.
4. `` code_alignment_factor `` is 4 bytes.
2018-02-24 07:01:06 +08:00
2020-02-20 15:08:59 +08:00
.. TODO ::
Add to :ref: `amdgpu-processor-table` table.
5. `` data_alignment_factor `` is 4 bytes.
.. TODO ::
Add to :ref: `amdgpu-processor-table` table.
6. `` return_address_register `` is `` PC_32 `` for 32-bit processes and `` PC_64 ``
for 64-bit processes defined in :ref: `amdgpu-dwarf-register-identifier` .
7. `` initial_instructions `` Since a subprogram X with fewer registers can be
called from subprogram Y that has more allocated, X will not change any of
the extra registers as it cannot access them. Therefore, the default rule
for all columns is `` same value `` .
For AMDGPU the register number follows the numbering defined in
:ref: `amdgpu-dwarf-register-identifier` .
For AMDGPU the instructions are variable size. A consumer can subtract 1 from
the return address to get the address of a byte within the call site
instructions. See DWARF Version 5 section 6.4.4.
Accelerated Access
------------------
See DWARF Version 5 section 6.1.
Lookup By Name Section Header
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
See DWARF Version 5 section 6.1.1.4.1 and :ref: `amdgpu-dwarf-lookup-by-name` .
For AMDGPU the lookup by name section header table:
`` augmentation_string_size `` (uword)
Set to the length of the `` augmentation_string `` value which is always a
multiple of 4.
`` augmentation_string `` (sequence of UTF-8 characters)
Contains the following UTF-8 string null padded to a multiple of 4 bytes:
::
[amdgpu:v0.0]
The "vX.Y" specifies the major X and minor Y version number of the AMDGPU
extensions used in the DWARF of this index. The version number conforms to
[SEMVER]_ .
.. note ::
This is different to the DWARF Version 5 definition that requires the first
4 characters to be the vendor ID. But this is consistent with the other
augmentation strings and does allow multiple vendor contributions. However,
backwards compatibility may be more desirable.
Lookup By Address Section Header
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
See DWARF Version 5 section 6.1.2.
For AMDGPU the lookup by address section header table:
`` address_size `` (ubyte)
Match the address size for the `` Global `` address space defined in
:ref: `amdgpu-dwarf-address-space-identifier` .
`` segment_selector_size `` (ubyte)
AMDGPU does not use a segment selector so this is 0. The entries in the
`` .debug_aranges `` do not have a segment selector.
Line Number Information
-----------------------
See DWARF Version 5 section 6.2 and :ref: `amdgpu-dwarf-line-number-information` .
AMDGPU does not use the `` isa `` state machine registers and always sets it to 0.
The instruction set must be obtained from the ELF file header `` e_flags `` field
in the `` EF_AMDGPU_MACH `` bit position (see :ref:`ELF Header
<amdgpu-elf-header> `). See DWARF Version 5 section 6.2.2.
.. TODO ::
Should the `` isa `` state machine register be used to indicate if the code is
in wavefront32 or wavefront64 mode? Or used to specify the architecture ISA?
For AMDGPU the line number program header fields have the following values (see
DWARF Version 5 section 6.2.4):
`` address_size `` (ubyte)
Matches the address size for the `` Global `` address space defined in
:ref: `amdgpu-dwarf-address-space-identifier` .
`` segment_selector_size `` (ubyte)
AMDGPU does not use a segment selector so this is 0.
`` minimum_instruction_length `` (ubyte)
For GFX9-GFX10 this is 4.
`` maximum_operations_per_instruction `` (ubyte)
For GFX9-GFX10 this is 1.
Source text for online-compiled programs (for example, those compiled by the
OpenCL language runtime) may be embedded into the DWARF Version 5 line table.
2020-07-04 06:31:53 +08:00
See DWARF Version 5 section 6.2.4.1 which is updated by *DWARF Extensions For
Heterogeneous Debugging* section :ref:`DW_LNCT_LLVM_source
2020-02-20 15:08:59 +08:00
<amdgpu-dwarf-line-number-information-dw-lnct-llvm-source> `.
The Clang option used to control source embedding in AMDGPU is defined in
:ref: `amdgpu-clang-debug-options-table` .
.. table :: AMDGPU Clang Debug Options
:name: amdgpu-clang-debug-options-table
2018-02-24 07:01:06 +08:00
==================== ==================================================
Debug Flag Description
==================== ==================================================
-g[no-]embed-source Enable/disable embedding source text in DWARF
debug sections. Useful for environments where
source cannot be written to disk, such as
when performing online compilation.
==================== ==================================================
2020-02-20 15:08:59 +08:00
For example:
`` -gembed-source ``
Enable the embedded source.
`` -gno-embed-source ``
Disable the embedded source.
32-Bit and 64-Bit DWARF Formats
-------------------------------
See DWARF Version 5 section 7.4 and
:ref: `amdgpu-dwarf-32-bit-and-64-bit-dwarf-formats` .
2018-02-24 07:01:06 +08:00
2020-02-20 15:08:59 +08:00
For AMDGPU:
2018-02-24 07:01:06 +08:00
2020-02-20 15:08:59 +08:00
* For the `` amdgcn `` target architecture only the 64-bit process address space
is supported.
2018-02-24 07:01:06 +08:00
2020-02-20 15:08:59 +08:00
* The producer can generate either 32-bit or 64-bit DWARF format. LLVM generates
the 32-bit DWARF format.
2018-02-24 07:01:06 +08:00
2020-02-20 15:08:59 +08:00
Unit Headers
------------
2018-02-24 07:01:06 +08:00
2020-02-20 15:08:59 +08:00
For AMDGPU the following values apply for each of the unit headers described in
DWARF Version 5 sections 7.5.1.1, 7.5.1.2, and 7.5.1.3:
2018-02-24 07:01:06 +08:00
2020-02-20 15:08:59 +08:00
`` address_size `` (ubyte)
Matches the address size for the `` Global `` address space defined in
:ref: `amdgpu-dwarf-address-space-identifier` .
2017-08-16 04:47:41 +08:00
.. _amdgpu-code-conventions:
Code Conventions
================
This section provides code conventions used for each supported target triple OS
(see :ref: `amdgpu-target-triples` ).
AMDHSA
------
This section provides code conventions used when the target triple OS is
`` amdhsa `` (see :ref: `amdgpu-target-triples` ).
2017-06-07 04:31:59 +08:00
2018-06-22 03:38:56 +08:00
.. _amdgpu-amdhsa-code-object-metadata:
2017-06-07 04:31:59 +08:00
Code Object Metadata
2017-08-16 04:47:41 +08:00
~~~~~~~~~~~~~~~~~~~~
2017-06-07 04:31:59 +08:00
2017-08-16 04:47:41 +08:00
The code object metadata specifies extensible metadata associated with the code
2020-12-01 12:12:04 +08:00
objects executed on HSA [HSA]_ compatible runtimes (see :ref: `amdgpu-os` ). The
encoding and semantics of this metadata depends on the code object version; see
:ref: `amdgpu-amdhsa-code-object-metadata-v2` ,
:ref: `amdgpu-amdhsa-code-object-metadata-v3` , and
:ref: `amdgpu-amdhsa-code-object-metadata-v4` .
2019-03-28 23:08:52 +08:00
Code object metadata is specified in a note record (see
:ref: `amdgpu-note-records` ) and is required when the target triple OS is
`` amdhsa `` (see :ref: `amdgpu-target-triples` ). It must contain the minimum
2020-12-01 12:12:04 +08:00
information necessary to support the HSA compatible runtime kernel queries. For
example, the segment sizes needed in a dispatch packet. In addition, a
high-level language runtime may require other information to be included. For
example, the AMD OpenCL runtime records kernel argument information.
2018-11-16 04:46:55 +08:00
.. _amdgpu-amdhsa-code-object-metadata-v2:
2020-12-01 12:12:04 +08:00
Code Object V2 Metadata
+++++++++++++++++++++++
2018-11-16 04:46:55 +08:00
2020-12-01 12:12:04 +08:00
.. warning ::
Code object V2 is not the default code object version emitted by this version
of LLVM.
2019-03-28 23:08:52 +08:00
2020-12-01 12:12:04 +08:00
Code object V2 metadata is specified by the `` NT_AMD_HSA_METADATA `` note record
(see :ref: `amdgpu-note-records-v2` ).
2017-06-07 04:31:59 +08:00
2017-06-26 10:45:39 +08:00
The metadata is specified as a YAML formatted string (see [YAML]_ and
2017-06-07 04:31:59 +08:00
:doc: `YamlIO` ).
2019-12-12 13:47:56 +08:00
.. TODO ::
Is the string null terminated? It probably should not if YAML allows it to
contain null characters, otherwise it should be.
2017-08-16 04:47:41 +08:00
2017-06-07 04:31:59 +08:00
The metadata is represented as a single YAML document comprised of the mapping
2020-12-01 12:12:04 +08:00
defined in table :ref: `amdgpu-amdhsa-code-object-metadata-map-v2-table` and
2017-06-07 04:31:59 +08:00
referenced tables.
For boolean values, the string values of `` false `` and `` true `` are used for
false and true respectively.
Additional information can be added to the mappings. To avoid conflicts, any
non-AMD key names should be prefixed by "*vendor-name* .".
2018-11-16 04:46:55 +08:00
.. table :: AMDHSA Code Object V2 Metadata Map
2020-12-01 12:12:04 +08:00
:name: amdgpu-amdhsa-code-object-metadata-map-v2-table
2017-06-07 04:31:59 +08:00
========== ============== ========= =======================================
String Key Value Type Required? Description
========== ============== ========= =======================================
"Version" sequence of Required - The first integer is the major
2 integers version. Currently 1.
- The second integer is the minor
version. Currently 0.
"Printf" sequence of Each string is encoded information
strings about a printf function call. The
encoded information is organized as
fields separated by colon (':'):
`` ID:N:S[0]:S[1]:...:S[N-1]:FormatString ``
where:
`` ID ``
2019-12-12 13:47:56 +08:00
A 32-bit integer as a unique id for
2017-06-07 04:31:59 +08:00
each printf function call
`` N ``
2019-12-12 13:47:56 +08:00
A 32-bit integer equal to the number
2017-06-07 04:31:59 +08:00
of arguments of printf function call
minus 1
`` S[i] `` (where i = 0, 1, ... , N-1)
2019-12-12 13:47:56 +08:00
32-bit integers for the size in bytes
2017-06-07 04:31:59 +08:00
of the i-th FormatString argument of
the printf function call
FormatString
The format string passed to the
printf function call.
"Kernels" sequence of Required Sequence of the mappings for each
mapping kernel in the code object. See
2020-12-01 12:12:04 +08:00
:ref: `amdgpu-amdhsa-code-object-kernel-metadata-map-v2-table`
2017-06-07 04:31:59 +08:00
for the definition of the mapping.
========== ============== ========= =======================================
..
2018-11-16 04:46:55 +08:00
.. table :: AMDHSA Code Object V2 Kernel Metadata Map
2020-12-01 12:12:04 +08:00
:name: amdgpu-amdhsa-code-object-kernel-metadata-map-v2-table
2017-06-07 04:31:59 +08:00
================= ============== ========= ================================
String Key Value Type Required? Description
================= ============== ========= ================================
"Name" string Required Source name of the kernel.
"SymbolName" string Required Name of the kernel
descriptor ELF symbol.
"Language" string Source language of the kernel.
Values include:
- "OpenCL C"
- "OpenCL C++"
- "HCC"
- "OpenMP"
"LanguageVersion" sequence of - The first integer is the major
2 integers version.
- The second integer is the
minor version.
"Attrs" mapping Mapping of kernel attributes.
See
2020-12-01 12:12:04 +08:00
:ref: `amdgpu-amdhsa-code-object-kernel-attribute-metadata-map-v2-table`
2017-06-07 04:31:59 +08:00
for the mapping definition.
2017-10-15 03:03:51 +08:00
"Args" sequence of Sequence of mappings of the
2017-06-07 04:31:59 +08:00
mapping kernel arguments. See
2020-12-01 12:12:04 +08:00
:ref: `amdgpu-amdhsa-code-object-kernel-argument-metadata-map-v2-table`
2017-06-07 04:31:59 +08:00
for the definition of the mapping.
"CodeProps" mapping Mapping of properties related to
the kernel code. See
2020-12-01 12:12:04 +08:00
:ref: `amdgpu-amdhsa-code-object-kernel-code-properties-metadata-map-v2-table`
2017-06-07 04:31:59 +08:00
for the mapping definition.
================= ============== ========= ================================
..
2018-11-16 04:46:55 +08:00
.. table :: AMDHSA Code Object V2 Kernel Attribute Metadata Map
2020-12-01 12:12:04 +08:00
:name: amdgpu-amdhsa-code-object-kernel-attribute-metadata-map-v2-table
2017-06-07 04:31:59 +08:00
=================== ============== ========= ==============================
String Key Value Type Required? Description
=================== ============== ========= ==============================
2018-01-31 07:07:10 +08:00
"ReqdWorkGroupSize" sequence of If not 0, 0, 0 then all values
3 integers must be >=1 and the dispatch
work-group size X, Y, Z must
correspond to the specified
values. Defaults to 0, 0, 0.
2017-06-07 04:31:59 +08:00
Corresponds to the OpenCL
`` reqd_work_group_size ``
attribute.
"WorkGroupSizeHint" sequence of The dispatch work-group size
3 integers X, Y, Z is likely to be the
specified values.
Corresponds to the OpenCL
`` work_group_size_hint ``
attribute.
"VecTypeHint" string The name of a scalar or vector
type.
Corresponds to the OpenCL
`` vec_type_hint `` attribute.
2017-10-11 03:39:48 +08:00
"RuntimeHandle" string The external symbol name
associated with a kernel.
OpenCL runtime allocates a
global buffer for the symbol
and saves the kernel's address
to it, which is used for
device side enqueueing. Only
available for device side
enqueued kernels.
2017-06-07 04:31:59 +08:00
=================== ============== ========= ==============================
..
2018-11-16 04:46:55 +08:00
.. table :: AMDHSA Code Object V2 Kernel Argument Metadata Map
2020-12-01 12:12:04 +08:00
:name: amdgpu-amdhsa-code-object-kernel-argument-metadata-map-v2-table
2017-06-07 04:31:59 +08:00
================= ============== ========= ================================
String Key Value Type Required? Description
================= ============== ========= ================================
"Name" string Kernel argument name.
"TypeName" string Kernel argument type name.
"Size" integer Required Kernel argument size in bytes.
"Align" integer Required Kernel argument alignment in
bytes. Must be a power of two.
"ValueKind" string Required Kernel argument kind that
specifies how to set up the
corresponding argument.
Values include:
"ByValue"
The argument is copied
directly into the kernarg.
"GlobalBuffer"
A global address space pointer
to the buffer data is passed
in the kernarg.
"DynamicSharedPointer"
A group address space pointer
to dynamically allocated LDS
is passed in the kernarg.
"Sampler"
A global address space
pointer to a S# is passed in
the kernarg.
"Image"
A global address space
pointer to a T# is passed in
the kernarg.
"Pipe"
A global address space pointer
to an OpenCL pipe is passed in
the kernarg.
"Queue"
A global address space pointer
to an OpenCL device enqueue
queue is passed in the
kernarg.
"HiddenGlobalOffsetX"
The OpenCL grid dispatch
global offset for the X
dimension is passed in the
kernarg.
"HiddenGlobalOffsetY"
The OpenCL grid dispatch
global offset for the Y
dimension is passed in the
kernarg.
"HiddenGlobalOffsetZ"
The OpenCL grid dispatch
global offset for the Z
dimension is passed in the
kernarg.
"HiddenNone"
An argument that is not used
by the kernel. Space needs to
be left for it, but it does
not need to be set up.
"HiddenPrintfBuffer"
A global address space pointer
to the runtime printf buffer
is passed in kernarg.
2019-11-19 16:35:39 +08:00
"HiddenHostcallBuffer"
A global address space pointer
to the runtime hostcall buffer
is passed in kernarg.
2017-06-07 04:31:59 +08:00
"HiddenDefaultQueue"
A global address space pointer
to the OpenCL device enqueue
queue that should be used by
the kernel by default is
passed in the kernarg.
"HiddenCompletionAction"
2017-10-30 22:30:28 +08:00
A global address space pointer
to help link enqueued kernels into
the ancestor tree for determining
when the parent kernel has finished.
2017-06-07 04:31:59 +08:00
2019-07-06 00:05:17 +08:00
"HiddenMultiGridSyncArg"
A global address space pointer for
multi-grid synchronization is
passed in the kernarg.
2020-06-30 05:02:10 +08:00
"ValueType" string Unused and deprecated. This should no longer
be emitted, but is accepted for compatibility.
2017-06-07 04:31:59 +08:00
"PointeeAlign" integer Alignment in bytes of pointee
type for pointer type kernel
argument. Must be a power
of 2. Only present if
"ValueKind" is
"DynamicSharedPointer".
"AddrSpaceQual" string Kernel argument address space
qualifier. Only present if
"ValueKind" is "GlobalBuffer" or
"DynamicSharedPointer". Values
are:
- "Private"
- "Global"
- "Constant"
- "Local"
- "Generic"
- "Region"
2019-12-12 13:47:56 +08:00
.. TODO ::
2020-10-22 15:15:41 +08:00
2017-06-07 04:31:59 +08:00
Is GlobalBuffer only Global
or Constant? Is
DynamicSharedPointer always
Local? Can HCC allow Generic?
How can Private or Region
ever happen?
2020-10-22 15:15:41 +08:00
2017-06-07 04:31:59 +08:00
"AccQual" string Kernel argument access
qualifier. Only present if
"ValueKind" is "Image" or
"Pipe". Values
are:
- "ReadOnly"
- "WriteOnly"
- "ReadWrite"
2019-12-12 13:47:56 +08:00
.. TODO ::
2020-10-22 15:15:41 +08:00
2017-06-07 04:31:59 +08:00
Does this apply to
GlobalBuffer?
2020-10-22 15:15:41 +08:00
2017-10-15 03:03:51 +08:00
"ActualAccQual" string The actual memory accesses
2017-06-07 04:31:59 +08:00
performed by the kernel on the
kernel argument. Only present if
"ValueKind" is "GlobalBuffer",
"Image", or "Pipe". This may be
more restrictive than indicated
by "AccQual" to reflect what the
kernel actual does. If not
present then the runtime must
assume what is implied by
"AccQual" and "IsConst". Values
are:
- "ReadOnly"
- "WriteOnly"
- "ReadWrite"
"IsConst" boolean Indicates if the kernel argument
is const qualified. Only present
if "ValueKind" is
"GlobalBuffer".
"IsRestrict" boolean Indicates if the kernel argument
is restrict qualified. Only
present if "ValueKind" is
"GlobalBuffer".
"IsVolatile" boolean Indicates if the kernel argument
is volatile qualified. Only
present if "ValueKind" is
"GlobalBuffer".
"IsPipe" boolean Indicates if the kernel argument
is pipe qualified. Only present
if "ValueKind" is "Pipe".
2019-12-12 13:47:56 +08:00
.. TODO ::
2020-10-22 15:15:41 +08:00
2017-06-07 04:31:59 +08:00
Can GlobalBuffer be pipe
qualified?
2020-10-22 15:15:41 +08:00
2017-06-07 04:31:59 +08:00
================= ============== ========= ================================
..
2018-11-16 04:46:55 +08:00
.. table :: AMDHSA Code Object V2 Kernel Code Properties Metadata Map
2020-12-01 12:12:04 +08:00
:name: amdgpu-amdhsa-code-object-kernel-code-properties-metadata-map-v2-table
2017-06-07 04:31:59 +08:00
============================ ============== ========= =====================
String Key Value Type Required? Description
============================ ============== ========= =====================
"KernargSegmentSize" integer Required The size in bytes of
the kernarg segment
that holds the values
of the arguments to
the kernel.
"GroupSegmentFixedSize" integer Required The amount of group
segment memory
required by a
work-group in
bytes. This does not
include any
dynamically allocated
group segment memory
that may be added
when the kernel is
dispatched.
"PrivateSegmentFixedSize" integer Required The amount of fixed
private address space
memory required for a
work-item in
2017-11-10 09:00:54 +08:00
bytes. If the kernel
uses a dynamic call
stack then additional
2017-06-07 04:31:59 +08:00
space must be added
to this value for the
call stack.
"KernargSegmentAlign" integer Required The maximum byte
alignment of
arguments in the
kernarg segment. Must
be a power of 2.
"WavefrontSize" integer Required Wavefront size. Must
be a power of 2.
2017-11-10 09:00:54 +08:00
"NumSGPRs" integer Required Number of scalar
2017-06-07 04:31:59 +08:00
registers used by a
wavefront for
2019-06-14 06:18:47 +08:00
GFX6-GFX10. This
2017-06-07 04:31:59 +08:00
includes the special
SGPRs for VCC, Flat
2019-06-14 06:18:47 +08:00
Scratch (GFX7-GFX10)
2017-06-07 04:31:59 +08:00
and XNACK (for
2019-06-14 06:18:47 +08:00
GFX8-GFX10). It does
2017-06-07 04:31:59 +08:00
not include the 16
SGPR added if a trap
handler is
enabled. It is not
rounded up to the
allocation
granularity.
2017-11-10 09:00:54 +08:00
"NumVGPRs" integer Required Number of vector
2017-06-07 04:31:59 +08:00
registers used by
each work-item for
2019-06-14 06:18:47 +08:00
GFX6-GFX10
2017-11-10 09:00:54 +08:00
"MaxFlatWorkGroupSize" integer Required Maximum flat
2017-06-07 04:31:59 +08:00
work-group size
supported by the
kernel in work-items.
2017-11-10 09:00:54 +08:00
Must be >=1 and
2018-01-31 07:07:10 +08:00
consistent with
ReqdWorkGroupSize if
not 0, 0, 0.
2017-11-29 01:51:08 +08:00
"NumSpilledSGPRs" integer Number of stores from
a scalar register to
a register allocator
created spill
location.
"NumSpilledVGPRs" integer Number of stores from
a vector register to
a register allocator
created spill
location.
2017-06-07 04:31:59 +08:00
============================ ============== ========= =====================
2018-11-16 04:46:55 +08:00
.. _amdgpu-amdhsa-code-object-metadata-v3:
2020-12-01 12:12:04 +08:00
Code Object V3 Metadata
+++++++++++++++++++++++
Code object V3 to V4 metadata is specified by the `` NT_AMDGPU_METADATA `` note
record (see :ref: `amdgpu-note-records-v3-v4` ).
2018-11-16 04:46:55 +08:00
The metadata is represented as Message Pack formatted binary data (see
[MsgPack]_ ). The top level is a Message Pack map that includes the
keys defined in table
:ref: `amdgpu-amdhsa-code-object-metadata-map-table-v3` and referenced
tables.
Additional information can be added to the maps. To avoid conflicts,
any key names should be prefixed by "*vendor-name* ." where
2020-01-07 23:06:14 +08:00
`` vendor-name `` can be the name of the vendor and specific vendor
2018-11-16 04:46:55 +08:00
tool that generates the information. The prefix is abbreviated to
simply "." when it appears within a map that has been added by the
same *vendor-name* .
.. table :: AMDHSA Code Object V3 Metadata Map
:name: amdgpu-amdhsa-code-object-metadata-map-table-v3
================= ============== ========= =======================================
String Key Value Type Required? Description
================= ============== ========= =======================================
"amdhsa.version" sequence of Required - The first integer is the major
2 integers version. Currently 1.
- The second integer is the minor
version. Currently 0.
"amdhsa.printf" sequence of Each string is encoded information
strings about a printf function call. The
encoded information is organized as
fields separated by colon (':'):
`` ID:N:S[0]:S[1]:...:S[N-1]:FormatString ``
where:
`` ID ``
2019-12-12 13:47:56 +08:00
A 32-bit integer as a unique id for
2018-11-16 04:46:55 +08:00
each printf function call
`` N ``
2019-12-12 13:47:56 +08:00
A 32-bit integer equal to the number
2018-11-16 04:46:55 +08:00
of arguments of printf function call
minus 1
`` S[i] `` (where i = 0, 1, ... , N-1)
2019-12-12 13:47:56 +08:00
32-bit integers for the size in bytes
2018-11-16 04:46:55 +08:00
of the i-th FormatString argument of
the printf function call
FormatString
The format string passed to the
printf function call.
"amdhsa.kernels" sequence of Required Sequence of the maps for each
map kernel in the code object. See
:ref: `amdgpu-amdhsa-code-object-kernel-metadata-map-table-v3`
for the definition of the keys included
in that map.
================= ============== ========= =======================================
..
.. table :: AMDHSA Code Object V3 Kernel Metadata Map
:name: amdgpu-amdhsa-code-object-kernel-metadata-map-table-v3
=================================== ============== ========= ================================
String Key Value Type Required? Description
=================================== ============== ========= ================================
".name" string Required Source name of the kernel.
".symbol" string Required Name of the kernel
descriptor ELF symbol.
".language" string Source language of the kernel.
Values include:
- "OpenCL C"
- "OpenCL C++"
- "HCC"
- "HIP"
- "OpenMP"
- "Assembler"
".language_version" sequence of - The first integer is the major
2 integers version.
- The second integer is the
minor version.
".args" sequence of Sequence of maps of the
map kernel arguments. See
:ref: `amdgpu-amdhsa-code-object-kernel-argument-metadata-map-table-v3`
for the definition of the keys
included in that map.
".reqd_workgroup_size" sequence of If not 0, 0, 0 then all values
3 integers must be >=1 and the dispatch
work-group size X, Y, Z must
correspond to the specified
values. Defaults to 0, 0, 0.
Corresponds to the OpenCL
`` reqd_work_group_size ``
attribute.
".workgroup_size_hint" sequence of The dispatch work-group size
3 integers X, Y, Z is likely to be the
specified values.
Corresponds to the OpenCL
`` work_group_size_hint ``
attribute.
".vec_type_hint" string The name of a scalar or vector
type.
Corresponds to the OpenCL
`` vec_type_hint `` attribute.
".device_enqueue_symbol" string The external symbol name
associated with a kernel.
OpenCL runtime allocates a
global buffer for the symbol
and saves the kernel's address
to it, which is used for
device side enqueueing. Only
available for device side
enqueued kernels.
".kernarg_segment_size" integer Required The size in bytes of
the kernarg segment
that holds the values
of the arguments to
the kernel.
".group_segment_fixed_size" integer Required The amount of group
segment memory
required by a
work-group in
bytes. This does not
include any
dynamically allocated
group segment memory
that may be added
when the kernel is
dispatched.
".private_segment_fixed_size" integer Required The amount of fixed
private address space
memory required for a
work-item in
bytes. If the kernel
uses a dynamic call
stack then additional
space must be added
to this value for the
call stack.
".kernarg_segment_align" integer Required The maximum byte
alignment of
arguments in the
kernarg segment. Must
be a power of 2.
".wavefront_size" integer Required Wavefront size. Must
be a power of 2.
".sgpr_count" integer Required Number of scalar
registers required by a
wavefront for
GFX6-GFX9. A register
is required if it is
used explicitly, or
if a higher numbered
register is used
explicitly. This
includes the special
SGPRs for VCC, Flat
Scratch (GFX7-GFX9)
and XNACK (for
GFX8-GFX9). It does
not include the 16
SGPR added if a trap
handler is
enabled. It is not
rounded up to the
allocation
granularity.
".vgpr_count" integer Required Number of vector
registers required by
each work-item for
GFX6-GFX9. A register
is required if it is
used explicitly, or
if a higher numbered
register is used
explicitly.
".max_flat_workgroup_size" integer Required Maximum flat
work-group size
supported by the
kernel in work-items.
Must be >=1 and
consistent with
ReqdWorkGroupSize if
not 0, 0, 0.
".sgpr_spill_count" integer Number of stores from
a scalar register to
a register allocator
created spill
location.
".vgpr_spill_count" integer Number of stores from
a vector register to
a register allocator
created spill
location.
=================================== ============== ========= ================================
..
.. table :: AMDHSA Code Object V3 Kernel Argument Metadata Map
:name: amdgpu-amdhsa-code-object-kernel-argument-metadata-map-table-v3
====================== ============== ========= ================================
String Key Value Type Required? Description
====================== ============== ========= ================================
".name" string Kernel argument name.
".type_name" string Kernel argument type name.
".size" integer Required Kernel argument size in bytes.
".offset" integer Required Kernel argument offset in
bytes. The offset must be a
multiple of the alignment
required by the argument.
".value_kind" string Required Kernel argument kind that
specifies how to set up the
corresponding argument.
Values include:
"by_value"
The argument is copied
directly into the kernarg.
"global_buffer"
A global address space pointer
to the buffer data is passed
in the kernarg.
"dynamic_shared_pointer"
A group address space pointer
to dynamically allocated LDS
is passed in the kernarg.
"sampler"
A global address space
pointer to a S# is passed in
the kernarg.
"image"
A global address space
pointer to a T# is passed in
the kernarg.
"pipe"
A global address space pointer
to an OpenCL pipe is passed in
the kernarg.
"queue"
A global address space pointer
to an OpenCL device enqueue
queue is passed in the
kernarg.
"hidden_global_offset_x"
The OpenCL grid dispatch
global offset for the X
dimension is passed in the
kernarg.
"hidden_global_offset_y"
The OpenCL grid dispatch
global offset for the Y
dimension is passed in the
kernarg.
"hidden_global_offset_z"
The OpenCL grid dispatch
global offset for the Z
dimension is passed in the
kernarg.
"hidden_none"
An argument that is not used
by the kernel. Space needs to
be left for it, but it does
not need to be set up.
"hidden_printf_buffer"
A global address space pointer
to the runtime printf buffer
is passed in kernarg.
2019-11-19 16:35:39 +08:00
"hidden_hostcall_buffer"
A global address space pointer
to the runtime hostcall buffer
is passed in kernarg.
2018-11-16 04:46:55 +08:00
"hidden_default_queue"
A global address space pointer
to the OpenCL device enqueue
queue that should be used by
the kernel by default is
passed in the kernarg.
"hidden_completion_action"
A global address space pointer
to help link enqueued kernels into
the ancestor tree for determining
when the parent kernel has finished.
2019-07-06 00:05:17 +08:00
"hidden_multigrid_sync_arg"
A global address space pointer for
multi-grid synchronization is
passed in the kernarg.
2020-06-30 05:02:10 +08:00
".value_type" string Unused and deprecated. This should no longer
be emitted, but is accepted for compatibility.
2018-11-16 04:46:55 +08:00
".pointee_align" integer Alignment in bytes of pointee
type for pointer type kernel
argument. Must be a power
of 2. Only present if
".value_kind" is
"dynamic_shared_pointer".
".address_space" string Kernel argument address space
qualifier. Only present if
".value_kind" is "global_buffer" or
"dynamic_shared_pointer". Values
are:
- "private"
- "global"
- "constant"
- "local"
- "generic"
- "region"
2019-12-12 13:47:56 +08:00
.. TODO ::
2020-10-22 15:15:41 +08:00
2018-11-16 04:46:55 +08:00
Is "global_buffer" only "global"
or "constant"? Is
"dynamic_shared_pointer" always
"local"? Can HCC allow "generic"?
How can "private" or "region"
ever happen?
2020-10-22 15:15:41 +08:00
2018-11-16 04:46:55 +08:00
".access" string Kernel argument access
qualifier. Only present if
".value_kind" is "image" or
"pipe". Values
are:
- "read_only"
- "write_only"
- "read_write"
2019-12-12 13:47:56 +08:00
.. TODO ::
2020-10-22 15:15:41 +08:00
2018-11-16 04:46:55 +08:00
Does this apply to
"global_buffer"?
2020-10-22 15:15:41 +08:00
2018-11-16 04:46:55 +08:00
".actual_access" string The actual memory accesses
performed by the kernel on the
kernel argument. Only present if
".value_kind" is "global_buffer",
"image", or "pipe". This may be
more restrictive than indicated
by ".access" to reflect what the
kernel actual does. If not
present then the runtime must
assume what is implied by
".access" and ".is_const" . Values
are:
- "read_only"
- "write_only"
- "read_write"
".is_const" boolean Indicates if the kernel argument
is const qualified. Only present
if ".value_kind" is
"global_buffer".
".is_restrict" boolean Indicates if the kernel argument
is restrict qualified. Only
present if ".value_kind" is
"global_buffer".
".is_volatile" boolean Indicates if the kernel argument
is volatile qualified. Only
present if ".value_kind" is
"global_buffer".
".is_pipe" boolean Indicates if the kernel argument
is pipe qualified. Only present
if ".value_kind" is "pipe".
2019-12-12 13:47:56 +08:00
.. TODO ::
2020-10-22 15:15:41 +08:00
2018-11-16 04:46:55 +08:00
Can "global_buffer" be pipe
qualified?
2020-10-22 15:15:41 +08:00
2018-11-16 04:46:55 +08:00
====================== ============== ========= ================================
2020-12-01 12:12:04 +08:00
.. _amdgpu-amdhsa-code-object-metadata-v4:
Code Object V4 Metadata
+++++++++++++++++++++++
2020-12-18 00:09:32 +08:00
.. warning ::
Code object V4 is not the default code object version emitted by this version
of LLVM.
2020-12-01 12:12:04 +08:00
Code object V4 metadata is the same as
:ref: `amdgpu-amdhsa-code-object-metadata-v3` with the changes and additions
defined in table :ref: `amdgpu-amdhsa-code-object-metadata-map-table-v3` .
.. table :: AMDHSA Code Object V4 Metadata Map Changes from :ref: `amdgpu-amdhsa-code-object-metadata-v3`
:name: amdgpu-amdhsa-code-object-metadata-map-table-v4
================= ============== ========= =======================================
String Key Value Type Required? Description
================= ============== ========= =======================================
"amdhsa.version" sequence of Required - The first integer is the major
2 integers version. Currently 1.
- The second integer is the minor
version. Currently 1.
"amdhsa.target" string Required The target name of the code using the syntax:
.. code ::
<target-triple> [ "-" <target-id> ]
A canonical target ID must be
used. See :ref: `amdgpu-target-triples`
and :ref: `amdgpu-target-id` .
================= ============== ========= =======================================
2017-06-07 04:31:59 +08:00
..
Kernel Dispatch
~~~~~~~~~~~~~~~
2020-10-22 15:15:41 +08:00
The HSA architected queuing language (AQL) defines a user space memory interface
that can be used to control the dispatch of kernels, in an agent independent
2020-12-01 12:12:04 +08:00
way. An agent can have zero or more AQL queues created for it using an HSA
compatible runtime (see :ref: `amdgpu-os` ), in which AQL packets (all of which
are 64 bytes) can be placed. See the *HSA Platform System Architecture
Specification* [HSA]_ for the AQL queue mechanics and packet layouts.
2017-06-07 04:31:59 +08:00
The packet processor of a kernel agent is responsible for detecting and
dispatching HSA kernels from the AQL queues associated with it. For AMD GPUs the
packet processor is implemented by the hardware command processor (CP),
asynchronous dispatch controller (ADC) and shader processor input controller
(SPI).
2020-12-01 12:12:04 +08:00
An HSA compatible runtime can be used to allocate an AQL queue object. It uses
the kernel mode driver to initialize and register the AQL queue with CP.
2017-06-07 04:31:59 +08:00
To dispatch a kernel the following actions are performed. This can occur in the
CPU host program, or from an HSA kernel executing on a GPU.
1. A pointer to an AQL queue for the kernel agent on which the kernel is to be
executed is obtained.
2. A pointer to the kernel descriptor (see
2019-12-12 13:47:56 +08:00
:ref: `amdgpu-amdhsa-kernel-descriptor` ) of the kernel to execute is obtained.
It must be for a kernel that is contained in a code object that that was
2020-12-01 12:12:04 +08:00
loaded by an HSA compatible runtime on the kernel agent with which the AQL
queue is associated.
3. Space is allocated for the kernel arguments using the HSA compatible runtime
allocator for a memory region with the kernarg property for the kernel agent
that will execute the kernel. It must be at least 16-byte aligned.
2017-06-07 04:31:59 +08:00
4. Kernel argument values are assigned to the kernel argument memory
2019-12-12 13:47:56 +08:00
allocation. The layout is defined in the *HSA Programmer's Language
Reference* [HSA]_ . For AMDGPU the kernel execution directly accesses the
kernel argument memory in the same way constant memory is accessed. (Note
that the HSA specification allows an implementation to copy the kernel
argument contents to another location that is accessed by the kernel.)
2020-12-01 12:12:04 +08:00
5. An AQL kernel dispatch packet is created on the AQL queue. The HSA compatible
runtime api uses 64-bit atomic operations to reserve space in the AQL queue
for the packet. The packet must be set up, and the final write must use an
atomic store release to set the packet kind to ensure the packet contents are
2017-06-07 04:31:59 +08:00
visible to the kernel agent. AQL defines a doorbell signal mechanism to
notify the kernel agent that the AQL queue has been updated. These rules, and
the layout of the AQL queue and kernel dispatch packet is defined in the *HSA
System Architecture Specification* [HSA]_ .
6. A kernel dispatch packet includes information about the actual dispatch,
such as grid and work-group size, together with information from the code
2020-12-01 12:12:04 +08:00
object about the kernel, such as segment sizes. The HSA compatible runtime
queries on the kernel symbol can be used to obtain the code object values
which are recorded in the :ref: `amdgpu-amdhsa-code-object-metadata` .
2017-06-07 04:31:59 +08:00
7. CP executes micro-code and is responsible for detecting and setting up the
GPU to execute the wavefronts of a kernel dispatch.
8. CP ensures that when the a wavefront starts executing the kernel machine
code, the scalar general purpose registers (SGPR) and vector general purpose
registers (VGPR) are set up as required by the machine code. The required
setup is defined in the :ref: `amdgpu-amdhsa-kernel-descriptor` . The initial
register state is defined in
:ref: `amdgpu-amdhsa-initial-kernel-execution-state` .
9. The prolog of the kernel machine code (see
:ref: `amdgpu-amdhsa-kernel-prolog` ) sets up the machine state as necessary
before continuing executing the machine code that corresponds to the kernel.
10. When the kernel dispatch has completed execution, CP signals the completion
signal specified in the kernel dispatch packet if not 0.
2020-10-22 15:15:41 +08:00
.. _amdgpu-amdhsa-memory-spaces:
Memory Spaces
~~~~~~~~~~~~~
The memory space properties are:
.. table :: AMDHSA Memory Spaces
:name: amdgpu-amdhsa-memory-spaces-table
================= =========== ======== ======= ==================
Memory Space Name HSA Segment Hardware Address NULL Value
Name Name Size
================= =========== ======== ======= ==================
Private private scratch 32 0x00000000
Local group LDS 32 0xFFFFFFFF
Global global global 64 0x0000000000000000
Constant constant *same as 64 0x0000000000000000
global*
Generic flat flat 64 0x0000000000000000
Region N/A GDS 32 *not implemented
for AMDHSA*
================= =========== ======== ======= ==================
The global and constant memory spaces both use global virtual addresses, which
are the same virtual address space used by the CPU. However, some virtual
addresses may only be accessible to the CPU, some only accessible by the GPU,
and some by both.
Using the constant memory space indicates that the data will not change during
the execution of the kernel. This allows scalar read instructions to be
used. The vector and scalar L1 caches are invalidated of volatile data before
each kernel dispatch execution to allow constant memory to change values between
kernel dispatches.
The local memory space uses the hardware Local Data Store (LDS) which is
automatically allocated when the hardware creates work-groups of wavefronts, and
freed when all the wavefronts of a work-group have terminated. The data store
(DS) instructions can be used to access it.
The private memory space uses the hardware scratch memory support. If the kernel
uses scratch, then the hardware allocates memory that is accessed using
wavefront lane dword (4 byte) interleaving. The mapping used from private
address to physical address is:
`` wavefront-scratch-base +
(private-address * wavefront-size * 4) +
(wavefront-lane-id * 4)``
There are different ways that the wavefront scratch base address is determined
by a wavefront (see :ref: `amdgpu-amdhsa-initial-kernel-execution-state` ). This
memory can be accessed in an interleaved manner using buffer instruction with
the scratch buffer descriptor and per wavefront scratch offset, by the scratch
instructions, or by flat instructions. If each lane of a wavefront accesses the
same private address, the interleaving results in adjacent dwords being accessed
and hence requires fewer cache lines to be fetched. Multi-dword access is not
supported except by flat and scratch instructions in GFX9-GFX10.
The generic address space uses the hardware flat address support available in
GFX7-GFX10. This uses two fixed ranges of virtual addresses (the private and
local apertures), that are outside the range of addressible global memory, to
map from a flat address to a private or local address.
FLAT instructions can take a flat address and access global, private (scratch)
and group (LDS) memory depending in if the address is within one of the
aperture ranges. Flat access to scratch requires hardware aperture setup and
setup in the kernel prologue (see
:ref: `amdgpu-amdhsa-kernel-prolog-flat-scratch` ). Flat access to LDS requires
hardware aperture setup and M0 (GFX7-GFX8) register setup (see
:ref: `amdgpu-amdhsa-kernel-prolog-m0` ).
To convert between a segment address and a flat address the base address of the
apertures address can be used. For GFX7-GFX8 these are available in the
:ref: `amdgpu-amdhsa-hsa-aql-queue` the address of which can be obtained with
Queue Ptr SGPR (see :ref: `amdgpu-amdhsa-initial-kernel-execution-state` ). For
GFX9-GFX10 the aperture base addresses are directly available as inline constant
registers `` SRC_SHARED_BASE/LIMIT `` and `` SRC_PRIVATE_BASE/LIMIT `` . In 64 bit
address mode the aperture sizes are 2^32 bytes and the base is aligned to 2^32
which makes it easier to convert from flat to segment or segment to flat.
2017-08-16 04:47:41 +08:00
Image and Samplers
~~~~~~~~~~~~~~~~~~
2017-06-07 04:31:59 +08:00
2020-12-01 12:12:04 +08:00
Image and sample handles created by an HSA compatible runtime (see
:ref: `amdgpu-os` ) are 64-bit addresses of a hardware 32-byte V# and 48 byte S#
object respectively. In order to support the HSA `` query_sampler `` operations
two extra dwords are used to store the HSA BRIG enumeration values for the
queries that are not trivially deducible from the S# representation.
2017-06-07 04:31:59 +08:00
HSA Signals
~~~~~~~~~~~
2020-12-01 12:12:04 +08:00
HSA signal handles created by an HSA compatible runtime (see :ref: `amdgpu-os` )
are 64-bit addresses of a structure allocated in memory accessible from both the
CPU and GPU. The structure is defined by the runtime and subject to change
between releases. For example, see [AMD-ROCm-github]_ .
2017-06-07 04:31:59 +08:00
.. _amdgpu-amdhsa-hsa-aql-queue:
HSA AQL Queue
~~~~~~~~~~~~~
2020-12-01 12:12:04 +08:00
The HSA AQL queue structure is defined by an HSA compatible runtime (see
:ref: `amdgpu-os` ) and subject to change between releases. For example, see
[AMD-ROCm-github]_ . For some processors it contains fields needed to implement
certain language features such as the flat address aperture bases. It also
contains fields used by CP such as managing the allocation of scratch memory.
2017-06-07 04:31:59 +08:00
.. _amdgpu-amdhsa-kernel-descriptor:
Kernel Descriptor
~~~~~~~~~~~~~~~~~
A kernel descriptor consists of the information needed by CP to initiate the
execution of a kernel, including the entry point address of the machine code
that implements the kernel.
2020-12-01 12:12:04 +08:00
Code Object V3 Kernel Descriptor
++++++++++++++++++++++++++++++++
2017-06-07 04:31:59 +08:00
2020-02-20 15:08:59 +08:00
CP microcode requires the Kernel descriptor to be allocated on 64-byte
2018-06-22 03:38:56 +08:00
alignment.
2017-06-07 04:31:59 +08:00
2020-11-14 05:08:37 +08:00
The fields used by CP for code objects before V3 also match those specified in
2020-12-01 12:12:04 +08:00
:ref: `amdgpu-amdhsa-kernel-descriptor-v3-table` .
2020-11-14 05:08:37 +08:00
2020-12-01 12:12:04 +08:00
.. table :: Code Object V3 Kernel Descriptor
:name: amdgpu-amdhsa-kernel-descriptor-v3-table
2017-06-07 04:31:59 +08:00
2017-10-19 06:16:55 +08:00
======= ======= =============================== ============================
2017-06-07 04:31:59 +08:00
Bits Size Field Name Description
2017-10-19 06:16:55 +08:00
======= ======= =============================== ============================
2018-06-13 02:02:46 +08:00
31:0 4 bytes GROUP_SEGMENT_FIXED_SIZE The amount of fixed local
2017-06-07 04:31:59 +08:00
address space memory
required for a work-group
in bytes. This does not
include any dynamically
allocated local address
space memory that may be
added when the kernel is
dispatched.
2018-06-13 02:02:46 +08:00
63:32 4 bytes PRIVATE_SEGMENT_FIXED_SIZE The amount of fixed
2017-06-07 04:31:59 +08:00
private address space
memory required for a
2020-12-01 12:12:04 +08:00
work-item in bytes.
Additional space may need to
be added to this value if
the call stack has
non-inlined function calls.
95:64 4 bytes KERNARG_SIZE The size of the kernarg
memory pointed to by the
AQL dispatch packet. The
kernarg memory is used to
pass arguments to the
kernel.
* If the kernarg pointer in
the dispatch packet is NULL
then there are no kernel
arguments.
* If the kernarg pointer in
the dispatch packet is
not NULL and this value
is 0 then the kernarg
memory size is
unspecified.
* If the kernarg pointer in
the dispatch packet is
not NULL and this value
is not 0 then the value
specifies the kernarg
memory size in bytes. It
is recommended to provide
a value as it may be used
by CP to optimize making
the kernarg memory
visible to the kernel
code.
127:96 4 bytes Reserved, must be 0.
2018-06-13 02:02:46 +08:00
191:128 8 bytes KERNEL_CODE_ENTRY_BYTE_OFFSET Byte offset (possibly
2017-06-07 04:31:59 +08:00
negative) from base
address of kernel
descriptor to kernel's
entry point instruction
which must be 256 byte
aligned.
2019-06-14 06:18:47 +08:00
351:272 20 Reserved, must be 0.
2017-06-07 04:31:59 +08:00
bytes
2021-01-27 01:45:41 +08:00
383:352 4 bytes COMPUTE_PGM_RSRC3 GFX6-GFX9
2019-06-14 06:18:47 +08:00
Reserved, must be 0.
2021-02-18 05:37:46 +08:00
GFX90A
Compute Shader (CS)
program settings used by
CP to set up
`` COMPUTE_PGM_RSRC3 ``
configuration
register. See
:ref: `amdgpu-amdhsa-compute_pgm_rsrc3-gfx90a-table` .
2019-06-14 06:18:47 +08:00
GFX10
Compute Shader (CS)
program settings used by
CP to set up
`` COMPUTE_PGM_RSRC3 ``
configuration
register. See
:ref: `amdgpu-amdhsa-compute_pgm_rsrc3-gfx10-table` .
2018-06-13 02:02:46 +08:00
415:384 4 bytes COMPUTE_PGM_RSRC1 Compute Shader (CS)
2017-06-07 04:31:59 +08:00
program settings used by
CP to set up
`` COMPUTE_PGM_RSRC1 ``
configuration
register. See
2019-06-14 06:18:47 +08:00
:ref: `amdgpu-amdhsa-compute_pgm_rsrc1-gfx6-gfx10-table` .
2018-06-13 02:02:46 +08:00
447:416 4 bytes COMPUTE_PGM_RSRC2 Compute Shader (CS)
2017-06-07 04:31:59 +08:00
program settings used by
CP to set up
`` COMPUTE_PGM_RSRC2 ``
configuration
register. See
2019-06-14 06:18:47 +08:00
:ref: `amdgpu-amdhsa-compute_pgm_rsrc2-gfx6-gfx10-table` .
2020-12-15 09:42:25 +08:00
458:448 7 bits *See separate bits below.* Enable the setup of the
SGPR user data registers
2017-06-07 04:31:59 +08:00
(see
:ref: `amdgpu-amdhsa-initial-kernel-execution-state` ).
The total number of SGPR
user data registers
requested must not exceed
16 and match value in
`` compute_pgm_rsrc2.user_sgpr.user_sgpr_count `` .
Any requests beyond 16
will be ignored.
2020-12-15 09:42:25 +08:00
>448 1 bit ENABLE_SGPR_PRIVATE_SEGMENT
_BUFFER
>449 1 bit ENABLE_SGPR_DISPATCH_PTR
>450 1 bit ENABLE_SGPR_QUEUE_PTR
>451 1 bit ENABLE_SGPR_KERNARG_SEGMENT_PTR
>452 1 bit ENABLE_SGPR_DISPATCH_ID
>453 1 bit ENABLE_SGPR_FLAT_SCRATCH_INIT
>454 1 bit ENABLE_SGPR_PRIVATE_SEGMENT
2018-06-13 02:02:46 +08:00
_SIZE
2019-06-14 06:18:47 +08:00
457:455 3 bits Reserved, must be 0.
2021-01-27 01:45:41 +08:00
458 1 bit ENABLE_WAVEFRONT_SIZE32 GFX6-GFX9
2019-06-14 06:18:47 +08:00
Reserved, must be 0.
GFX10
- If 0 execute in
wavefront size 64 mode.
- If 1 execute in
native wavefront size
32 mode.
2020-11-14 05:08:37 +08:00
463:459 1 bit Reserved, must be 0.
464 1 bit RESERVED_464 Deprecated, must be 0.
467:465 3 bits Reserved, must be 0.
468 1 bit RESERVED_468 Deprecated, must be 0.
469:471 3 bits Reserved, must be 0.
511:472 5 bytes Reserved, must be 0.
2017-06-07 04:31:59 +08:00
512 **Total size 64 bytes.**
2017-10-19 06:16:55 +08:00
======= ====================================================================
2017-06-07 04:31:59 +08:00
..
2019-06-14 06:18:47 +08:00
.. table :: compute_pgm_rsrc1 for GFX6-GFX10
:name: amdgpu-amdhsa-compute_pgm_rsrc1-gfx6-gfx10-table
2017-06-07 04:31:59 +08:00
2017-06-07 08:46:08 +08:00
======= ======= =============================== ===========================================================================
2017-06-07 04:31:59 +08:00
Bits Size Field Name Description
2017-06-07 08:46:08 +08:00
======= ======= =============================== ===========================================================================
2018-06-22 03:38:56 +08:00
5:0 6 bits GRANULATED_WORKITEM_VGPR_COUNT Number of vector register
blocks used by each work-item;
2017-06-07 04:31:59 +08:00
granularity is device
specific:
2017-11-10 09:00:54 +08:00
GFX6-GFX9
2018-06-22 03:38:56 +08:00
- vgprs_used 0..256
- max(0, ceil(vgprs_used / 4) - 1)
2021-02-18 05:37:46 +08:00
GFX90A
- vgprs_used 0..512
- vgprs_used = align(arch_vgprs, 4)
+ acc_vgprs
- max(0, ceil(vgprs_used / 8) - 1)
2019-06-14 06:18:47 +08:00
GFX10 (wavefront size 64)
- max_vgpr 1..256
- max(0, ceil(vgprs_used / 4) - 1)
GFX10 (wavefront size 32)
- max_vgpr 1..256
- max(0, ceil(vgprs_used / 8) - 1)
2018-06-22 03:38:56 +08:00
Where vgprs_used is defined
as the highest VGPR number
explicitly referenced plus
one.
2017-06-07 04:31:59 +08:00
Used by CP to set up
`` COMPUTE_PGM_RSRC1.VGPRS `` .
2018-06-22 03:38:56 +08:00
The
:ref: `amdgpu-assembler`
calculates this
automatically for the
selected processor from
values provided to the
`.amdhsa_kernel`
by the
`.amdhsa_next_free_vgpr`
nested directive (see
:ref: `amdhsa-kernel-directives-table` ).
9:6 4 bits GRANULATED_WAVEFRONT_SGPR_COUNT Number of scalar register
blocks used by a wavefront;
2017-06-07 04:31:59 +08:00
granularity is device
specific:
2017-11-10 09:00:54 +08:00
GFX6-GFX8
2018-06-22 03:38:56 +08:00
- sgprs_used 0..112
- max(0, ceil(sgprs_used / 8) - 1)
2017-06-07 04:31:59 +08:00
GFX9
2018-06-22 03:38:56 +08:00
- sgprs_used 0..112
- 2 * max(0, ceil(sgprs_used / 16) - 1)
2019-06-14 06:18:47 +08:00
GFX10
Reserved, must be 0.
(128 SGPRs always
allocated.)
2018-06-22 03:38:56 +08:00
Where sgprs_used is
defined as the highest
SGPR number explicitly
referenced plus one, plus
2020-02-20 15:08:59 +08:00
a target specific number
2018-06-22 03:38:56 +08:00
of additional special
SGPRs for VCC,
FLAT_SCRATCH (GFX7+) and
XNACK_MASK (GFX8+), and
any additional
2020-02-20 15:08:59 +08:00
target specific
2018-06-22 03:38:56 +08:00
limitations. It does not
include the 16 SGPRs added
if a trap handler is
2017-06-07 04:31:59 +08:00
enabled.
2020-02-20 15:08:59 +08:00
The target specific
2018-06-22 03:38:56 +08:00
limitations and special
SGPR layout are defined in
the hardware
documentation, which can
be found in the
:ref: `amdgpu-processors`
table.
2017-06-07 04:31:59 +08:00
Used by CP to set up
`` COMPUTE_PGM_RSRC1.SGPRS `` .
2018-06-22 03:38:56 +08:00
The
:ref: `amdgpu-assembler`
calculates this
automatically for the
selected processor from
values provided to the
`.amdhsa_kernel`
by the
`.amdhsa_next_free_sgpr`
and `.amdhsa_reserve_*`
nested directives (see
:ref: `amdhsa-kernel-directives-table` ).
2017-10-15 03:17:08 +08:00
11:10 2 bits PRIORITY Must be 0.
2017-06-07 04:31:59 +08:00
Start executing wavefront
at the specified priority.
CP is responsible for
filling in
`` COMPUTE_PGM_RSRC1.PRIORITY `` .
2017-10-15 03:17:08 +08:00
13:12 2 bits FLOAT_ROUND_MODE_32 Wavefront starts execution
2017-06-07 04:31:59 +08:00
with specified rounding
mode for single (32
bit) floating point
precision floating point
operations.
Floating point rounding
mode values are defined in
:ref: `amdgpu-amdhsa-floating-point-rounding-mode-enumeration-values-table` .
Used by CP to set up
`` COMPUTE_PGM_RSRC1.FLOAT_MODE `` .
2017-10-15 03:17:08 +08:00
15:14 2 bits FLOAT_ROUND_MODE_16_64 Wavefront starts execution
2017-06-07 04:31:59 +08:00
with specified rounding
denorm mode for half/double (16
2019-12-12 13:47:56 +08:00
and 64-bit) floating point
2017-06-07 04:31:59 +08:00
precision floating point
operations.
Floating point rounding
mode values are defined in
:ref: `amdgpu-amdhsa-floating-point-rounding-mode-enumeration-values-table` .
Used by CP to set up
`` COMPUTE_PGM_RSRC1.FLOAT_MODE `` .
2017-10-15 03:17:08 +08:00
17:16 2 bits FLOAT_DENORM_MODE_32 Wavefront starts execution
2017-06-07 04:31:59 +08:00
with specified denorm mode
for single (32
bit) floating point
precision floating point
operations.
Floating point denorm mode
values are defined in
:ref: `amdgpu-amdhsa-floating-point-denorm-mode-enumeration-values-table` .
Used by CP to set up
`` COMPUTE_PGM_RSRC1.FLOAT_MODE `` .
2017-10-15 03:17:08 +08:00
19:18 2 bits FLOAT_DENORM_MODE_16_64 Wavefront starts execution
2017-06-07 04:31:59 +08:00
with specified denorm mode
for half/double (16
2019-12-12 13:47:56 +08:00
and 64-bit) floating point
2017-06-07 04:31:59 +08:00
precision floating point
operations.
Floating point denorm mode
values are defined in
:ref: `amdgpu-amdhsa-floating-point-denorm-mode-enumeration-values-table` .
Used by CP to set up
`` COMPUTE_PGM_RSRC1.FLOAT_MODE `` .
2017-10-15 03:17:08 +08:00
20 1 bit PRIV Must be 0.
2017-06-07 04:31:59 +08:00
Start executing wavefront
in privilege trap handler
mode.
CP is responsible for
filling in
`` COMPUTE_PGM_RSRC1.PRIV `` .
2017-10-15 03:17:08 +08:00
21 1 bit ENABLE_DX10_CLAMP Wavefront starts execution
2017-06-07 04:31:59 +08:00
with DX10 clamp mode
enabled. Used by the vector
2017-10-19 06:16:55 +08:00
ALU to force DX10 style
2017-06-07 04:31:59 +08:00
treatment of NaN's (when
set, clamp NaN to zero,
otherwise pass NaN
through).
Used by CP to set up
`` COMPUTE_PGM_RSRC1.DX10_CLAMP `` .
2017-10-15 03:17:08 +08:00
22 1 bit DEBUG_MODE Must be 0.
2017-06-07 04:31:59 +08:00
Start executing wavefront
in single step mode.
CP is responsible for
filling in
`` COMPUTE_PGM_RSRC1.DEBUG_MODE `` .
2017-10-15 03:17:08 +08:00
23 1 bit ENABLE_IEEE_MODE Wavefront starts execution
2017-06-07 04:31:59 +08:00
with IEEE mode
enabled. Floating point
opcodes that support
exception flag gathering
will quiet and propagate
signaling-NaN inputs per
IEEE 754-2008. Min_dx10 and
max_dx10 become IEEE
754-2008 compliant due to
signaling-NaN propagation
and quieting.
Used by CP to set up
`` COMPUTE_PGM_RSRC1.IEEE_MODE `` .
2017-10-15 03:17:08 +08:00
24 1 bit BULKY Must be 0.
2017-06-07 04:31:59 +08:00
Only one work-group allowed
to execute on a compute
unit.
CP is responsible for
filling in
`` COMPUTE_PGM_RSRC1.BULKY `` .
2017-10-15 03:17:08 +08:00
25 1 bit CDBG_USER Must be 0.
2017-06-07 04:31:59 +08:00
Flag that can be used to
control debugging code.
CP is responsible for
filling in
`` COMPUTE_PGM_RSRC1.CDBG_USER `` .
2017-11-10 09:00:54 +08:00
26 1 bit FP16_OVFL GFX6-GFX8
2017-10-19 06:16:55 +08:00
Reserved, must be 0.
2019-06-14 06:18:47 +08:00
GFX9-GFX10
2017-10-19 06:16:55 +08:00
Wavefront starts execution
with specified fp16 overflow
mode.
- If 0, fp16 overflow generates
2017-10-15 03:17:08 +08:00
+/-INF values.
2017-10-19 06:16:55 +08:00
- If 1, fp16 overflow that is the
result of an +/-INF input value
or divide by 0 produces a +/-INF,
otherwise clamps computed
overflow to +/-MAX_FP16 as
appropriate.
2017-10-15 03:17:08 +08:00
Used by CP to set up
`` COMPUTE_PGM_RSRC1.FP16_OVFL `` .
2019-06-14 06:18:47 +08:00
28:27 2 bits Reserved, must be 0.
29 1 bit WGP_MODE GFX6-GFX9
Reserved, must be 0.
GFX10
- If 0 execute work-groups in
CU wavefront execution mode.
- If 1 execute work-groups on
in WGP wavefront execution mode.
See :ref: `amdgpu-amdhsa-memory-model` .
Used by CP to set up
`` COMPUTE_PGM_RSRC1.WGP_MODE `` .
2021-01-27 01:45:41 +08:00
30 1 bit MEM_ORDERED GFX6-GFX9
2019-06-14 06:18:47 +08:00
Reserved, must be 0.
GFX10
Controls the behavior of the
2020-10-31 12:43:55 +08:00
s_waitcnt's vmcnt and vscnt
2019-06-14 06:18:47 +08:00
counters.
- If 0 vmcnt reports completion
of load and atomic with return
out of order with sample
instructions, and the vscnt
reports the completion of
store and atomic without
return in order.
- If 1 vmcnt reports completion
of load, atomic with return
and sample instructions in
order, and the vscnt reports
the completion of store and
atomic without return in order.
Used by CP to set up
`` COMPUTE_PGM_RSRC1.MEM_ORDERED `` .
2021-01-27 01:45:41 +08:00
31 1 bit FWD_PROGRESS GFX6-GFX9
2019-06-14 06:18:47 +08:00
Reserved, must be 0.
GFX10
- If 0 execute SIMD wavefronts
using oldest first policy.
- If 1 execute SIMD wavefronts to
ensure wavefronts will make some
forward progress.
Used by CP to set up
`` COMPUTE_PGM_RSRC1.FWD_PROGRESS `` .
2017-06-07 04:31:59 +08:00
32 **Total size 4 bytes**
2017-06-07 08:46:08 +08:00
======= ===================================================================================================================
2017-06-07 04:31:59 +08:00
..
2019-06-14 06:18:47 +08:00
.. table :: compute_pgm_rsrc2 for GFX6-GFX10
:name: amdgpu-amdhsa-compute_pgm_rsrc2-gfx6-gfx10-table
2017-06-07 04:31:59 +08:00
2017-06-07 08:46:08 +08:00
======= ======= =============================== ===========================================================================
2017-06-07 04:31:59 +08:00
Bits Size Field Name Description
2017-06-07 08:46:08 +08:00
======= ======= =============================== ===========================================================================
2020-12-15 09:42:25 +08:00
0 1 bit ENABLE_PRIVATE_SEGMENT Enable the setup of the
private segment.
In addition, enable the
setup of the SGPR
wavefront scratch offset
2017-06-07 04:31:59 +08:00
system register (see
:ref: `amdgpu-amdhsa-initial-kernel-execution-state` ).
Used by CP to set up
`` COMPUTE_PGM_RSRC2.SCRATCH_EN `` .
2017-10-15 03:17:08 +08:00
5:1 5 bits USER_SGPR_COUNT The total number of SGPR
2017-06-07 04:31:59 +08:00
user data registers
requested. This number must
match the number of user
data registers enabled.
Used by CP to set up
`` COMPUTE_PGM_RSRC2.USER_SGPR `` .
2018-05-30 03:09:13 +08:00
6 1 bit ENABLE_TRAP_HANDLER Must be 0.
This bit represents
`` COMPUTE_PGM_RSRC2.TRAP_PRESENT `` ,
which is set by the CP if
the runtime has installed a
trap handler.
2017-10-15 03:17:08 +08:00
7 1 bit ENABLE_SGPR_WORKGROUP_ID_X Enable the setup of the
2017-06-07 04:31:59 +08:00
system SGPR register for
the work-group id in the X
dimension (see
:ref: `amdgpu-amdhsa-initial-kernel-execution-state` ).
Used by CP to set up
`` COMPUTE_PGM_RSRC2.TGID_X_EN `` .
2017-10-15 03:17:08 +08:00
8 1 bit ENABLE_SGPR_WORKGROUP_ID_Y Enable the setup of the
2017-06-07 04:31:59 +08:00
system SGPR register for
the work-group id in the Y
dimension (see
:ref: `amdgpu-amdhsa-initial-kernel-execution-state` ).
Used by CP to set up
`` COMPUTE_PGM_RSRC2.TGID_Y_EN `` .
2017-10-15 03:17:08 +08:00
9 1 bit ENABLE_SGPR_WORKGROUP_ID_Z Enable the setup of the
2017-06-07 04:31:59 +08:00
system SGPR register for
the work-group id in the Z
dimension (see
:ref: `amdgpu-amdhsa-initial-kernel-execution-state` ).
Used by CP to set up
`` COMPUTE_PGM_RSRC2.TGID_Z_EN `` .
2017-10-15 03:17:08 +08:00
10 1 bit ENABLE_SGPR_WORKGROUP_INFO Enable the setup of the
2017-06-07 04:31:59 +08:00
system SGPR register for
work-group information (see
:ref: `amdgpu-amdhsa-initial-kernel-execution-state` ).
Used by CP to set up
`` COMPUTE_PGM_RSRC2.TGID_SIZE_EN `` .
2017-10-15 03:17:08 +08:00
12:11 2 bits ENABLE_VGPR_WORKITEM_ID Enable the setup of the
2017-06-07 04:31:59 +08:00
VGPR system registers used
for the work-item ID.
:ref: `amdgpu-amdhsa-system-vgpr-work-item-id-enumeration-values-table`
defines the values.
Used by CP to set up
`` COMPUTE_PGM_RSRC2.TIDIG_CMP_CNT `` .
2017-10-15 03:17:08 +08:00
13 1 bit ENABLE_EXCEPTION_ADDRESS_WATCH Must be 0.
2017-06-07 04:31:59 +08:00
Wavefront starts execution
with address watch
exceptions enabled which
are generated when L1 has
witnessed a thread access
an *address of
interest*.
CP is responsible for
filling in the address
watch bit in
`` COMPUTE_PGM_RSRC2.EXCP_EN_MSB ``
according to what the
runtime requests.
2017-10-15 03:17:08 +08:00
14 1 bit ENABLE_EXCEPTION_MEMORY Must be 0.
2017-06-07 04:31:59 +08:00
Wavefront starts execution
with memory violation
exceptions exceptions
enabled which are generated
when a memory violation has
2018-03-08 13:46:01 +08:00
occurred for this wavefront from
2017-06-07 04:31:59 +08:00
L1 or LDS
(write-to-read-only-memory,
mis-aligned atomic, LDS
address out of range,
illegal address, etc.).
CP sets the memory
violation bit in
`` COMPUTE_PGM_RSRC2.EXCP_EN_MSB ``
according to what the
runtime requests.
2017-10-15 03:17:08 +08:00
23:15 9 bits GRANULATED_LDS_SIZE Must be 0.
2017-06-07 04:31:59 +08:00
CP uses the rounded value
from the dispatch packet,
not this value, as the
dispatch may contain
dynamically allocated group
segment memory. CP writes
directly to
`` COMPUTE_PGM_RSRC2.LDS_SIZE `` .
Amount of group segment
(LDS) to allocate for each
work-group. Granularity is
device specific:
2021-03-23 06:28:11 +08:00
GFX6
2017-06-07 04:31:59 +08:00
roundup(lds-size / (64 * 4))
2021-03-23 06:28:11 +08:00
GFX7-GFX10
2017-06-07 04:31:59 +08:00
roundup(lds-size / (128 * 4))
2017-10-15 03:17:08 +08:00
24 1 bit ENABLE_EXCEPTION_IEEE_754_FP Wavefront starts execution
_INVALID_OPERATION with specified exceptions
2017-06-07 04:31:59 +08:00
enabled.
Used by CP to set up
`` COMPUTE_PGM_RSRC2.EXCP_EN ``
(set from bits 0..6).
IEEE 754 FP Invalid
Operation
2017-10-15 03:17:08 +08:00
25 1 bit ENABLE_EXCEPTION_FP_DENORMAL FP Denormal one or more
_SOURCE input operands is a
2017-06-07 04:31:59 +08:00
denormal number
2017-10-15 03:17:08 +08:00
26 1 bit ENABLE_EXCEPTION_IEEE_754_FP IEEE 754 FP Division by
_DIVISION_BY_ZERO Zero
27 1 bit ENABLE_EXCEPTION_IEEE_754_FP IEEE 754 FP FP Overflow
_OVERFLOW
28 1 bit ENABLE_EXCEPTION_IEEE_754_FP IEEE 754 FP Underflow
_UNDERFLOW
29 1 bit ENABLE_EXCEPTION_IEEE_754_FP IEEE 754 FP Inexact
_INEXACT
30 1 bit ENABLE_EXCEPTION_INT_DIVIDE_BY Integer Division by Zero
_ZERO (rcp_iflag_f32 instruction
2017-06-07 04:31:59 +08:00
only)
2017-10-19 06:16:55 +08:00
31 1 bit Reserved, must be 0.
2017-06-07 04:31:59 +08:00
32 **Total size 4 bytes.**
2017-06-07 08:46:08 +08:00
======= ===================================================================================================================
2017-06-07 04:31:59 +08:00
2021-02-18 05:37:46 +08:00
..
.. table :: compute_pgm_rsrc3 for GFX90A
:name: amdgpu-amdhsa-compute_pgm_rsrc3-gfx90a-table
======= ======= =============================== ===========================================================================
Bits Size Field Name Description
======= ======= =============================== ===========================================================================
5:0 6 bits ACCUM_OFFSET Offset of a first AccVGPR in the unified register file. Granularity 4.
Value 0-63. 0 - accum-offset = 4, 1 - accum-offset = 8, ...,
63 - accum-offset = 256.
6:15 10 Reserved, must be 0.
bits
16 1 bit TG_SPLIT - If 0 the waves of a work-group are
launched in the same CU.
- If 1 the waves of a work-group can be
launched in different CUs. The waves
cannot use S_BARRIER or LDS.
17:31 15 Reserved, must be 0.
bits
32 **Total size 4 bytes.**
======= ===================================================================================================================
2019-06-14 06:18:47 +08:00
..
.. table :: compute_pgm_rsrc3 for GFX10
:name: amdgpu-amdhsa-compute_pgm_rsrc3-gfx10-table
======= ======= =============================== ===========================================================================
Bits Size Field Name Description
======= ======= =============================== ===========================================================================
3:0 4 bits SHARED_VGPR_COUNT Number of shared VGPRs for wavefront size 64. Granularity 8. Value 0-120.
compute_pgm_rsrc1.vgprs + shared_vgpr_cnt cannot exceed 64.
31:4 28 Reserved, must be 0.
bits
32 **Total size 4 bytes.**
======= ===================================================================================================================
2017-06-07 04:31:59 +08:00
..
.. table :: Floating Point Rounding Mode Enumeration Values
:name: amdgpu-amdhsa-floating-point-rounding-mode-enumeration-values-table
2017-10-15 03:17:08 +08:00
====================================== ===== ==============================
Enumeration Name Value Description
====================================== ===== ==============================
2018-06-13 02:02:46 +08:00
FLOAT_ROUND_MODE_NEAR_EVEN 0 Round Ties To Even
FLOAT_ROUND_MODE_PLUS_INFINITY 1 Round Toward +infinity
FLOAT_ROUND_MODE_MINUS_INFINITY 2 Round Toward -infinity
FLOAT_ROUND_MODE_ZERO 3 Round Toward 0
2017-10-15 03:17:08 +08:00
====================================== ===== ==============================
2017-06-07 04:31:59 +08:00
..
.. table :: Floating Point Denorm Mode Enumeration Values
:name: amdgpu-amdhsa-floating-point-denorm-mode-enumeration-values-table
2017-10-15 03:17:08 +08:00
====================================== ===== ==============================
Enumeration Name Value Description
====================================== ===== ==============================
2018-06-13 02:02:46 +08:00
FLOAT_DENORM_MODE_FLUSH_SRC_DST 0 Flush Source and Destination
2017-10-15 03:17:08 +08:00
Denorms
2018-06-13 02:02:46 +08:00
FLOAT_DENORM_MODE_FLUSH_DST 1 Flush Output Denorms
FLOAT_DENORM_MODE_FLUSH_SRC 2 Flush Source Denorms
FLOAT_DENORM_MODE_FLUSH_NONE 3 No Flush
2017-10-15 03:17:08 +08:00
====================================== ===== ==============================
2017-06-07 04:31:59 +08:00
..
.. table :: System VGPR Work-Item ID Enumeration Values
:name: amdgpu-amdhsa-system-vgpr-work-item-id-enumeration-values-table
2017-10-15 03:17:08 +08:00
======================================== ===== ============================
Enumeration Name Value Description
======================================== ===== ============================
2018-06-13 02:02:46 +08:00
SYSTEM_VGPR_WORKITEM_ID_X 0 Set work-item X dimension
2017-10-15 03:17:08 +08:00
ID.
2018-06-13 02:02:46 +08:00
SYSTEM_VGPR_WORKITEM_ID_X_Y 1 Set work-item X and Y
2017-10-15 03:17:08 +08:00
dimensions ID.
2018-06-13 02:02:46 +08:00
SYSTEM_VGPR_WORKITEM_ID_X_Y_Z 2 Set work-item X, Y and Z
2017-10-15 03:17:08 +08:00
dimensions ID.
2018-06-13 02:02:46 +08:00
SYSTEM_VGPR_WORKITEM_ID_UNDEFINED 3 Undefined.
2017-10-15 03:17:08 +08:00
======================================== ===== ============================
2017-06-07 04:31:59 +08:00
.. _amdgpu-amdhsa-initial-kernel-execution-state:
Initial Kernel Execution State
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
This section defines the register state that will be set up by the packet
processor prior to the start of execution of every wavefront. This is limited by
the constraints of the hardware controllers of CP/ADC/SPI.
The order of the SGPR registers is defined, but the compiler can specify which
ones are actually setup in the kernel descriptor using the `` enable_sgpr_* `` bit
fields (see :ref: `amdgpu-amdhsa-kernel-descriptor` ). The register numbers used
for enabled registers are dense starting at SGPR0: the first enabled register is
SGPR0, the next enabled register is SGPR1 etc.; disabled registers do not have
an SGPR number.
The initial SGPRs comprise up to 16 User SRGPs that are set by CP and apply to
2019-12-12 13:47:56 +08:00
all wavefronts of the grid. It is possible to specify more than 16 User SGPRs
using the `` enable_sgpr_* `` bit fields, in which case only the first 16 are
actually initialized. These are then immediately followed by the System SGPRs
that are set up by ADC/SPI and can have different values for each wavefront of
the grid dispatch.
2017-06-07 04:31:59 +08:00
SGPR register initial state is defined in
:ref: `amdgpu-amdhsa-sgpr-register-set-up-order-table` .
.. table :: SGPR Register Set Up Order
:name: amdgpu-amdhsa-sgpr-register-set-up-order-table
========== ========================== ====== ==============================
SGPR Order Name Number Description
(kernel descriptor enable of
field) SGPRs
========== ========================== ====== ==============================
2020-12-17 10:45:47 +08:00
First Private Segment Buffer 4 See
(enable_sgpr_private :ref: `amdgpu-amdhsa-kernel-prolog-private-segment-buffer` .
2020-12-15 09:42:25 +08:00
_segment_buffer)
2019-12-12 13:47:56 +08:00
then Dispatch Ptr 2 64-bit address of AQL dispatch
2017-06-07 04:31:59 +08:00
(enable_sgpr_dispatch_ptr) packet for kernel dispatch
actually executing.
2019-12-12 13:47:56 +08:00
then Queue Ptr 2 64-bit address of amd_queue_t
2017-06-07 04:31:59 +08:00
(enable_sgpr_queue_ptr) object for AQL queue on which
the dispatch packet was
queued.
2019-12-12 13:47:56 +08:00
then Kernarg Segment Ptr 2 64-bit address of Kernarg
2017-06-07 04:31:59 +08:00
(enable_sgpr_kernarg segment. This is directly
_segment_ptr) copied from the
kernarg_address in the kernel
dispatch packet.
Having CP load it once avoids
loading it at the beginning of
every wavefront.
2019-12-12 13:47:56 +08:00
then Dispatch Id 2 64-bit Dispatch ID of the
2017-06-07 04:31:59 +08:00
(enable_sgpr_dispatch_id) dispatch packet being
executed.
2020-12-17 10:45:47 +08:00
then Flat Scratch Init 2 See
2021-03-24 06:38:10 +08:00
(enable_sgpr_flat_scratch :ref: `amdgpu-amdhsa-kernel-prolog-flat-scratch` .
_init)
2019-12-12 13:47:56 +08:00
then Private Segment Size 1 The 32-bit byte size of a
2021-03-24 06:38:10 +08:00
(enable_sgpr_private single work-item's memory
_segment_size) allocation. This is the
2017-11-11 04:51:43 +08:00
value from the kernel
dispatch packet Private
Segment Byte Size rounded up
by CP to a multiple of
DWORD.
2017-06-07 04:31:59 +08:00
Having CP load it once avoids
loading it at the beginning of
every wavefront.
This is not used for
GFX7-GFX8 since it is the same
value as the second SGPR of
Flat Scratch Init. However, it
2019-06-14 06:18:47 +08:00
may be needed for GFX9-GFX10 which
2017-06-07 04:31:59 +08:00
changes the meaning of the
Flat Scratch Init value.
2019-12-12 13:47:56 +08:00
then Work-Group Id X 1 32-bit work-group id in X
2017-06-07 04:31:59 +08:00
(enable_sgpr_workgroup_id dimension of grid for
_X) wavefront.
2019-12-12 13:47:56 +08:00
then Work-Group Id Y 1 32-bit work-group id in Y
2017-06-07 04:31:59 +08:00
(enable_sgpr_workgroup_id dimension of grid for
_Y) wavefront.
2019-12-12 13:47:56 +08:00
then Work-Group Id Z 1 32-bit work-group id in Z
2017-06-07 04:31:59 +08:00
(enable_sgpr_workgroup_id dimension of grid for
_Z) wavefront.
2018-03-08 13:46:01 +08:00
then Work-Group Info 1 {first_wavefront, 14'b0000,
2017-06-07 04:31:59 +08:00
(enable_sgpr_workgroup ordered_append_term[10:0],
2018-03-08 13:46:01 +08:00
_info) threadgroup_size_in_wavefronts[5:0]}
2020-12-17 10:45:47 +08:00
then Scratch Wavefront Offset 1 See
(enable_sgpr_private :ref: `amdgpu-amdhsa-kernel-prolog-flat-scratch` .
_segment_wavefront_offset) and
:ref: `amdgpu-amdhsa-kernel-prolog-private-segment-buffer` .
2017-06-07 04:31:59 +08:00
========== ========================== ====== ==============================
The order of the VGPR registers is defined, but the compiler can specify which
ones are actually setup in the kernel descriptor using the `` enable_vgpr* `` bit
fields (see :ref: `amdgpu-amdhsa-kernel-descriptor` ). The register numbers used
for enabled registers are dense starting at VGPR0: the first enabled register is
VGPR0, the next enabled register is VGPR1 etc.; disabled registers do not have a
VGPR number.
2021-02-18 05:37:46 +08:00
There are different methods used for the VGPR initial state:
2017-06-07 04:31:59 +08:00
2021-02-18 05:37:46 +08:00
* Unless the *Target Properties* column of :ref: `amdgpu-processor-table`
specifies otherwise, a separate VGPR register is used per work-item ID. The
VGPR register initial state for this method is defined in
:ref: `amdgpu-amdhsa-vgpr-register-set-up-order-for-unpacked-work-item-id-method-table` .
* If *Target Properties* column of :ref: `amdgpu-processor-table`
specifies *Packed work-item IDs* , the initial value of VGPR0 register is used
for all work-item IDs. The register layout for this method is defined in
:ref: `amdgpu-amdhsa-register-layout-for-packed-work-item-id-method-table` .
.. table :: VGPR Register Set Up Order for Unpacked Work-Item ID Method
:name: amdgpu-amdhsa-vgpr-register-set-up-order-for-unpacked-work-item-id-method-table
2017-06-07 04:31:59 +08:00
========== ========================== ====== ==============================
VGPR Order Name Number Description
(kernel descriptor enable of
field) VGPRs
========== ========================== ====== ==============================
2020-10-22 15:15:41 +08:00
First Work-Item Id X 1 32-bit work-item id in X
2017-06-07 04:31:59 +08:00
(Always initialized) dimension of work-group for
wavefront lane.
2020-10-22 15:15:41 +08:00
then Work-Item Id Y 1 32-bit work-item id in Y
2017-06-07 04:31:59 +08:00
(enable_vgpr_workitem_id dimension of work-group for
> 0) wavefront lane.
2020-10-22 15:15:41 +08:00
then Work-Item Id Z 1 32-bit work-item id in Z
2017-06-07 04:31:59 +08:00
(enable_vgpr_workitem_id dimension of work-group for
> 1) wavefront lane.
========== ========================== ====== ==============================
2021-02-18 05:37:46 +08:00
..
.. table :: Register Layout for Packed Work-Item ID Method
:name: amdgpu-amdhsa-register-layout-for-packed-work-item-id-method-table
======= ======= ================ =========================================
Bits Size Field Name Description
======= ======= ================ =========================================
0:9 10 bits Work-Item Id X Work-item id in X
dimension of work-group for
wavefront lane.
Always initialized.
10:19 10 bits Work-Item Id Y Work-item id in Y
dimension of work-group for
wavefront lane.
Initialized if enable_vgpr_workitem_id >
0, otherwise set to 0.
20:29 10 bits Work-Item Id Z Work-item id in Z
dimension of work-group for
wavefront lane.
Initialized if enable_vgpr_workitem_id >
1, otherwise set to 0.
30:31 2 bits Reserved, set to 0.
======= ======= ================ =========================================
2018-04-12 13:53:20 +08:00
The setting of registers is done by GPU CP/ADC/SPI hardware as follows:
2017-06-07 04:31:59 +08:00
1. SGPRs before the Work-Group Ids are set by CP using the 16 User Data
registers.
2. Work-group Id registers X, Y, Z are set by ADC which supports any
combination including none.
2018-03-08 13:46:01 +08:00
3. Scratch Wavefront Offset is set by SPI in a per wavefront basis which is why
2020-02-20 15:08:59 +08:00
its value cannot be included with the flat scratch init value which is per
2020-12-17 10:45:47 +08:00
queue (see :ref: `amdgpu-amdhsa-kernel-prolog-flat-scratch` ).
2017-06-07 04:31:59 +08:00
4. The VGPRs are set by SPI which only supports specifying either (X), (X, Y)
or (X, Y, Z).
2020-12-17 10:45:47 +08:00
5. Flat Scratch register pair initialization is described in
:ref: `amdgpu-amdhsa-kernel-prolog-flat-scratch` .
2017-06-07 04:31:59 +08:00
The global segment can be accessed either using buffer instructions (GFX6 which
2019-12-12 13:47:56 +08:00
has V# 64-bit address support), flat instructions (GFX7-GFX10), or global
2019-06-14 06:18:47 +08:00
instructions (GFX9-GFX10).
2017-06-07 04:31:59 +08:00
2020-02-20 15:08:59 +08:00
If buffer operations are used, then the compiler can generate a V# with the
2017-06-07 04:31:59 +08:00
following properties:
* base address of 0
* no swizzle
* ATC: 1 if IOMMU present (such as APU)
* ptr64: 1
* MTYPE set to support memory coherence that matches the runtime (such as CC for
APU and NC for dGPU).
.. _amdgpu-amdhsa-kernel-prolog:
Kernel Prolog
~~~~~~~~~~~~~
2020-01-22 06:27:57 +08:00
The compiler performs initialization in the kernel prologue depending on the
target and information about things like stack usage in the kernel and called
functions. Some of this initialization requires the compiler to request certain
User and System SGPRs be present in the
:ref: `amdgpu-amdhsa-initial-kernel-execution-state` via the
:ref: `amdgpu-amdhsa-kernel-descriptor` .
.. _amdgpu-amdhsa-kernel-prolog-cfi:
CFI
+++
2020-02-20 15:08:59 +08:00
1. The CFI return address is undefined.
2. The CFI CFA is defined using an expression which evaluates to a location
description that comprises one memory location description for the
`` DW_ASPACE_AMDGPU_private_lane `` address space address `` 0 `` .
2020-01-22 06:27:57 +08:00
.. _amdgpu-amdhsa-kernel-prolog-m0:
2017-06-07 04:31:59 +08:00
M0
++
GFX6-GFX8
The M0 register must be initialized with a value at least the total LDS size
if the kernel may access LDS via DS or flat operations. Total LDS size is
available in dispatch packet. For M0, it is also possible to use maximum
possible value of LDS for given target (0x7FFF for GFX6 and 0xFFFF for
GFX7-GFX8).
2019-06-14 06:18:47 +08:00
GFX9-GFX10
2017-06-07 04:31:59 +08:00
The M0 register is not used for range checking LDS accesses and so does not
need to be initialized in the prolog.
2020-01-22 06:27:57 +08:00
.. _amdgpu-amdhsa-kernel-prolog-stack-pointer:
Stack Pointer
+++++++++++++
If the kernel has function calls it must set up the ABI stack pointer described
2020-02-20 15:08:59 +08:00
in :ref: `amdgpu-amdhsa-function-call-convention-non-kernel-functions` by setting
SGPR32 to the unswizzled scratch offset of the address past the last local
allocation.
2020-01-22 06:27:57 +08:00
.. _amdgpu-amdhsa-kernel-prolog-frame-pointer:
Frame Pointer
+++++++++++++
If the kernel needs a frame pointer for the reasons defined in
2020-03-04 09:39:47 +08:00
`` SIFrameLowering `` then SGPR33 is used and is always set to `` 0 `` in the
2020-01-22 06:27:57 +08:00
kernel prolog. If a frame pointer is not required then all uses of the frame
pointer are replaced with immediate `` 0 `` offsets.
.. _amdgpu-amdhsa-kernel-prolog-flat-scratch:
2017-06-07 04:31:59 +08:00
Flat Scratch
++++++++++++
2020-12-17 10:45:47 +08:00
There are different methods used for initializing flat scratch:
* If the *Target Properties* column of :ref: `amdgpu-processor-table`
specifies *Does not support generic address space* :
Flat scratch is not supported and there is no flat scratch register pair.
* If the *Target Properties* column of :ref: `amdgpu-processor-table`
specifies *Offset flat scratch* :
If the kernel or any function it calls may use flat operations to access
scratch memory, the prolog code must set up the FLAT_SCRATCH register pair
(FLAT_SCRATCH_LO/FLAT_SCRATCH_HI). Initialization uses Flat Scratch Init and
Scratch Wavefront Offset SGPR registers (see
:ref: `amdgpu-amdhsa-initial-kernel-execution-state` ):
1. The low word of Flat Scratch Init is the 32-bit byte offset from
`` SH_HIDDEN_PRIVATE_BASE_VIMID `` to the base of scratch backing memory
being managed by SPI for the queue executing the kernel dispatch. This is
the same value used in the Scratch Segment Buffer V# base address.
CP obtains this from the runtime. (The Scratch Segment Buffer base address
is `` SH_HIDDEN_PRIVATE_BASE_VIMID `` plus this offset.)
The prolog must add the value of Scratch Wavefront Offset to get the
wavefront's byte scratch backing memory offset from
`` SH_HIDDEN_PRIVATE_BASE_VIMID `` .
The Scratch Wavefront Offset must also be used as an offset with Private
segment address when using the Scratch Segment Buffer.
Since FLAT_SCRATCH_LO is in units of 256 bytes, the offset must be right
shifted by 8 before moving into FLAT_SCRATCH_HI.
FLAT_SCRATCH_HI corresponds to SGPRn-4 on GFX7, and SGPRn-6 on GFX8 (where
SGPRn is the highest numbered SGPR allocated to the wavefront).
FLAT_SCRATCH_HI is multiplied by 256 (as it is in units of 256 bytes) and
added to `` SH_HIDDEN_PRIVATE_BASE_VIMID `` to calculate the per wavefront
FLAT SCRATCH BASE in flat memory instructions that access the scratch
aperture.
2. The second word of Flat Scratch Init is 32-bit byte size of a single
work-items scratch memory usage.
CP obtains this from the runtime, and it is always a multiple of DWORD. CP
checks that the value in the kernel dispatch packet Private Segment Byte
Size is not larger and requests the runtime to increase the queue's scratch
size if necessary.
CP directly loads from the kernel dispatch packet Private Segment Byte Size
field and rounds up to a multiple of DWORD. Having CP load it once avoids
loading it at the beginning of every wavefront.
The kernel prolog code must move it to FLAT_SCRATCH_LO which is SGPRn-3 on
GFX7 and SGPRn-5 on GFX8. FLAT_SCRATCH_LO is used as the FLAT SCRATCH SIZE
in flat memory instructions.
* If the *Target Properties* column of :ref: `amdgpu-processor-table`
specifies *Absolute flat scratch* :
2017-06-07 04:31:59 +08:00
2020-12-15 09:42:25 +08:00
If the kernel or any function it calls may use flat operations to access
scratch memory, the prolog code must set up the FLAT_SCRATCH register pair
(FLAT_SCRATCH_LO/FLAT_SCRATCH_HI which are in SGPRn-4/SGPRn-3). Initialization
uses Flat Scratch Init and Scratch Wavefront Offset SGPR registers (see
:ref: `amdgpu-amdhsa-initial-kernel-execution-state` ):
2020-12-17 10:45:47 +08:00
The Flat Scratch Init is the 64-bit address of the base of scratch backing
memory being managed by SPI for the queue executing the kernel dispatch.
CP obtains this from the runtime.
The kernel prolog must add the value of the wave's Scratch Wavefront Offset
and move the result as a 64-bit value to the FLAT_SCRATCH SGPR register pair
which is SGPRn-6 and SGPRn-5. It is used as the FLAT SCRATCH BASE in flat
memory instructions.
The Scratch Wavefront Offset must also be used as an offset with Private
segment address when using the Scratch Segment Buffer (see
:ref: `amdgpu-amdhsa-kernel-prolog-private-segment-buffer` ).
2017-06-07 04:31:59 +08:00
2020-01-22 06:27:57 +08:00
.. _amdgpu-amdhsa-kernel-prolog-private-segment-buffer:
Private Segment Buffer
++++++++++++++++++++++
2021-02-28 02:09:23 +08:00
Private Segment Buffer SGPR register is used to initialize 4 SGPRs
2020-12-17 10:45:47 +08:00
that are used as a V# to access scratch. CP uses the value provided by the
runtime. It is used, together with Scratch Wavefront Offset as an offset, to
access the private memory space using a segment address. See
:ref: `amdgpu-amdhsa-initial-kernel-execution-state` .
The scratch V# is a four-aligned SGPR and always selected for the kernel as
follows:
2020-01-22 06:27:57 +08:00
2020-02-20 15:08:59 +08:00
- If it is known during instruction selection that there is stack usage,
2020-01-22 06:27:57 +08:00
SGPR0-3 is reserved for use as the scratch V#. Stack usage is assumed if
2020-02-20 15:08:59 +08:00
optimizations are disabled (`` -O0 `` ), if stack objects already exist (for
2020-01-22 06:27:57 +08:00
locals, etc.), or if there are any function calls.
- Otherwise, four high numbered SGPRs beginning at a four-aligned SGPR index
are reserved for the tentative scratch V#. These will be used if it is
determined that spilling is needed.
2020-02-20 15:08:59 +08:00
- If no use is made of the tentative scratch V#, then it is unreserved,
2020-01-22 06:27:57 +08:00
and the register count is determined ignoring it.
2020-02-20 15:08:59 +08:00
- If use is made of the tentative scratch V#, then its register numbers
2020-01-22 06:27:57 +08:00
are shifted to the first four-aligned SGPR index after the highest one
allocated by the register allocator, and all uses are updated. The
register count includes them in the shifted location.
- In either case, if the processor has the SGPR allocation bug, the
tentative allocation is not shifted or unreserved in order to ensure
the register count is higher to workaround the bug.
.. note ::
This approach of using a tentative scratch V# and shifting the register
numbers if used avoids having to perform register allocation a second
time if the tentative V# is eliminated. This is more efficient and
avoids the problem that the second register allocation may perform
spilling which will fail as there is no longer a scratch V#.
When the kernel prolog code is being emitted it is known whether the scratch V#
described above is actually used. If it is, the prolog code must set it up by
copying the Private Segment Buffer to the scratch V# registers and then adding
the Private Segment Wavefront Offset to the queue base address in the V#. The
result is a V# with a base address pointing to the beginning of the wavefront
scratch backing memory.
The Private Segment Buffer is always requested, but the Private Segment
Wavefront Offset is only requested if it is used (see
:ref: `amdgpu-amdhsa-initial-kernel-execution-state` ).
2017-06-07 04:31:59 +08:00
.. _amdgpu-amdhsa-memory-model:
Memory Model
~~~~~~~~~~~~
2020-10-31 12:43:55 +08:00
This section describes the mapping of the LLVM memory model onto AMDGPU machine
code (see :ref: `memmodel` ).
2017-06-07 04:31:59 +08:00
The AMDGPU backend supports the memory synchronization scopes specified in
:ref: `amdgpu-memory-scopes` .
2020-10-31 12:43:55 +08:00
The code sequences used to implement the memory model specify the order of
instructions that a single thread must execute. The `` s_waitcnt `` and cache
management instructions such as `` buffer_wbinvl1_vol `` are defined with respect
2017-06-07 04:31:59 +08:00
to other memory instructions executed by the same thread. This allows them to be
moved earlier or later which can allow them to be combined with other instances
2020-10-31 12:43:55 +08:00
of the same instruction, or hoisted/sunk out of loops to improve performance.
Only the instructions related to the memory model are given; additional
`` s_waitcnt `` instructions are required to ensure registers are defined before
being used. These may be able to be combined with the memory model `` s_waitcnt ``
instructions as described above.
2017-06-07 04:31:59 +08:00
2017-10-19 06:16:55 +08:00
The AMDGPU backend supports the following memory models:
HSA Memory Model [HSA]_
The HSA memory model uses a single happens-before relation for all address
spaces (see :ref: `amdgpu-address-spaces` ).
OpenCL Memory Model [OpenCL]_
The OpenCL memory model which has separate happens-before relations for the
global and local address spaces. Only a fence specifying both global and
local address space, and seq_cst instructions join the relationships. Since
the LLVM `` memfence `` instruction does not allow an address space to be
2019-12-12 13:47:56 +08:00
specified the OpenCL fence has to conservatively assume both local and
2017-10-19 06:16:55 +08:00
global address space was specified. However, optimizations can often be
done to eliminate the additional `` s_waitcnt `` instructions when there are
no intervening memory instructions which access the corresponding address
space. The code sequences in the table indicate what can be omitted for the
OpenCL memory. The target triple environment is used to determine if the
source language is OpenCL (see :ref: `amdgpu-opencl` ).
2017-06-07 04:31:59 +08:00
`` ds/flat_load/store/atomic `` instructions to local memory are termed LDS
operations.
`` buffer/global/flat_load/store/atomic `` instructions to global memory are
termed vector memory operations.
2020-10-31 12:43:55 +08:00
Private address space uses `` buffer_load/store `` using the scratch V#
(GFX6-GFX8), or `` scratch_load/store `` (GFX9-GFX10). Since only a single thread
is accessing the memory, atomic memory orderings are not meaningful, and all
accesses are treated as non-atomic.
Constant address space uses `` buffer/global_load `` instructions (or equivalent
scalar memory instructions). Since the constant address space contents do not
change during the execution of a kernel dispatch it is not legal to perform
stores, and atomic memory orderings are not meaningful, and all accesses are
treated as non-atomic.
A memory synchronization scope wider than work-group is not meaningful for the
group (LDS) address space and is treated as work-group.
The memory model does not support the region address space which is treated as
non-atomic.
Acquire memory ordering is not meaningful on store atomic instructions and is
treated as non-atomic.
Release memory ordering is not meaningful on load atomic instructions and is
treated a non-atomic.
Acquire-release memory ordering is not meaningful on load or store atomic
instructions and is treated as acquire and release respectively.
The memory order also adds the single thread optimization constraints defined in
table
:ref: `amdgpu-amdhsa-memory-model-single-thread-optimization-constraints-table` .
.. table :: AMDHSA Memory Model Single Thread Optimization Constraints
:name: amdgpu-amdhsa-memory-model-single-thread-optimization-constraints-table
============ ==============================================================
LLVM Memory Optimization Constraints
Ordering
============ ==============================================================
unordered *none*
monotonic *none*
acquire - If a load atomic/atomicrmw then no following load/load
atomic/store/store atomic/atomicrmw/fence instruction can be
moved before the acquire.
- If a fence then same as load atomic, plus no preceding
associated fence-paired-atomic can be moved after the fence.
release - If a store atomic/atomicrmw then no preceding load/load
atomic/store/store atomic/atomicrmw/fence instruction can be
moved after the release.
- If a fence then same as store atomic, plus no following
associated fence-paired-atomic can be moved before the
fence.
acq_rel Same constraints as both acquire and release.
seq_cst - If a load atomic then same constraints as acquire, plus no
preceding sequentially consistent load atomic/store
atomic/atomicrmw/fence instruction can be moved after the
seq_cst.
- If a store atomic then the same constraints as release, plus
no following sequentially consistent load atomic/store
atomic/atomicrmw/fence instruction can be moved before the
seq_cst.
- If an atomicrmw/fence then same constraints as acq_rel.
============ ==============================================================
The code sequences used to implement the memory model are defined in the
following sections:
* :ref: `amdgpu-amdhsa-memory-model-gfx6-gfx9`
2021-02-18 05:37:46 +08:00
* :ref: `amdgpu-amdhsa-memory-model-gfx90a`
2020-10-31 12:43:55 +08:00
* :ref: `amdgpu-amdhsa-memory-model-gfx10`
.. _amdgpu-amdhsa-memory-model-gfx6-gfx9:
Memory Model GFX6-GFX9
++++++++++++++++++++++
2017-06-07 04:31:59 +08:00
For GFX6-GFX9:
2019-06-14 06:18:47 +08:00
* Each agent has multiple shader arrays (SA).
* Each SA has multiple compute units (CU).
2017-06-07 04:31:59 +08:00
* Each CU has multiple SIMDs that execute wavefronts.
* The wavefronts for a single work-group are executed in the same CU but may be
executed by different SIMDs.
* Each CU has a single LDS memory shared by the wavefronts of the work-groups
executing on it.
* All LDS operations of a CU are performed as wavefront wide operations in a
global order and involve no caching. Completion is reported to a wavefront in
execution order.
* The LDS memory has multiple request queues shared by the SIMDs of a
2019-12-12 13:47:56 +08:00
CU. Therefore, the LDS operations performed by different wavefronts of a
work-group can be reordered relative to each other, which can result in
reordering the visibility of vector memory operations with respect to LDS
operations of other wavefronts in the same work-group. A `` s_waitcnt
lgkmcnt(0)`` is required to ensure synchronization between LDS operations and
vector memory operations between wavefronts of a work-group, but not between
operations performed by the same wavefront.
2017-06-07 04:31:59 +08:00
* The vector memory operations are performed as wavefront wide operations and
completion is reported to a wavefront in execution order. The exception is
2017-11-10 09:00:54 +08:00
that for GFX7-GFX9 `` flat_load/store/atomic `` instructions can report out of
2017-06-07 04:31:59 +08:00
vector memory order if they access LDS memory, and out of LDS operation order
if they access global memory.
2017-10-19 06:16:55 +08:00
* The vector memory operations access a single vector L1 cache shared by all
SIMDs a CU. Therefore, no special action is required for coherence between the
lanes of a single wavefront, or for coherence between wavefronts in the same
2019-12-12 13:47:56 +08:00
work-group. A `` buffer_wbinvl1_vol `` is required for coherence between
wavefronts executing in different work-groups as they may be executing on
different CUs.
2017-06-07 04:31:59 +08:00
* The scalar memory operations access a scalar L1 cache shared by all wavefronts
on a group of CUs. The scalar and vector L1 caches are not coherent. However,
scalar operations are used in a restricted way so do not impact the memory
2020-10-22 15:15:41 +08:00
model. See :ref: `amdgpu-amdhsa-memory-spaces` .
2017-06-07 04:31:59 +08:00
* The vector and scalar memory operations use an L2 cache shared by all CUs on
the same agent.
* The L2 cache has independent channels to service disjoint ranges of virtual
addresses.
* Each CU has a separate request queue per channel. Therefore, the vector and
2019-12-12 13:47:56 +08:00
scalar memory operations performed by wavefronts executing in different
work-groups (which may be executing on different CUs) of an agent can be
reordered relative to each other. A `` s_waitcnt vmcnt(0) `` is required to
ensure synchronization between vector memory operations of different CUs. It
ensures a previous vector memory operation has completed before executing a
subsequent vector memory or LDS operation and so can be used to meet the
requirements of acquire and release.
2017-06-07 04:31:59 +08:00
* The L2 cache can be kept coherent with other agents on some targets, or ranges
of virtual addresses can be set up to bypass it to ensure system coherence.
2020-10-31 12:43:55 +08:00
Scalar memory operations are only used to access memory that is proven to not
change during the execution of the kernel dispatch. This includes constant
2020-11-03 10:14:45 +08:00
address space and global address space for program scope `` const `` variables.
2020-10-31 12:43:55 +08:00
Therefore, the kernel machine code does not have to maintain the scalar cache to
ensure it is coherent with the vector caches. The scalar and vector caches are
invalidated between kernel dispatches by CP since constant address space data
may change between kernel dispatch executions. See
2020-10-22 15:15:41 +08:00
:ref: `amdgpu-amdhsa-memory-spaces` .
2017-06-07 04:31:59 +08:00
2019-12-12 13:47:56 +08:00
The one exception is if scalar writes are used to spill SGPR registers. In this
2017-06-07 04:31:59 +08:00
case the AMDGPU backend ensures the memory location used to spill is never
accessed by vector memory operations at the same time. If scalar writes are used
then a `` s_dcache_wb `` is inserted before the `` s_endpgm `` and before a function
return since the locations may be used for vector memory instructions by a
2019-12-12 13:47:56 +08:00
future wavefront that uses the same scratch area, or a function call that
creates a frame at the same address, respectively. There is no need for a
`` s_dcache_inv `` as all scalar writes are write-before-read in the same thread.
2017-06-07 04:31:59 +08:00
2020-10-31 12:43:55 +08:00
For kernarg backing memory:
2019-06-14 06:18:47 +08:00
2020-10-31 12:43:55 +08:00
* CP invalidates the L1 cache at the start of each kernel dispatch.
* On dGPU the kernarg backing memory is allocated in host memory accessed as
MTYPE UC (uncached) to avoid needing to invalidate the L2 cache. This also
causes it to be treated as non-volatile and so is not invalidated by
`` *_vol `` .
* On APU the kernarg backing memory it is accessed as MTYPE CC (cache coherent)
and so the L2 cache will be coherent with the CPU and other agents.
2019-06-14 06:18:47 +08:00
2020-10-31 12:43:55 +08:00
Scratch backing memory (which is used for the private address space) is accessed
with MTYPE NC_NV (non-coherent non-volatile). Since the private address space is
only accessed by a single thread, and is always write-before-read, there is
never a need to invalidate these entries from the L1 cache. Hence all cache
invalidates are done as `` *_vol `` to only invalidate the volatile cache lines.
2019-06-14 06:18:47 +08:00
2020-10-31 12:43:55 +08:00
The code sequences used to implement the memory model for GFX6-GFX9 are defined
in table :ref: `amdgpu-amdhsa-memory-model-code-sequences-gfx6-gfx9-table` .
2017-06-07 04:31:59 +08:00
2020-10-31 12:43:55 +08:00
.. table :: AMDHSA Memory Model Code Sequences GFX6-GFX9
:name: amdgpu-amdhsa-memory-model-code-sequences-gfx6-gfx9-table
2017-06-07 04:31:59 +08:00
2020-10-31 12:43:55 +08:00
============ ============ ============== ========== ================================
LLVM Instr LLVM Memory LLVM Memory AMDGPU AMDGPU Machine Code
2021-01-27 01:45:41 +08:00
Ordering Sync Scope Address GFX6-GFX9
2017-06-07 04:31:59 +08:00
Space
2020-10-31 12:43:55 +08:00
============ ============ ============== ========== ================================
2017-06-07 04:31:59 +08:00
**Non-Atomic**
2020-10-31 12:43:55 +08:00
------------------------------------------------------------------------------------
load *none* *none* - global - !volatile & !nontemporal
2017-10-19 06:16:55 +08:00
- generic
2020-10-31 12:43:55 +08:00
- private 1. buffer/global/flat_load
2017-10-19 06:16:55 +08:00
- constant
2021-01-09 18:20:42 +08:00
- !volatile & nontemporal
1. buffer/global/flat_load
glc=1 slc=1
- volatile
2017-10-19 06:16:55 +08:00
2020-10-31 12:43:55 +08:00
1. buffer/global/flat_load
glc=1
2020-12-19 10:05:11 +08:00
2. s_waitcnt vmcnt(0)
- Must happen before
any following volatile
global/generic
load/store.
- Ensures that
volatile
operations to
different
addresses will not
be reordered by
hardware.
2020-10-31 12:43:55 +08:00
load *none* *none* - local 1. ds_load
2020-12-19 10:05:11 +08:00
store *none* *none* - global - !volatile & !nontemporal
2017-06-07 04:31:59 +08:00
- generic
2020-10-31 12:43:55 +08:00
- private 1. buffer/global/flat_store
2017-10-19 06:16:55 +08:00
- constant
2021-01-09 18:20:42 +08:00
- !volatile & nontemporal
1. buffer/global/flat_store
glc=1 slc=1
- volatile
2020-12-19 10:05:11 +08:00
1. buffer/global/flat_store
2. s_waitcnt vmcnt(0)
- Must happen before
any following volatile
global/generic
load/store.
- Ensures that
volatile
operations to
different
addresses will not
be reordered by
hardware.
2020-10-31 12:43:55 +08:00
store *none* *none* - local 1. ds_store
2017-06-07 04:31:59 +08:00
**Unordered Atomic**
2020-10-31 12:43:55 +08:00
------------------------------------------------------------------------------------
load atomic unordered *any* *any* *Same as non-atomic* .
store atomic unordered *any* *any* *Same as non-atomic* .
2020-11-03 10:14:45 +08:00
atomicrmw unordered *any* *any* *Same as monotonic atomic* .
2017-06-07 04:31:59 +08:00
**Monotonic Atomic**
2020-10-31 12:43:55 +08:00
------------------------------------------------------------------------------------
load atomic monotonic - singlethread - global 1. buffer/global/ds/flat_load
- wavefront - local
- workgroup - generic
load atomic monotonic - agent - global 1. buffer/global/flat_load
- system - generic glc=1
store atomic monotonic - singlethread - global 1. buffer/global/flat_store
2017-06-07 04:31:59 +08:00
- wavefront - generic
- workgroup
- agent
- system
2020-10-31 12:43:55 +08:00
store atomic monotonic - singlethread - local 1. ds_store
2017-06-07 04:31:59 +08:00
- wavefront
- workgroup
2020-10-31 12:43:55 +08:00
atomicrmw monotonic - singlethread - global 1. buffer/global/flat_atomic
2017-06-07 04:31:59 +08:00
- wavefront - generic
- workgroup
- agent
- system
2020-10-31 12:43:55 +08:00
atomicrmw monotonic - singlethread - local 1. ds_atomic
2017-06-07 04:31:59 +08:00
- wavefront
- workgroup
**Acquire Atomic**
2020-10-31 12:43:55 +08:00
------------------------------------------------------------------------------------
load atomic acquire - singlethread - global 1. buffer/global/ds/flat_load
2017-06-07 04:31:59 +08:00
- wavefront - local
- generic
2020-10-31 12:43:55 +08:00
load atomic acquire - workgroup - global 1. buffer/global_load
load atomic acquire - workgroup - local 1. ds/flat_load
- generic 2. s_waitcnt lgkmcnt(0)
- If OpenCL, omit.
- Must happen before
any following
global/generic
load/load
atomic/store/store
atomic/atomicrmw.
- Ensures any
following global
data read is no
older than a local load
atomic value being
acquired.
load atomic acquire - agent - global 1. buffer/global_load
- system glc=1
2. s_waitcnt vmcnt(0)
- Must happen before
following
buffer_wbinvl1_vol.
- Ensures the load
has completed
before invalidating
the cache.
3. buffer_wbinvl1_vol
- Must happen before
any following
global/generic
load/load
atomic/atomicrmw.
- Ensures that
following
loads will not see
stale global data.
load atomic acquire - agent - generic 1. flat_load glc=1
- system 2. s_waitcnt vmcnt(0) &
lgkmcnt(0)
- If OpenCL omit
lgkmcnt(0).
- Must happen before
following
buffer_wbinvl1_vol.
- Ensures the flat_load
has completed
before invalidating
the cache.
3. buffer_wbinvl1_vol
- Must happen before
any following
global/generic
load/load
atomic/atomicrmw.
- Ensures that
following loads
will not see stale
global data.
atomicrmw acquire - singlethread - global 1. buffer/global/ds/flat_atomic
2017-06-07 04:31:59 +08:00
- wavefront - local
- generic
2020-10-31 12:43:55 +08:00
atomicrmw acquire - workgroup - global 1. buffer/global_atomic
atomicrmw acquire - workgroup - local 1. ds/flat_atomic
- generic 2. s_waitcnt lgkmcnt(0)
- If OpenCL, omit.
- Must happen before
any following
global/generic
2017-06-07 04:31:59 +08:00
load/load
atomic/store/store
atomic/atomicrmw.
2020-10-31 12:43:55 +08:00
- Ensures any
following global
data read is no
older than a local
atomicrmw value
being acquired.
atomicrmw acquire - agent - global 1. buffer/global_atomic
- system 2. s_waitcnt vmcnt(0)
- Must happen before
following
buffer_wbinvl1_vol.
- Ensures the
atomicrmw has
completed before
invalidating the
cache.
3. buffer_wbinvl1_vol
- Must happen before
any following
global/generic
load/load
atomic/atomicrmw.
- Ensures that
following loads
will not see stale
global data.
atomicrmw acquire - agent - generic 1. flat_atomic
- system 2. s_waitcnt vmcnt(0) &
lgkmcnt(0)
- If OpenCL, omit
lgkmcnt(0).
- Must happen before
following
buffer_wbinvl1_vol.
- Ensures the
atomicrmw has
completed before
invalidating the
cache.
3. buffer_wbinvl1_vol
- Must happen before
any following
global/generic
2017-10-19 06:16:55 +08:00
load/load
atomic/atomicrmw.
2020-10-31 12:43:55 +08:00
- Ensures that
following loads
will not see stale
global data.
fence acquire - singlethread *none* *none*
2017-06-07 04:31:59 +08:00
- wavefront
2020-10-31 12:43:55 +08:00
fence acquire - workgroup *none* 1. s_waitcnt lgkmcnt(0)
- If OpenCL and
address space is
not generic, omit.
- However, since LLVM
currently has no
address space on
the fence need to
conservatively
always generate. If
fence had an
address space then
set to address
space of OpenCL
fence flag, or to
generic if both
local and global
flags are
specified.
2017-06-07 04:31:59 +08:00
- Must happen after
any preceding
local/generic load
atomic/atomicrmw
with an equal or
wider sync scope
and memory ordering
stronger than
unordered (this is
termed the
fence-paired-atomic).
- Must happen before
any following
global/generic
load/load
atomic/store/store
atomic/atomicrmw.
- Ensures any
following global
data read is no
older than the
value read by the
fence-paired-atomic.
2020-10-31 12:43:55 +08:00
fence acquire - agent *none* 1. s_waitcnt lgkmcnt(0) &
- system vmcnt(0)
- If OpenCL and
address space is
not generic, omit
lgkmcnt(0).
- However, since LLVM
currently has no
address space on
the fence need to
conservatively
always generate
(see comment for
previous fence).
2017-06-07 08:08:35 +08:00
- Could be split into
2017-06-07 04:31:59 +08:00
separate s_waitcnt
vmcnt(0) and
s_waitcnt
lgkmcnt(0) to allow
them to be
independently moved
according to the
following rules.
- s_waitcnt vmcnt(0)
must happen after
any preceding
global/generic load
atomic/atomicrmw
with an equal or
wider sync scope
and memory ordering
stronger than
unordered (this is
termed the
fence-paired-atomic).
- s_waitcnt lgkmcnt(0)
must happen after
any preceding
2017-10-19 06:16:55 +08:00
local/generic load
2017-06-07 04:31:59 +08:00
atomic/atomicrmw
with an equal or
wider sync scope
and memory ordering
stronger than
unordered (this is
termed the
fence-paired-atomic).
- Must happen before
the following
buffer_wbinvl1_vol.
- Ensures that the
fence-paired atomic
has completed
before invalidating
the
cache. Therefore
any following
locations read must
be no older than
the value read by
the
fence-paired-atomic.
2020-10-31 12:43:55 +08:00
2. buffer_wbinvl1_vol
- Must happen before any
following global/generic
load/load
atomic/store/store
atomic/atomicrmw.
- Ensures that
following loads
will not see stale
global data.
2017-06-07 04:31:59 +08:00
**Release Atomic**
2020-10-31 12:43:55 +08:00
------------------------------------------------------------------------------------
store atomic release - singlethread - global 1. buffer/global/ds/flat_store
2017-06-07 04:31:59 +08:00
- wavefront - local
- generic
2020-10-31 12:43:55 +08:00
store atomic release - workgroup - global 1. s_waitcnt lgkmcnt(0)
- generic
- If OpenCL, omit.
2017-06-07 04:31:59 +08:00
- Must happen after
any preceding
local/generic
load/store/load
atomic/store
atomic/atomicrmw.
2020-10-31 12:43:55 +08:00
- Must happen before
the following
store.
- Ensures that all
memory operations
to local have
completed before
performing the
store that is being
released.
2. buffer/global/flat_store
store atomic release - workgroup - local 1. ds_store
store atomic release - agent - global 1. s_waitcnt lgkmcnt(0) &
- system - generic vmcnt(0)
- If OpenCL and
address space is
not generic, omit
lgkmcnt(0).
- Could be split into
separate s_waitcnt
vmcnt(0) and
s_waitcnt
lgkmcnt(0) to allow
them to be
independently moved
according to the
following rules.
- s_waitcnt vmcnt(0)
must happen after
any preceding
global/generic
load/store/load
atomic/store
atomic/atomicrmw.
- s_waitcnt lgkmcnt(0)
must happen after
2017-10-19 06:16:55 +08:00
any preceding
local/generic
load/store/load
atomic/store
atomic/atomicrmw.
2020-10-31 12:43:55 +08:00
- Must happen before
the following
store.
- Ensures that all
memory operations
to memory have
completed before
performing the
store that is being
released.
2. buffer/global/flat_store
atomicrmw release - singlethread - global 1. buffer/global/ds/flat_atomic
2017-06-07 04:31:59 +08:00
- wavefront - local
- generic
2020-10-31 12:43:55 +08:00
atomicrmw release - workgroup - global 1. s_waitcnt lgkmcnt(0)
- generic
2017-10-19 06:16:55 +08:00
- If OpenCL, omit.
2017-06-07 04:31:59 +08:00
- Must happen after
any preceding
local/generic
load/store/load
atomic/store
atomic/atomicrmw.
2020-10-31 12:43:55 +08:00
- Must happen before
the following
atomicrmw.
- Ensures that all
memory operations
to local have
completed before
performing the
atomicrmw that is
being released.
2. buffer/global/flat_atomic
atomicrmw release - workgroup - local 1. ds_atomic
atomicrmw release - agent - global 1. s_waitcnt lgkmcnt(0) &
- system - generic vmcnt(0)
- If OpenCL, omit
lgkmcnt(0).
- Could be split into
separate s_waitcnt
vmcnt(0) and
s_waitcnt
lgkmcnt(0) to allow
them to be
independently moved
according to the
following rules.
- s_waitcnt vmcnt(0)
must happen after
2017-10-19 06:16:55 +08:00
any preceding
2020-10-31 12:43:55 +08:00
global/generic
2017-10-19 06:16:55 +08:00
load/store/load
atomic/store
atomic/atomicrmw.
2020-10-31 12:43:55 +08:00
- s_waitcnt lgkmcnt(0)
must happen after
any preceding
local/generic
load/store/load
atomic/store
2017-06-07 04:31:59 +08:00
atomic/atomicrmw.
2020-10-31 12:43:55 +08:00
- Must happen before
the following
atomicrmw.
- Ensures that all
memory operations
to global and local
have completed
before performing
the atomicrmw that
is being released.
2. buffer/global/flat_atomic
fence release - singlethread *none* *none*
2017-06-07 04:31:59 +08:00
- wavefront
2020-10-31 12:43:55 +08:00
fence release - workgroup *none* 1. s_waitcnt lgkmcnt(0)
- If OpenCL and
address space is
not generic, omit.
- However, since LLVM
currently has no
address space on
the fence need to
conservatively
always generate. If
fence had an
address space then
set to address
space of OpenCL
fence flag, or to
generic if both
local and global
flags are
specified.
2017-06-07 04:31:59 +08:00
- Must happen after
any preceding
local/generic
load/load
atomic/store/store
atomic/atomicrmw.
2020-10-31 12:43:55 +08:00
- Must happen before
any following store
atomic/atomicrmw
with an equal or
wider sync scope
and memory ordering
stronger than
unordered (this is
termed the
fence-paired-atomic).
- Ensures that all
memory operations
to local have
completed before
performing the
following
fence-paired-atomic.
fence release - agent *none* 1. s_waitcnt lgkmcnt(0) &
- system vmcnt(0)
- If OpenCL and
address space is
not generic, omit
lgkmcnt(0).
- If OpenCL and
address space is
local, omit
vmcnt(0).
- However, since LLVM
currently has no
address space on
the fence need to
conservatively
always generate. If
fence had an
address space then
set to address
space of OpenCL
fence flag, or to
generic if both
local and global
flags are
specified.
- Could be split into
separate s_waitcnt
vmcnt(0) and
s_waitcnt
lgkmcnt(0) to allow
them to be
independently moved
according to the
following rules.
- s_waitcnt vmcnt(0)
must happen after
any preceding
global/generic
load/store/load
atomic/store
atomic/atomicrmw.
- s_waitcnt lgkmcnt(0)
must happen after
any preceding
local/generic
load/store/load
atomic/store
2017-06-07 04:31:59 +08:00
atomic/atomicrmw.
2020-10-31 12:43:55 +08:00
- Must happen before
any following store
atomic/atomicrmw
with an equal or
wider sync scope
and memory ordering
stronger than
unordered (this is
termed the
fence-paired-atomic).
- Ensures that all
memory operations
have
completed before
performing the
following
fence-paired-atomic.
2017-06-07 04:31:59 +08:00
**Acquire-Release Atomic**
2020-10-31 12:43:55 +08:00
------------------------------------------------------------------------------------
atomicrmw acq_rel - singlethread - global 1. buffer/global/ds/flat_atomic
2017-06-07 04:31:59 +08:00
- wavefront - local
- generic
2020-10-31 12:43:55 +08:00
atomicrmw acq_rel - workgroup - global 1. s_waitcnt lgkmcnt(0)
- If OpenCL, omit.
- Must happen after
any preceding
local/generic
load/store/load
atomic/store
atomic/atomicrmw.
- Must happen before
the following
atomicrmw.
- Ensures that all
memory operations
to local have
completed before
performing the
atomicrmw that is
being released.
2. buffer/global_atomic
atomicrmw acq_rel - workgroup - local 1. ds_atomic
2. s_waitcnt lgkmcnt(0)
- If OpenCL, omit.
- Must happen before
any following
global/generic
2017-06-07 04:31:59 +08:00
load/load
atomic/store/store
atomic/atomicrmw.
2020-10-31 12:43:55 +08:00
- Ensures any
following global
data read is no
older than the local load
atomic value being
acquired.
atomicrmw acq_rel - workgroup - generic 1. s_waitcnt lgkmcnt(0)
- If OpenCL, omit.
2017-06-07 04:31:59 +08:00
- Must happen after
any preceding
local/generic
load/store/load
atomic/store
atomic/atomicrmw.
2020-10-31 12:43:55 +08:00
- Must happen before
the following
atomicrmw.
- Ensures that all
memory operations
to local have
completed before
performing the
atomicrmw that is
being released.
2. flat_atomic
3. s_waitcnt lgkmcnt(0)
- If OpenCL, omit.
- Must happen before
any following
global/generic
2017-06-07 04:31:59 +08:00
load/load
atomic/store/store
atomic/atomicrmw.
2020-10-31 12:43:55 +08:00
- Ensures any
following global
data read is no
older than a local load
atomic value being
acquired.
atomicrmw acq_rel - agent - global 1. s_waitcnt lgkmcnt(0) &
- system vmcnt(0)
- If OpenCL, omit
lgkmcnt(0).
- Could be split into
separate s_waitcnt
vmcnt(0) and
s_waitcnt
lgkmcnt(0) to allow
them to be
independently moved
according to the
following rules.
- s_waitcnt vmcnt(0)
must happen after
any preceding
global/generic
load/store/load
atomic/store
2017-06-07 04:31:59 +08:00
atomic/atomicrmw.
2020-10-31 12:43:55 +08:00
- s_waitcnt lgkmcnt(0)
must happen after
2017-06-07 04:31:59 +08:00
any preceding
local/generic
2020-10-31 12:43:55 +08:00
load/store/load
atomic/store
atomic/atomicrmw.
- Must happen before
the following
atomicrmw.
- Ensures that all
memory operations
to global have
completed before
performing the
atomicrmw that is
being released.
2. buffer/global_atomic
3. s_waitcnt vmcnt(0)
- Must happen before
following
buffer_wbinvl1_vol.
- Ensures the
atomicrmw has
completed before
invalidating the
cache.
4. buffer_wbinvl1_vol
- Must happen before
any following
global/generic
2017-06-07 04:31:59 +08:00
load/load
atomic/atomicrmw.
2020-10-31 12:43:55 +08:00
- Ensures that
following loads
will not see stale
global data.
2017-06-07 04:31:59 +08:00
2020-10-31 12:43:55 +08:00
atomicrmw acq_rel - agent - generic 1. s_waitcnt lgkmcnt(0) &
- system vmcnt(0)
2017-10-19 06:16:55 +08:00
2020-10-31 12:43:55 +08:00
- If OpenCL, omit
lgkmcnt(0).
- Could be split into
separate s_waitcnt
vmcnt(0) and
s_waitcnt
lgkmcnt(0) to allow
them to be
independently moved
according to the
following rules.
- s_waitcnt vmcnt(0)
must happen after
any preceding
global/generic
load/store/load
atomic/store
atomic/atomicrmw.
- s_waitcnt lgkmcnt(0)
must happen after
any preceding
local/generic
load/store/load
atomic/store
atomic/atomicrmw.
- Must happen before
the following
atomicrmw.
- Ensures that all
memory operations
to global have
completed before
performing the
atomicrmw that is
being released.
2. flat_atomic
3. s_waitcnt vmcnt(0) &
lgkmcnt(0)
- If OpenCL, omit
lgkmcnt(0).
- Must happen before
following
buffer_wbinvl1_vol.
- Ensures the
atomicrmw has
completed before
invalidating the
cache.
2017-06-07 04:31:59 +08:00
2020-10-31 12:43:55 +08:00
4. buffer_wbinvl1_vol
2017-06-07 04:31:59 +08:00
2020-10-31 12:43:55 +08:00
- Must happen before
any following
global/generic
load/load
atomic/atomicrmw.
- Ensures that
following loads
will not see stale
global data.
fence acq_rel - singlethread *none* *none*
- wavefront
fence acq_rel - workgroup *none* 1. s_waitcnt lgkmcnt(0)
- If OpenCL and
address space is
not generic, omit.
- However,
since LLVM
currently has no
address space on
the fence need to
conservatively
always generate
(see comment for
previous fence).
- Must happen after
any preceding
local/generic
load/load
atomic/store/store
atomic/atomicrmw.
- Must happen before
any following
global/generic
load/load
atomic/store/store
atomic/atomicrmw.
- Ensures that all
memory operations
to local have
completed before
performing any
following global
memory operations.
- Ensures that the
preceding
local/generic load
atomic/atomicrmw
with an equal or
wider sync scope
and memory ordering
stronger than
unordered (this is
termed the
acquire-fence-paired-atomic)
has completed
before following
global memory
operations. This
satisfies the
requirements of
acquire.
- Ensures that all
previous memory
operations have
completed before a
following
local/generic store
atomic/atomicrmw
with an equal or
wider sync scope
and memory ordering
stronger than
unordered (this is
termed the
release-fence-paired-atomic).
This satisfies the
requirements of
release.
fence acq_rel - agent *none* 1. s_waitcnt lgkmcnt(0) &
- system vmcnt(0)
- If OpenCL and
address space is
not generic, omit
lgkmcnt(0).
- However, since LLVM
currently has no
address space on
the fence need to
conservatively
always generate
(see comment for
previous fence).
- Could be split into
separate s_waitcnt
vmcnt(0) and
s_waitcnt
lgkmcnt(0) to allow
them to be
independently moved
according to the
following rules.
- s_waitcnt vmcnt(0)
must happen after
any preceding
global/generic
load/store/load
atomic/store
atomic/atomicrmw.
- s_waitcnt lgkmcnt(0)
must happen after
any preceding
local/generic
load/store/load
atomic/store
atomic/atomicrmw.
- Must happen before
the following
buffer_wbinvl1_vol.
- Ensures that the
preceding
global/local/generic
load
atomic/atomicrmw
with an equal or
wider sync scope
and memory ordering
stronger than
unordered (this is
termed the
acquire-fence-paired-atomic)
has completed
before invalidating
the cache. This
satisfies the
requirements of
acquire.
- Ensures that all
previous memory
operations have
completed before a
following
global/local/generic
store
atomic/atomicrmw
with an equal or
wider sync scope
and memory ordering
stronger than
unordered (this is
termed the
release-fence-paired-atomic).
This satisfies the
requirements of
release.
2. buffer_wbinvl1_vol
- Must happen before
any following
global/generic
load/load
atomic/store/store
atomic/atomicrmw.
- Ensures that
following loads
will not see stale
global data. This
satisfies the
requirements of
acquire.
**Sequential Consistent Atomic**
------------------------------------------------------------------------------------
load atomic seq_cst - singlethread - global *Same as corresponding
- wavefront - local load atomic acquire,
- generic except must generated
all instructions even
for OpenCL.*
load atomic seq_cst - workgroup - global 1. s_waitcnt lgkmcnt(0)
- generic
- Must
happen after
preceding
local/generic load
atomic/store
atomic/atomicrmw
with memory
ordering of seq_cst
and with equal or
wider sync scope.
(Note that seq_cst
fences have their
own s_waitcnt
lgkmcnt(0) and so do
not need to be
considered.)
- Ensures any
preceding
sequential
consistent local
memory instructions
have completed
before executing
this sequentially
consistent
instruction. This
prevents reordering
a seq_cst store
followed by a
seq_cst load. (Note
that seq_cst is
stronger than
acquire/release as
the reordering of
load acquire
followed by a store
release is
prevented by the
s_waitcnt of
the release, but
there is nothing
preventing a store
release followed by
load acquire from
completing out of
order. The s_waitcnt
could be placed after
seq_store or before
the seq_load. We
choose the load to
make the s_waitcnt be
as late as possible
so that the store
may have already
completed.)
2. *Following
instructions same as
corresponding load
atomic acquire,
except must generated
all instructions even
for OpenCL.*
load atomic seq_cst - workgroup - local *Same as corresponding
load atomic acquire,
except must generated
all instructions even
for OpenCL.*
load atomic seq_cst - agent - global 1. s_waitcnt lgkmcnt(0) &
- system - generic vmcnt(0)
- Could be split into
separate s_waitcnt
vmcnt(0)
and s_waitcnt
lgkmcnt(0) to allow
them to be
independently moved
according to the
following rules.
- s_waitcnt lgkmcnt(0)
must happen after
preceding
global/generic load
atomic/store
atomic/atomicrmw
with memory
ordering of seq_cst
and with equal or
wider sync scope.
(Note that seq_cst
fences have their
own s_waitcnt
lgkmcnt(0) and so do
not need to be
considered.)
- s_waitcnt vmcnt(0)
must happen after
preceding
global/generic load
atomic/store
atomic/atomicrmw
with memory
ordering of seq_cst
and with equal or
wider sync scope.
(Note that seq_cst
fences have their
own s_waitcnt
vmcnt(0) and so do
not need to be
considered.)
- Ensures any
preceding
sequential
consistent global
memory instructions
have completed
before executing
this sequentially
consistent
instruction. This
prevents reordering
a seq_cst store
followed by a
seq_cst load. (Note
that seq_cst is
stronger than
acquire/release as
the reordering of
load acquire
followed by a store
release is
prevented by the
s_waitcnt of
the release, but
there is nothing
preventing a store
release followed by
load acquire from
completing out of
order. The s_waitcnt
could be placed after
seq_store or before
the seq_load. We
choose the load to
make the s_waitcnt be
as late as possible
so that the store
may have already
completed.)
2. *Following
instructions same as
corresponding load
atomic acquire,
except must generated
all instructions even
for OpenCL.*
store atomic seq_cst - singlethread - global *Same as corresponding
- wavefront - local store atomic release,
- workgroup - generic except must generated
- agent all instructions even
- system for OpenCL.*
atomicrmw seq_cst - singlethread - global *Same as corresponding
- wavefront - local atomicrmw acq_rel,
- workgroup - generic except must generated
- agent all instructions even
- system for OpenCL.*
fence seq_cst - singlethread *none* *Same as corresponding
- wavefront fence acq_rel,
- workgroup except must generated
- agent all instructions even
- system for OpenCL.*
============ ============ ============== ========== ================================
2021-02-18 05:37:46 +08:00
.. _amdgpu-amdhsa-memory-model-gfx90a:
2020-10-31 12:43:55 +08:00
2021-02-18 05:37:46 +08:00
Memory Model GFX90A
+++++++++++++++++++
2020-10-31 12:43:55 +08:00
2021-02-18 05:37:46 +08:00
For GFX90A:
2020-10-31 12:43:55 +08:00
* Each agent has multiple shader arrays (SA).
2021-02-18 05:37:46 +08:00
* Each SA has multiple compute units (CU).
2020-10-31 12:43:55 +08:00
* Each CU has multiple SIMDs that execute wavefronts.
2021-02-18 05:37:46 +08:00
* The wavefronts for a single work-group are executed in the same CU but may be
executed by different SIMDs. The exception is when in tgsplit execution mode
when the wavefronts may be executed by different SIMDs in different CUs.
* Each CU has a single LDS memory shared by the wavefronts of the work-groups
executing on it. The exception is when in tgsplit execution mode when no LDS
is allocated as wavefronts of the same work-group can be in different CUs.
* All LDS operations of a CU are performed as wavefront wide operations in a
2020-10-31 12:43:55 +08:00
global order and involve no caching. Completion is reported to a wavefront in
execution order.
* The LDS memory has multiple request queues shared by the SIMDs of a
2021-02-18 05:37:46 +08:00
CU. Therefore, the LDS operations performed by different wavefronts of a
2020-10-31 12:43:55 +08:00
work-group can be reordered relative to each other, which can result in
reordering the visibility of vector memory operations with respect to LDS
operations of other wavefronts in the same work-group. A `` s_waitcnt
lgkmcnt(0)`` is required to ensure synchronization between LDS operations and
vector memory operations between wavefronts of a work-group, but not between
operations performed by the same wavefront.
2021-02-18 05:37:46 +08:00
* The vector memory operations are performed as wavefront wide operations and
completion is reported to a wavefront in execution order. The exception is
that `` flat_load/store/atomic `` instructions can report out of vector memory
order if they access LDS memory, and out of LDS operation order if they access
global memory.
* The vector memory operations access a single vector L1 cache shared by all
SIMDs a CU. Therefore:
* No special action is required for coherence between the lanes of a single
wavefront.
* No special action is required for coherence between wavefronts in the same
2021-02-28 02:09:23 +08:00
work-group since they execute on the same CU. The exception is when in
2021-02-18 05:37:46 +08:00
tgsplit execution mode as wavefronts of the same work-group can be in
different CUs and so a `` buffer_wbinvl1_vol `` is required as described in
the following item.
* A `` buffer_wbinvl1_vol `` is required for coherence between wavefronts
executing in different work-groups as they may be executing on different
CUs.
* The scalar memory operations access a scalar L1 cache shared by all wavefronts
on a group of CUs. The scalar and vector L1 caches are not coherent. However,
scalar operations are used in a restricted way so do not impact the memory
model. See :ref: `amdgpu-amdhsa-memory-spaces` .
* The vector and scalar memory operations use an L2 cache shared by all CUs on
the same agent.
* The L2 cache has independent channels to service disjoint ranges of virtual
addresses.
* Each CU has a separate request queue per channel. Therefore, the vector and
scalar memory operations performed by wavefronts executing in different
work-groups (which may be executing on different CUs), or the same
work-group if executing in tgsplit mode, of an agent can be reordered
relative to each other. A `` s_waitcnt vmcnt(0) `` is required to ensure
synchronization between vector memory operations of different CUs. It
ensures a previous vector memory operation has completed before executing a
subsequent vector memory or LDS operation and so can be used to meet the
requirements of acquire and release.
* The L2 cache of one agent can be kept coherent with other agents by using
2021-03-31 06:38:19 +08:00
the MTYPE CC (cache-coherent) with the PTE C-bit for memory local to the L2,
and MTYPE UC (uncached) with the PTE C-bit set for memory not local to the
L2.
2021-02-18 05:37:46 +08:00
* Any local memory cache lines will be automatically invalidated by writes
from CUs associated with other L2 caches, or writes from the CPU, due to
2021-03-31 06:38:19 +08:00
the cache probe caused by coherent requests. Coherent requests are caused
by GPU accesses to pages with the PTE C-bit set, by CPU accesses over
XGMI, and by PCIe requests that are configured to be coherent requests.
2021-04-08 08:58:02 +08:00
* XGMI accesses from the CPU to local memory may be cached on the CPU.
2021-02-18 05:37:46 +08:00
Subsequent access from the GPU will automatically invalidate or writeback
2021-03-31 06:38:19 +08:00
the CPU cache due to the L2 probe filter and and the PTE C-bit being set.
2021-02-18 05:37:46 +08:00
* Since all work-groups on the same agent share the same L2, no L2
invalidation or writeback is required for coherence.
2021-03-31 06:38:19 +08:00
* Since local memory reads and writes of work-groups in different agents
access memory using MTYPE CC, no L2 invalidate or writeback is required
for coherence. MTYPE CC causes write through to DRAM and local reads to be
invalidated by remote writes with with the PTE C-bit.
* Since remote memory reads and writes of work-groups in different agents
access memory using MTYPE UC, no L2 invalidate or writeback is required
for coherence. MTYPE UC causes direct accesses to DRAM.
* PCIe access from the GPU to the CPU memory is kept coherent by using the
2021-02-18 05:37:46 +08:00
MTYPE UC (uncached) which bypasses the L2.
2020-10-31 12:43:55 +08:00
Scalar memory operations are only used to access memory that is proven to not
change during the execution of the kernel dispatch. This includes constant
2020-11-03 10:14:45 +08:00
address space and global address space for program scope `` const `` variables.
2020-10-31 12:43:55 +08:00
Therefore, the kernel machine code does not have to maintain the scalar cache to
ensure it is coherent with the vector caches. The scalar and vector caches are
invalidated between kernel dispatches by CP since constant address space data
may change between kernel dispatch executions. See
:ref: `amdgpu-amdhsa-memory-spaces` .
The one exception is if scalar writes are used to spill SGPR registers. In this
case the AMDGPU backend ensures the memory location used to spill is never
accessed by vector memory operations at the same time. If scalar writes are used
then a `` s_dcache_wb `` is inserted before the `` s_endpgm `` and before a function
return since the locations may be used for vector memory instructions by a
future wavefront that uses the same scratch area, or a function call that
creates a frame at the same address, respectively. There is no need for a
`` s_dcache_inv `` as all scalar writes are write-before-read in the same thread.
For kernarg backing memory:
2021-02-18 05:37:46 +08:00
* CP invalidates the L1 cache at the start of each kernel dispatch.
* On dGPU over XGMI or PCIe the kernarg backing memory is allocated in host
memory accessed as MTYPE UC (uncached) to avoid needing to invalidate the L2
cache. This also causes it to be treated as non-volatile and so is not
invalidated by `` *_vol `` .
2020-10-31 12:43:55 +08:00
* On APU the kernarg backing memory is accessed as MTYPE CC (cache coherent) and
so the L2 cache will be coherent with the CPU and other agents.
Scratch backing memory (which is used for the private address space) is accessed
2021-02-18 05:37:46 +08:00
with MTYPE NC_NV (non-coherent non-volatile). Since the private address space is
only accessed by a single thread, and is always write-before-read, there is
never a need to invalidate these entries from the L1 cache. Hence all cache
invalidates are done as `` *_vol `` to only invalidate the volatile cache lines.
2020-10-31 12:43:55 +08:00
2021-02-18 05:37:46 +08:00
The code sequences used to implement the memory model for GFX90A are defined
in table :ref: `amdgpu-amdhsa-memory-model-code-sequences-gfx90a-table` .
2020-10-31 12:43:55 +08:00
2021-02-18 05:37:46 +08:00
.. table :: AMDHSA Memory Model Code Sequences GFX90A
:name: amdgpu-amdhsa-memory-model-code-sequences-gfx90a-table
2020-10-31 12:43:55 +08:00
============ ============ ============== ========== ================================
LLVM Instr LLVM Memory LLVM Memory AMDGPU AMDGPU Machine Code
2021-02-18 05:37:46 +08:00
Ordering Sync Scope Address GFX90A
2020-10-31 12:43:55 +08:00
Space
============ ============ ============== ========== ================================
**Non-Atomic**
------------------------------------------------------------------------------------
load *none* *none* - global - !volatile & !nontemporal
- generic
- private 1. buffer/global/flat_load
- constant
2021-01-09 18:20:42 +08:00
- !volatile & nontemporal
1. buffer/global/flat_load
2021-02-18 05:37:46 +08:00
glc=1 slc=1
2021-01-09 18:20:42 +08:00
- volatile
2020-10-31 12:43:55 +08:00
1. buffer/global/flat_load
2021-03-31 06:38:19 +08:00
glc=1
2020-12-19 10:05:11 +08:00
2. s_waitcnt vmcnt(0)
- Must happen before
any following volatile
global/generic
load/store.
- Ensures that
volatile
operations to
different
addresses will not
be reordered by
hardware.
2020-10-31 12:43:55 +08:00
load *none* *none* - local 1. ds_load
2020-12-19 10:05:11 +08:00
store *none* *none* - global - !volatile & !nontemporal
2020-10-31 12:43:55 +08:00
- generic
- private 1. buffer/global/flat_store
- constant
2021-01-09 18:20:42 +08:00
- !volatile & nontemporal
2021-02-18 05:37:46 +08:00
1. buffer/global/flat_store
glc=1 slc=1
2021-01-09 18:20:42 +08:00
- volatile
2020-12-19 10:05:11 +08:00
1. buffer/global/flat_store
2021-02-18 05:37:46 +08:00
2. s_waitcnt vmcnt(0)
2020-12-19 10:05:11 +08:00
- Must happen before
any following volatile
global/generic
load/store.
- Ensures that
volatile
operations to
different
addresses will not
be reordered by
hardware.
2020-10-31 12:43:55 +08:00
store *none* *none* - local 1. ds_store
**Unordered Atomic**
------------------------------------------------------------------------------------
load atomic unordered *any* *any* *Same as non-atomic* .
store atomic unordered *any* *any* *Same as non-atomic* .
2020-11-03 10:14:45 +08:00
atomicrmw unordered *any* *any* *Same as monotonic atomic* .
2020-10-31 12:43:55 +08:00
**Monotonic Atomic**
------------------------------------------------------------------------------------
load atomic monotonic - singlethread - global 1. buffer/global/flat_load
- wavefront - generic
load atomic monotonic - workgroup - global 1. buffer/global/flat_load
- generic glc=1
2021-02-18 05:37:46 +08:00
- If not TgSplit execution
2020-10-31 12:43:55 +08:00
mode, omit glc=1.
2021-02-18 05:37:46 +08:00
load atomic monotonic - singlethread - local *If TgSplit execution mode,
- wavefront local address space cannot
- workgroup be used.*
1. ds_load
2020-10-31 12:43:55 +08:00
load atomic monotonic - agent - global 1. buffer/global/flat_load
2021-02-18 05:37:46 +08:00
- generic glc=1
load atomic monotonic - system - global 1. buffer/global/flat_load
2021-03-31 06:38:19 +08:00
- generic glc=1
2020-10-31 12:43:55 +08:00
store atomic monotonic - singlethread - global 1. buffer/global/flat_store
- wavefront - generic
- workgroup
- agent
2021-02-18 05:37:46 +08:00
store atomic monotonic - system - global 1. buffer/global/flat_store
2021-03-31 06:38:19 +08:00
- generic
2021-02-18 05:37:46 +08:00
store atomic monotonic - singlethread - local *If TgSplit execution mode,
- wavefront local address space cannot
- workgroup be used.*
1. ds_store
2020-10-31 12:43:55 +08:00
atomicrmw monotonic - singlethread - global 1. buffer/global/flat_atomic
- wavefront - generic
- workgroup
- agent
2021-02-18 05:37:46 +08:00
atomicrmw monotonic - system - global 1. buffer/global/flat_atomic
2021-03-31 06:38:19 +08:00
- generic
2021-02-18 05:37:46 +08:00
atomicrmw monotonic - singlethread - local *If TgSplit execution mode,
- wavefront local address space cannot
- workgroup be used.*
1. ds_atomic
2020-10-31 12:43:55 +08:00
**Acquire Atomic**
------------------------------------------------------------------------------------
load atomic acquire - singlethread - global 1. buffer/global/ds/flat_load
- wavefront - local
- generic
load atomic acquire - workgroup - global 1. buffer/global_load glc=1
2021-02-18 05:37:46 +08:00
- If not TgSplit execution
2020-10-31 12:43:55 +08:00
mode, omit glc=1.
2. s_waitcnt vmcnt(0)
2021-02-18 05:37:46 +08:00
- If not TgSplit execution
mode, omit.
- Must happen before the
following buffer_wbinvl1_vol.
3. buffer_wbinvl1_vol
- If not TgSplit execution
2020-10-31 12:43:55 +08:00
mode, omit.
- Must happen before
2021-02-18 05:37:46 +08:00
any following
2020-10-31 12:43:55 +08:00
global/generic
load/load
atomic/store/store
atomic/atomicrmw.
- Ensures that
following
loads will not see
stale data.
2021-02-18 05:37:46 +08:00
load atomic acquire - workgroup - local *If TgSplit execution mode,
local address space cannot
be used.*
1. ds_load
2020-10-31 12:43:55 +08:00
2. s_waitcnt lgkmcnt(0)
- If OpenCL, omit.
- Must happen before
2021-02-18 05:37:46 +08:00
any following
global/generic
load/load
2020-10-31 12:43:55 +08:00
atomic/store/store
atomic/atomicrmw.
- Ensures any
following global
data read is no
older than the local load
atomic value being
acquired.
load atomic acquire - workgroup - generic 1. flat_load glc=1
2021-02-18 05:37:46 +08:00
- If not TgSplit execution
2020-10-31 12:43:55 +08:00
mode, omit glc=1.
2021-02-18 05:37:46 +08:00
2. s_waitcnt lgkm/vmcnt(0)
2020-10-31 12:43:55 +08:00
2021-02-18 05:37:46 +08:00
- Use lgkmcnt(0) if not
TgSplit execution mode
and vmcnt(0) if TgSplit
execution mode.
- If OpenCL, omit lgkmcnt(0).
2020-10-31 12:43:55 +08:00
- Must happen before
the following
2021-02-18 05:37:46 +08:00
buffer_wbinvl1_vol and any
2020-10-31 12:43:55 +08:00
following global/generic
load/load
atomic/store/store
atomic/atomicrmw.
- Ensures any
following global
data read is no
older than a local load
atomic value being
acquired.
2021-02-18 05:37:46 +08:00
3. buffer_wbinvl1_vol
2020-10-31 12:43:55 +08:00
2021-02-18 05:37:46 +08:00
- If not TgSplit execution
2020-10-31 12:43:55 +08:00
mode, omit.
- Ensures that
following
loads will not see
stale data.
load atomic acquire - agent - global 1. buffer/global_load
2021-02-18 05:37:46 +08:00
glc=1
2020-10-31 12:43:55 +08:00
2. s_waitcnt vmcnt(0)
- Must happen before
following
2021-02-18 05:37:46 +08:00
buffer_wbinvl1_vol.
2020-10-31 12:43:55 +08:00
- Ensures the load
has completed
before invalidating
2021-02-18 05:37:46 +08:00
the cache.
2020-10-31 12:43:55 +08:00
2021-02-18 05:37:46 +08:00
3. buffer_wbinvl1_vol
2020-10-31 12:43:55 +08:00
- Must happen before
any following
global/generic
load/load
atomic/atomicrmw.
- Ensures that
following
loads will not see
stale global data.
2021-02-18 05:37:46 +08:00
load atomic acquire - system - global 1. buffer/global/flat_load
2021-03-31 06:38:19 +08:00
glc=1
2021-02-18 05:37:46 +08:00
2. s_waitcnt vmcnt(0)
- Must happen before
2021-03-31 06:38:19 +08:00
following
2021-02-18 05:37:46 +08:00
buffer_wbinvl1_vol.
- Ensures the load
has completed
before invalidating
the cache.
2021-03-31 06:38:19 +08:00
3. buffer_wbinvl1_vol
2021-02-18 05:37:46 +08:00
- Must happen before
any following
global/generic
load/load
atomic/atomicrmw.
- Ensures that
following
loads will not see
2021-03-31 06:38:19 +08:00
stale L1 global data.
2021-02-18 05:37:46 +08:00
MTYPE RW and CC memory will
2021-03-31 06:38:19 +08:00
never be stale in L2 due to
the memory probes.
2021-02-18 05:37:46 +08:00
load atomic acquire - agent - generic 1. flat_load glc=1
2. s_waitcnt vmcnt(0) &
2020-10-31 12:43:55 +08:00
lgkmcnt(0)
2021-02-18 05:37:46 +08:00
- If TgSplit execution mode,
omit lgkmcnt(0).
2020-10-31 12:43:55 +08:00
- If OpenCL omit
lgkmcnt(0).
- Must happen before
following
2021-02-18 05:37:46 +08:00
buffer_wbinvl1_vol.
2020-10-31 12:43:55 +08:00
- Ensures the flat_load
has completed
before invalidating
2021-02-18 05:37:46 +08:00
the cache.
2020-10-31 12:43:55 +08:00
2021-02-18 05:37:46 +08:00
3. buffer_wbinvl1_vol
2020-10-31 12:43:55 +08:00
- Must happen before
any following
global/generic
load/load
atomic/atomicrmw.
- Ensures that
following loads
will not see stale
global data.
2021-03-31 06:38:19 +08:00
load atomic acquire - system - generic 1. flat_load glc=1
2021-02-18 05:37:46 +08:00
2. s_waitcnt vmcnt(0) &
lgkmcnt(0)
- If TgSplit execution mode,
omit lgkmcnt(0).
- If OpenCL omit
lgkmcnt(0).
- Must happen before
following
buffer_wbinvl1_vol.
- Ensures the flat_load
has completed
before invalidating
the caches.
2021-03-31 06:38:19 +08:00
3. buffer_wbinvl1_vol
2021-02-18 05:37:46 +08:00
- Must happen before
any following
global/generic
load/load
atomic/atomicrmw.
- Ensures that
following
2021-03-31 06:38:19 +08:00
L1 loads will not see
stale global data.
2021-02-18 05:37:46 +08:00
MTYPE RW and CC memory will
2021-03-31 06:38:19 +08:00
never be stale in L2 due to
the memory probes.
2021-02-18 05:37:46 +08:00
atomicrmw acquire - singlethread - global 1. buffer/global/flat_atomic
- wavefront - generic
atomicrmw acquire - singlethread - local *If TgSplit execution mode,
- wavefront local address space cannot
be used.*
1. ds_atomic
atomicrmw acquire - workgroup - global 1. buffer/global_atomic
2. s_waitcnt vmcnt(0)
- If not TgSplit execution
mode, omit.
- Must happen before the
following buffer_wbinvl1_vol.
- Ensures the atomicrmw
has completed
before invalidating
the cache.
3. buffer_wbinvl1_vol
- If not TgSplit execution
mode, omit.
- Must happen before
any following
global/generic
load/load
atomic/atomicrmw.
- Ensures that
following loads
will not see stale
global data.
atomicrmw acquire - workgroup - local *If TgSplit execution mode,
local address space cannot
be used.*
1. ds_atomic
2. s_waitcnt lgkmcnt(0)
- If OpenCL, omit.
- Must happen before
any following
global/generic
load/load
atomic/store/store
atomic/atomicrmw.
- Ensures any
following global
data read is no
older than the local
atomicrmw value
being acquired.
atomicrmw acquire - workgroup - generic 1. flat_atomic
2. s_waitcnt lgkm/vmcnt(0)
- Use lgkmcnt(0) if not
TgSplit execution mode
and vmcnt(0) if TgSplit
execution mode.
- If OpenCL, omit lgkmcnt(0).
- Must happen before
the following
buffer_wbinvl1_vol and
any following
global/generic
load/load
atomic/store/store
atomic/atomicrmw.
- Ensures any
following global
data read is no
older than a local
atomicrmw value
being acquired.
3. buffer_wbinvl1_vol
- If not TgSplit execution
mode, omit.
- Ensures that
following
loads will not see
stale data.
atomicrmw acquire - agent - global 1. buffer/global_atomic
2. s_waitcnt vmcnt(0)
- Must happen before
following
buffer_wbinvl1_vol.
- Ensures the
atomicrmw has
completed before
invalidating the
cache.
3. buffer_wbinvl1_vol
- Must happen before
any following
global/generic
load/load
atomic/atomicrmw.
- Ensures that
following loads
will not see stale
global data.
atomicrmw acquire - system - global 1. buffer/global_atomic
2. s_waitcnt vmcnt(0)
- Must happen before
2021-03-31 06:38:19 +08:00
following
2021-02-18 05:37:46 +08:00
buffer_wbinvl1_vol.
- Ensures the
atomicrmw has
completed before
invalidating the
caches.
2021-03-31 06:38:19 +08:00
3. buffer_wbinvl1_vol
2021-02-18 05:37:46 +08:00
- Must happen before
any following
global/generic
load/load
atomic/atomicrmw.
- Ensures that
following
loads will not see
2021-03-31 06:38:19 +08:00
stale L1 global data.
MTYPE RW and CC L2 memory
never be stale in L2 due to
the memory probes.
2021-02-18 05:37:46 +08:00
atomicrmw acquire - agent - generic 1. flat_atomic
2. s_waitcnt vmcnt(0) &
lgkmcnt(0)
- If TgSplit execution mode,
omit lgkmcnt(0).
- If OpenCL, omit
lgkmcnt(0).
- Must happen before
following
buffer_wbinvl1_vol.
- Ensures the
atomicrmw has
completed before
invalidating the
cache.
3. buffer_wbinvl1_vol
- Must happen before
any following
global/generic
load/load
atomic/atomicrmw.
- Ensures that
following loads
will not see stale
global data.
2021-03-31 06:38:19 +08:00
atomicrmw acquire - system - generic 1. flat_atomic
2021-02-18 05:37:46 +08:00
2. s_waitcnt vmcnt(0) &
lgkmcnt(0)
- If TgSplit execution mode,
omit lgkmcnt(0).
- If OpenCL, omit
lgkmcnt(0).
- Must happen before
following
buffer_wbinvl1_vol.
- Ensures the
atomicrmw has
completed before
invalidating the
caches.
2021-03-31 06:38:19 +08:00
3. buffer_wbinvl1_vol
2021-02-18 05:37:46 +08:00
- Must happen before
any following
global/generic
load/load
atomic/atomicrmw.
- Ensures that
following
loads will not see
2021-03-31 06:38:19 +08:00
stale L1 global data.
2021-02-18 05:37:46 +08:00
MTYPE RW and CC memory will
2021-03-31 06:38:19 +08:00
never be stale in L2 due to
the memory probes.
2021-02-18 05:37:46 +08:00
fence acquire - singlethread *none* *none*
- wavefront
fence acquire - workgroup *none* 1. s_waitcnt lgkm/vmcnt(0)
- Use lgkmcnt(0) if not
TgSplit execution mode
and vmcnt(0) if TgSplit
execution mode.
- If OpenCL and
address space is
not generic, omit
lgkmcnt(0).
- If OpenCL and
address space is
local, omit
vmcnt(0).
- However, since LLVM
currently has no
address space on
the fence need to
conservatively
always generate. If
fence had an
address space then
set to address
space of OpenCL
fence flag, or to
generic if both
local and global
flags are
specified.
- s_waitcnt vmcnt(0)
must happen after
any preceding
global/generic load
atomic/
atomicrmw
with an equal or
wider sync scope
and memory ordering
stronger than
unordered (this is
termed the
fence-paired-atomic).
- s_waitcnt lgkmcnt(0)
must happen after
any preceding
local/generic load
atomic/atomicrmw
with an equal or
wider sync scope
and memory ordering
stronger than
unordered (this is
termed the
fence-paired-atomic).
- Must happen before
the following
buffer_wbinvl1_vol and
any following
global/generic
load/load
atomic/store/store
atomic/atomicrmw.
- Ensures any
following global
data read is no
older than the
value read by the
fence-paired-atomic.
3. buffer_wbinvl1_vol
- If not TgSplit execution
mode, omit.
- Ensures that
following
loads will not see
stale data.
fence acquire - agent *none* 1. s_waitcnt lgkmcnt(0) &
vmcnt(0)
- If TgSplit execution mode,
omit lgkmcnt(0).
- If OpenCL and
address space is
not generic, omit
lgkmcnt(0).
- However, since LLVM
currently has no
address space on
the fence need to
conservatively
always generate
(see comment for
previous fence).
- Could be split into
separate s_waitcnt
vmcnt(0) and
s_waitcnt
lgkmcnt(0) to allow
them to be
independently moved
according to the
following rules.
- s_waitcnt vmcnt(0)
must happen after
any preceding
global/generic load
atomic/atomicrmw
with an equal or
wider sync scope
and memory ordering
stronger than
unordered (this is
termed the
fence-paired-atomic).
- s_waitcnt lgkmcnt(0)
must happen after
any preceding
local/generic load
atomic/atomicrmw
with an equal or
wider sync scope
and memory ordering
stronger than
unordered (this is
termed the
fence-paired-atomic).
- Must happen before
the following
buffer_wbinvl1_vol.
- Ensures that the
fence-paired atomic
has completed
before invalidating
the
cache. Therefore
any following
locations read must
be no older than
the value read by
the
fence-paired-atomic.
2. buffer_wbinvl1_vol
- Must happen before any
following global/generic
load/load
atomic/store/store
atomic/atomicrmw.
- Ensures that
following loads
will not see stale
global data.
fence acquire - system *none* 1. s_waitcnt lgkmcnt(0) &
vmcnt(0)
- If TgSplit execution mode,
omit lgkmcnt(0).
- If OpenCL and
address space is
not generic, omit
lgkmcnt(0).
- However, since LLVM
currently has no
address space on
the fence need to
conservatively
always generate
(see comment for
previous fence).
- Could be split into
separate s_waitcnt
vmcnt(0) and
s_waitcnt
lgkmcnt(0) to allow
them to be
independently moved
according to the
following rules.
- s_waitcnt vmcnt(0)
must happen after
any preceding
global/generic load
atomic/atomicrmw
with an equal or
wider sync scope
and memory ordering
stronger than
unordered (this is
termed the
fence-paired-atomic).
- s_waitcnt lgkmcnt(0)
must happen after
any preceding
local/generic load
atomic/atomicrmw
with an equal or
wider sync scope
and memory ordering
stronger than
unordered (this is
termed the
fence-paired-atomic).
- Must happen before
2021-03-31 06:38:19 +08:00
the following
2021-02-18 05:37:46 +08:00
buffer_wbinvl1_vol.
- Ensures that the
fence-paired atomic
has completed
before invalidating
the
cache. Therefore
any following
locations read must
be no older than
the value read by
the
fence-paired-atomic.
2021-03-31 06:38:19 +08:00
2. buffer_wbinvl1_vol
2021-02-18 05:37:46 +08:00
- Must happen before any
following global/generic
load/load
atomic/store/store
atomic/atomicrmw.
- Ensures that
2021-03-31 06:38:19 +08:00
following
loads will not see
stale L1 global data.
MTYPE RW and CC memory will
never be stale in L2 due to
the memory probes.
2021-02-18 05:37:46 +08:00
**Release Atomic**
------------------------------------------------------------------------------------
store atomic release - singlethread - global 1. buffer/global/flat_store
- wavefront - generic
store atomic release - singlethread - local *If TgSplit execution mode,
- wavefront local address space cannot
be used.*
1. ds_store
store atomic release - workgroup - global 1. s_waitcnt lgkm/vmcnt(0)
- generic
- Use lgkmcnt(0) if not
TgSplit execution mode
and vmcnt(0) if TgSplit
execution mode.
- If OpenCL, omit lgkmcnt(0).
- s_waitcnt vmcnt(0)
must happen after
any preceding
global/generic load/store/
load atomic/store atomic/
atomicrmw.
- s_waitcnt lgkmcnt(0)
must happen after
any preceding
local/generic
load/store/load
atomic/store
atomic/atomicrmw.
- Must happen before
the following
store.
- Ensures that all
memory operations
have
completed before
performing the
store that is being
released.
2. buffer/global/flat_store
store atomic release - workgroup - local *If TgSplit execution mode,
local address space cannot
be used.*
1. ds_store
store atomic release - agent - global 1. s_waitcnt lgkmcnt(0) &
- generic vmcnt(0)
- If TgSplit execution mode,
omit lgkmcnt(0).
- If OpenCL and
address space is
not generic, omit
lgkmcnt(0).
- Could be split into
separate s_waitcnt
vmcnt(0) and
s_waitcnt
lgkmcnt(0) to allow
them to be
independently moved
according to the
following rules.
- s_waitcnt vmcnt(0)
must happen after
any preceding
global/generic
load/store/load
atomic/store
atomic/atomicrmw.
- s_waitcnt lgkmcnt(0)
must happen after
any preceding
local/generic
load/store/load
atomic/store
atomic/atomicrmw.
- Must happen before
the following
store.
- Ensures that all
memory operations
to memory have
completed before
performing the
store that is being
released.
2. buffer/global/flat_store
2021-03-31 06:38:19 +08:00
store atomic release - system - global 1. s_waitcnt lgkmcnt(0) &
- generic vmcnt(0)
2021-02-18 05:37:46 +08:00
- If TgSplit execution mode,
omit lgkmcnt(0).
- If OpenCL and
address space is
not generic, omit
lgkmcnt(0).
- Could be split into
separate s_waitcnt
vmcnt(0) and
s_waitcnt
lgkmcnt(0) to allow
them to be
independently moved
according to the
following rules.
- s_waitcnt vmcnt(0)
must happen after any
preceding
global/generic
load/store/load
atomic/store
atomic/atomicrmw.
- s_waitcnt lgkmcnt(0)
must happen after any
preceding
local/generic
load/store/load
atomic/store
atomic/atomicrmw.
- Must happen before
the following
store.
- Ensures that all
memory operations
to memory and the L2
writeback have
completed before
performing the
store that is being
released.
2. buffer/global/flat_store
atomicrmw release - singlethread - global 1. buffer/global/flat_atomic
- wavefront - generic
atomicrmw release - singlethread - local *If TgSplit execution mode,
- wavefront local address space cannot
be used.*
1. ds_atomic
atomicrmw release - workgroup - global 1. s_waitcnt lgkm/vmcnt(0)
- generic
- Use lgkmcnt(0) if not
TgSplit execution mode
and vmcnt(0) if TgSplit
execution mode.
- If OpenCL, omit
lgkmcnt(0).
- s_waitcnt vmcnt(0)
must happen after
any preceding
global/generic load/store/
load atomic/store atomic/
atomicrmw.
- s_waitcnt lgkmcnt(0)
must happen after
any preceding
local/generic
load/store/load
atomic/store
atomic/atomicrmw.
- Must happen before
the following
atomicrmw.
- Ensures that all
memory operations
have
completed before
performing the
atomicrmw that is
being released.
2. buffer/global/flat_atomic
atomicrmw release - workgroup - local *If TgSplit execution mode,
local address space cannot
be used.*
1. ds_atomic
atomicrmw release - agent - global 1. s_waitcnt lgkmcnt(0) &
- generic vmcnt(0)
- If TgSplit execution mode,
omit lgkmcnt(0).
- If OpenCL, omit
lgkmcnt(0).
- Could be split into
separate s_waitcnt
vmcnt(0) and
s_waitcnt
lgkmcnt(0) to allow
them to be
independently moved
according to the
following rules.
- s_waitcnt vmcnt(0)
must happen after
any preceding
global/generic
load/store/load
atomic/store
atomic/atomicrmw.
- s_waitcnt lgkmcnt(0)
must happen after
any preceding
local/generic
load/store/load
atomic/store
atomic/atomicrmw.
- Must happen before
the following
atomicrmw.
- Ensures that all
memory operations
to global and local
have completed
before performing
the atomicrmw that
is being released.
2. buffer/global/flat_atomic
2021-03-31 06:38:19 +08:00
atomicrmw release - system - global 1. s_waitcnt lgkmcnt(0) &
- generic vmcnt(0)
2021-02-18 05:37:46 +08:00
- If TgSplit execution mode,
omit lgkmcnt(0).
- If OpenCL, omit
lgkmcnt(0).
- Could be split into
separate s_waitcnt
vmcnt(0) and
s_waitcnt
lgkmcnt(0) to allow
them to be
independently moved
according to the
following rules.
- s_waitcnt vmcnt(0)
must happen after
any preceding
global/generic
load/store/load
atomic/store
atomic/atomicrmw.
- s_waitcnt lgkmcnt(0)
must happen after
any preceding
local/generic
load/store/load
atomic/store
atomic/atomicrmw.
- Must happen before
the following
atomicrmw.
- Ensures that all
memory operations
to memory and the L2
writeback have
completed before
performing the
store that is being
released.
2021-03-31 06:38:19 +08:00
2. buffer/global/flat_atomic
2021-02-18 05:37:46 +08:00
fence release - singlethread *none* *none*
- wavefront
fence release - workgroup *none* 1. s_waitcnt lgkm/vmcnt(0)
- Use lgkmcnt(0) if not
TgSplit execution mode
and vmcnt(0) if TgSplit
execution mode.
- If OpenCL and
address space is
not generic, omit
lgkmcnt(0).
- If OpenCL and
address space is
local, omit
vmcnt(0).
- However, since LLVM
currently has no
address space on
the fence need to
conservatively
always generate. If
fence had an
address space then
set to address
space of OpenCL
fence flag, or to
generic if both
local and global
flags are
specified.
- s_waitcnt vmcnt(0)
must happen after
any preceding
global/generic
load/store/
load atomic/store atomic/
atomicrmw.
- s_waitcnt lgkmcnt(0)
must happen after
any preceding
local/generic
load/load
atomic/store/store
atomic/atomicrmw.
- Must happen before
any following store
atomic/atomicrmw
with an equal or
wider sync scope
and memory ordering
stronger than
unordered (this is
termed the
fence-paired-atomic).
- Ensures that all
memory operations
have
completed before
performing the
following
fence-paired-atomic.
fence release - agent *none* 1. s_waitcnt lgkmcnt(0) &
vmcnt(0)
- If TgSplit execution mode,
omit lgkmcnt(0).
- If OpenCL and
address space is
not generic, omit
lgkmcnt(0).
- If OpenCL and
address space is
local, omit
vmcnt(0).
- However, since LLVM
currently has no
address space on
the fence need to
conservatively
always generate. If
fence had an
address space then
set to address
space of OpenCL
fence flag, or to
generic if both
local and global
flags are
specified.
- Could be split into
separate s_waitcnt
vmcnt(0) and
s_waitcnt
lgkmcnt(0) to allow
them to be
independently moved
according to the
following rules.
- s_waitcnt vmcnt(0)
must happen after
any preceding
global/generic
load/store/load
atomic/store
atomic/atomicrmw.
- s_waitcnt lgkmcnt(0)
must happen after
any preceding
local/generic
load/store/load
atomic/store
atomic/atomicrmw.
- Must happen before
any following store
atomic/atomicrmw
with an equal or
wider sync scope
and memory ordering
stronger than
unordered (this is
termed the
fence-paired-atomic).
- Ensures that all
memory operations
have
completed before
performing the
following
fence-paired-atomic.
2021-03-31 06:38:19 +08:00
fence release - system *none* 1. s_waitcnt lgkmcnt(0) &
2021-02-18 05:37:46 +08:00
vmcnt(0)
- If TgSplit execution mode,
omit lgkmcnt(0).
- If OpenCL and
address space is
not generic, omit
lgkmcnt(0).
- If OpenCL and
address space is
local, omit
vmcnt(0).
- However, since LLVM
currently has no
address space on
the fence need to
conservatively
always generate. If
fence had an
address space then
set to address
space of OpenCL
fence flag, or to
generic if both
local and global
flags are
specified.
- Could be split into
separate s_waitcnt
vmcnt(0) and
s_waitcnt
lgkmcnt(0) to allow
them to be
independently moved
according to the
following rules.
- s_waitcnt vmcnt(0)
must happen after
any preceding
global/generic
load/store/load
atomic/store
atomic/atomicrmw.
- s_waitcnt lgkmcnt(0)
must happen after
any preceding
local/generic
load/store/load
atomic/store
atomic/atomicrmw.
- Must happen before
any following store
atomic/atomicrmw
with an equal or
wider sync scope
and memory ordering
stronger than
unordered (this is
termed the
fence-paired-atomic).
- Ensures that all
memory operations
have
completed before
performing the
following
fence-paired-atomic.
**Acquire-Release Atomic**
------------------------------------------------------------------------------------
atomicrmw acq_rel - singlethread - global 1. buffer/global/flat_atomic
- wavefront - generic
atomicrmw acq_rel - singlethread - local *If TgSplit execution mode,
- wavefront local address space cannot
be used.*
1. ds_atomic
atomicrmw acq_rel - workgroup - global 1. s_waitcnt lgkm/vmcnt(0)
- Use lgkmcnt(0) if not
TgSplit execution mode
and vmcnt(0) if TgSplit
execution mode.
- If OpenCL, omit
lgkmcnt(0).
- Must happen after
any preceding
local/generic
load/store/load
atomic/store
atomic/atomicrmw.
- s_waitcnt vmcnt(0)
must happen after
any preceding
global/generic load/store/
load atomic/store atomic/
atomicrmw.
- s_waitcnt lgkmcnt(0)
must happen after
any preceding
local/generic
load/store/load
atomic/store
atomic/atomicrmw.
- Must happen before
the following
atomicrmw.
- Ensures that all
memory operations
have
completed before
performing the
atomicrmw that is
being released.
2. buffer/global_atomic
3. s_waitcnt vmcnt(0)
- If not TgSplit execution
mode, omit.
- Must happen before
the following
buffer_wbinvl1_vol.
- Ensures any
following global
data read is no
older than the
atomicrmw value
being acquired.
4. buffer_wbinvl1_vol
- If not TgSplit execution
mode, omit.
- Ensures that
following
loads will not see
stale data.
atomicrmw acq_rel - workgroup - local *If TgSplit execution mode,
local address space cannot
be used.*
1. ds_atomic
2. s_waitcnt lgkmcnt(0)
- If OpenCL, omit.
- Must happen before
any following
global/generic
load/load
atomic/store/store
atomic/atomicrmw.
- Ensures any
following global
data read is no
older than the local load
atomic value being
acquired.
atomicrmw acq_rel - workgroup - generic 1. s_waitcnt lgkm/vmcnt(0)
- Use lgkmcnt(0) if not
TgSplit execution mode
and vmcnt(0) if TgSplit
execution mode.
- If OpenCL, omit
lgkmcnt(0).
- s_waitcnt vmcnt(0)
must happen after
any preceding
global/generic load/store/
load atomic/store atomic/
atomicrmw.
- s_waitcnt lgkmcnt(0)
must happen after
any preceding
local/generic
load/store/load
atomic/store
atomic/atomicrmw.
- Must happen before
the following
atomicrmw.
- Ensures that all
memory operations
have
completed before
performing the
atomicrmw that is
being released.
2. flat_atomic
3. s_waitcnt lgkmcnt(0) &
vmcnt(0)
- If not TgSplit execution
mode, omit vmcnt(0).
- If OpenCL, omit
lgkmcnt(0).
- Must happen before
the following
buffer_wbinvl1_vol and
any following
global/generic
load/load
atomic/store/store
atomic/atomicrmw.
- Ensures any
following global
data read is no
older than a local load
atomic value being
acquired.
3. buffer_wbinvl1_vol
- If not TgSplit execution
mode, omit.
- Ensures that
following
loads will not see
stale data.
atomicrmw acq_rel - agent - global 1. s_waitcnt lgkmcnt(0) &
vmcnt(0)
- If TgSplit execution mode,
omit lgkmcnt(0).
- If OpenCL, omit
lgkmcnt(0).
- Could be split into
separate s_waitcnt
vmcnt(0) and
s_waitcnt
lgkmcnt(0) to allow
them to be
independently moved
according to the
following rules.
- s_waitcnt vmcnt(0)
must happen after
any preceding
global/generic
load/store/load
atomic/store
atomic/atomicrmw.
- s_waitcnt lgkmcnt(0)
must happen after
any preceding
local/generic
load/store/load
atomic/store
atomic/atomicrmw.
- Must happen before
the following
atomicrmw.
- Ensures that all
memory operations
to global have
completed before
performing the
atomicrmw that is
being released.
2. buffer/global_atomic
3. s_waitcnt vmcnt(0)
- Must happen before
following
buffer_wbinvl1_vol.
- Ensures the
atomicrmw has
completed before
invalidating the
cache.
4. buffer_wbinvl1_vol
- Must happen before
any following
global/generic
load/load
atomic/atomicrmw.
- Ensures that
following loads
will not see stale
global data.
2021-03-31 06:38:19 +08:00
atomicrmw acq_rel - system - global 1. s_waitcnt lgkmcnt(0) &
2021-02-18 05:37:46 +08:00
vmcnt(0)
- If TgSplit execution mode,
omit lgkmcnt(0).
- If OpenCL, omit
lgkmcnt(0).
- Could be split into
separate s_waitcnt
vmcnt(0) and
s_waitcnt
lgkmcnt(0) to allow
them to be
independently moved
according to the
following rules.
- s_waitcnt vmcnt(0)
must happen after
any preceding
global/generic
load/store/load
atomic/store
atomic/atomicrmw.
- s_waitcnt lgkmcnt(0)
must happen after
any preceding
local/generic
load/store/load
atomic/store
atomic/atomicrmw.
- Must happen before
the following
atomicrmw.
- Ensures that all
memory operations
to global and L2 writeback
have completed before
performing the
atomicrmw that is
being released.
2021-03-31 06:38:19 +08:00
2. buffer/global_atomic
3. s_waitcnt vmcnt(0)
2021-02-18 05:37:46 +08:00
- Must happen before
2021-03-31 06:38:19 +08:00
following
2021-02-18 05:37:46 +08:00
buffer_wbinvl1_vol.
- Ensures the
atomicrmw has
completed before
invalidating the
caches.
2021-03-31 06:38:19 +08:00
4. buffer_wbinvl1_vol
2021-02-18 05:37:46 +08:00
- Must happen before
any following
global/generic
load/load
atomic/atomicrmw.
- Ensures that
2021-03-31 06:38:19 +08:00
following
loads will not see
stale L1 global data.
2021-02-18 05:37:46 +08:00
MTYPE RW and CC memory will
2021-03-31 06:38:19 +08:00
never be stale in L2 due to
the memory probes.
2021-02-18 05:37:46 +08:00
atomicrmw acq_rel - agent - generic 1. s_waitcnt lgkmcnt(0) &
vmcnt(0)
- If TgSplit execution mode,
omit lgkmcnt(0).
- If OpenCL, omit
lgkmcnt(0).
- Could be split into
separate s_waitcnt
vmcnt(0) and
s_waitcnt
lgkmcnt(0) to allow
them to be
independently moved
according to the
following rules.
- s_waitcnt vmcnt(0)
must happen after
any preceding
global/generic
load/store/load
atomic/store
atomic/atomicrmw.
- s_waitcnt lgkmcnt(0)
must happen after
any preceding
local/generic
load/store/load
atomic/store
atomic/atomicrmw.
- Must happen before
the following
atomicrmw.
- Ensures that all
memory operations
to global have
completed before
performing the
atomicrmw that is
being released.
2. flat_atomic
3. s_waitcnt vmcnt(0) &
lgkmcnt(0)
- If TgSplit execution mode,
omit lgkmcnt(0).
- If OpenCL, omit
lgkmcnt(0).
- Must happen before
following
buffer_wbinvl1_vol.
- Ensures the
atomicrmw has
completed before
invalidating the
cache.
4. buffer_wbinvl1_vol
- Must happen before
any following
global/generic
load/load
atomic/atomicrmw.
- Ensures that
following loads
will not see stale
global data.
2021-03-31 06:38:19 +08:00
atomicrmw acq_rel - system - generic 1. s_waitcnt lgkmcnt(0) &
2021-02-18 05:37:46 +08:00
vmcnt(0)
- If TgSplit execution mode,
omit lgkmcnt(0).
- If OpenCL, omit
lgkmcnt(0).
- Could be split into
separate s_waitcnt
vmcnt(0) and
s_waitcnt
lgkmcnt(0) to allow
them to be
independently moved
according to the
following rules.
- s_waitcnt vmcnt(0)
must happen after
any preceding
global/generic
load/store/load
atomic/store
atomic/atomicrmw.
- s_waitcnt lgkmcnt(0)
must happen after
any preceding
local/generic
load/store/load
atomic/store
atomic/atomicrmw.
- Must happen before
the following
atomicrmw.
- Ensures that all
memory operations
to global and L2 writeback
have completed before
performing the
atomicrmw that is
being released.
2021-03-31 06:38:19 +08:00
2. flat_atomic
3. s_waitcnt vmcnt(0) &
2021-02-18 05:37:46 +08:00
lgkmcnt(0)
- If TgSplit execution mode,
omit lgkmcnt(0).
- If OpenCL, omit
lgkmcnt(0).
- Must happen before
2021-03-31 06:38:19 +08:00
following
2021-02-18 05:37:46 +08:00
buffer_wbinvl1_vol.
- Ensures the
atomicrmw has
completed before
invalidating the
caches.
2021-03-31 06:38:19 +08:00
4. buffer_wbinvl1_vol
2021-02-18 05:37:46 +08:00
- Must happen before
any following
global/generic
load/load
atomic/atomicrmw.
- Ensures that
2021-03-31 06:38:19 +08:00
following
loads will not see
stale L1 global data.
2021-02-18 05:37:46 +08:00
MTYPE RW and CC memory will
2021-03-31 06:38:19 +08:00
never be stale in L2 due to
the memory probes.
2021-02-18 05:37:46 +08:00
fence acq_rel - singlethread *none* *none*
- wavefront
fence acq_rel - workgroup *none* 1. s_waitcnt lgkm/vmcnt(0)
- Use lgkmcnt(0) if not
TgSplit execution mode
and vmcnt(0) if TgSplit
execution mode.
- If OpenCL and
address space is
not generic, omit
lgkmcnt(0).
- If OpenCL and
address space is
local, omit
vmcnt(0).
- However,
since LLVM
currently has no
address space on
the fence need to
conservatively
always generate
(see comment for
previous fence).
- s_waitcnt vmcnt(0)
must happen after
any preceding
global/generic
load/store/
load atomic/store atomic/
atomicrmw.
- s_waitcnt lgkmcnt(0)
must happen after
any preceding
local/generic
load/load
atomic/store/store
atomic/atomicrmw.
- Must happen before
any following
global/generic
load/load
atomic/store/store
atomic/atomicrmw.
- Ensures that all
memory operations
have
completed before
performing any
following global
memory operations.
- Ensures that the
preceding
local/generic load
atomic/atomicrmw
with an equal or
wider sync scope
and memory ordering
stronger than
unordered (this is
termed the
acquire-fence-paired-atomic)
has completed
before following
global memory
operations. This
satisfies the
requirements of
acquire.
- Ensures that all
previous memory
operations have
completed before a
following
local/generic store
atomic/atomicrmw
with an equal or
wider sync scope
and memory ordering
stronger than
unordered (this is
termed the
release-fence-paired-atomic).
This satisfies the
requirements of
release.
- Must happen before
the following
buffer_wbinvl1_vol.
- Ensures that the
acquire-fence-paired
atomic has completed
before invalidating
the
cache. Therefore
any following
locations read must
be no older than
the value read by
the
acquire-fence-paired-atomic.
3. buffer_wbinvl1_vol
- If not TgSplit execution
mode, omit.
- Ensures that
following
loads will not see
stale data.
fence acq_rel - agent *none* 1. s_waitcnt lgkmcnt(0) &
vmcnt(0)
- If TgSplit execution mode,
omit lgkmcnt(0).
- If OpenCL and
address space is
not generic, omit
lgkmcnt(0).
- However, since LLVM
currently has no
address space on
the fence need to
conservatively
always generate
(see comment for
previous fence).
- Could be split into
separate s_waitcnt
vmcnt(0) and
s_waitcnt
lgkmcnt(0) to allow
them to be
independently moved
according to the
following rules.
- s_waitcnt vmcnt(0)
must happen after
any preceding
global/generic
load/store/load
atomic/store
atomic/atomicrmw.
- s_waitcnt lgkmcnt(0)
must happen after
any preceding
local/generic
load/store/load
atomic/store
atomic/atomicrmw.
- Must happen before
the following
buffer_wbinvl1_vol.
- Ensures that the
preceding
global/local/generic
load
atomic/atomicrmw
with an equal or
wider sync scope
and memory ordering
stronger than
unordered (this is
termed the
acquire-fence-paired-atomic)
has completed
before invalidating
the cache. This
satisfies the
requirements of
acquire.
- Ensures that all
previous memory
operations have
completed before a
following
global/local/generic
store
atomic/atomicrmw
with an equal or
wider sync scope
and memory ordering
stronger than
unordered (this is
termed the
release-fence-paired-atomic).
This satisfies the
requirements of
release.
2. buffer_wbinvl1_vol
- Must happen before
any following
global/generic
load/load
atomic/store/store
atomic/atomicrmw.
- Ensures that
following loads
will not see stale
global data. This
satisfies the
requirements of
acquire.
2021-03-31 06:38:19 +08:00
fence acq_rel - system *none* 1. s_waitcnt lgkmcnt(0) &
2021-02-18 05:37:46 +08:00
vmcnt(0)
- If TgSplit execution mode,
omit lgkmcnt(0).
- If OpenCL and
address space is
not generic, omit
lgkmcnt(0).
- However, since LLVM
currently has no
address space on
the fence need to
conservatively
always generate
(see comment for
previous fence).
- Could be split into
separate s_waitcnt
vmcnt(0) and
s_waitcnt
lgkmcnt(0) to allow
them to be
independently moved
according to the
following rules.
- s_waitcnt vmcnt(0)
must happen after
any preceding
global/generic
load/store/load
atomic/store
atomic/atomicrmw.
- s_waitcnt lgkmcnt(0)
must happen after
any preceding
local/generic
load/store/load
atomic/store
atomic/atomicrmw.
- Must happen before
2021-03-31 06:38:19 +08:00
the following
2021-02-18 05:37:46 +08:00
buffer_wbinvl1_vol.
- Ensures that the
preceding
global/local/generic
load
atomic/atomicrmw
with an equal or
wider sync scope
and memory ordering
stronger than
unordered (this is
termed the
acquire-fence-paired-atomic)
has completed
before invalidating
the cache. This
satisfies the
requirements of
acquire.
- Ensures that all
previous memory
operations have
completed before a
following
global/local/generic
store
atomic/atomicrmw
with an equal or
wider sync scope
and memory ordering
stronger than
unordered (this is
termed the
release-fence-paired-atomic).
This satisfies the
requirements of
release.
2021-03-31 06:38:19 +08:00
2. buffer_wbinvl1_vol
2021-02-18 05:37:46 +08:00
- Must happen before
any following
global/generic
load/load
atomic/store/store
atomic/atomicrmw.
- Ensures that
2021-03-31 06:38:19 +08:00
following
loads will not see
stale L1 global data.
2021-02-18 05:37:46 +08:00
MTYPE RW and CC memory will
2021-03-31 06:38:19 +08:00
never be stale in L2 due to
the memory probes.
2021-02-18 05:37:46 +08:00
**Sequential Consistent Atomic**
------------------------------------------------------------------------------------
load atomic seq_cst - singlethread - global *Same as corresponding
- wavefront - local load atomic acquire,
- generic except must generated
all instructions even
for OpenCL.*
load atomic seq_cst - workgroup - global 1. s_waitcnt lgkm/vmcnt(0)
- generic
- Use lgkmcnt(0) if not
TgSplit execution mode
and vmcnt(0) if TgSplit
execution mode.
- s_waitcnt lgkmcnt(0) must
happen after
preceding
local/generic load
atomic/store
atomic/atomicrmw
with memory
ordering of seq_cst
and with equal or
wider sync scope.
(Note that seq_cst
fences have their
own s_waitcnt
lgkmcnt(0) and so do
not need to be
considered.)
- s_waitcnt vmcnt(0)
must happen after
preceding
global/generic load
atomic/store
atomic/atomicrmw
with memory
ordering of seq_cst
and with equal or
wider sync scope.
(Note that seq_cst
fences have their
own s_waitcnt
vmcnt(0) and so do
not need to be
considered.)
- Ensures any
preceding
sequential
consistent global/local
memory instructions
have completed
before executing
this sequentially
consistent
instruction. This
prevents reordering
a seq_cst store
followed by a
seq_cst load. (Note
that seq_cst is
stronger than
acquire/release as
the reordering of
load acquire
followed by a store
release is
prevented by the
s_waitcnt of
the release, but
there is nothing
preventing a store
release followed by
load acquire from
completing out of
order. The s_waitcnt
could be placed after
seq_store or before
the seq_load. We
choose the load to
make the s_waitcnt be
as late as possible
so that the store
may have already
completed.)
2. *Following
instructions same as
corresponding load
atomic acquire,
except must generated
all instructions even
for OpenCL.*
load atomic seq_cst - workgroup - local *If TgSplit execution mode,
local address space cannot
be used.*
*Same as corresponding
load atomic acquire,
except must generated
all instructions even
for OpenCL.*
load atomic seq_cst - agent - global 1. s_waitcnt lgkmcnt(0) &
- system - generic vmcnt(0)
- If TgSplit execution mode,
omit lgkmcnt(0).
- Could be split into
separate s_waitcnt
vmcnt(0)
and s_waitcnt
lgkmcnt(0) to allow
them to be
independently moved
according to the
following rules.
- s_waitcnt lgkmcnt(0)
must happen after
preceding
global/generic load
atomic/store
atomic/atomicrmw
with memory
ordering of seq_cst
and with equal or
wider sync scope.
(Note that seq_cst
fences have their
own s_waitcnt
lgkmcnt(0) and so do
not need to be
considered.)
- s_waitcnt vmcnt(0)
must happen after
preceding
global/generic load
atomic/store
atomic/atomicrmw
with memory
ordering of seq_cst
and with equal or
wider sync scope.
(Note that seq_cst
fences have their
own s_waitcnt
vmcnt(0) and so do
not need to be
considered.)
- Ensures any
preceding
sequential
consistent global
memory instructions
have completed
before executing
this sequentially
consistent
instruction. This
prevents reordering
a seq_cst store
followed by a
seq_cst load. (Note
that seq_cst is
stronger than
acquire/release as
the reordering of
load acquire
followed by a store
release is
prevented by the
s_waitcnt of
the release, but
there is nothing
preventing a store
release followed by
load acquire from
completing out of
order. The s_waitcnt
could be placed after
seq_store or before
the seq_load. We
choose the load to
make the s_waitcnt be
as late as possible
so that the store
may have already
completed.)
2. *Following
instructions same as
corresponding load
atomic acquire,
except must generated
all instructions even
for OpenCL.*
store atomic seq_cst - singlethread - global *Same as corresponding
- wavefront - local store atomic release,
- workgroup - generic except must generated
- agent all instructions even
- system for OpenCL.*
atomicrmw seq_cst - singlethread - global *Same as corresponding
- wavefront - local atomicrmw acq_rel,
- workgroup - generic except must generated
- agent all instructions even
- system for OpenCL.*
fence seq_cst - singlethread *none* *Same as corresponding
- wavefront fence acq_rel,
- workgroup except must generated
- agent all instructions even
- system for OpenCL.*
============ ============ ============== ========== ================================
.. _amdgpu-amdhsa-memory-model-gfx10:
Memory Model GFX10
++++++++++++++++++
For GFX10:
* Each agent has multiple shader arrays (SA).
* Each SA has multiple work-group processors (WGP).
* Each WGP has multiple compute units (CU).
* Each CU has multiple SIMDs that execute wavefronts.
* The wavefronts for a single work-group are executed in the same
WGP. In CU wavefront execution mode the wavefronts may be executed by
different SIMDs in the same CU. In WGP wavefront execution mode the
wavefronts may be executed by different SIMDs in different CUs in the same
WGP.
* Each WGP has a single LDS memory shared by the wavefronts of the work-groups
executing on it.
* All LDS operations of a WGP are performed as wavefront wide operations in a
global order and involve no caching. Completion is reported to a wavefront in
execution order.
* The LDS memory has multiple request queues shared by the SIMDs of a
WGP. Therefore, the LDS operations performed by different wavefronts of a
work-group can be reordered relative to each other, which can result in
reordering the visibility of vector memory operations with respect to LDS
operations of other wavefronts in the same work-group. A `` s_waitcnt
lgkmcnt(0)`` is required to ensure synchronization between LDS operations and
vector memory operations between wavefronts of a work-group, but not between
operations performed by the same wavefront.
* The vector memory operations are performed as wavefront wide operations.
Completion of load/store/sample operations are reported to a wavefront in
execution order of other load/store/sample operations performed by that
wavefront.
* The vector memory operations access a vector L0 cache. There is a single L0
cache per CU. Each SIMD of a CU accesses the same L0 cache. Therefore, no
special action is required for coherence between the lanes of a single
wavefront. However, a `` buffer_gl0_inv `` is required for coherence between
wavefronts executing in the same work-group as they may be executing on SIMDs
of different CUs that access different L0s. A `` buffer_gl0_inv `` is also
required for coherence between wavefronts executing in different work-groups
as they may be executing on different WGPs.
* The scalar memory operations access a scalar L0 cache shared by all wavefronts
on a WGP. The scalar and vector L0 caches are not coherent. However, scalar
operations are used in a restricted way so do not impact the memory model. See
:ref: `amdgpu-amdhsa-memory-spaces` .
* The vector and scalar memory L0 caches use an L1 cache shared by all WGPs on
the same SA. Therefore, no special action is required for coherence between
the wavefronts of a single work-group. However, a `` buffer_gl1_inv `` is
required for coherence between wavefronts executing in different work-groups
as they may be executing on different SAs that access different L1s.
* The L1 caches have independent quadrants to service disjoint ranges of virtual
addresses.
* Each L0 cache has a separate request queue per L1 quadrant. Therefore, the
vector and scalar memory operations performed by different wavefronts, whether
executing in the same or different work-groups (which may be executing on
different CUs accessing different L0s), can be reordered relative to each
other. A `` s_waitcnt vmcnt(0) & vscnt(0) `` is required to ensure
synchronization between vector memory operations of different wavefronts. It
ensures a previous vector memory operation has completed before executing a
subsequent vector memory or LDS operation and so can be used to meet the
requirements of acquire, release and sequential consistency.
* The L1 caches use an L2 cache shared by all SAs on the same agent.
* The L2 cache has independent channels to service disjoint ranges of virtual
addresses.
* Each L1 quadrant of a single SA accesses a different L2 channel. Each L1
quadrant has a separate request queue per L2 channel. Therefore, the vector
and scalar memory operations performed by wavefronts executing in different
work-groups (which may be executing on different SAs) of an agent can be
reordered relative to each other. A `` s_waitcnt vmcnt(0) & vscnt(0) `` is
required to ensure synchronization between vector memory operations of
different SAs. It ensures a previous vector memory operation has completed
before executing a subsequent vector memory and so can be used to meet the
requirements of acquire, release and sequential consistency.
* The L2 cache can be kept coherent with other agents on some targets, or ranges
of virtual addresses can be set up to bypass it to ensure system coherence.
Scalar memory operations are only used to access memory that is proven to not
change during the execution of the kernel dispatch. This includes constant
address space and global address space for program scope `` const `` variables.
Therefore, the kernel machine code does not have to maintain the scalar cache to
ensure it is coherent with the vector caches. The scalar and vector caches are
invalidated between kernel dispatches by CP since constant address space data
may change between kernel dispatch executions. See
:ref: `amdgpu-amdhsa-memory-spaces` .
The one exception is if scalar writes are used to spill SGPR registers. In this
case the AMDGPU backend ensures the memory location used to spill is never
accessed by vector memory operations at the same time. If scalar writes are used
then a `` s_dcache_wb `` is inserted before the `` s_endpgm `` and before a function
return since the locations may be used for vector memory instructions by a
future wavefront that uses the same scratch area, or a function call that
creates a frame at the same address, respectively. There is no need for a
`` s_dcache_inv `` as all scalar writes are write-before-read in the same thread.
For kernarg backing memory:
* CP invalidates the L0 and L1 caches at the start of each kernel dispatch.
* On dGPU the kernarg backing memory is accessed as MTYPE UC (uncached) to avoid
needing to invalidate the L2 cache.
* On APU the kernarg backing memory is accessed as MTYPE CC (cache coherent) and
so the L2 cache will be coherent with the CPU and other agents.
Scratch backing memory (which is used for the private address space) is accessed
with MTYPE NC (non-coherent). Since the private address space is only accessed
by a single thread, and is always write-before-read, there is never a need to
invalidate these entries from the L0 or L1 caches.
Wavefronts are executed in native mode with in-order reporting of loads and
sample instructions. In this mode vmcnt reports completion of load, atomic with
return and sample instructions in order, and the vscnt reports the completion of
store and atomic without return in order. See `` MEM_ORDERED `` field in
:ref: `amdgpu-amdhsa-compute_pgm_rsrc1-gfx6-gfx10-table` .
Wavefronts can be executed in WGP or CU wavefront execution mode:
* In WGP wavefront execution mode the wavefronts of a work-group are executed
on the SIMDs of both CUs of the WGP. Therefore, explicit management of the per
CU L0 caches is required for work-group synchronization. Also accesses to L1
at work-group scope need to be explicitly ordered as the accesses from
different CUs are not ordered.
* In CU wavefront execution mode the wavefronts of a work-group are executed on
the SIMDs of a single CU of the WGP. Therefore, all global memory access by
the work-group access the same L0 which in turn ensures L1 accesses are
ordered and so do not require explicit management of the caches for
work-group synchronization.
See `` WGP_MODE `` field in
:ref: `amdgpu-amdhsa-compute_pgm_rsrc1-gfx6-gfx10-table` and
:ref: `amdgpu-target-features` .
The code sequences used to implement the memory model for GFX10 are defined in
table :ref: `amdgpu-amdhsa-memory-model-code-sequences-gfx10-table` .
.. table :: AMDHSA Memory Model Code Sequences GFX10
:name: amdgpu-amdhsa-memory-model-code-sequences-gfx10-table
============ ============ ============== ========== ================================
LLVM Instr LLVM Memory LLVM Memory AMDGPU AMDGPU Machine Code
Ordering Sync Scope Address GFX10
Space
============ ============ ============== ========== ================================
**Non-Atomic**
------------------------------------------------------------------------------------
load *none* *none* - global - !volatile & !nontemporal
- generic
- private 1. buffer/global/flat_load
- constant
- !volatile & nontemporal
1. buffer/global/flat_load
slc=1
- volatile
1. buffer/global/flat_load
glc=1 dlc=1
2. s_waitcnt vmcnt(0)
- Must happen before
any following volatile
global/generic
load/store.
- Ensures that
volatile
operations to
different
addresses will not
be reordered by
hardware.
load *none* *none* - local 1. ds_load
store *none* *none* - global - !volatile & !nontemporal
- generic
- private 1. buffer/global/flat_store
- constant
- !volatile & nontemporal
1. buffer/global/flat_store
slc=1
- volatile
1. buffer/global/flat_store
2. s_waitcnt vscnt(0)
- Must happen before
any following volatile
global/generic
load/store.
- Ensures that
volatile
operations to
different
addresses will not
be reordered by
hardware.
store *none* *none* - local 1. ds_store
**Unordered Atomic**
------------------------------------------------------------------------------------
load atomic unordered *any* *any* *Same as non-atomic* .
store atomic unordered *any* *any* *Same as non-atomic* .
atomicrmw unordered *any* *any* *Same as monotonic atomic* .
**Monotonic Atomic**
------------------------------------------------------------------------------------
load atomic monotonic - singlethread - global 1. buffer/global/flat_load
- wavefront - generic
load atomic monotonic - workgroup - global 1. buffer/global/flat_load
- generic glc=1
- If CU wavefront execution
mode, omit glc=1.
load atomic monotonic - singlethread - local 1. ds_load
- wavefront
- workgroup
load atomic monotonic - agent - global 1. buffer/global/flat_load
- system - generic glc=1 dlc=1
store atomic monotonic - singlethread - global 1. buffer/global/flat_store
- wavefront - generic
- workgroup
- agent
- system
store atomic monotonic - singlethread - local 1. ds_store
- wavefront
- workgroup
atomicrmw monotonic - singlethread - global 1. buffer/global/flat_atomic
- wavefront - generic
- workgroup
- agent
- system
atomicrmw monotonic - singlethread - local 1. ds_atomic
- wavefront
- workgroup
**Acquire Atomic**
------------------------------------------------------------------------------------
load atomic acquire - singlethread - global 1. buffer/global/ds/flat_load
- wavefront - local
- generic
load atomic acquire - workgroup - global 1. buffer/global_load glc=1
- If CU wavefront execution
mode, omit glc=1.
2. s_waitcnt vmcnt(0)
- If CU wavefront execution
mode, omit.
- Must happen before
the following buffer_gl0_inv
and before any following
global/generic
load/load
atomic/store/store
atomic/atomicrmw.
3. buffer_gl0_inv
- If CU wavefront execution
mode, omit.
- Ensures that
following
loads will not see
stale data.
load atomic acquire - workgroup - local 1. ds_load
2. s_waitcnt lgkmcnt(0)
- If OpenCL, omit.
- Must happen before
the following buffer_gl0_inv
and before any following
global/generic load/load
atomic/store/store
atomic/atomicrmw.
- Ensures any
following global
data read is no
older than the local load
atomic value being
acquired.
3. buffer_gl0_inv
- If CU wavefront execution
mode, omit.
- If OpenCL, omit.
- Ensures that
following
loads will not see
stale data.
load atomic acquire - workgroup - generic 1. flat_load glc=1
- If CU wavefront execution
mode, omit glc=1.
2. s_waitcnt lgkmcnt(0) &
vmcnt(0)
- If CU wavefront execution
mode, omit vmcnt(0).
- If OpenCL, omit
lgkmcnt(0).
- Must happen before
the following
buffer_gl0_inv and any
following global/generic
load/load
atomic/store/store
atomic/atomicrmw.
- Ensures any
following global
data read is no
older than a local load
atomic value being
acquired.
3. buffer_gl0_inv
- If CU wavefront execution
mode, omit.
- Ensures that
following
loads will not see
stale data.
load atomic acquire - agent - global 1. buffer/global_load
- system glc=1 dlc=1
2. s_waitcnt vmcnt(0)
- Must happen before
following
buffer_gl*_inv.
- Ensures the load
has completed
before invalidating
the caches.
3. buffer_gl0_inv;
buffer_gl1_inv
- Must happen before
any following
global/generic
load/load
atomic/atomicrmw.
- Ensures that
following
loads will not see
stale global data.
load atomic acquire - agent - generic 1. flat_load glc=1 dlc=1
- system 2. s_waitcnt vmcnt(0) &
lgkmcnt(0)
- If OpenCL omit
lgkmcnt(0).
- Must happen before
following
buffer_gl*_invl.
- Ensures the flat_load
has completed
before invalidating
the caches.
3. buffer_gl0_inv;
buffer_gl1_inv
- Must happen before
any following
global/generic
load/load
atomic/atomicrmw.
- Ensures that
following loads
will not see stale
global data.
atomicrmw acquire - singlethread - global 1. buffer/global/ds/flat_atomic
- wavefront - local
2020-10-31 12:43:55 +08:00
- generic
atomicrmw acquire - workgroup - global 1. buffer/global_atomic
2. s_waitcnt vm/vscnt(0)
- If CU wavefront execution
mode, omit.
- Use vmcnt(0) if atomic with
return and vscnt(0) if
atomic with no-return.
- Must happen before
the following buffer_gl0_inv
and before any following
global/generic
load/load
atomic/store/store
atomic/atomicrmw.
3. buffer_gl0_inv
- If CU wavefront execution
mode, omit.
- Ensures that
following
loads will not see
stale data.
atomicrmw acquire - workgroup - local 1. ds_atomic
2. s_waitcnt lgkmcnt(0)
- If OpenCL, omit.
- Must happen before
the following
buffer_gl0_inv.
- Ensures any
following global
data read is no
older than the local
atomicrmw value
being acquired.
3. buffer_gl0_inv
- If OpenCL omit.
- Ensures that
following
loads will not see
stale data.
atomicrmw acquire - workgroup - generic 1. flat_atomic
2. s_waitcnt lgkmcnt(0) &
vm/vscnt(0)
- If CU wavefront execution
mode, omit vm/vscnt(0).
- If OpenCL, omit lgkmcnt(0).
- Use vmcnt(0) if atomic with
return and vscnt(0) if
atomic with no-return.
- Must happen before
the following
buffer_gl0_inv.
- Ensures any
following global
data read is no
older than a local
atomicrmw value
being acquired.
3. buffer_gl0_inv
- If CU wavefront execution
mode, omit.
- Ensures that
following
loads will not see
stale data.
atomicrmw acquire - agent - global 1. buffer/global_atomic
- system 2. s_waitcnt vm/vscnt(0)
- Use vmcnt(0) if atomic with
return and vscnt(0) if
atomic with no-return.
- Must happen before
following
buffer_gl*_inv.
- Ensures the
atomicrmw has
completed before
invalidating the
caches.
3. buffer_gl0_inv;
buffer_gl1_inv
- Must happen before
any following
global/generic
load/load
atomic/atomicrmw.
- Ensures that
following loads
will not see stale
global data.
atomicrmw acquire - agent - generic 1. flat_atomic
- system 2. s_waitcnt vm/vscnt(0) &
lgkmcnt(0)
- If OpenCL, omit
lgkmcnt(0).
- Use vmcnt(0) if atomic with
return and vscnt(0) if
atomic with no-return.
- Must happen before
following
buffer_gl*_inv.
- Ensures the
atomicrmw has
completed before
invalidating the
caches.
3. buffer_gl0_inv;
buffer_gl1_inv
- Must happen before
any following
global/generic
load/load
atomic/atomicrmw.
- Ensures that
following loads
will not see stale
global data.
fence acquire - singlethread *none* *none*
- wavefront
fence acquire - workgroup *none* 1. s_waitcnt lgkmcnt(0) &
vmcnt(0) & vscnt(0)
- If CU wavefront execution
mode, omit vmcnt(0) and
vscnt(0).
- If OpenCL and
address space is
not generic, omit
lgkmcnt(0).
- If OpenCL and
address space is
local, omit
vmcnt(0) and vscnt(0).
- However, since LLVM
currently has no
address space on
the fence need to
conservatively
always generate. If
fence had an
address space then
set to address
space of OpenCL
fence flag, or to
generic if both
local and global
flags are
specified.
- Could be split into
separate s_waitcnt
vmcnt(0), s_waitcnt
vscnt(0) and s_waitcnt
lgkmcnt(0) to allow
them to be
independently moved
according to the
following rules.
- s_waitcnt vmcnt(0)
must happen after
any preceding
global/generic load
atomic/
atomicrmw-with-return-value
with an equal or
wider sync scope
and memory ordering
stronger than
unordered (this is
termed the
fence-paired-atomic).
- s_waitcnt vscnt(0)
must happen after
any preceding
global/generic
atomicrmw-no-return-value
with an equal or
wider sync scope
and memory ordering
stronger than
unordered (this is
termed the
fence-paired-atomic).
- s_waitcnt lgkmcnt(0)
must happen after
any preceding
local/generic load
atomic/atomicrmw
with an equal or
wider sync scope
and memory ordering
stronger than
unordered (this is
termed the
fence-paired-atomic).
- Must happen before
the following
buffer_gl0_inv.
- Ensures that the
fence-paired atomic
has completed
before invalidating
the
cache. Therefore
any following
locations read must
be no older than
the value read by
the
fence-paired-atomic.
3. buffer_gl0_inv
- If CU wavefront execution
mode, omit.
- Ensures that
following
loads will not see
stale data.
fence acquire - agent *none* 1. s_waitcnt lgkmcnt(0) &
- system vmcnt(0) & vscnt(0)
- If OpenCL and
address space is
not generic, omit
lgkmcnt(0).
- If OpenCL and
address space is
local, omit
vmcnt(0) and vscnt(0).
- However, since LLVM
currently has no
address space on
the fence need to
conservatively
always generate
(see comment for
previous fence).
- Could be split into
separate s_waitcnt
vmcnt(0), s_waitcnt
vscnt(0) and s_waitcnt
lgkmcnt(0) to allow
them to be
independently moved
according to the
following rules.
- s_waitcnt vmcnt(0)
must happen after
any preceding
global/generic load
atomic/
atomicrmw-with-return-value
with an equal or
wider sync scope
and memory ordering
stronger than
unordered (this is
termed the
fence-paired-atomic).
- s_waitcnt vscnt(0)
must happen after
any preceding
global/generic
atomicrmw-no-return-value
with an equal or
wider sync scope
and memory ordering
stronger than
unordered (this is
termed the
fence-paired-atomic).
- s_waitcnt lgkmcnt(0)
must happen after
any preceding
local/generic load
atomic/atomicrmw
with an equal or
wider sync scope
and memory ordering
stronger than
unordered (this is
termed the
fence-paired-atomic).
- Must happen before
the following
buffer_gl*_inv.
- Ensures that the
fence-paired atomic
has completed
before invalidating
the
caches. Therefore
any following
locations read must
be no older than
the value read by
the
fence-paired-atomic.
2. buffer_gl0_inv;
buffer_gl1_inv
- Must happen before any
following global/generic
load/load
atomic/store/store
atomic/atomicrmw.
- Ensures that
following loads
will not see stale
global data.
**Release Atomic**
------------------------------------------------------------------------------------
store atomic release - singlethread - global 1. buffer/global/ds/flat_store
- wavefront - local
- generic
store atomic release - workgroup - global 1. s_waitcnt lgkmcnt(0) &
- generic vmcnt(0) & vscnt(0)
- If CU wavefront execution
mode, omit vmcnt(0) and
vscnt(0).
- If OpenCL, omit
lgkmcnt(0).
- Could be split into
separate s_waitcnt
vmcnt(0), s_waitcnt
vscnt(0) and s_waitcnt
lgkmcnt(0) to allow
them to be
independently moved
according to the
following rules.
- s_waitcnt vmcnt(0)
must happen after
any preceding
global/generic load/load
atomic/
atomicrmw-with-return-value.
- s_waitcnt vscnt(0)
must happen after
any preceding
global/generic
store/store
atomic/
atomicrmw-no-return-value.
- s_waitcnt lgkmcnt(0)
must happen after
any preceding
local/generic
load/store/load
atomic/store
atomic/atomicrmw.
- Must happen before
the following
store.
- Ensures that all
memory operations
have
completed before
performing the
store that is being
released.
2. buffer/global/flat_store
store atomic release - workgroup - local 1. s_waitcnt vmcnt(0) & vscnt(0)
- If CU wavefront execution
mode, omit.
- If OpenCL, omit.
- Could be split into
separate s_waitcnt
vmcnt(0) and s_waitcnt
vscnt(0) to allow
them to be
independently moved
according to the
following rules.
- s_waitcnt vmcnt(0)
must happen after
any preceding
global/generic load/load
atomic/
atomicrmw-with-return-value.
- s_waitcnt vscnt(0)
must happen after
any preceding
global/generic
store/store atomic/
atomicrmw-no-return-value.
- Must happen before
the following
store.
- Ensures that all
global memory
operations have
completed before
performing the
store that is being
released.
2. ds_store
store atomic release - agent - global 1. s_waitcnt lgkmcnt(0) &
- system - generic vmcnt(0) & vscnt(0)
- If OpenCL and
address space is
not generic, omit
lgkmcnt(0).
- Could be split into
separate s_waitcnt
vmcnt(0), s_waitcnt vscnt(0)
and s_waitcnt
lgkmcnt(0) to allow
them to be
independently moved
according to the
following rules.
- s_waitcnt vmcnt(0)
must happen after
any preceding
global/generic
load/load
atomic/
atomicrmw-with-return-value.
- s_waitcnt vscnt(0)
must happen after
any preceding
global/generic
store/store atomic/
atomicrmw-no-return-value.
- s_waitcnt lgkmcnt(0)
must happen after
any preceding
local/generic
load/store/load
atomic/store
atomic/atomicrmw.
- Must happen before
the following
store.
- Ensures that all
memory operations
have
completed before
performing the
store that is being
released.
2. buffer/global/flat_store
atomicrmw release - singlethread - global 1. buffer/global/ds/flat_atomic
- wavefront - local
- generic
atomicrmw release - workgroup - global 1. s_waitcnt lgkmcnt(0) &
- generic vmcnt(0) & vscnt(0)
- If CU wavefront execution
mode, omit vmcnt(0) and
vscnt(0).
- If OpenCL, omit lgkmcnt(0).
- Could be split into
separate s_waitcnt
vmcnt(0), s_waitcnt
vscnt(0) and s_waitcnt
lgkmcnt(0) to allow
them to be
independently moved
according to the
following rules.
- s_waitcnt vmcnt(0)
must happen after
any preceding
global/generic load/load
atomic/
atomicrmw-with-return-value.
- s_waitcnt vscnt(0)
must happen after
any preceding
global/generic
store/store
atomic/
atomicrmw-no-return-value.
- s_waitcnt lgkmcnt(0)
must happen after
any preceding
local/generic
load/store/load
atomic/store
atomic/atomicrmw.
- Must happen before
the following
atomicrmw.
- Ensures that all
memory operations
have
completed before
performing the
atomicrmw that is
being released.
2. buffer/global/flat_atomic
atomicrmw release - workgroup - local 1. s_waitcnt vmcnt(0) & vscnt(0)
- If CU wavefront execution
mode, omit.
- If OpenCL, omit.
- Could be split into
separate s_waitcnt
vmcnt(0) and s_waitcnt
vscnt(0) to allow
them to be
independently moved
according to the
following rules.
- s_waitcnt vmcnt(0)
must happen after
any preceding
global/generic load/load
atomic/
atomicrmw-with-return-value.
- s_waitcnt vscnt(0)
must happen after
any preceding
global/generic
store/store atomic/
atomicrmw-no-return-value.
- Must happen before
the following
store.
- Ensures that all
global memory
operations have
completed before
performing the
store that is being
released.
2. ds_atomic
atomicrmw release - agent - global 1. s_waitcnt lgkmcnt(0) &
- system - generic vmcnt(0) & vscnt(0)
- If OpenCL, omit
lgkmcnt(0).
- Could be split into
separate s_waitcnt
vmcnt(0), s_waitcnt
vscnt(0) and s_waitcnt
lgkmcnt(0) to allow
them to be
independently moved
according to the
following rules.
- s_waitcnt vmcnt(0)
must happen after
any preceding
global/generic
load/load atomic/
atomicrmw-with-return-value.
- s_waitcnt vscnt(0)
must happen after
any preceding
global/generic
store/store atomic/
atomicrmw-no-return-value.
- s_waitcnt lgkmcnt(0)
must happen after
any preceding
local/generic
load/store/load
atomic/store
atomic/atomicrmw.
- Must happen before
the following
atomicrmw.
- Ensures that all
memory operations
to global and local
have completed
before performing
the atomicrmw that
is being released.
2. buffer/global/flat_atomic
fence release - singlethread *none* *none*
- wavefront
fence release - workgroup *none* 1. s_waitcnt lgkmcnt(0) &
vmcnt(0) & vscnt(0)
- If CU wavefront execution
mode, omit vmcnt(0) and
vscnt(0).
- If OpenCL and
address space is
not generic, omit
lgkmcnt(0).
- If OpenCL and
address space is
local, omit
vmcnt(0) and vscnt(0).
- However, since LLVM
currently has no
address space on
the fence need to
conservatively
always generate. If
fence had an
address space then
set to address
space of OpenCL
fence flag, or to
generic if both
local and global
flags are
specified.
- Could be split into
separate s_waitcnt
vmcnt(0), s_waitcnt
vscnt(0) and s_waitcnt
lgkmcnt(0) to allow
them to be
independently moved
according to the
following rules.
- s_waitcnt vmcnt(0)
must happen after
any preceding
global/generic
load/load
atomic/
atomicrmw-with-return-value.
- s_waitcnt vscnt(0)
must happen after
any preceding
global/generic
store/store atomic/
atomicrmw-no-return-value.
- s_waitcnt lgkmcnt(0)
must happen after
any preceding
local/generic
load/store/load
atomic/store atomic/
atomicrmw.
- Must happen before
any following store
atomic/atomicrmw
with an equal or
wider sync scope
and memory ordering
stronger than
unordered (this is
termed the
fence-paired-atomic).
- Ensures that all
memory operations
have
completed before
performing the
following
fence-paired-atomic.
fence release - agent *none* 1. s_waitcnt lgkmcnt(0) &
- system vmcnt(0) & vscnt(0)
- If OpenCL and
address space is
not generic, omit
lgkmcnt(0).
- If OpenCL and
address space is
local, omit
vmcnt(0) and vscnt(0).
- However, since LLVM
currently has no
address space on
the fence need to
conservatively
always generate. If
fence had an
address space then
set to address
space of OpenCL
fence flag, or to
generic if both
local and global
flags are
specified.
- Could be split into
separate s_waitcnt
vmcnt(0), s_waitcnt
vscnt(0) and s_waitcnt
lgkmcnt(0) to allow
them to be
independently moved
according to the
following rules.
- s_waitcnt vmcnt(0)
must happen after
any preceding
global/generic
load/load atomic/
atomicrmw-with-return-value.
- s_waitcnt vscnt(0)
must happen after
any preceding
global/generic
store/store atomic/
atomicrmw-no-return-value.
- s_waitcnt lgkmcnt(0)
must happen after
any preceding
local/generic
load/store/load
atomic/store
atomic/atomicrmw.
- Must happen before
any following store
atomic/atomicrmw
with an equal or
wider sync scope
and memory ordering
stronger than
unordered (this is
termed the
fence-paired-atomic).
- Ensures that all
memory operations
have
completed before
performing the
following
fence-paired-atomic.
**Acquire-Release Atomic**
------------------------------------------------------------------------------------
atomicrmw acq_rel - singlethread - global 1. buffer/global/ds/flat_atomic
- wavefront - local
- generic
atomicrmw acq_rel - workgroup - global 1. s_waitcnt lgkmcnt(0) &
vmcnt(0) & vscnt(0)
- If CU wavefront execution
mode, omit vmcnt(0) and
vscnt(0).
- If OpenCL, omit
lgkmcnt(0).
- Must happen after
any preceding
local/generic
load/store/load
atomic/store
atomic/atomicrmw.
- Could be split into
separate s_waitcnt
vmcnt(0), s_waitcnt
vscnt(0), and s_waitcnt
lgkmcnt(0) to allow
them to be
independently moved
according to the
following rules.
- s_waitcnt vmcnt(0)
must happen after
any preceding
global/generic load/load
atomic/
atomicrmw-with-return-value.
- s_waitcnt vscnt(0)
must happen after
any preceding
global/generic
store/store
atomic/
atomicrmw-no-return-value.
- s_waitcnt lgkmcnt(0)
must happen after
any preceding
local/generic
load/store/load
atomic/store
atomic/atomicrmw.
- Must happen before
the following
atomicrmw.
- Ensures that all
memory operations
have
completed before
performing the
atomicrmw that is
being released.
2. buffer/global_atomic
3. s_waitcnt vm/vscnt(0)
- If CU wavefront execution
mode, omit.
- Use vmcnt(0) if atomic with
return and vscnt(0) if
atomic with no-return.
- Must happen before
the following
buffer_gl0_inv.
- Ensures any
following global
data read is no
older than the
atomicrmw value
being acquired.
4. buffer_gl0_inv
- If CU wavefront execution
mode, omit.
- Ensures that
following
loads will not see
stale data.
atomicrmw acq_rel - workgroup - local 1. s_waitcnt vmcnt(0) & vscnt(0)
- If CU wavefront execution
mode, omit.
- If OpenCL, omit.
- Could be split into
separate s_waitcnt
vmcnt(0) and s_waitcnt
vscnt(0) to allow
them to be
independently moved
according to the
following rules.
- s_waitcnt vmcnt(0)
must happen after
any preceding
global/generic load/load
atomic/
atomicrmw-with-return-value.
- s_waitcnt vscnt(0)
must happen after
any preceding
global/generic
store/store atomic/
atomicrmw-no-return-value.
- Must happen before
the following
store.
- Ensures that all
global memory
operations have
completed before
performing the
store that is being
released.
2. ds_atomic
3. s_waitcnt lgkmcnt(0)
- If OpenCL, omit.
- Must happen before
the following
buffer_gl0_inv.
- Ensures any
following global
data read is no
older than the local load
atomic value being
acquired.
4. buffer_gl0_inv
- If CU wavefront execution
mode, omit.
- If OpenCL omit.
- Ensures that
following
loads will not see
stale data.
atomicrmw acq_rel - workgroup - generic 1. s_waitcnt lgkmcnt(0) &
vmcnt(0) & vscnt(0)
- If CU wavefront execution
mode, omit vmcnt(0) and
vscnt(0).
- If OpenCL, omit lgkmcnt(0).
- Could be split into
separate s_waitcnt
vmcnt(0), s_waitcnt
vscnt(0) and s_waitcnt
lgkmcnt(0) to allow
them to be
independently moved
according to the
following rules.
- s_waitcnt vmcnt(0)
must happen after
any preceding
global/generic load/load
atomic/
atomicrmw-with-return-value.
- s_waitcnt vscnt(0)
must happen after
any preceding
global/generic
store/store
atomic/
atomicrmw-no-return-value.
- s_waitcnt lgkmcnt(0)
must happen after
any preceding
local/generic
load/store/load
atomic/store
atomic/atomicrmw.
- Must happen before
the following
atomicrmw.
- Ensures that all
memory operations
have
completed before
performing the
atomicrmw that is
being released.
2. flat_atomic
3. s_waitcnt lgkmcnt(0) &
vmcnt(0) & vscnt(0)
- If CU wavefront execution
mode, omit vmcnt(0) and
vscnt(0).
- If OpenCL, omit lgkmcnt(0).
- Must happen before
the following
buffer_gl0_inv.
- Ensures any
following global
data read is no
older than the load
atomic value being
acquired.
3. buffer_gl0_inv
- If CU wavefront execution
mode, omit.
- Ensures that
following
loads will not see
stale data.
atomicrmw acq_rel - agent - global 1. s_waitcnt lgkmcnt(0) &
- system vmcnt(0) & vscnt(0)
- If OpenCL, omit
lgkmcnt(0).
- Could be split into
separate s_waitcnt
vmcnt(0), s_waitcnt
vscnt(0) and s_waitcnt
lgkmcnt(0) to allow
them to be
independently moved
according to the
following rules.
- s_waitcnt vmcnt(0)
must happen after
any preceding
global/generic
load/load atomic/
atomicrmw-with-return-value.
- s_waitcnt vscnt(0)
must happen after
any preceding
global/generic
store/store atomic/
atomicrmw-no-return-value.
- s_waitcnt lgkmcnt(0)
must happen after
any preceding
local/generic
load/store/load
atomic/store
atomic/atomicrmw.
- Must happen before
the following
atomicrmw.
- Ensures that all
memory operations
to global have
completed before
performing the
atomicrmw that is
being released.
2. buffer/global_atomic
3. s_waitcnt vm/vscnt(0)
- Use vmcnt(0) if atomic with
return and vscnt(0) if
atomic with no-return.
- Must happen before
following
buffer_gl*_inv.
- Ensures the
atomicrmw has
completed before
invalidating the
caches.
4. buffer_gl0_inv;
buffer_gl1_inv
- Must happen before
any following
global/generic
load/load
atomic/atomicrmw.
- Ensures that
following loads
will not see stale
global data.
atomicrmw acq_rel - agent - generic 1. s_waitcnt lgkmcnt(0) &
- system vmcnt(0) & vscnt(0)
- If OpenCL, omit
lgkmcnt(0).
- Could be split into
separate s_waitcnt
vmcnt(0), s_waitcnt
vscnt(0), and s_waitcnt
lgkmcnt(0) to allow
them to be
independently moved
according to the
following rules.
- s_waitcnt vmcnt(0)
must happen after
any preceding
global/generic
load/load atomic
atomicrmw-with-return-value.
- s_waitcnt vscnt(0)
must happen after
any preceding
global/generic
store/store atomic/
atomicrmw-no-return-value.
- s_waitcnt lgkmcnt(0)
must happen after
any preceding
local/generic
load/store/load
atomic/store
atomic/atomicrmw.
- Must happen before
the following
atomicrmw.
- Ensures that all
memory operations
have
completed before
performing the
atomicrmw that is
being released.
2. flat_atomic
3. s_waitcnt vm/vscnt(0) &
lgkmcnt(0)
- If OpenCL, omit
lgkmcnt(0).
- Use vmcnt(0) if atomic with
return and vscnt(0) if
atomic with no-return.
- Must happen before
following
buffer_gl*_inv.
- Ensures the
atomicrmw has
completed before
invalidating the
caches.
4. buffer_gl0_inv;
buffer_gl1_inv
- Must happen before
any following
global/generic
load/load
atomic/atomicrmw.
- Ensures that
following loads
will not see stale
global data.
fence acq_rel - singlethread *none* *none*
- wavefront
fence acq_rel - workgroup *none* 1. s_waitcnt lgkmcnt(0) &
vmcnt(0) & vscnt(0)
- If CU wavefront execution
mode, omit vmcnt(0) and
vscnt(0).
- If OpenCL and
address space is
not generic, omit
lgkmcnt(0).
- If OpenCL and
address space is
local, omit
vmcnt(0) and vscnt(0).
- However,
since LLVM
currently has no
address space on
the fence need to
conservatively
always generate
(see comment for
previous fence).
- Could be split into
separate s_waitcnt
vmcnt(0), s_waitcnt
vscnt(0) and s_waitcnt
lgkmcnt(0) to allow
them to be
independently moved
according to the
following rules.
- s_waitcnt vmcnt(0)
must happen after
any preceding
global/generic
load/load
atomic/
atomicrmw-with-return-value.
- s_waitcnt vscnt(0)
must happen after
any preceding
global/generic
store/store atomic/
atomicrmw-no-return-value.
- s_waitcnt lgkmcnt(0)
must happen after
any preceding
local/generic
load/store/load
atomic/store atomic/
atomicrmw.
- Must happen before
any following
global/generic
load/load
atomic/store/store
atomic/atomicrmw.
- Ensures that all
memory operations
have
completed before
performing any
following global
memory operations.
- Ensures that the
preceding
local/generic load
atomic/atomicrmw
with an equal or
wider sync scope
and memory ordering
stronger than
unordered (this is
termed the
acquire-fence-paired-atomic)
has completed
before following
global memory
operations. This
satisfies the
requirements of
acquire.
- Ensures that all
previous memory
operations have
completed before a
following
local/generic store
atomic/atomicrmw
with an equal or
wider sync scope
and memory ordering
stronger than
unordered (this is
termed the
release-fence-paired-atomic).
This satisfies the
requirements of
release.
- Must happen before
the following
buffer_gl0_inv.
- Ensures that the
acquire-fence-paired
atomic has completed
before invalidating
the
cache. Therefore
any following
locations read must
be no older than
the value read by
the
acquire-fence-paired-atomic.
3. buffer_gl0_inv
- If CU wavefront execution
mode, omit.
- Ensures that
following
loads will not see
stale data.
fence acq_rel - agent *none* 1. s_waitcnt lgkmcnt(0) &
- system vmcnt(0) & vscnt(0)
- If OpenCL and
address space is
not generic, omit
lgkmcnt(0).
- If OpenCL and
address space is
local, omit
vmcnt(0) and vscnt(0).
- However, since LLVM
currently has no
address space on
the fence need to
conservatively
always generate
(see comment for
previous fence).
- Could be split into
separate s_waitcnt
vmcnt(0), s_waitcnt
vscnt(0) and s_waitcnt
lgkmcnt(0) to allow
them to be
independently moved
according to the
following rules.
- s_waitcnt vmcnt(0)
must happen after
any preceding
global/generic
load/load
atomic/
atomicrmw-with-return-value.
- s_waitcnt vscnt(0)
must happen after
any preceding
global/generic
store/store atomic/
atomicrmw-no-return-value.
- s_waitcnt lgkmcnt(0)
must happen after
any preceding
local/generic
load/store/load
atomic/store
atomic/atomicrmw.
- Must happen before
the following
buffer_gl*_inv.
- Ensures that the
preceding
global/local/generic
load
atomic/atomicrmw
with an equal or
wider sync scope
and memory ordering
stronger than
unordered (this is
termed the
acquire-fence-paired-atomic)
has completed
before invalidating
the caches. This
satisfies the
requirements of
acquire.
- Ensures that all
previous memory
operations have
completed before a
following
global/local/generic
store
atomic/atomicrmw
with an equal or
wider sync scope
and memory ordering
stronger than
unordered (this is
termed the
release-fence-paired-atomic).
This satisfies the
requirements of
release.
2. buffer_gl0_inv;
buffer_gl1_inv
- Must happen before
any following
global/generic
load/load
atomic/store/store
atomic/atomicrmw.
- Ensures that
following loads
will not see stale
global data. This
satisfies the
requirements of
acquire.
**Sequential Consistent Atomic**
------------------------------------------------------------------------------------
load atomic seq_cst - singlethread - global *Same as corresponding
- wavefront - local load atomic acquire,
- generic except must generated
all instructions even
for OpenCL.*
load atomic seq_cst - workgroup - global 1. s_waitcnt lgkmcnt(0) &
- generic vmcnt(0) & vscnt(0)
- If CU wavefront execution
mode, omit vmcnt(0) and
vscnt(0).
- Could be split into
separate s_waitcnt
vmcnt(0), s_waitcnt
vscnt(0), and s_waitcnt
lgkmcnt(0) to allow
them to be
independently moved
according to the
following rules.
- s_waitcnt lgkmcnt(0) must
happen after
preceding
local/generic load
atomic/store
atomic/atomicrmw
with memory
ordering of seq_cst
and with equal or
wider sync scope.
(Note that seq_cst
fences have their
own s_waitcnt
lgkmcnt(0) and so do
not need to be
considered.)
- s_waitcnt vmcnt(0)
must happen after
preceding
global/generic load
atomic/
atomicrmw-with-return-value
with memory
ordering of seq_cst
and with equal or
wider sync scope.
(Note that seq_cst
fences have their
own s_waitcnt
vmcnt(0) and so do
not need to be
considered.)
- s_waitcnt vscnt(0)
Must happen after
preceding
global/generic store
atomic/
atomicrmw-no-return-value
with memory
ordering of seq_cst
and with equal or
wider sync scope.
(Note that seq_cst
fences have their
own s_waitcnt
vscnt(0) and so do
not need to be
considered.)
- Ensures any
preceding
sequential
consistent global/local
memory instructions
have completed
before executing
this sequentially
consistent
instruction. This
prevents reordering
a seq_cst store
followed by a
seq_cst load. (Note
that seq_cst is
stronger than
acquire/release as
the reordering of
load acquire
followed by a store
release is
prevented by the
s_waitcnt of
the release, but
there is nothing
preventing a store
release followed by
load acquire from
completing out of
order. The s_waitcnt
could be placed after
seq_store or before
the seq_load. We
choose the load to
make the s_waitcnt be
as late as possible
so that the store
may have already
completed.)
2. *Following
instructions same as
corresponding load
atomic acquire,
except must generated
all instructions even
for OpenCL.*
load atomic seq_cst - workgroup - local
1. s_waitcnt vmcnt(0) & vscnt(0)
- If CU wavefront execution
mode, omit.
- Could be split into
separate s_waitcnt
vmcnt(0) and s_waitcnt
vscnt(0) to allow
them to be
independently moved
according to the
following rules.
- s_waitcnt vmcnt(0)
Must happen after
preceding
global/generic load
atomic/
atomicrmw-with-return-value
with memory
ordering of seq_cst
and with equal or
wider sync scope.
(Note that seq_cst
fences have their
own s_waitcnt
vmcnt(0) and so do
not need to be
considered.)
- s_waitcnt vscnt(0)
Must happen after
preceding
global/generic store
atomic/
atomicrmw-no-return-value
with memory
ordering of seq_cst
and with equal or
wider sync scope.
(Note that seq_cst
fences have their
own s_waitcnt
vscnt(0) and so do
not need to be
considered.)
- Ensures any
preceding
sequential
consistent global
memory instructions
have completed
before executing
this sequentially
consistent
instruction. This
prevents reordering
a seq_cst store
followed by a
seq_cst load. (Note
that seq_cst is
stronger than
acquire/release as
the reordering of
load acquire
followed by a store
release is
prevented by the
s_waitcnt of
the release, but
there is nothing
preventing a store
release followed by
load acquire from
completing out of
order. The s_waitcnt
could be placed after
seq_store or before
the seq_load. We
choose the load to
make the s_waitcnt be
as late as possible
so that the store
may have already
completed.)
2. *Following
instructions same as
corresponding load
atomic acquire,
except must generated
all instructions even
for OpenCL.*
load atomic seq_cst - agent - global 1. s_waitcnt lgkmcnt(0) &
- system - generic vmcnt(0) & vscnt(0)
- Could be split into
separate s_waitcnt
vmcnt(0), s_waitcnt
vscnt(0) and s_waitcnt
lgkmcnt(0) to allow
them to be
independently moved
according to the
following rules.
- s_waitcnt lgkmcnt(0)
must happen after
preceding
local load
atomic/store
atomic/atomicrmw
with memory
ordering of seq_cst
and with equal or
wider sync scope.
(Note that seq_cst
fences have their
own s_waitcnt
lgkmcnt(0) and so do
not need to be
considered.)
- s_waitcnt vmcnt(0)
must happen after
preceding
global/generic load
atomic/
atomicrmw-with-return-value
with memory
ordering of seq_cst
and with equal or
wider sync scope.
(Note that seq_cst
fences have their
own s_waitcnt
vmcnt(0) and so do
not need to be
considered.)
- s_waitcnt vscnt(0)
Must happen after
preceding
global/generic store
atomic/
atomicrmw-no-return-value
with memory
ordering of seq_cst
and with equal or
wider sync scope.
(Note that seq_cst
fences have their
own s_waitcnt
vscnt(0) and so do
not need to be
considered.)
- Ensures any
preceding
sequential
consistent global
memory instructions
have completed
before executing
this sequentially
consistent
instruction. This
prevents reordering
a seq_cst store
followed by a
seq_cst load. (Note
that seq_cst is
stronger than
acquire/release as
the reordering of
load acquire
followed by a store
release is
prevented by the
s_waitcnt of
the release, but
there is nothing
preventing a store
release followed by
load acquire from
completing out of
order. The s_waitcnt
could be placed after
seq_store or before
the seq_load. We
choose the load to
make the s_waitcnt be
as late as possible
so that the store
may have already
completed.)
2. *Following
instructions same as
corresponding load
atomic acquire,
except must generated
all instructions even
for OpenCL.*
store atomic seq_cst - singlethread - global *Same as corresponding
- wavefront - local store atomic release,
- workgroup - generic except must generated
- agent all instructions even
- system for OpenCL.*
atomicrmw seq_cst - singlethread - global *Same as corresponding
- wavefront - local atomicrmw acq_rel,
- workgroup - generic except must generated
- agent all instructions even
- system for OpenCL.*
fence seq_cst - singlethread *none* *Same as corresponding
- wavefront fence acq_rel,
- workgroup except must generated
- agent all instructions even
- system for OpenCL.*
============ ============ ============== ========== ================================
2017-06-07 04:31:59 +08:00
Trap Handler ABI
~~~~~~~~~~~~~~~~
2020-12-01 12:12:04 +08:00
For code objects generated by the AMDGPU backend for HSA [HSA]_ compatible
runtimes (see :ref: `amdgpu-os` ), the runtime installs a trap handler that
supports the `` s_trap `` instruction. For usage see:
- :ref: `amdgpu-trap-handler-for-amdhsa-os-v2-table`
- :ref: `amdgpu-trap-handler-for-amdhsa-os-v3-table`
- :ref: `amdgpu-trap-handler-for-amdhsa-os-v4-table`
.. table :: AMDGPU Trap Handler for AMDHSA OS Code Object V2
:name: amdgpu-trap-handler-for-amdhsa-os-v2-table
2017-06-07 04:31:59 +08:00
2020-12-01 12:12:04 +08:00
=================== =============== =============== =======================================
Usage Code Sequence Trap Handler Description
Inputs
=================== =============== =============== =======================================
reserved `` s_trap 0x00 `` Reserved by hardware.
`` debugtrap(arg) `` `` s_trap 0x01 `` `` SGPR0-1 `` : Reserved for Finalizer HSA `` debugtrap ``
`` queue_ptr `` intrinsic (not implemented).
`` VGPR0 `` :
`` arg ``
`` llvm.trap `` `` s_trap 0x02 `` `` SGPR0-1 `` : Causes wave to be halted with the PC at
`` queue_ptr `` the trap instruction. The associated
queue is signalled to put it into the
error state. When the queue is put in
the error state, the waves executing
dispatches on the queue will be
terminated.
`` llvm.debugtrap `` `` s_trap 0x03 `` *none* - If debugger not enabled then behaves
as a no-operation. The trap handler
is entered and immediately returns to
continue execution of the wavefront.
- If the debugger is enabled, causes
the debug trap to be reported by the
debugger and the wavefront is put in
the halt state with the PC at the
instruction. The debugger must
increment the PC and resume the wave.
reserved `` s_trap 0x04 `` Reserved.
reserved `` s_trap 0x05 `` Reserved.
reserved `` s_trap 0x06 `` Reserved.
reserved `` s_trap 0x07 `` Reserved.
reserved `` s_trap 0x08 `` Reserved.
reserved `` s_trap 0xfe `` Reserved.
reserved `` s_trap 0xff `` Reserved.
=================== =============== =============== =======================================
..
.. table :: AMDGPU Trap Handler for AMDHSA OS Code Object V3
:name: amdgpu-trap-handler-for-amdhsa-os-v3-table
2017-06-07 04:31:59 +08:00
2020-12-01 12:12:04 +08:00
=================== =============== =============== =======================================
2017-06-07 04:31:59 +08:00
Usage Code Sequence Trap Handler Description
Inputs
2020-12-01 12:12:04 +08:00
=================== =============== =============== =======================================
2017-06-07 04:31:59 +08:00
reserved `` s_trap 0x00 `` Reserved by hardware.
2020-12-01 12:12:04 +08:00
debugger breakpoint `` s_trap 0x01 `` *none* Reserved for debugger to use for
breakpoints. Causes wave to be halted
with the PC at the trap instruction.
The debugger is responsible to resume
the wave, including the instruction
that the breakpoint overwrote.
`` llvm.trap `` `` s_trap 0x02 `` `` SGPR0-1 `` : Causes wave to be halted with the PC at
`` queue_ptr `` the trap instruction. The associated
queue is signalled to put it into the
error state. When the queue is put in
the error state, the waves executing
dispatches on the queue will be
terminated.
`` llvm.debugtrap `` `` s_trap 0x03 `` *none* - If debugger not enabled then behaves
as a no-operation. The trap handler
is entered and immediately returns to
continue execution of the wavefront.
- If the debugger is enabled, causes
the debug trap to be reported by the
debugger and the wavefront is put in
the halt state with the PC at the
instruction. The debugger must
increment the PC and resume the wave.
2018-05-17 00:19:34 +08:00
reserved `` s_trap 0x04 `` Reserved.
reserved `` s_trap 0x05 `` Reserved.
reserved `` s_trap 0x06 `` Reserved.
2020-12-01 12:12:04 +08:00
reserved `` s_trap 0x07 `` Reserved.
2018-05-17 00:19:34 +08:00
reserved `` s_trap 0x08 `` Reserved.
reserved `` s_trap 0xfe `` Reserved.
reserved `` s_trap 0xff `` Reserved.
2020-12-01 12:12:04 +08:00
=================== =============== =============== =======================================
..
.. table :: AMDGPU Trap Handler for AMDHSA OS Code Object V4
:name: amdgpu-trap-handler-for-amdhsa-os-v4-table
2021-01-27 01:45:41 +08:00
=================== =============== ================ ================= =======================================
Usage Code Sequence GFX6-GFX8 Inputs GFX9-GFX10 Inputs Description
=================== =============== ================ ================= =======================================
reserved `` s_trap 0x00 `` Reserved by hardware.
debugger breakpoint `` s_trap 0x01 `` *none* *none* Reserved for debugger to use for
breakpoints. Causes wave to be halted
with the PC at the trap instruction.
The debugger is responsible to resume
the wave, including the instruction
that the breakpoint overwrote.
`` llvm.trap `` `` s_trap 0x02 `` `` SGPR0-1 `` : *none* Causes wave to be halted with the PC at
`` queue_ptr `` the trap instruction. The associated
queue is signalled to put it into the
error state. When the queue is put in
the error state, the waves executing
dispatches on the queue will be
terminated.
`` llvm.debugtrap `` `` s_trap 0x03 `` *none* *none* - If debugger not enabled then behaves
as a no-operation. The trap handler
is entered and immediately returns to
continue execution of the wavefront.
- If the debugger is enabled, causes
the debug trap to be reported by the
debugger and the wavefront is put in
the halt state with the PC at the
instruction. The debugger must
increment the PC and resume the wave.
reserved `` s_trap 0x04 `` Reserved.
reserved `` s_trap 0x05 `` Reserved.
reserved `` s_trap 0x06 `` Reserved.
reserved `` s_trap 0x07 `` Reserved.
reserved `` s_trap 0x08 `` Reserved.
reserved `` s_trap 0xfe `` Reserved.
reserved `` s_trap 0xff `` Reserved.
=================== =============== ================ ================= =======================================
2017-06-07 04:31:59 +08:00
[AMDGPU] AMDGPUUsage define call convention ABI
Reviewers: scott.linder, arsenm, b-sumner
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, dstuttard, tpr, kerbowa, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D74861
2020-02-13 14:19:25 +08:00
.. _amdgpu-amdhsa-function-call-convention:
Call Convention
~~~~~~~~~~~~~~~
.. note ::
2020-08-13 16:40:24 +08:00
This section is currently incomplete and has inaccuracies. It is WIP that will
[AMDGPU] AMDGPUUsage define call convention ABI
Reviewers: scott.linder, arsenm, b-sumner
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, dstuttard, tpr, kerbowa, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D74861
2020-02-13 14:19:25 +08:00
be updated as information is determined.
2020-02-20 15:08:59 +08:00
See :ref: `amdgpu-dwarf-address-space-identifier` for information on swizzled
[AMDGPU] AMDGPUUsage define call convention ABI
Reviewers: scott.linder, arsenm, b-sumner
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, dstuttard, tpr, kerbowa, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D74861
2020-02-13 14:19:25 +08:00
addresses. Unswizzled addresses are normal linear addresses.
2020-01-22 06:27:57 +08:00
.. _amdgpu-amdhsa-function-call-convention-kernel-functions:
[AMDGPU] AMDGPUUsage define call convention ABI
Reviewers: scott.linder, arsenm, b-sumner
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, dstuttard, tpr, kerbowa, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D74861
2020-02-13 14:19:25 +08:00
Kernel Functions
++++++++++++++++
This section describes the call convention ABI for the outer kernel function.
See :ref: `amdgpu-amdhsa-initial-kernel-execution-state` for the kernel call
convention.
The following is not part of the AMDGPU kernel calling convention but describes
how the AMDGPU implements function calls:
1. Clang decides the kernarg layout to match the *HSA Programmer's Language
Reference* [HSA]_ .
- All structs are passed directly.
- Lambda values are passed *TBA* .
.. TODO ::
- Does this really follow HSA rules? Or are structs >16 bytes passed
by-value struct?
- What is ABI for lambda values?
2020-01-22 06:27:57 +08:00
4. The kernel performs certain setup in its prolog, as described in
:ref: `amdgpu-amdhsa-kernel-prolog` .
[AMDGPU] AMDGPUUsage define call convention ABI
Reviewers: scott.linder, arsenm, b-sumner
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, dstuttard, tpr, kerbowa, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D74861
2020-02-13 14:19:25 +08:00
2020-01-22 06:27:57 +08:00
.. _amdgpu-amdhsa-function-call-convention-non-kernel-functions:
[AMDGPU] AMDGPUUsage define call convention ABI
Reviewers: scott.linder, arsenm, b-sumner
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, dstuttard, tpr, kerbowa, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D74861
2020-02-13 14:19:25 +08:00
Non-Kernel Functions
++++++++++++++++++++
This section describes the call convention ABI for functions other than the
outer kernel function.
2020-01-22 06:27:57 +08:00
If a kernel has function calls then scratch is always allocated and used for
the call stack which grows from low address to high address using the swizzled
[AMDGPU] AMDGPUUsage define call convention ABI
Reviewers: scott.linder, arsenm, b-sumner
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, dstuttard, tpr, kerbowa, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D74861
2020-02-13 14:19:25 +08:00
scratch address space.
On entry to a function:
2020-01-22 06:27:57 +08:00
1. SGPR0-3 contain a V# with the following properties (see
:ref: `amdgpu-amdhsa-kernel-prolog-private-segment-buffer` ):
[AMDGPU] AMDGPUUsage define call convention ABI
Reviewers: scott.linder, arsenm, b-sumner
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, dstuttard, tpr, kerbowa, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D74861
2020-02-13 14:19:25 +08:00
2020-01-22 06:27:57 +08:00
* Base address pointing to the beginning of the wavefront scratch backing
memory.
[AMDGPU] AMDGPUUsage define call convention ABI
Reviewers: scott.linder, arsenm, b-sumner
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, dstuttard, tpr, kerbowa, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D74861
2020-02-13 14:19:25 +08:00
* Swizzled with dword element size and stride of wavefront size elements.
2. The FLAT_SCRATCH register pair is setup. See
2020-01-22 06:27:57 +08:00
:ref: `amdgpu-amdhsa-kernel-prolog-flat-scratch` .
2021-01-27 01:45:41 +08:00
3. GFX6-GFX8: M0 register set to the size of LDS in bytes. See
2020-01-22 06:27:57 +08:00
:ref: `amdgpu-amdhsa-kernel-prolog-m0` .
[AMDGPU] AMDGPUUsage define call convention ABI
Reviewers: scott.linder, arsenm, b-sumner
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, dstuttard, tpr, kerbowa, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D74861
2020-02-13 14:19:25 +08:00
4. The EXEC register is set to the lanes active on entry to the function.
5. MODE register: *TBD*
6. VGPR0-31 and SGPR4-29 are used to pass function input arguments as described
below.
7. SGPR30-31 return address (RA). The code address that the function must
return to when it completes. The value is undefined if the function is *no
return*.
2020-01-22 06:27:57 +08:00
8. SGPR32 is used for the stack pointer (SP). It is an unswizzled scratch
offset relative to the beginning of the wavefront scratch backing memory.
[AMDGPU] AMDGPUUsage define call convention ABI
Reviewers: scott.linder, arsenm, b-sumner
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, dstuttard, tpr, kerbowa, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D74861
2020-02-13 14:19:25 +08:00
The unswizzled SP can be used with buffer instructions as an unswizzled SGPR
offset with the scratch V# in SGPR0-3 to access the stack in a swizzled
manner.
2020-01-22 06:27:57 +08:00
The unswizzled SP value can be converted into the swizzled SP value by:
| swizzled SP = unswizzled SP / wavefront size
This may be used to obtain the private address space address of stack
objects and to convert this address to a flat address by adding the flat
scratch aperture base address.
[AMDGPU] AMDGPUUsage define call convention ABI
Reviewers: scott.linder, arsenm, b-sumner
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, dstuttard, tpr, kerbowa, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D74861
2020-02-13 14:19:25 +08:00
The swizzled SP value is always 4 bytes aligned for the `` r600 ``
architecture and 16 byte aligned for the `` amdgcn `` architecture.
.. note ::
The `` amdgcn `` value is selected to avoid dynamic stack alignment for the
OpenCL language which has the largest base type defined as 16 bytes.
On entry, the swizzled SP value is the address of the first function
argument passed on the stack. Other stack passed arguments are positive
offsets from the entry swizzled SP value.
The function may use positive offsets beyond the last stack passed argument
2020-02-20 15:08:59 +08:00
for stack allocated local variables and register spill slots. If necessary,
[AMDGPU] AMDGPUUsage define call convention ABI
Reviewers: scott.linder, arsenm, b-sumner
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, dstuttard, tpr, kerbowa, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D74861
2020-02-13 14:19:25 +08:00
the function may align these to greater alignment than 16 bytes. After these
the function may dynamically allocate space for such things as runtime sized
`` alloca `` local allocations.
If the function calls another function, it will place any stack allocated
arguments after the last local allocation and adjust SGPR32 to the address
after the last local allocation.
2020-01-22 06:27:57 +08:00
9. All other registers are unspecified.
2020-10-31 12:43:55 +08:00
10. Any necessary `` s_waitcnt `` has been performed to ensure memory is available
[AMDGPU] AMDGPUUsage define call convention ABI
Reviewers: scott.linder, arsenm, b-sumner
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, dstuttard, tpr, kerbowa, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D74861
2020-02-13 14:19:25 +08:00
to the function.
On exit from a function:
1. VGPR0-31 and SGPR4-29 are used to pass function result arguments as
2020-01-22 06:27:57 +08:00
described below. Any registers used are considered clobbered registers.
[AMDGPU] AMDGPUUsage define call convention ABI
Reviewers: scott.linder, arsenm, b-sumner
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, dstuttard, tpr, kerbowa, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D74861
2020-02-13 14:19:25 +08:00
2. The following registers are preserved and have the same value as on entry:
* FLAT_SCRATCH
* EXEC
2021-01-27 01:45:41 +08:00
* GFX6-GFX8: M0
2020-03-27 15:46:51 +08:00
* All SGPR registers except the clobbered registers of SGPR4-31.
* VGPR40-47
2021-01-28 21:53:22 +08:00
* VGPR56-63
* VGPR72-79
* VGPR88-95
* VGPR104-111
* VGPR120-127
* VGPR136-143
* VGPR152-159
* VGPR168-175
* VGPR184-191
* VGPR200-207
* VGPR216-223
* VGPR232-239
* VGPR248-255
.. note ::
Except the argument registers, the VGPRs clobbered and the preserved
registers are intermixed at regular intervals in order to keep a
similar ratio independent of the number of allocated VGPRs.
* Lanes of all VGPRs that are inactive at the call site.
[AMDGPU] AMDGPUUsage define call convention ABI
Reviewers: scott.linder, arsenm, b-sumner
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, dstuttard, tpr, kerbowa, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D74861
2020-02-13 14:19:25 +08:00
For the AMDGPU backend, an inter-procedural register allocation (IPRA)
2020-03-27 15:46:51 +08:00
optimization may mark some of clobbered SGPR and VGPR registers as
[AMDGPU] AMDGPUUsage define call convention ABI
Reviewers: scott.linder, arsenm, b-sumner
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, dstuttard, tpr, kerbowa, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D74861
2020-02-13 14:19:25 +08:00
preserved if it can be determined that the called function does not change
their value.
2. The PC is set to the RA provided on entry.
3. MODE register: *TBD* .
4. All other registers are clobbered.
2020-10-31 12:43:55 +08:00
5. Any necessary `` s_waitcnt `` has been performed to ensure memory accessed by
[AMDGPU] AMDGPUUsage define call convention ABI
Reviewers: scott.linder, arsenm, b-sumner
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, dstuttard, tpr, kerbowa, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D74861
2020-02-13 14:19:25 +08:00
function is available to the caller.
.. TODO ::
- On gfx908 are all ACC registers clobbered?
- How are function results returned? The address of structured types is passed
by reference, but what about other types?
The function input arguments are made up of the formal arguments explicitly
declared by the source language function plus the implicit input arguments used
by the implementation.
The source language input arguments are:
1. Any source language implicit `` this `` or `` self `` argument comes first as a
pointer type.
2. Followed by the function formal arguments in left to right source order.
The source language result arguments are:
1. The function result argument.
The source language input or result struct type arguments that are less than or
equal to 16 bytes, are decomposed recursively into their base type fields, and
each field is passed as if a separate argument. For input arguments, if the
called function requires the struct to be in memory, for example because its
address is taken, then the function body is responsible for allocating a stack
location and copying the field arguments into it. Clang terms this *direct
struct*.
The source language input struct type arguments that are greater than 16 bytes,
are passed by reference. The caller is responsible for allocating a stack
location to make a copy of the struct value and pass the address as the input
argument. The called function is responsible to perform the dereference when
accessing the input argument. Clang terms this *by-value struct* .
A source language result struct type argument that is greater than 16 bytes, is
returned by reference. The caller is responsible for allocating a stack location
to hold the result value and passes the address as the last input argument
(before the implicit input arguments). In this case there are no result
arguments. The called function is responsible to perform the dereference when
storing the result value. Clang terms this *structured return (sret)* .
2020-02-20 15:08:59 +08:00
*TODO: correct the ``sret`` definition.*
[AMDGPU] AMDGPUUsage define call convention ABI
Reviewers: scott.linder, arsenm, b-sumner
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, dstuttard, tpr, kerbowa, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D74861
2020-02-13 14:19:25 +08:00
.. TODO ::
2020-02-20 15:08:59 +08:00
Is this definition correct? Or is `` sret `` only used if passing in registers, and
[AMDGPU] AMDGPUUsage define call convention ABI
Reviewers: scott.linder, arsenm, b-sumner
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, dstuttard, tpr, kerbowa, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D74861
2020-02-13 14:19:25 +08:00
pass as non-decomposed struct as stack argument? Or something else? Is the
memory location in the caller stack frame, or a stack memory argument and so
no address is passed as the caller can directly write to the argument stack
2020-02-20 15:08:59 +08:00
location? But then the stack location is still live after return. If an
[AMDGPU] AMDGPUUsage define call convention ABI
Reviewers: scott.linder, arsenm, b-sumner
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, dstuttard, tpr, kerbowa, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D74861
2020-02-13 14:19:25 +08:00
argument stack location is it the first stack argument or the last one?
Lambda argument types are treated as struct types with an implementation defined
set of fields.
.. TODO ::
Need to specify the ABI for lambda types for AMDGPU.
For AMDGPU backend all source language arguments (including the decomposed
struct type arguments) are passed in VGPRs unless marked `` inreg `` in which case
they are passed in SGPRs.
The AMDGPU backend walks the function call graph from the leaves to determine
which implicit input arguments are used, propagating to each caller of the
function. The used implicit arguments are appended to the function arguments
after the source language arguments in the following order:
.. TODO ::
Is recursion or external functions supported?
1. Work-Item ID (1 VGPR)
The X, Y and Z work-item ID are packed into a single VGRP with the following
layout. Only fields actually used by the function are set. The other bits
are undefined.
The values come from the initial kernel execution state. See
2020-12-01 12:12:04 +08:00
:ref: `amdgpu-amdhsa-initial-kernel-execution-state` .
[AMDGPU] AMDGPUUsage define call convention ABI
Reviewers: scott.linder, arsenm, b-sumner
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, dstuttard, tpr, kerbowa, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D74861
2020-02-13 14:19:25 +08:00
2020-04-23 13:26:07 +08:00
.. table :: Work-item implicit argument layout
:name: amdgpu-amdhsa-workitem-implicit-argument-layout-table
[AMDGPU] AMDGPUUsage define call convention ABI
Reviewers: scott.linder, arsenm, b-sumner
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, dstuttard, tpr, kerbowa, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D74861
2020-02-13 14:19:25 +08:00
======= ======= ==============
Bits Size Field Name
======= ======= ==============
9:0 10 bits X Work-Item ID
19:10 10 bits Y Work-Item ID
29:20 10 bits Z Work-Item ID
31:30 2 bits Unused
======= ======= ==============
2. Dispatch Ptr (2 SGPRs)
The value comes from the initial kernel execution state. See
:ref: `amdgpu-amdhsa-sgpr-register-set-up-order-table` .
3. Queue Ptr (2 SGPRs)
The value comes from the initial kernel execution state. See
:ref: `amdgpu-amdhsa-sgpr-register-set-up-order-table` .
4. Kernarg Segment Ptr (2 SGPRs)
The value comes from the initial kernel execution state. See
:ref: `amdgpu-amdhsa-sgpr-register-set-up-order-table` .
5. Dispatch id (2 SGPRs)
The value comes from the initial kernel execution state. See
:ref: `amdgpu-amdhsa-sgpr-register-set-up-order-table` .
6. Work-Group ID X (1 SGPR)
The value comes from the initial kernel execution state. See
:ref: `amdgpu-amdhsa-sgpr-register-set-up-order-table` .
7. Work-Group ID Y (1 SGPR)
The value comes from the initial kernel execution state. See
:ref: `amdgpu-amdhsa-sgpr-register-set-up-order-table` .
8. Work-Group ID Z (1 SGPR)
The value comes from the initial kernel execution state. See
:ref: `amdgpu-amdhsa-sgpr-register-set-up-order-table` .
9. Implicit Argument Ptr (2 SGPRs)
The value is computed by adding an offset to Kernarg Segment Ptr to get the
global address space pointer to the first kernarg implicit argument.
The input and result arguments are assigned in order in the following manner:
2020-05-22 05:20:06 +08:00
.. note ::
[AMDGPU] AMDGPUUsage define call convention ABI
Reviewers: scott.linder, arsenm, b-sumner
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, dstuttard, tpr, kerbowa, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D74861
2020-02-13 14:19:25 +08:00
2020-02-20 15:08:59 +08:00
There are likely some errors and omissions in the following description that
[AMDGPU] AMDGPUUsage define call convention ABI
Reviewers: scott.linder, arsenm, b-sumner
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, dstuttard, tpr, kerbowa, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D74861
2020-02-13 14:19:25 +08:00
need correction.
2020-05-22 05:20:06 +08:00
.. TODO ::
[AMDGPU] AMDGPUUsage define call convention ABI
Reviewers: scott.linder, arsenm, b-sumner
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, dstuttard, tpr, kerbowa, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D74861
2020-02-13 14:19:25 +08:00
2020-12-01 12:12:04 +08:00
Check the Clang source code to decipher how function arguments and return
[AMDGPU] AMDGPUUsage define call convention ABI
Reviewers: scott.linder, arsenm, b-sumner
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, dstuttard, tpr, kerbowa, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D74861
2020-02-13 14:19:25 +08:00
results are handled. Also see the AMDGPU specific values used.
* VGPR arguments are assigned to consecutive VGPRs starting at VGPR0 up to
VGPR31.
If there are more arguments than will fit in these registers, the remaining
arguments are allocated on the stack in order on naturally aligned
addresses.
.. TODO ::
How are overly aligned structures allocated on the stack?
* SGPR arguments are assigned to consecutive SGPRs starting at SGPR0 up to
SGPR29.
If there are more arguments than will fit in these registers, the remaining
arguments are allocated on the stack in order on naturally aligned
addresses.
Note that decomposed struct type arguments may have some fields passed in
registers and some in memory.
2020-02-20 15:08:59 +08:00
.. TODO ::
[AMDGPU] AMDGPUUsage define call convention ABI
Reviewers: scott.linder, arsenm, b-sumner
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, dstuttard, tpr, kerbowa, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D74861
2020-02-13 14:19:25 +08:00
2020-02-20 15:08:59 +08:00
So, a struct which can pass some fields as decomposed register arguments, will
pass the rest as decomposed stack elements? But an argument that will not start
[AMDGPU] AMDGPUUsage define call convention ABI
Reviewers: scott.linder, arsenm, b-sumner
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, dstuttard, tpr, kerbowa, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D74861
2020-02-13 14:19:25 +08:00
in registers will not be decomposed and will be passed as a non-decomposed
stack value?
The following is not part of the AMDGPU function calling convention but
describes how the AMDGPU implements function calls:
2020-03-04 09:39:47 +08:00
1. SGPR33 is used as a frame pointer (FP) if necessary. Like the SP it is an
[AMDGPU] AMDGPUUsage define call convention ABI
Reviewers: scott.linder, arsenm, b-sumner
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, dstuttard, tpr, kerbowa, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D74861
2020-02-13 14:19:25 +08:00
unswizzled scratch address. It is only needed if runtime sized `` alloca ``
2020-01-22 06:27:57 +08:00
are used, or for the reasons defined in `` SIFrameLowering `` .
2020-04-21 17:34:33 +08:00
2. Runtime stack alignment is supported. SGPR34 is used as a base pointer (BP)
to access the incoming stack arguments in the function. The BP is needed
only when the function requires the runtime stack alignment.
[AMDGPU] AMDGPUUsage define call convention ABI
Reviewers: scott.linder, arsenm, b-sumner
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, dstuttard, tpr, kerbowa, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D74861
2020-02-13 14:19:25 +08:00
2020-04-21 17:34:33 +08:00
3. Allocating SGPR arguments on the stack are not supported.
[AMDGPU] AMDGPUUsage define call convention ABI
Reviewers: scott.linder, arsenm, b-sumner
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, dstuttard, tpr, kerbowa, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D74861
2020-02-13 14:19:25 +08:00
2020-04-21 17:34:33 +08:00
4. No CFI is currently generated. See
2020-02-20 15:08:59 +08:00
:ref: `amdgpu-dwarf-call-frame-information` .
[AMDGPU] AMDGPUUsage define call convention ABI
Reviewers: scott.linder, arsenm, b-sumner
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, dstuttard, tpr, kerbowa, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D74861
2020-02-13 14:19:25 +08:00
2020-05-22 05:20:06 +08:00
.. note ::
[AMDGPU] AMDGPUUsage define call convention ABI
Reviewers: scott.linder, arsenm, b-sumner
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, dstuttard, tpr, kerbowa, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D74861
2020-02-13 14:19:25 +08:00
CFI will be generated that defines the CFA as the unswizzled address
relative to the wave scratch base in the unswizzled private address space
of the lowest address stack allocated local variable.
2020-01-22 06:27:57 +08:00
`` DW_AT_frame_base `` will be defined as the swizzled address in the
[AMDGPU] AMDGPUUsage define call convention ABI
Reviewers: scott.linder, arsenm, b-sumner
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, dstuttard, tpr, kerbowa, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D74861
2020-02-13 14:19:25 +08:00
swizzled private address space by dividing the CFA by the wavefront size
(since CFA is always at least dword aligned which matches the scratch
swizzle element size).
If no dynamic stack alignment was performed, the stack allocated arguments
are accessed as negative offsets relative to `` DW_AT_frame_base `` , and the
local variables and register spill slots are accessed as positive offsets
relative to `` DW_AT_frame_base `` .
2020-04-21 17:34:33 +08:00
5. Function argument passing is implemented by copying the input physical
[AMDGPU] AMDGPUUsage define call convention ABI
Reviewers: scott.linder, arsenm, b-sumner
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, dstuttard, tpr, kerbowa, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D74861
2020-02-13 14:19:25 +08:00
registers to virtual registers on entry. The register allocator can spill if
necessary. These are copied back to physical registers at call sites. The
net effect is that each function call can have these values in entirely
distinct locations. The IPRA can help avoid shuffling argument registers.
2020-04-21 17:34:33 +08:00
6. Call sites are implemented by setting up the arguments at positive offsets
[AMDGPU] AMDGPUUsage define call convention ABI
Reviewers: scott.linder, arsenm, b-sumner
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, dstuttard, tpr, kerbowa, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D74861
2020-02-13 14:19:25 +08:00
from SP. Then SP is incremented to account for the known frame size before
the call and decremented after the call.
2020-05-22 05:20:06 +08:00
.. note ::
[AMDGPU] AMDGPUUsage define call convention ABI
Reviewers: scott.linder, arsenm, b-sumner
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, dstuttard, tpr, kerbowa, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D74861
2020-02-13 14:19:25 +08:00
The CFI will reflect the changed calculation needed to compute the CFA
from SP.
2020-04-21 17:34:33 +08:00
7. 4 byte spill slots are used in the stack frame. One slot is allocated for an
[AMDGPU] AMDGPUUsage define call convention ABI
Reviewers: scott.linder, arsenm, b-sumner
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, dstuttard, tpr, kerbowa, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D74861
2020-02-13 14:19:25 +08:00
emergency spill slot. Buffer instructions are used for stack accesses and
not the `` flat_scratch `` instruction.
2020-05-22 05:20:06 +08:00
.. TODO ::
[AMDGPU] AMDGPUUsage define call convention ABI
Reviewers: scott.linder, arsenm, b-sumner
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, dstuttard, tpr, kerbowa, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D74861
2020-02-13 14:19:25 +08:00
Explain when the emergency spill slot is used.
.. TODO ::
Possible broken issues:
- Stack arguments must be aligned to required alignment.
- Stack is aligned to max(16, max formal argument alignment)
- Direct argument < 64 bits should check register budget.
- Register budget calculation should respect `` inreg `` for SGPR.
- SGPR overflow is not handled.
- struct with 1 member unpeeling is not checking size of member.
- `` sret `` is after `` this `` pointer.
- Caller is not implementing stack realignment: need an extra pointer.
- Should say AMDGPU passes FP rather than SP.
- Should CFI define CFA as address of locals or arguments. Difference is
apparent when have implemented dynamic alignment.
2020-02-20 15:08:59 +08:00
- If `` SCRATCH `` instruction could allow negative offsets, then can make FP be
[AMDGPU] AMDGPUUsage define call convention ABI
Reviewers: scott.linder, arsenm, b-sumner
Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, dstuttard, tpr, kerbowa, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D74861
2020-02-13 14:19:25 +08:00
highest address of stack frame and use negative offset for locals. Would
allow SP to be the same as FP and could support signal-handler-like as now
have a real SP for the top of the stack.
- How is `` sret `` passed on the stack? In argument stack area? Can it overlay
arguments?
2018-04-04 21:02:09 +08:00
AMDPAL
------
This section provides code conventions used when the target triple OS is
2021-03-30 15:33:07 +08:00
`` amdpal `` (see :ref: `amdgpu-target-triples` ).
2018-04-04 21:02:09 +08:00
2021-03-30 15:33:07 +08:00
.. _amdgpu-amdpal-code-object-metadata-section:
2018-04-04 21:02:09 +08:00
2021-03-30 15:33:07 +08:00
Code Object Metadata
~~~~~~~~~~~~~~~~~~~~
2018-04-04 21:02:09 +08:00
2021-03-30 15:33:07 +08:00
.. note ::
2018-04-04 21:02:09 +08:00
2021-03-30 15:33:07 +08:00
The metadata is currently in development and is subject to major
changes. Only the current version is supported. *When this document
was generated the version was 2.6.*
2018-04-04 21:02:09 +08:00
2021-03-30 15:33:07 +08:00
Code object metadata is specified by the `` NT_AMDGPU_METADATA `` note
record (see :ref: `amdgpu-note-records-v3-v4` ).
2018-04-04 21:02:09 +08:00
2021-03-30 15:33:07 +08:00
The metadata is represented as Message Pack formatted binary data (see
[MsgPack]_ ). The top level is a Message Pack map that includes the keys
defined in table :ref: `amdgpu-amdpal-code-object-metadata-map-table`
and referenced tables.
2018-04-04 21:02:09 +08:00
2021-03-30 15:33:07 +08:00
Additional information can be added to the maps. To avoid conflicts, any
key names should be prefixed by "*vendor-name* ." where `` vendor-name ``
can be the name of the vendor and specific vendor tool that generates the
information. The prefix is abbreviated to simply "." when it appears
within a map that has been added by the same *vendor-name* .
2018-04-04 21:02:09 +08:00
2021-03-30 15:33:07 +08:00
.. table :: AMDPAL Code Object Metadata Map
:name: amdgpu-amdpal-code-object-metadata-map-table
2018-04-04 21:02:09 +08:00
2021-03-30 15:33:07 +08:00
=================== ============== ========= ======================================================================
String Key Value Type Required? Description
=================== ============== ========= ======================================================================
"amdpal.version" sequence of Required PAL code object metadata (major, minor) version. The current values
2 integers are defined by *Util::Abi::PipelineMetadata(Major|Minor)Version* .
"amdpal.pipelines" sequence of Required Per-pipeline metadata. See
map :ref: `amdgpu-amdpal-code-object-pipeline-metadata-map-table` for the
definition of the keys included in that map.
=================== ============== ========= ======================================================================
2020-09-16 20:38:54 +08:00
2021-03-30 15:33:07 +08:00
..
.. table :: AMDPAL Code Object Pipeline Metadata Map
:name: amdgpu-amdpal-code-object-pipeline-metadata-map-table
====================================== ============== ========= ===================================================
String Key Value Type Required? Description
====================================== ============== ========= ===================================================
".name" string Source name of the pipeline.
".type" string Pipeline type, e.g. VsPs. Values include:
- "VsPs"
- "Gs"
- "Cs"
- "Ngg"
- "Tess"
- "GsTess"
- "NggTess"
".internal_pipeline_hash" sequence of Required Internal compiler hash for this pipeline. Lower
2 integers 64 bits is the "stable" portion of the hash, used
for e.g. shader replacement lookup. Upper 64 bits
is the "unique" portion of the hash, used for
e.g. pipeline cache lookup. The value is
implementation defined, and can not be relied on
between different builds of the compiler.
".shaders" map Per-API shader metadata. See
:ref: `amdgpu-amdpal-code-object-shader-map-table`
for the definition of the keys included in that
map.
".hardware_stages" map Per-hardware stage metadata. See
:ref: `amdgpu-amdpal-code-object-hardware-stage-map-table`
for the definition of the keys included in that
map.
".shader_functions" map Per-shader function metadata. See
:ref: `amdgpu-amdpal-code-object-shader-function-map-table`
for the definition of the keys included in that
map.
".registers" map Required Hardware register configuration. See
:ref: `amdgpu-amdpal-code-object-register-map-table`
for the definition of the keys included in that
map.
".user_data_limit" integer Number of user data entries accessed by this
pipeline.
".spill_threshold" integer The user data spill threshold. 0xFFFF for
NoUserDataSpilling.
".uses_viewport_array_index" boolean Indicates whether or not the pipeline uses the
viewport array index feature. Pipelines which use
this feature can render into all 16 viewports,
whereas pipelines which do not use it are
restricted to viewport #0.
".es_gs_lds_size" integer Size in bytes of LDS space used internally for
handling data-passing between the ES and GS
shader stages. This can be zero if the data is
passed using off-chip buffers. This value should
be used to program all user-SGPRs which have been
marked with "UserDataMapping::EsGsLdsSize"
(typically only the GS and VS HW stages will ever
have a user-SGPR so marked).
".nggSubgroupSize" integer Explicit maximum subgroup size for NGG shaders
(maximum number of threads in a subgroup).
".num_interpolants" integer Graphics only. Number of PS interpolants.
".mesh_scratch_memory_size" integer Max mesh shader scratch memory used.
".api" string Name of the client graphics API.
".api_create_info" binary Graphics API shader create info binary blob. Can
be defined by the driver using the compiler if
they want to be able to correlate API-specific
information used during creation at a later time.
====================================== ============== ========= ===================================================
..
.. table :: AMDPAL Code Object Shader Map
:name: amdgpu-amdpal-code-object-shader-map-table
+-------------+--------------+-------------------------------------------------------------------+
|String Key |Value Type |Description |
+=============+==============+===================================================================+
|- ".compute" |map |See :ref: `amdgpu-amdpal-code-object-api-shader-metadata-map-table` |
|- ".vertex" | |for the definition of the keys included in that map. |
|- ".hull" | | |
|- ".domain" | | |
|- ".geometry"| | |
|- ".pixel" | | |
+-------------+--------------+-------------------------------------------------------------------+
..
.. table :: AMDPAL Code Object API Shader Metadata Map
:name: amdgpu-amdpal-code-object-api-shader-metadata-map-table
==================== ============== ========= =====================================================================
String Key Value Type Required? Description
==================== ============== ========= =====================================================================
".api_shader_hash" sequence of Required Input shader hash, typically passed in from the client. The value
2 integers is implementation defined, and can not be relied on between
different builds of the compiler.
".hardware_mapping" sequence of Required Flags indicating the HW stages this API shader maps to. Values
string include:
- ".ls"
- ".hs"
- ".es"
- ".gs"
- ".vs"
- ".ps"
- ".cs"
2020-09-16 20:38:54 +08:00
2021-03-30 15:33:07 +08:00
==================== ============== ========= =====================================================================
..
.. table :: AMDPAL Code Object Hardware Stage Map
:name: amdgpu-amdpal-code-object-hardware-stage-map-table
+-------------+--------------+-----------------------------------------------------------------------+
|String Key |Value Type |Description |
+=============+==============+=======================================================================+
|- ".ls" |map |See :ref: `amdgpu-amdpal-code-object-hardware-stage-metadata-map-table` |
|- ".hs" | |for the definition of the keys included in that map. |
|- ".es" | | |
|- ".gs" | | |
|- ".vs" | | |
|- ".ps" | | |
|- ".cs" | | |
+-------------+--------------+-----------------------------------------------------------------------+
..
.. table :: AMDPAL Code Object Hardware Stage Metadata Map
:name: amdgpu-amdpal-code-object-hardware-stage-metadata-map-table
========================== ============== ========= ===============================================================
String Key Value Type Required? Description
========================== ============== ========= ===============================================================
".entry_point" string The ELF symbol pointing to this pipeline's stage entry point.
".scratch_memory_size" integer Scratch memory size in bytes.
".lds_size" integer Local Data Share size in bytes.
".perf_data_buffer_size" integer Performance data buffer size in bytes.
".vgpr_count" integer Number of VGPRs used.
".sgpr_count" integer Number of SGPRs used.
".vgpr_limit" integer If non-zero, indicates the shader was compiled with a
directive to instruct the compiler to limit the VGPR usage to
be less than or equal to the specified value (only set if
different from HW default).
".sgpr_limit" integer SGPR count upper limit (only set if different from HW
default).
".threadgroup_dimensions" sequence of Thread-group X/Y/Z dimensions (Compute only).
3 integers
".wavefront_size" integer Wavefront size (only set if different from HW default).
".uses_uavs" boolean The shader reads or writes UAVs.
".uses_rovs" boolean The shader reads or writes ROVs.
".writes_uavs" boolean The shader writes to one or more UAVs.
".writes_depth" boolean The shader writes out a depth value.
".uses_append_consume" boolean The shader uses append and/or consume operations, either
memory or GDS.
".uses_prim_id" boolean The shader uses PrimID.
========================== ============== ========= ===============================================================
..
.. table :: AMDPAL Code Object Shader Function Map
:name: amdgpu-amdpal-code-object-shader-function-map-table
=============== ============== ====================================================================
String Key Value Type Description
=============== ============== ====================================================================
*symbol name* map *symbol name* is the ELF symbol name of the shader function code
entry address. The value is the function's metadata. See
:ref: `amdgpu-amdpal-code-object-shader-function-metadata-map-table` .
=============== ============== ====================================================================
..
.. table :: AMDPAL Code Object Shader Function Metadata Map
:name: amdgpu-amdpal-code-object-shader-function-metadata-map-table
============================= ============== =================================================================
String Key Value Type Description
============================= ============== =================================================================
".api_shader_hash" sequence of Input shader hash, typically passed in from the client. The value
2 integers is implementation defined, and can not be relied on between
different builds of the compiler.
".scratch_memory_size" sequence of Size in bytes of scratch memory used by the shader.
2 integers
".lds_size" sequence of Size in bytes of LDS memory.
2 integers
".vgpr_count" integer Number of VGPRs used by the shader.
".sgpr_count" integer Number of SGPRs used by the shader.
".stack_frame_size_in_bytes" integer Amount of stack size used by the shader.
".shader_subtype" string Shader subtype/kind. Values include:
- "Unknown"
============================= ============== =================================================================
..
.. table :: AMDPAL Code Object Register Map
:name: amdgpu-amdpal-code-object-register-map-table
========================== ============== ====================================================================
32-bit Integer Key Value Type Description
========================== ============== ====================================================================
`` reg offset `` 32-bit integer `` reg offset `` is the dword offset into the GFXIP register space of
a GRBM register (i.e., driver accessible GPU register number, not
shader GPR register number). The driver is required to program each
specified register to the corresponding specified value when
executing this pipeline. Typically, the `` reg offsets `` are the
`` uint16_t `` offsets to each register as defined by the hardware
chip headers. The register is set to the provided value. However, a
`` reg offset `` that specifies a user data register (e.g.,
COMPUTE_USER_DATA_0) needs special treatment. See
:ref: `amdgpu-amdpal-code-object-user-data-section` section for more
information.
========================== ============== ====================================================================
.. _amdgpu-amdpal-code-object-user-data-section:
User Data
+++++++++
Each hardware stage has a set of 32-bit physical SPI *user data registers*
(either 16 or 32 based on graphics IP and the stage) which can be
written from a command buffer and then loaded into SGPRs when waves are
launched via a subsequent dispatch or draw operation. This is the way
most arguments are passed from the application/runtime to a hardware
shader.
PAL abstracts this functionality by exposing a set of 128 *user data
entries* per pipeline a client can use to pass arguments from a command
buffer to one or more shaders in that pipeline. The ELF code object must
specify a mapping from virtualized *user data entries* to physical *user
data registers*, and PAL is responsible for implementing that mapping,
including spilling overflow *user data entries* to memory if needed.
Since the *user data registers* are GRBM-accessible SPI registers, this
mapping is actually embedded in the `` .registers `` metadata entry. For
most registers, the value in that map is a literal 32-bit value that
should be written to the register by the driver. However, when the
register is a *user data register* (any USER_DATA register e.g.,
SPI_SHADER_USER_DATA_PS_5), the value is instead an encoding that tells
the driver to write either a *user data entry* value or one of several
driver-internal values to the register. This encoding is described in
the following table:
2020-09-16 20:38:54 +08:00
.. note ::
2021-03-30 15:33:07 +08:00
Currently, *user data registers* 0 and 1 (e.g., SPI_SHADER_USER_DATA_PS_0,
and SPI_SHADER_USER_DATA_PS_1) are reserved. *User data register* 0 must
always be programmed to the address of the GlobalTable, and *user data
register* 1 must always be programmed to the address of the PerShaderTable.
2020-09-16 20:38:54 +08:00
2021-03-30 15:33:07 +08:00
..
2020-09-16 20:38:54 +08:00
2021-03-30 15:33:07 +08:00
.. table :: AMDPAL User Data Mapping
:name: amdgpu-amdpal-code-object-metadata-user-data-mapping-table
========== ================= ===============================================================================
Value Name Description
========== ================= ===============================================================================
0..127 *User Data Entry* 32-bit value of user_data_entry[N] as specified via *CmdSetUserData()*
0x10000000 GlobalTable 32-bit pointer to GPU memory containing the global internal table (should
always point to *user data register* 0).
0x10000001 PerShaderTable 32-bit pointer to GPU memory containing the per-shader internal table. See
:ref: `amdgpu-amdpal-code-object-metadata-user-data-per-shader-table-section`
for more detail (should always point to *user data register* 1).
0x10000002 SpillTable 32-bit pointer to GPU memory containing the user data spill table. See
:ref: `amdgpu-amdpal-code-object-metadata-user-data-spill-table-section` for
more detail.
0x10000003 BaseVertex Vertex offset (32-bit unsigned integer). Not needed if the pipeline doesn't
reference the draw index in the vertex shader. Only supported by the first
stage in a graphics pipeline.
0x10000004 BaseInstance Instance offset (32-bit unsigned integer). Only supported by the first stage in
a graphics pipeline.
0x10000005 DrawIndex Draw index (32-bit unsigned integer). Only supported by the first stage in a
graphics pipeline.
0x10000006 Workgroup Thread group count (32-bit unsigned integer). Low half of a 64-bit address of
a buffer containing the grid dimensions for a Compute dispatch operation. The
high half of the address is stored in the next sequential user-SGPR. Only
supported by compute pipelines.
0x1000000A EsGsLdsSize Indicates that PAL will program this user-SGPR to contain the amount of LDS
space used for the ES/GS pseudo-ring-buffer for passing data between shader
stages.
0x1000000B ViewId View id (32-bit unsigned integer) identifies a view of graphic
pipeline instancing.
0x1000000C StreamOutTable 32-bit pointer to GPU memory containing the stream out target SRD table. This
can only appear for one shader stage per pipeline.
0x1000000D PerShaderPerfData 32-bit pointer to GPU memory containing the per-shader performance data buffer.
0x1000000F VertexBufferTable 32-bit pointer to GPU memory containing the vertex buffer SRD table. This can
only appear for one shader stage per pipeline.
0x10000010 UavExportTable 32-bit pointer to GPU memory containing the UAV export SRD table. This can
only appear for one shader stage per pipeline (PS). These replace color targets
and are completely separate from any UAVs used by the shader. This is optional,
and only used by the PS when UAV exports are used to replace color-target
exports to optimize specific shaders.
0x10000011 NggCullingData 64-bit pointer to GPU memory containing the hardware register data needed by
some NGG pipelines to perform culling. This value contains the address of the
first of two consecutive registers which provide the full GPU address.
0x10000015 FetchShaderPtr 64-bit pointer to GPU memory containing the fetch shader subroutine.
========== ================= ===============================================================================
.. _amdgpu-amdpal-code-object-metadata-user-data-per-shader-table-section:
Per-Shader Table
################
Low 32 bits of the GPU address for an optional buffer in the `` .data ``
section of the ELF. The high 32 bits of the address match the high 32 bits
of the shader's program counter.
The buffer can be anything the shader compiler needs it for, and
allows each shader to have its own region of the `` .data `` section.
Typically, this could be a table of buffer SRD's and the data pointed to
by the buffer SRD's, but it could be a flat-address region of memory as
well. Its layout and usage are defined by the shader compiler.
Each shader's table in the `` .data `` section is referenced by the symbol
`` _amdgpu_ ` ` \ *xs*\ ` ` _shdr_intrl_data `` where *xs* corresponds with the
hardware shader stage the data is for. E.g.,
`` _amdgpu_cs_shdr_intrl_data `` for the compute shader hardware stage.
.. _amdgpu-amdpal-code-object-metadata-user-data-spill-table-section:
Spill Table
###########
It is possible for a hardware shader to need access to more *user data
entries* than there are slots available in user data registers for one
or more hardware shader stages. In that case, the PAL runtime expects
the necessary *user data entries* to be spilled to GPU memory and use
one user data register to point to the spilled user data memory. The
value of the *user data entry* must then represent the location where
a shader expects to read the low 32-bits of the table's GPU virtual
address. The *spill table* itself represents a set of 32-bit values
managed by the PAL runtime in GPU-accessible memory that can be made
indirectly accessible to a hardware shader.
2020-09-16 20:38:54 +08:00
2017-08-16 04:47:41 +08:00
Unspecified OS
--------------
This section provides code conventions used when the target triple OS is
empty (see :ref: `amdgpu-target-triples` ).
2017-03-09 07:55:44 +08:00
2017-02-22 02:48:01 +08:00
Trap Handler ABI
2017-06-07 04:31:59 +08:00
~~~~~~~~~~~~~~~~
For code objects generated by AMDGPU backend for non-amdhsa OS, the runtime does
not install a trap handler. The `` llvm.trap `` and `` llvm.debugtrap ``
instructions are handled as follows:
.. table :: AMDGPU Trap Handler for Non-AMDHSA OS
:name: amdgpu-trap-handler-for-non-amdhsa-os-table
=============== =============== ===========================================
Usage Code Sequence Description
=============== =============== ===========================================
llvm.trap s_endpgm Causes wavefront to be terminated.
llvm.debugtrap *none* Compiler warning given that there is no
trap handler installed.
=============== =============== ===========================================
Source Languages
================
.. _amdgpu-opencl:
OpenCL
------
When the language is OpenCL the following differences occur:
1. The OpenCL memory model is used (see :ref: `amdgpu-amdhsa-memory-model` ).
2018-03-24 02:45:18 +08:00
2. The AMDGPU backend appends additional arguments to the kernel's explicit
arguments for the AMDHSA OS (see
:ref: `opencl-kernel-implicit-arguments-appended-for-amdhsa-os-table` ).
2017-08-16 04:47:41 +08:00
3. Additional metadata is generated
2018-06-22 03:38:56 +08:00
(see :ref: `amdgpu-amdhsa-code-object-metadata` ).
2017-06-07 04:31:59 +08:00
2018-03-24 02:45:18 +08:00
.. table :: OpenCL kernel implicit arguments appended for AMDHSA OS
:name: opencl-kernel-implicit-arguments-appended-for-amdhsa-os-table
======== ==== ========= ===========================================
Position Byte Byte Description
Size Alignment
======== ==== ========= ===========================================
2018-03-24 02:58:47 +08:00
1 8 8 OpenCL Global Offset X
2 8 8 OpenCL Global Offset Y
3 8 8 OpenCL Global Offset Z
4 8 8 OpenCL address of printf buffer
5 8 8 OpenCL address of virtual queue used by
enqueue_kernel.
6 8 8 OpenCL address of AqlWrap struct used by
enqueue_kernel.
2019-07-06 00:05:17 +08:00
7 8 8 Pointer argument used for Multi-gird
synchronization.
2018-03-24 02:45:18 +08:00
======== ==== ========= ===========================================
2017-06-07 04:31:59 +08:00
.. _amdgpu-hcc:
HCC
---
2018-03-24 02:45:18 +08:00
When the language is HCC the following differences occur:
2017-06-07 04:31:59 +08:00
1. The HSA memory model is used (see :ref: `amdgpu-amdhsa-memory-model` ).
2018-06-22 03:38:56 +08:00
.. _amdgpu-assembler:
2014-11-14 22:08:00 +08:00
Assembler
2017-06-07 04:31:59 +08:00
---------
2014-11-14 22:08:00 +08:00
2016-09-20 17:04:51 +08:00
AMDGPU backend has LLVM-MC based assembler which is currently in development.
2019-06-14 06:18:47 +08:00
It supports AMDGCN GFX6-GFX10.
2015-04-08 09:09:26 +08:00
2018-03-12 23:55:08 +08:00
This section describes general syntax for instructions and operands.
Instructions
~~~~~~~~~~~~
2018-12-18 01:38:11 +08:00
An instruction has the following :doc: `syntax<AMDGPUInstructionSyntax>` :
2018-03-12 23:55:08 +08:00
2019-12-12 13:47:56 +08:00
| `` < ` ` \ *opcode*\ `` > <`` \ *operand0*\ `` >, <`` \ *operand1*\ `` >,...
<`` \ *modifier0*\ `` > <`` \ *modifier1*\ `` >...``
2018-03-12 23:55:08 +08:00
2019-12-25 22:51:53 +08:00
:doc: `Operands<AMDGPUOperandSyntax>` are comma-separated while
2018-12-18 01:38:11 +08:00
:doc: `modifiers<AMDGPUModifierSyntax>` are space-separated.
2018-03-12 23:55:08 +08:00
2019-12-25 22:51:53 +08:00
The order of operands and modifiers is fixed.
Most modifiers are optional and may be omitted.
2015-04-08 09:09:26 +08:00
2019-12-25 22:51:53 +08:00
Links to detailed instruction syntax description may be found in the following
table. Note that features under development are not included
in this description.
2015-04-08 09:09:26 +08:00
2020-02-07 21:20:37 +08:00
=================================== =======================================
Core ISA ISA Extensions
=================================== =======================================
:doc: `GFX7<AMDGPU/AMDGPUAsmGFX7>` \-
:doc: `GFX8<AMDGPU/AMDGPUAsmGFX8>` \-
:doc: `GFX9<AMDGPU/AMDGPUAsmGFX9>` :doc: `gfx900<AMDGPU/AMDGPUAsmGFX900>`
2019-12-25 22:51:53 +08:00
2020-02-07 21:20:37 +08:00
:doc: `gfx902<AMDGPU/AMDGPUAsmGFX900>`
2019-12-25 22:51:53 +08:00
2020-02-07 21:20:37 +08:00
:doc: `gfx904<AMDGPU/AMDGPUAsmGFX904>`
2019-12-25 22:51:53 +08:00
2020-02-07 21:20:37 +08:00
:doc: `gfx906<AMDGPU/AMDGPUAsmGFX906>`
2019-12-25 22:51:53 +08:00
2020-02-07 21:20:37 +08:00
:doc: `gfx908<AMDGPU/AMDGPUAsmGFX908>`
2019-12-25 22:51:53 +08:00
2020-02-07 21:20:37 +08:00
:doc: `gfx909<AMDGPU/AMDGPUAsmGFX900>`
2019-12-25 22:51:53 +08:00
2020-02-07 21:20:37 +08:00
:doc: `GFX10<AMDGPU/AMDGPUAsmGFX10>` :doc: `gfx1011<AMDGPU/AMDGPUAsmGFX1011>`
2019-12-25 22:51:53 +08:00
2020-02-07 21:20:37 +08:00
:doc: `gfx1012<AMDGPU/AMDGPUAsmGFX1011>`
=================================== =======================================
2018-03-12 23:55:08 +08:00
2019-12-12 13:47:56 +08:00
For more information about instructions, their semantics and supported
combinations of operands, refer to one of instruction set architecture manuals
2020-12-13 09:06:18 +08:00
[AMD-GCN-GFX6]_ , [AMD-GCN-GFX7]_ , [AMD-GCN-GFX8]_ , [AMD-GCN-GFX9]_ ,
[AMD-GCN-GFX10-RDNA1]_ and [AMD-GCN-GFX10-RDNA2]_ .
2015-04-08 09:09:26 +08:00
2016-09-20 17:04:51 +08:00
Operands
2017-06-07 04:31:59 +08:00
~~~~~~~~
2015-04-08 09:09:26 +08:00
2018-12-18 01:38:11 +08:00
Detailed description of operands may be found :doc: `here<AMDGPUOperandSyntax>` .
2015-04-08 09:09:26 +08:00
2018-03-12 23:55:08 +08:00
Modifiers
~~~~~~~~~
2014-11-14 22:08:00 +08:00
2019-12-12 13:47:56 +08:00
Detailed description of modifiers may be found
:doc: `here<AMDGPUModifierSyntax>` .
2016-09-20 17:04:51 +08:00
2017-06-07 04:31:59 +08:00
Instruction Examples
~~~~~~~~~~~~~~~~~~~~
DS
2018-03-12 23:55:08 +08:00
++
2014-11-14 22:08:00 +08:00
.. code-block :: nasm
2016-09-20 17:04:51 +08:00
ds_add_u32 v2, v4 offset:16
ds_write_src2_b64 v2 offset0:4 offset1:8
ds_cmpst_f32 v2, v4, v6
ds_min_rtn_f64 v[8:9], v2, v[4:5]
2014-11-14 22:08:00 +08:00
2019-12-12 13:47:56 +08:00
For full list of supported instructions, refer to "LDS/GDS instructions" in ISA
Manual.
2014-11-14 22:08:00 +08:00
2017-06-07 04:31:59 +08:00
FLAT
++++
2015-04-08 09:09:26 +08:00
2016-09-20 17:04:51 +08:00
.. code-block :: nasm
2015-04-08 09:09:26 +08:00
2016-09-20 17:04:51 +08:00
flat_load_dword v1, v[3:4]
flat_store_dwordx3 v[3:4], v[5:7]
flat_atomic_swap v1, v[3:4], v5 glc
flat_atomic_cmpswap v1, v[3:4], v[5:6] glc slc
flat_atomic_fmax_x2 v[1:2], v[3:4], v[5:6] glc
2019-12-12 13:47:56 +08:00
For full list of supported instructions, refer to "FLAT instructions" in ISA
Manual.
2016-09-20 17:04:51 +08:00
2017-06-07 04:31:59 +08:00
MUBUF
+++++
2015-04-08 09:09:26 +08:00
.. code-block :: nasm
2016-09-20 17:04:51 +08:00
buffer_load_dword v1, off, s[4:7], s1
buffer_store_dwordx4 v[1:4], v2, ttmp[4:7], s1 offen offset:4 glc tfe
buffer_store_format_xy v[1:2], off, s[4:7], s1
buffer_wbinvl1
buffer_atomic_inc v1, v2, s[8:11], s4 idxen offset:4 slc
2019-12-12 13:47:56 +08:00
For full list of supported instructions, refer to "MUBUF Instructions" in ISA
Manual.
2016-09-20 17:04:51 +08:00
2017-06-07 04:31:59 +08:00
SMRD/SMEM
+++++++++
2016-09-20 17:04:51 +08:00
.. code-block :: nasm
s_load_dword s1, s[2:3], 0xfc
s_load_dwordx8 s[8:15], s[2:3], s4
s_load_dwordx16 s[88:103], s[2:3], s4
s_dcache_inv_vol
s_memtime s[4:5]
2019-12-12 13:47:56 +08:00
For full list of supported instructions, refer to "Scalar Memory Operations" in
ISA Manual.
2016-09-20 17:04:51 +08:00
2017-06-07 04:31:59 +08:00
SOP1
++++
2016-09-20 17:04:51 +08:00
.. code-block :: nasm
s_mov_b32 s1, s2
s_mov_b64 s[0:1], 0x80000000
s_cmov_b32 s1, 200
s_wqm_b64 s[2:3], s[4:5]
s_bcnt0_i32_b64 s1, s[2:3]
s_swappc_b64 s[2:3], s[4:5]
s_cbranch_join s[4:5]
2019-12-12 13:47:56 +08:00
For full list of supported instructions, refer to "SOP1 Instructions" in ISA
Manual.
2016-09-20 17:04:51 +08:00
2017-06-07 04:31:59 +08:00
SOP2
++++
2016-09-20 17:04:51 +08:00
.. code-block :: nasm
s_add_u32 s1, s2, s3
s_and_b64 s[2:3], s[4:5], s[6:7]
s_cselect_b32 s1, s2, s3
s_andn2_b32 s2, s4, s6
s_lshr_b64 s[2:3], s[4:5], s6
s_ashr_i32 s2, s4, s6
s_bfm_b64 s[2:3], s4, s6
s_bfe_i64 s[2:3], s[4:5], s6
s_cbranch_g_fork s[4:5], s[6:7]
2019-12-12 13:47:56 +08:00
For full list of supported instructions, refer to "SOP2 Instructions" in ISA
Manual.
2016-09-20 17:04:51 +08:00
2017-06-07 04:31:59 +08:00
SOPC
++++
2016-09-20 17:04:51 +08:00
.. code-block :: nasm
2015-06-27 05:15:07 +08:00
2016-09-20 17:04:51 +08:00
s_cmp_eq_i32 s1, s2
s_bitcmp1_b32 s1, s2
s_bitcmp0_b64 s[2:3], s4
s_setvskip s3, s5
2019-12-12 13:47:56 +08:00
For full list of supported instructions, refer to "SOPC Instructions" in ISA
Manual.
2016-09-20 17:04:51 +08:00
2017-06-07 04:31:59 +08:00
SOPP
++++
2016-09-20 17:04:51 +08:00
.. code-block :: nasm
s_barrier
s_nop 2
s_endpgm
s_waitcnt 0 ; Wait for all counters to be 0
s_waitcnt vmcnt(0) & expcnt(0) & lgkmcnt(0) ; Equivalent to above
s_waitcnt vmcnt(1) ; Wait for vmcnt counter to be 1.
s_sethalt 9
s_sleep 10
s_sendmsg 0x1
s_sendmsg sendmsg(MSG_INTERRUPT)
s_trap 1
2019-12-12 13:47:56 +08:00
For full list of supported instructions, refer to "SOPP Instructions" in ISA
Manual.
2016-09-20 17:04:51 +08:00
Unless otherwise mentioned, little verification is performed on the operands
2017-01-14 19:37:01 +08:00
of SOPP Instructions, so it is up to the programmer to be familiar with the
2016-09-20 17:04:51 +08:00
range or acceptable values.
2017-06-07 04:31:59 +08:00
VALU
++++
2016-09-20 17:04:51 +08:00
For vector ALU instruction opcodes (VOP1, VOP2, VOP3, VOPC, VOP_DPP, VOP_SDWA),
2019-12-12 13:47:56 +08:00
the assembler will automatically use optimal encoding based on its operands. To
force specific encoding, one can add a suffix to the opcode of the instruction:
2016-09-20 17:04:51 +08:00
* _e32 for 32-bit VOP1/VOP2/VOPC
* _e64 for 64-bit VOP3
* _dpp for VOP_DPP
* _sdwa for VOP_SDWA
VOP1/VOP2/VOP3/VOPC examples:
.. code-block :: nasm
v_mov_b32 v1, v2
v_mov_b32_e32 v1, v2
v_nop
v_cvt_f64_i32_e32 v[1:2], v2
v_floor_f32_e32 v1, v2
v_bfrev_b32_e32 v1, v2
v_add_f32_e32 v1, v2, v3
v_mul_i32_i24_e64 v1, v2, 3
v_mul_i32_i24_e32 v1, -3, v3
v_mul_i32_i24_e32 v1, -100, v3
v_addc_u32 v1, s[0:1], v2, v3, s[2:3]
v_max_f16_e32 v1, v2, v3
VOP_DPP examples:
.. code-block :: nasm
v_mov_b32 v0, v0 quad_perm:[0,2,1,1]
v_sin_f32 v0, v0 row_shl:1 row_mask:0xa bank_mask:0x1 bound_ctrl:0
v_mov_b32 v0, v0 wave_shl:1
v_mov_b32 v0, v0 row_mirror
v_mov_b32 v0, v0 row_bcast:31
v_mov_b32 v0, v0 quad_perm:[1,3,0,1] row_mask:0xa bank_mask:0x1 bound_ctrl:0
v_add_f32 v0, v0, |v0| row_shl:1 row_mask:0xa bank_mask:0x1 bound_ctrl:0
v_max_f16 v1, v2, v3 row_shl:1 row_mask:0xa bank_mask:0x1 bound_ctrl:0
VOP_SDWA examples:
.. code-block :: nasm
v_mov_b32 v1, v2 dst_sel:BYTE_0 dst_unused:UNUSED_PRESERVE src0_sel:DWORD
v_min_u32 v200, v200, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_1 src1_sel:DWORD
v_sin_f32 v0, v0 dst_unused:UNUSED_PAD src0_sel:WORD_1
v_fract_f32 v0, |v0| dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
v_cmpx_le_u32 vcc, v1, v2 src0_sel:BYTE_2 src1_sel:WORD_0
For full list of supported instructions, refer to "Vector ALU instructions".
2019-03-28 23:08:52 +08:00
.. _amdgpu-amdhsa-assembler-predefined-symbols-v2:
2020-12-01 12:12:04 +08:00
Code Object V2 Predefined Symbols
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2019-03-28 23:08:52 +08:00
2020-12-01 12:12:04 +08:00
.. warning ::
Code object V2 is not the default code object version emitted by
this version of LLVM.
2019-03-28 23:08:52 +08:00
The AMDGPU assembler defines and updates some symbols automatically. These
symbols do not affect code generation.
.option.machine_version_major
+++++++++++++++++++++++++++++
Set to the GFX major generation number of the target being assembled for. For
example, when assembling for a "GFX9" target this will be set to the integer
value "9". The possible GFX major generation numbers are presented in
:ref: `amdgpu-processors` .
.option.machine_version_minor
+++++++++++++++++++++++++++++
Set to the GFX minor generation number of the target being assembled for. For
example, when assembling for a "GFX810" target this will be set to the integer
value "1". The possible GFX minor generation numbers are presented in
:ref: `amdgpu-processors` .
.option.machine_version_stepping
++++++++++++++++++++++++++++++++
Set to the GFX stepping generation number of the target being assembled for.
For example, when assembling for a "GFX704" target this will be set to the
integer value "4". The possible GFX stepping generation numbers are presented
in :ref: `amdgpu-processors` .
.kernel.vgpr_count
++++++++++++++++++
Set to zero each time a
:ref: `amdgpu-amdhsa-assembler-directive-amdgpu_hsa_kernel` directive is
encountered. At each instruction, if the current value of this symbol is less
2020-08-19 19:18:52 +08:00
than or equal to the maximum VGPR number explicitly referenced within that
2019-03-28 23:08:52 +08:00
instruction then the symbol value is updated to equal that VGPR number plus
one.
.kernel.sgpr_count
++++++++++++++++++
Set to zero each time a
:ref: `amdgpu-amdhsa-assembler-directive-amdgpu_hsa_kernel` directive is
encountered. At each instruction, if the current value of this symbol is less
2020-08-19 19:18:52 +08:00
than or equal to the maximum VGPR number explicitly referenced within that
2019-03-28 23:08:52 +08:00
instruction then the symbol value is updated to equal that SGPR number plus
one.
.. _amdgpu-amdhsa-assembler-directives-v2:
2020-12-01 12:12:04 +08:00
Code Object V2 Directives
~~~~~~~~~~~~~~~~~~~~~~~~~
2019-03-28 23:08:52 +08:00
2020-12-01 12:12:04 +08:00
.. warning ::
Code object V2 is not the default code object version emitted by
this version of LLVM.
2018-06-23 03:23:18 +08:00
AMDGPU ABI defines auxiliary data in output code object. In assembly source,
one can specify them with assembler directives.
.hsa_code_object_version major, minor
+++++++++++++++++++++++++++++++++++++
*major* and *minor* are integers that specify the version of the HSA code
object that will be generated by the assembler.
.hsa_code_object_isa [major, minor, stepping, vendor, arch]
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
*major* , *minor* , and *stepping* are all integers that describe the instruction
set architecture (ISA) version of the assembly program.
2019-12-12 13:47:56 +08:00
*vendor* and *arch* are quoted strings. *vendor* should always be equal to
2018-06-23 03:23:18 +08:00
"AMD" and *arch* should always be equal to "AMDGPU".
By default, the assembler will derive the ISA version, *vendor* , and *arch*
from the value of the -mcpu option that is passed to the assembler.
2019-03-28 23:08:52 +08:00
.. _amdgpu-amdhsa-assembler-directive-amdgpu_hsa_kernel:
2018-06-23 03:23:18 +08:00
.amdgpu_hsa_kernel (name)
+++++++++++++++++++++++++
2019-12-12 13:47:56 +08:00
This directives specifies that the symbol with given name is a kernel entry
point (label) and the object should contain corresponding symbol of type
STT_AMDGPU_HSA_KERNEL.
2018-06-23 03:23:18 +08:00
.amd_kernel_code_t
++++++++++++++++++
This directive marks the beginning of a list of key / value pairs that are used
to specify the amd_kernel_code_t object that will be emitted by the assembler.
2019-12-12 13:47:56 +08:00
The list must be terminated by the *.end_amd_kernel_code_t* directive. For any
amd_kernel_code_t values that are unspecified a default value will be used. The
default value for all keys is 0, with the following exceptions:
2018-06-23 03:23:18 +08:00
2019-06-14 06:18:47 +08:00
- *amd_code_version_major* defaults to 1.
- *amd_kernel_code_version_minor* defaults to 2.
- *amd_machine_kind* defaults to 1.
- *amd_machine_version_major* , *machine_version_minor* , and
*amd_machine_version_stepping* are derived from the value of the -mcpu option
2018-06-23 03:23:18 +08:00
that is passed to the assembler.
- *kernel_code_entry_byte_offset* defaults to 256.
2019-06-14 06:18:47 +08:00
- *wavefront_size* defaults 6 for all targets before GFX10. For GFX10 onwards
defaults to 6 if target feature `` wavefrontsize64 `` is enabled, otherwise 5.
Note that wavefront size is specified as a power of two, so a value of **n**
means a size of 2^ **n** .
- *call_convention* defaults to -1.
2018-06-23 03:23:18 +08:00
- *kernarg_segment_alignment* , *group_segment_alignment* , and
*private_segment_alignment* default to 4. Note that alignments are specified
2018-11-16 04:46:55 +08:00
as a power of 2, so a value of **n** means an alignment of 2^ **n** .
2021-02-18 05:37:46 +08:00
- *enable_tg_split* defaults to 1 if target feature `` tgsplit `` is enabled for
GFX90A onwards.
2019-06-14 06:18:47 +08:00
- *enable_wgp_mode* defaults to 1 if target feature `` cumode `` is disabled for
GFX10 onwards.
- *enable_mem_ordered* defaults to 1 for GFX10 onwards.
2018-06-23 03:23:18 +08:00
The *.amd_kernel_code_t* directive must be placed immediately after the
function label and before any instructions.
For a full list of amd_kernel_code_t keys, refer to AMDGPU ABI document,
comments in lib/Target/AMDGPU/AmdKernelCodeT.h and test/CodeGen/AMDGPU/hsa.s.
2019-03-28 23:08:52 +08:00
.. _amdgpu-amdhsa-assembler-example-v2:
2020-12-01 12:12:04 +08:00
Code Object V2 Example Source Code
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2019-03-28 23:08:52 +08:00
2020-12-01 12:12:04 +08:00
.. warning ::
Code Object V2 is not the default code object version emitted by
this version of LLVM.
2019-03-28 23:08:52 +08:00
Here is an example of a minimal assembly source file, defining one HSA kernel:
2018-06-23 03:23:18 +08:00
2019-12-12 13:47:56 +08:00
.. code ::
:number-lines:
2018-06-23 03:23:18 +08:00
.hsa_code_object_version 1,0
.hsa_code_object_isa
.hsatext
.globl hello_world
.p2align 8
.amdgpu_hsa_kernel hello_world
hello_world:
.amd_kernel_code_t
enable_sgpr_kernarg_segment_ptr = 1
is_ptr64 = 1
compute_pgm_rsrc1_vgprs = 0
compute_pgm_rsrc1_sgprs = 0
compute_pgm_rsrc2_user_sgpr = 2
2019-06-14 06:18:47 +08:00
compute_pgm_rsrc1_wgp_mode = 0
compute_pgm_rsrc1_mem_ordered = 0
compute_pgm_rsrc1_fwd_progress = 1
2018-06-23 03:23:18 +08:00
.end_amd_kernel_code_t
s_load_dwordx2 s[0:1], s[0:1] 0x0
v_mov_b32 v0, 3.14159
s_waitcnt lgkmcnt(0)
v_mov_b32 v1, s0
v_mov_b32 v2, s1
flat_store_dword v[1:2], v0
s_endpgm
.Lfunc_end0:
.size hello_world, .Lfunc_end0-hello_world
2020-12-01 12:12:04 +08:00
.. _amdgpu-amdhsa-assembler-predefined-symbols-v3-v4:
2019-03-28 23:08:52 +08:00
2020-12-01 12:12:04 +08:00
Code Object V3 to V4 Predefined Symbols
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2015-06-27 05:58:31 +08:00
2018-06-22 03:38:56 +08:00
The AMDGPU assembler defines and updates some symbols automatically. These
symbols do not affect code generation.
.amdgcn.gfx_generation_number
+++++++++++++++++++++++++++++
2019-02-08 21:51:31 +08:00
Set to the GFX major generation number of the target being assembled for. For
2018-06-22 03:38:56 +08:00
example, when assembling for a "GFX9" target this will be set to the integer
2019-02-08 21:51:31 +08:00
value "9". The possible GFX major generation numbers are presented in
2018-06-22 03:38:56 +08:00
:ref: `amdgpu-processors` .
2019-02-08 21:51:31 +08:00
.amdgcn.gfx_generation_minor
++++++++++++++++++++++++++++
Set to the GFX minor generation number of the target being assembled for. For
example, when assembling for a "GFX810" target this will be set to the integer
value "1". The possible GFX minor generation numbers are presented in
:ref: `amdgpu-processors` .
.amdgcn.gfx_generation_stepping
+++++++++++++++++++++++++++++++
Set to the GFX stepping generation number of the target being assembled for.
For example, when assembling for a "GFX704" target this will be set to the
integer value "4". The possible GFX stepping generation numbers are presented
in :ref: `amdgpu-processors` .
2019-03-30 01:49:51 +08:00
.. _amdgpu-amdhsa-assembler-symbol-next_free_vgpr:
2018-06-22 03:38:56 +08:00
.amdgcn.next_free_vgpr
++++++++++++++++++++++
Set to zero before assembly begins. At each instruction, if the current value
of this symbol is less than or equal to the maximum VGPR number explicitly
referenced within that instruction then the symbol value is updated to equal
that VGPR number plus one.
2020-08-19 19:18:52 +08:00
May be used to set the `.amdhsa_next_free_vgpr`
2018-06-22 03:38:56 +08:00
:ref: `amdhsa-kernel-directives-table` .
May be set at any time, e.g. manually set to zero at the start of each kernel.
2019-03-30 01:49:51 +08:00
.. _amdgpu-amdhsa-assembler-symbol-next_free_sgpr:
2018-06-22 03:38:56 +08:00
.amdgcn.next_free_sgpr
++++++++++++++++++++++
Set to zero before assembly begins. At each instruction, if the current value
of this symbol is less than or equal the maximum SGPR number explicitly
referenced within that instruction then the symbol value is updated to equal
that SGPR number plus one.
May be used to set the `.amdhsa_next_free_spgr`
:ref: `amdhsa-kernel-directives-table` .
May be set at any time, e.g. manually set to zero at the start of each kernel.
2020-12-01 12:12:04 +08:00
.. _amdgpu-amdhsa-assembler-directives-v3-v4:
2019-03-28 23:08:52 +08:00
2020-12-01 12:12:04 +08:00
Code Object V3 to V4 Directives
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2018-06-22 03:38:56 +08:00
Directives which begin with `` .amdgcn `` are valid for all `` amdgcn ``
architecture processors, and are not OS-specific. Directives which begin with
`` .amdhsa `` are specific to `` amdgcn `` architecture processors when the
`` amdhsa `` OS is specified. See :ref: `amdgpu-target-triples` and
:ref: `amdgpu-processors` .
2021-01-20 11:38:55 +08:00
.. _amdgpu-assembler-directive-amdgcn-target:
2020-12-01 12:12:04 +08:00
.amdgcn_target <target-triple> "-" <target-id>
++++++++++++++++++++++++++++++++++++++++++++++
2018-06-22 03:38:56 +08:00
2020-12-01 12:12:04 +08:00
Optional directive which declares the `` <target-triple>-<target-id> `` supported
by the containing assembler source file. Used by the assembler to validate
command-line options such as `` -triple `` , `` -mcpu `` , and
`` --offload-arch=<target-id> `` . A non-canonical target ID is allowed. See
:ref: `amdgpu-target-triples` and :ref: `amdgpu-target-id` .
2018-06-22 03:38:56 +08:00
2021-01-20 11:38:55 +08:00
.. note ::
The target ID syntax used for code object V2 to V3 for this directive differs
from that used elsewhere. See :ref: `amdgpu-target-id-v2-v3` .
2018-06-22 03:38:56 +08:00
.amdhsa_kernel <name>
+++++++++++++++++++++
Creates a correctly aligned AMDHSA kernel descriptor and a symbol,
`` <name>.kd `` , in the current location of the current section. Only valid when
the OS is `` amdhsa `` . `` <name> `` must be a symbol that labels the first
instruction to execute, and does not need to be previously defined.
Marks the beginning of a list of directives used to generate the bytes of a
kernel descriptor, as described in :ref: `amdgpu-amdhsa-kernel-descriptor` .
Directives which may appear in this list are described in
:ref: `amdhsa-kernel-directives-table` . Directives may appear in any order, must
be valid for the target being assembled for, and cannot be repeated. Directives
support the range of values specified by the field they reference in
:ref: `amdgpu-amdhsa-kernel-descriptor` . If a directive is not specified, it is
assumed to have its default value, unless it is marked as "Required", in which
case it is an error to omit the directive. This list of directives is
terminated by an `` .end_amdhsa_kernel `` directive.
.. table :: AMDHSA Kernel Assembler Directives
:name: amdhsa-kernel-directives-table
2019-06-14 06:18:47 +08:00
======================================================== =================== ============ ===================
Directive Default Supported On Description
======================================================== =================== ============ ===================
`` .amdhsa_group_segment_fixed_size `` 0 GFX6-GFX10 Controls GROUP_SEGMENT_FIXED_SIZE in
2020-12-01 12:12:04 +08:00
:ref: `amdgpu-amdhsa-kernel-descriptor-v3-table` .
2019-06-14 06:18:47 +08:00
`` .amdhsa_private_segment_fixed_size `` 0 GFX6-GFX10 Controls PRIVATE_SEGMENT_FIXED_SIZE in
2020-12-01 12:12:04 +08:00
:ref: `amdgpu-amdhsa-kernel-descriptor-v3-table` .
`` .amdhsa_kernarg_size `` 0 GFX6-GFX10 Controls KERNARG_SIZE in
:ref: `amdgpu-amdhsa-kernel-descriptor-v3-table` .
2019-06-14 06:18:47 +08:00
`` .amdhsa_user_sgpr_private_segment_buffer `` 0 GFX6-GFX10 Controls ENABLE_SGPR_PRIVATE_SEGMENT_BUFFER in
2020-12-01 12:12:04 +08:00
:ref: `amdgpu-amdhsa-kernel-descriptor-v3-table` .
2019-06-14 06:18:47 +08:00
`` .amdhsa_user_sgpr_dispatch_ptr `` 0 GFX6-GFX10 Controls ENABLE_SGPR_DISPATCH_PTR in
2020-12-01 12:12:04 +08:00
:ref: `amdgpu-amdhsa-kernel-descriptor-v3-table` .
2019-06-14 06:18:47 +08:00
`` .amdhsa_user_sgpr_queue_ptr `` 0 GFX6-GFX10 Controls ENABLE_SGPR_QUEUE_PTR in
2020-12-01 12:12:04 +08:00
:ref: `amdgpu-amdhsa-kernel-descriptor-v3-table` .
2019-06-14 06:18:47 +08:00
`` .amdhsa_user_sgpr_kernarg_segment_ptr `` 0 GFX6-GFX10 Controls ENABLE_SGPR_KERNARG_SEGMENT_PTR in
2020-12-01 12:12:04 +08:00
:ref: `amdgpu-amdhsa-kernel-descriptor-v3-table` .
2019-06-14 06:18:47 +08:00
`` .amdhsa_user_sgpr_dispatch_id `` 0 GFX6-GFX10 Controls ENABLE_SGPR_DISPATCH_ID in
2020-12-01 12:12:04 +08:00
:ref: `amdgpu-amdhsa-kernel-descriptor-v3-table` .
2019-06-14 06:18:47 +08:00
`` .amdhsa_user_sgpr_flat_scratch_init `` 0 GFX6-GFX10 Controls ENABLE_SGPR_FLAT_SCRATCH_INIT in
2020-12-01 12:12:04 +08:00
:ref: `amdgpu-amdhsa-kernel-descriptor-v3-table` .
2019-06-14 06:18:47 +08:00
`` .amdhsa_user_sgpr_private_segment_size `` 0 GFX6-GFX10 Controls ENABLE_SGPR_PRIVATE_SEGMENT_SIZE in
2020-12-01 12:12:04 +08:00
:ref: `amdgpu-amdhsa-kernel-descriptor-v3-table` .
2019-06-14 06:18:47 +08:00
`` .amdhsa_wavefront_size32 `` Target GFX10 Controls ENABLE_WAVEFRONT_SIZE32 in
2020-12-01 12:12:04 +08:00
Feature :ref: `amdgpu-amdhsa-kernel-descriptor-v3-table` .
2019-06-14 06:18:47 +08:00
Specific
2020-12-01 12:12:04 +08:00
(wavefrontsize64)
2020-12-15 09:42:25 +08:00
`` .amdhsa_system_sgpr_private_segment_wavefront_offset `` 0 GFX6-GFX10 Controls ENABLE_PRIVATE_SEGMENT in
2019-06-14 06:18:47 +08:00
:ref: `amdgpu-amdhsa-compute_pgm_rsrc2-gfx6-gfx10-table` .
`` .amdhsa_system_sgpr_workgroup_id_x `` 1 GFX6-GFX10 Controls ENABLE_SGPR_WORKGROUP_ID_X in
:ref: `amdgpu-amdhsa-compute_pgm_rsrc2-gfx6-gfx10-table` .
`` .amdhsa_system_sgpr_workgroup_id_y `` 0 GFX6-GFX10 Controls ENABLE_SGPR_WORKGROUP_ID_Y in
:ref: `amdgpu-amdhsa-compute_pgm_rsrc2-gfx6-gfx10-table` .
`` .amdhsa_system_sgpr_workgroup_id_z `` 0 GFX6-GFX10 Controls ENABLE_SGPR_WORKGROUP_ID_Z in
:ref: `amdgpu-amdhsa-compute_pgm_rsrc2-gfx6-gfx10-table` .
`` .amdhsa_system_sgpr_workgroup_info `` 0 GFX6-GFX10 Controls ENABLE_SGPR_WORKGROUP_INFO in
:ref: `amdgpu-amdhsa-compute_pgm_rsrc2-gfx6-gfx10-table` .
`` .amdhsa_system_vgpr_workitem_id `` 0 GFX6-GFX10 Controls ENABLE_VGPR_WORKITEM_ID in
:ref: `amdgpu-amdhsa-compute_pgm_rsrc2-gfx6-gfx10-table` .
Possible values are defined in
:ref: `amdgpu-amdhsa-system-vgpr-work-item-id-enumeration-values-table` .
`` .amdhsa_next_free_vgpr `` Required GFX6-GFX10 Maximum VGPR number explicitly referenced, plus one.
Used to calculate GRANULATED_WORKITEM_VGPR_COUNT in
:ref: `amdgpu-amdhsa-compute_pgm_rsrc1-gfx6-gfx10-table` .
`` .amdhsa_next_free_sgpr `` Required GFX6-GFX10 Maximum SGPR number explicitly referenced, plus one.
Used to calculate GRANULATED_WAVEFRONT_SGPR_COUNT in
:ref: `amdgpu-amdhsa-compute_pgm_rsrc1-gfx6-gfx10-table` .
2021-02-18 05:37:46 +08:00
`` .amdhsa_accum_offset `` Required GFX90A Offset of a first AccVGPR in the unified register file.
Used to calculate ACCUM_OFFSET in
:ref: `amdgpu-amdhsa-compute_pgm_rsrc3-gfx90a-table` .
2019-06-14 06:18:47 +08:00
`` .amdhsa_reserve_vcc `` 1 GFX6-GFX10 Whether the kernel may use the special VCC SGPR.
Used to calculate GRANULATED_WAVEFRONT_SGPR_COUNT in
:ref: `amdgpu-amdhsa-compute_pgm_rsrc1-gfx6-gfx10-table` .
`` .amdhsa_reserve_flat_scratch `` 1 GFX7-GFX10 Whether the kernel may use flat instructions to access
scratch memory. Used to calculate
GRANULATED_WAVEFRONT_SGPR_COUNT in
:ref: `amdgpu-amdhsa-compute_pgm_rsrc1-gfx6-gfx10-table` .
`` .amdhsa_reserve_xnack_mask `` Target GFX8-GFX10 Whether the kernel may trigger XNACK replay.
Feature Used to calculate GRANULATED_WAVEFRONT_SGPR_COUNT in
Specific :ref: `amdgpu-amdhsa-compute_pgm_rsrc1-gfx6-gfx10-table` .
2020-12-01 12:12:04 +08:00
(xnack)
2019-06-14 06:18:47 +08:00
`` .amdhsa_float_round_mode_32 `` 0 GFX6-GFX10 Controls FLOAT_ROUND_MODE_32 in
:ref: `amdgpu-amdhsa-compute_pgm_rsrc1-gfx6-gfx10-table` .
Possible values are defined in
:ref: `amdgpu-amdhsa-floating-point-rounding-mode-enumeration-values-table` .
`` .amdhsa_float_round_mode_16_64 `` 0 GFX6-GFX10 Controls FLOAT_ROUND_MODE_16_64 in
:ref: `amdgpu-amdhsa-compute_pgm_rsrc1-gfx6-gfx10-table` .
Possible values are defined in
:ref: `amdgpu-amdhsa-floating-point-rounding-mode-enumeration-values-table` .
`` .amdhsa_float_denorm_mode_32 `` 0 GFX6-GFX10 Controls FLOAT_DENORM_MODE_32 in
:ref: `amdgpu-amdhsa-compute_pgm_rsrc1-gfx6-gfx10-table` .
Possible values are defined in
:ref: `amdgpu-amdhsa-floating-point-denorm-mode-enumeration-values-table` .
`` .amdhsa_float_denorm_mode_16_64 `` 3 GFX6-GFX10 Controls FLOAT_DENORM_MODE_16_64 in
:ref: `amdgpu-amdhsa-compute_pgm_rsrc1-gfx6-gfx10-table` .
Possible values are defined in
:ref: `amdgpu-amdhsa-floating-point-denorm-mode-enumeration-values-table` .
`` .amdhsa_dx10_clamp `` 1 GFX6-GFX10 Controls ENABLE_DX10_CLAMP in
:ref: `amdgpu-amdhsa-compute_pgm_rsrc1-gfx6-gfx10-table` .
`` .amdhsa_ieee_mode `` 1 GFX6-GFX10 Controls ENABLE_IEEE_MODE in
:ref: `amdgpu-amdhsa-compute_pgm_rsrc1-gfx6-gfx10-table` .
`` .amdhsa_fp16_overflow `` 0 GFX9-GFX10 Controls FP16_OVFL in
:ref: `amdgpu-amdhsa-compute_pgm_rsrc1-gfx6-gfx10-table` .
2021-02-18 05:37:46 +08:00
`` .amdhsa_tg_split `` Target GFX90A Controls TG_SPLIT in
Feature :ref: `amdgpu-amdhsa-compute_pgm_rsrc3-gfx90a-table` .
Specific
(tgsplit)
2019-06-14 06:18:47 +08:00
`` .amdhsa_workgroup_processor_mode `` Target GFX10 Controls ENABLE_WGP_MODE in
2020-12-01 12:12:04 +08:00
Feature :ref: `amdgpu-amdhsa-kernel-descriptor-v3-table` .
2019-06-14 06:18:47 +08:00
Specific
2020-12-01 12:12:04 +08:00
(cumode)
2019-06-14 06:18:47 +08:00
`` .amdhsa_memory_ordered `` 1 GFX10 Controls MEM_ORDERED in
:ref: `amdgpu-amdhsa-compute_pgm_rsrc1-gfx6-gfx10-table` .
`` .amdhsa_forward_progress `` 0 GFX10 Controls FWD_PROGRESS in
:ref: `amdgpu-amdhsa-compute_pgm_rsrc1-gfx6-gfx10-table` .
`` .amdhsa_exception_fp_ieee_invalid_op `` 0 GFX6-GFX10 Controls ENABLE_EXCEPTION_IEEE_754_FP_INVALID_OPERATION in
:ref: `amdgpu-amdhsa-compute_pgm_rsrc2-gfx6-gfx10-table` .
`` .amdhsa_exception_fp_denorm_src `` 0 GFX6-GFX10 Controls ENABLE_EXCEPTION_FP_DENORMAL_SOURCE in
:ref: `amdgpu-amdhsa-compute_pgm_rsrc2-gfx6-gfx10-table` .
`` .amdhsa_exception_fp_ieee_div_zero `` 0 GFX6-GFX10 Controls ENABLE_EXCEPTION_IEEE_754_FP_DIVISION_BY_ZERO in
:ref: `amdgpu-amdhsa-compute_pgm_rsrc2-gfx6-gfx10-table` .
`` .amdhsa_exception_fp_ieee_overflow `` 0 GFX6-GFX10 Controls ENABLE_EXCEPTION_IEEE_754_FP_OVERFLOW in
:ref: `amdgpu-amdhsa-compute_pgm_rsrc2-gfx6-gfx10-table` .
`` .amdhsa_exception_fp_ieee_underflow `` 0 GFX6-GFX10 Controls ENABLE_EXCEPTION_IEEE_754_FP_UNDERFLOW in
:ref: `amdgpu-amdhsa-compute_pgm_rsrc2-gfx6-gfx10-table` .
`` .amdhsa_exception_fp_ieee_inexact `` 0 GFX6-GFX10 Controls ENABLE_EXCEPTION_IEEE_754_FP_INEXACT in
:ref: `amdgpu-amdhsa-compute_pgm_rsrc2-gfx6-gfx10-table` .
`` .amdhsa_exception_int_div_zero `` 0 GFX6-GFX10 Controls ENABLE_EXCEPTION_INT_DIVIDE_BY_ZERO in
:ref: `amdgpu-amdhsa-compute_pgm_rsrc2-gfx6-gfx10-table` .
======================================================== =================== ============ ===================
2018-06-22 03:38:56 +08:00
2018-11-16 04:46:55 +08:00
.amdgpu_metadata
++++++++++++++++
Optional directive which declares the contents of the `` NT_AMDGPU_METADATA ``
2020-12-01 12:12:04 +08:00
note record (see :ref: `amdgpu-elf-note-records-table-v3-v4` ).
2018-11-16 04:46:55 +08:00
The contents must be in the [YAML]_ markup format, with the same structure and
2020-12-01 12:12:04 +08:00
semantics described in :ref: `amdgpu-amdhsa-code-object-metadata-v3` or
:ref: `amdgpu-amdhsa-code-object-metadata-v4` .
2018-11-16 04:46:55 +08:00
This directive is terminated by an `` .end_amdgpu_metadata `` directive.
2020-12-01 12:12:04 +08:00
.. _amdgpu-amdhsa-assembler-example-v3-v4:
2019-03-28 23:08:52 +08:00
2020-12-01 12:12:04 +08:00
Code Object V3 to V4 Example Source Code
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2018-06-22 03:38:56 +08:00
Here is an example of a minimal assembly source file, defining one HSA kernel:
2015-06-27 05:58:31 +08:00
2019-12-12 13:47:56 +08:00
.. code ::
:number-lines:
.amdgcn_target "amdgcn-amd-amdhsa--gfx900+xnack" // optional
.text
.globl hello_world
.p2align 8
.type hello_world,@function
hello_world:
s_load_dwordx2 s[0:1], s[0:1] 0x0
v_mov_b32 v0, 3.14159
s_waitcnt lgkmcnt(0)
v_mov_b32 v1, s0
v_mov_b32 v2, s1
flat_store_dword v[1:2], v0
s_endpgm
.Lfunc_end0:
.size hello_world, .Lfunc_end0-hello_world
.rodata
.p2align 6
.amdhsa_kernel hello_world
.amdhsa_user_sgpr_kernarg_segment_ptr 1
.amdhsa_next_free_vgpr .amdgcn.next_free_vgpr
.amdhsa_next_free_sgpr .amdgcn.next_free_sgpr
.end_amdhsa_kernel
.amdgpu_metadata
---
amdhsa.version:
- 1
- 0
amdhsa.kernels:
- .name: hello_world
.symbol: hello_world.kd
.kernarg_segment_size: 48
.group_segment_fixed_size: 0
.private_segment_fixed_size: 0
.kernarg_segment_align: 4
.wavefront_size: 64
.sgpr_count: 2
.vgpr_count: 3
.max_flat_workgroup_size: 256
...
.end_amdgpu_metadata
2017-06-07 04:31:59 +08:00
2019-03-30 01:49:51 +08:00
If an assembly source file contains multiple kernels and/or functions, the
:ref: `amdgpu-amdhsa-assembler-symbol-next_free_vgpr` and
:ref: `amdgpu-amdhsa-assembler-symbol-next_free_sgpr` symbols may be reset using
the `` .set <symbol>, <expression> `` directive. For example, in the case of two
kernels, where `` function1 `` is only called from `` kernel1 `` it is sufficient
to group the function with the kernel that calls it and reset the symbols
between the two connected components:
2019-12-12 13:47:56 +08:00
.. code ::
:number-lines:
.amdgcn_target "amdgcn-amd-amdhsa--gfx900+xnack" // optional
// gpr tracking symbols are implicitly set to zero
.text
.globl kern0
.p2align 8
.type kern0,@function
kern0:
// ...
s_endpgm
.Lkern0_end:
.size kern0, .Lkern0_end-kern0
.rodata
.p2align 6
.amdhsa_kernel kern0
// ...
.amdhsa_next_free_vgpr .amdgcn.next_free_vgpr
.amdhsa_next_free_sgpr .amdgcn.next_free_sgpr
.end_amdhsa_kernel
// reset symbols to begin tracking usage in func1 and kern1
.set .amdgcn.next_free_vgpr, 0
.set .amdgcn.next_free_sgpr, 0
.text
.hidden func1
.global func1
.p2align 2
.type func1,@function
func1:
// ...
s_setpc_b64 s[30:31]
.Lfunc1_end:
.size func1, .Lfunc1_end-func1
.globl kern1
.p2align 8
.type kern1,@function
kern1:
// ...
s_getpc_b64 s[4:5]
s_add_u32 s4, s4, func1@rel32@lo+4
s_addc_u32 s5, s5, func1@rel32@lo+4
s_swappc_b64 s[30:31], s[4:5]
// ...
s_endpgm
.Lkern1_end:
.size kern1, .Lkern1_end-kern1
.rodata
.p2align 6
.amdhsa_kernel kern1
// ...
.amdhsa_next_free_vgpr .amdgcn.next_free_vgpr
.amdhsa_next_free_sgpr .amdgcn.next_free_sgpr
.end_amdhsa_kernel
2019-03-30 01:49:51 +08:00
These symbols cannot identify connected components in order to automatically
track the usage for each kernel. However, in some cases careful organization of
the kernels and functions in the source file means there is minimal additional
effort required to accurately calculate GPR usage.
2017-06-07 04:31:59 +08:00
Additional Documentation
========================
2017-10-19 01:59:20 +08:00
.. [AMD-GCN-GFX6] `AMD Southern Islands Series ISA <http://developer.amd.com/wordpress/media/2012/12/AMD_Southern_Islands_Instruction_Set_Architecture.pdf> `__
.. [AMD-GCN-GFX7] `AMD Sea Islands Series ISA <http://developer.amd.com/wordpress/media/2013/07/AMD_Sea_Islands_Instruction_Set_Architecture.pdf> `_
.. [AMD-GCN-GFX8] `AMD GCN3 Instruction Set Architecture <http://amd-dev.wpengine.netdna-cdn.com/wordpress/media/2013/12/AMD_GCN3_Instruction_Set_Architecture_rev1.1.pdf> `__
.. [AMD-GCN-GFX9] `AMD "Vega" Instruction Set Architecture <http://developer.amd.com/wordpress/media/2013/12/Vega_Shader_ISA_28July2017.pdf> `__
2020-12-13 09:06:18 +08:00
.. [AMD-GCN-GFX10-RDNA1] `AMD "RDNA 1.0" Instruction Set Architecture <https://gpuopen.com/wp-content/uploads/2019/08/RDNA_Shader_ISA_5August2019.pdf> `__
.. [AMD-GCN-GFX10-RDNA2] `AMD "RDNA 2" Instruction Set Architecture <https://developer.amd.com/wp-content/resources/RDNA2_Shader_ISA_November2020.pdf> `__
2020-04-14 12:55:43 +08:00
.. [AMD-RADEON-HD-2000-3000] `AMD R6xx shader ISA <http://developer.amd.com/wordpress/media/2012/10/R600_Instruction_Set_Architecture.pdf> `__
.. [AMD-RADEON-HD-4000] `AMD R7xx shader ISA <http://developer.amd.com/wordpress/media/2012/10/R700-Family_Instruction_Set_Architecture.pdf> `__
.. [AMD-RADEON-HD-5000] `AMD Evergreen shader ISA <http://developer.amd.com/wordpress/media/2012/10/AMD_Evergreen-Family_Instruction_Set_Architecture.pdf> `__
.. [AMD-RADEON-HD-6000] `AMD Cayman/Trinity shader ISA <http://developer.amd.com/wordpress/media/2012/10/AMD_HD_6900_Series_Instruction_Set_Architecture.pdf> `__
2020-12-10 11:38:32 +08:00
.. [AMD-ROCm] `AMD ROCm™ Platform <https://rocmdocs.amd.com/> `__
.. [AMD-ROCm-github] `AMD ROCm™ github <http://github.com/RadeonOpenCompute> `__
2020-12-01 12:12:04 +08:00
.. [AMD-ROCm-Release-Notes] `AMD ROCm Release Notes <https://github.com/RadeonOpenCompute/ROCm> `__
2020-04-14 12:55:43 +08:00
.. [CLANG-ATTR] `Attributes in Clang <https://clang.llvm.org/docs/AttributeReference.html> `__
2017-06-07 04:31:59 +08:00
.. [DWARF] `DWARF Debugging Information Format <http://dwarfstd.org/> `__
2020-04-14 12:55:43 +08:00
.. [ELF] `Executable and Linkable Format (ELF) <http://www.sco.com/developers/gabi/> `__
.. [HRF] `Heterogeneous-race-free Memory Models <http://benedictgaster.org/wp-content/uploads/2014/01/asplos269-FINAL.pdf> `__
.. [HSA] `Heterogeneous System Architecture (HSA) Foundation <http://www.hsafoundation.com/> `__
2018-11-16 04:46:55 +08:00
.. [MsgPack] `Message Pack <http://www.msgpack.org/> `__
2017-06-07 04:31:59 +08:00
.. [OpenCL] `The OpenCL Specification Version 2.0 <http://www.khronos.org/registry/cl/specs/opencl-2.0.pdf> `__
2020-04-14 12:55:43 +08:00
.. [SEMVER] `Semantic Versioning <https://semver.org/> `__
.. [YAML] `YAML Ain't Markup Language (YAML™) Version 1.2 <http://www.yaml.org/spec/1.2/spec.html> `__
