Originally, we were attaching attributes containing CUBIN blobs to the kernel
function called by `gpu.launch_func`. This kernel is now contained in a nested
module that is used as a compilation unit. Attach compiled CUBIN blobs to the
module rather than to the function since we were compiling the module. This
also avoids duplication of the attribute on multiple kernels within the same
module.
PiperOrigin-RevId: 273497303
Originally, the CUBIN getter function was introduced as a mechanism to
circumvent the absence of globals in the LLVM dialect. It would allocate memory
and populate it with the CUBIN data. LLVM dialect now supports globals and they
are already used to store CUBIN data, making the getter function a trivial
address computation of a global. Emit the address computation directly at the
place of `gpu.launch_func` instead of putting it in a function and calling it.
This simplifies the conversion flow and prepares it for using the
DialectConversion infrastructure.
PiperOrigin-RevId: 273496221
Now that the accessor function is a trivial getter of the global variable, it
makes less sense to have the getter generation as a separate pass. Move the
getter generation into the lowering of `gpu.launch_func` to CUDA calls. This
change is mostly code motion, but the process can be simplified further by
generating the addressof inplace instead of using a call. This is will be done
in a follow-up.
PiperOrigin-RevId: 273492517
The kernel function called by gpu.launch_func is now placed into an isolated
nested module during the outlining stage to simplify separate compilation.
Until recently, modules did not have names and could not be referenced. This
limitation was circumvented by introducing a stub kernel at the same name at
the same nesting level as the module containing the actual kernel. This
relation is only effective in one direction: from actual kernel function to its
launch_func "caller".
Leverage the recently introduced symbol name attributes on modules to refer to
a specific nested module from `gpu.launch_func`. This removes the implicit
connection between the identically named stub and kernel functions. It also
enables support for `gpu.launch_func`s to call different kernels located in the
same module.
PiperOrigin-RevId: 273491891
The GPUIndexIntrinsicOpLowering template is currently used by the code in both the GPUToNVVM and GPUToROCDL dirs.
Moving it to a common location to remove code duplication.
Closestensorflow/mlir#163
COPYBARA_INTEGRATE_REVIEW=https://github.com/tensorflow/mlir/pull/163 from deven-amd:deven-refactor-gpu-index-ops-lowering b8dc2a5f5353df196039b6ff2ad42106028693ed
PiperOrigin-RevId: 272863297
This makes the name of the conversion pass more consistent with the naming
scheme, since it actually converts from the Loop dialect to the Standard
dialect rather than working with arbitrary control flow operations.
PiperOrigin-RevId: 272612112
This also adds coverage with a missing test, which uncovered a bug in the conditional for testing whether an offset is dynamic or not.
PiperOrigin-RevId: 272505798
This is a follow-up to the PRtensorflow/mlir#146 which introduced the ROCDL Dialect. This PR introduces a pass to lower GPU Dialect to the ROCDL Dialect. As with the previous PR, this one builds on the work done by @whchung, and addresses most of the review comments in the original PR.
Closestensorflow/mlir#154
COPYBARA_INTEGRATE_REVIEW=https://github.com/tensorflow/mlir/pull/154 from deven-amd:deven-lower-gpu-to-rocdl 809893e08236da5ab6a38e3459692fa04247773d
PiperOrigin-RevId: 272390729
A recent ABI compatibility change affected the conversion from standard
CallOp/CallIndirectOp to LLVM::CallOp by changing its signature. In order to
analyze the signature, the code was looking up the callee symbol in the module.
This is incorrect since, during the conversion, the module may contain both the
original and the converted function op that have the same symbol name. There is
no strict guarantee on which of the two symbols will be found by the lookup.
The conversion was not failing because the type legalizer converts the LLVM
types to themselves making the original and the converted function signatures
ultimately produce the same type.
Instead of looking up the function signature to get the list of result types,
use the types of the CallOp/CallIndirectOp results which must match those of
the function in valid IR. These types are guaranteed to be the original,
unconverted types when converting the operation. Furthermore, this avoids the
need to perform a lookup of a symbol name in the module which may be expensive.
Finally, propagate attributes as-is from the original op to the converted op
since they share the attribute name for the callee of direct calls and the rest
of attributes are not affected by the conversion. This removes the need for
additional contorsions between direct and indirect calls to extract the name of
the optional callee attribute only to insert it back. This also prevents the
conversion from unintentionally dropping the other attributes of the op.
PiperOrigin-RevId: 272218871
This CL finishes the implementation of the Linalg + Affine type unification of the [strided memref RFC](https://groups.google.com/a/tensorflow.org/forum/#!topic/mlir/MaL8m2nXuio).
As a consequence, the !linalg.view type, linalg::DimOp, linalg::LoadOp and linalg::StoreOp can now disappear and Linalg can use standard types everywhere.
PiperOrigin-RevId: 272187165
Perform second reduce only with first warp. This requires an additional __sync_threads(), but doesn't need special handling when the last warp is small. This simplifies support for block sizes that are not multiple of 32.
Supporting partial warp reduce will be done in a separate CL.
PiperOrigin-RevId: 272168917
This CL finishes the implementation of the lowering part of the [strided memref RFC](https://groups.google.com/a/tensorflow.org/forum/#!topic/mlir/MaL8m2nXuio).
Strided memrefs correspond conceptually to the following templated C++ struct:
```
template <typename Elem, size_t Rank>
struct {
Elem *ptr;
int64_t offset;
int64_t sizes[Rank];
int64_t strides[Rank];
};
```
The linearization procedure for address calculation for strided memrefs is the same as for linalg views:
`base_offset + SUM_i index_i * stride_i`.
The following CL will unify Linalg and Standard by removing !linalg.view in favor of strided memrefs.
PiperOrigin-RevId: 272033399
The strided MemRef RFC discusses a normalized descriptor and interaction with library calls (https://groups.google.com/a/tensorflow.org/forum/#!topic/mlir/MaL8m2nXuio).
Lowering of nested LLVM structs as value types does not play nicely with externally compiled C/C++ functions due to ABI issues.
Solving the ABI problem generally is a very complex problem and most likely involves taking
a dependence on clang that we do not want atm.
A simple workaround is to pass pointers to memref descriptors at function boundaries, which this CL implement.
PiperOrigin-RevId: 271591708
The reduction operation is currently fixed to "add", and the scope is fixed to "workgroup".
The implementation is currently limited to sizes that are multiple 32 (warp size) and no larger than 1024.
PiperOrigin-RevId: 271290265
- introduce splat op in standard dialect (currently for int/float/index input
type, output type can be vector or statically shaped tensor)
- implement LLVM lowering (when result type is 1-d vector)
- add constant folding hook for it
- while on Ops.cpp, fix some stale names
Signed-off-by: Uday Bondhugula <uday@polymagelabs.com>
Closestensorflow/mlir#141
COPYBARA_INTEGRATE_REVIEW=https://github.com/tensorflow/mlir/pull/141 from bondhugula:splat 48976a6aa0a75be6d91187db6418de989e03eb51
PiperOrigin-RevId: 270965304
The RFC for unifying Linalg and Affine compilation passes into an end-to-end flow with a predictable ABI and linkage to external function calls raised the question of why we have variable sized descriptors for memrefs depending on whether they have static or dynamic dimensions (https://groups.google.com/a/tensorflow.org/forum/#!topic/mlir/MaL8m2nXuio).
This CL standardizes the ABI on the rank of the memrefs.
The LLVM struct for a memref becomes equivalent to:
```
template <typename Elem, size_t Rank>
struct {
Elem *ptr;
int64_t sizes[Rank];
};
```
PiperOrigin-RevId: 270947276
Roll forward of commit 5684a12.
When outlining GPU kernels, put the kernel function inside a nested module. Then use a nested pipeline to generate the cubins, independently per kernel. In a final pass, move the cubins back to the parent module.
PiperOrigin-RevId: 270639748
This adds sign- and zero-extension and truncation of integer types to the
standard dialects. This allows to perform integer type conversions without
having to go to the LLVM dialect and introduce custom type casts (between
standard and LLVM integer types).
Closestensorflow/mlir#134
COPYBARA_INTEGRATE_REVIEW=https://github.com/tensorflow/mlir/pull/134 from ombre5733:sext-zext-trunc-in-std c7657bc84c0ca66b304e53ec03797e09152e4d31
PiperOrigin-RevId: 270479722
When outlining GPU kernels, put the kernel function inside a nested module. Then use a nested pipeline to generate the cubins, independently per kernel. In a final pass, move the cubins back to the parent module.
PiperOrigin-RevId: 269987720
The helper functions makePositionAttr() and positionAttr() were originally
introduced in the lowering-to-LLVM-dialect pass to construct integer array
attributes that are used for static positions in extract/insertelement.
Constructing an integer array attribute being fairly common, a utility function
Builder::getI64ArrayAttr was later introduced into the Builder API. Drop
makePositionAttr and similar homegrown functions and use that API instead.
PiperOrigin-RevId: 269295836
Add support for specifying extended instructions sets. The operations
in SPIR-V dialect are named as 'spv.<extension-name>.<op-name>'. Use
this mechanism to define a 'Exp' operation from GLSL(450)
instructions.
Later CLs will add support for (de)serialization of these operations,
and update the dialect generation scripts to auto-generate the
specification using the spec directly.
Additional changes:
Add a Type Constraint to OpBase.td to check for vector of specified
lengths. This is used to check that the vector type used in SPIR-V
dialect are of lengths 2, 3 or 4.
Update SPIRVBase.td to use this Type constraints for vectors.
PiperOrigin-RevId: 269234377
SPIR-V recently publishes v1.5, which brings a bunch of symbols
into core. So the suffix "KHR"/"EXT"/etc. is removed from the
symbols. We use a script to pull information from the spec
directly.
Also changed conversion and tests to use GLSL450 instead of
VulkanKHR memory model. GLSL450 is still the main memory model
supported by Vulkan shaders and it does not require extra
capability to enable.
PiperOrigin-RevId: 268992661
This will allow us to use MLIR's folding infrastructure to deduplicate
SPIR-V constants.
This CL also changed isValidSPIRVType in SPIRVDialect to a static method.
PiperOrigin-RevId: 266984403
- the list of passes run by mlir-cpu-runner included -lower-affine and
-lower-to-llvm but was missing -lower-to-cfg (because -lower-affine at
some point used to lower straight to CFG); add -lower-to-cfg in
between. IR with affine ops can now be run by mlir-cpu-runner.
- update -lower-to-cfg to be consistent with other passes (create*Pass methods
were changed to return unique ptrs, but -lower-to-cfg appears to have been
missed).
- mlir-cpu-runner was unable to parse custom form of affine op's - fix
link options
- drop unnecessary run options from test/mlir-cpu-runner/simple.mlir
(none of the test cases had loops)
- -convert-to-llvmir was changed to -lower-to-llvm at some point, but the
create pass method name wasn't updated (this pass converts/lowers to LLVM
dialect as opposed to LLVM IR). Fix this.
(If we prefer "convert", the cmd-line options could be changed to
"-convert-to-llvm/cfg" then.)
Signed-off-by: Uday Bondhugula <uday@polymagelabs.com>
Closestensorflow/mlir#115
PiperOrigin-RevId: 266666909
This change refactors and cleans up the implementation of the operation walk methods. After this refactoring is that the explicit template parameter for the operation type is no longer needed for the explicit op walks. For example:
op->walk<AffineForOp>([](AffineForOp op) { ... });
is now accomplished via:
op->walk([](AffineForOp op) { ... });
PiperOrigin-RevId: 266209552
Instead of lowering the program in two steps (Standard->LLVM followed
by GPU->NVVM), leading to invalid IR inbetween, the runner now uses
one pattern based rewrite step to go directly from Standard+GPU to
LLVM+NVVM.
PiperOrigin-RevId: 265861934
To support a conversion of a simple load-compute-store kernel from GPU
dialect to SPIR-V dialect, the conversion of operations like
"gpu.block_dim", "gpu.thread_id" which allow threads to get the launch
conversion is needed. In SPIR-V these are specified as global
variables with builin attributes. This CL adds support to specify
builtin variables in SPIR-V conversion framework. This is used to
convert the relevant operations from GPU dialect to SPIR-V dialect.
Also add support for conversion of load/store operation in Standard
dialect to SPIR-V dialect.
To simplify the conversion add a method to build a spv.AccessChain
operation that automatically determines the return type based on the
base pointer type and the indices provided.
PiperOrigin-RevId: 265718525
This conversion has been using a stack-allocated array of i8 to store the
null-terminated kernel name in order to pass it to the CUDA wrappers expecting
a C string because the LLVM dialect was missing support for globals. Now that
the suport is introduced, use a global instead.
Refactor global string construction from GenerateCubinAccessors into a common
utility function living in the LLVM namespace.
PiperOrigin-RevId: 264382489
This CL allows binary operations on n-D vector types to be lowered to LLVMIR by performing an (n-1)-D extractvalue, 1-D vector operation and an (n-1)-D insertvalue.
PiperOrigin-RevId: 264339118
The linalg.view type used to be lowered to a struct containing a data pointer, offset, sizes/strides information. This was problematic when passing to external functions due to ABI, struct padding and alignment issues.
The linalg.view type is now lowered to LLVMIR as a *pointer* to a struct containing the data pointer, offset and sizes/strides. This simplifies the interfacing with external library functions and makes it trivial to add new functions without creating a shim that would go from a value type struct to a pointer type.
The consequences are that:
1. lowering explicitly uses llvm.alloca in lieu of llvm.undef and performs the proper llvm.load/llvm.store where relevant.
2. the shim creation function `getLLVMLibraryCallDefinition` disappears.
3. views are passed by pointer, scalars are passed by value. In the future, other structs will be passed by pointer (on a per-need basis).
PiperOrigin-RevId: 264183671
Switch to C++14 standard method as llvm::make_unique has been removed (
https://reviews.llvm.org/D66259). Also mark some targets as c++14 to ease next
integrates.
PiperOrigin-RevId: 263953918
Alex Zinenko