maxjhandsome
/
llvm-project
forked from OSchip/llvm-project
1
0
Fork 0
Code Issues Pull Requests Packages Releases Wiki Activity
llvm-project/llvm/test/CodeGen/AMDGPU/scratch-simple.ll

Ignoring revisions in .git-blame-ignore-revs. Click here to bypass and see the normal blame view.

245 lines
31 KiB
LLVM
Raw Normal View History

[AMDGPU] Use flat scratch instructions where available The support is disabled by default. So far there is instruction selection, spilling, and frame elimination. It also changes SP from unswizzled to swizzled as used by flat scratch instructions, so it cannot be mixed with MUBUF stack access. At the very least missing: - GlobalISel; - Some optimizations in frame elimination in between vector and scalar ALU; - It shall finally allow to always materialize frame index as an SGPR, but that is not implemented and frame elimination cannot handle it yet; - Unaligned and/or multidword flat scratch shall work, but it is legalized now for MUBUF; - Operand folding cannot optimize FI like with MUBUF yet; - It will need scaling the value of the SP/FP in the DWARF expression to recover the unswizzled scratch address; Differential Revision: https://reviews.llvm.org/D89170
2020-10-22 05:27:03 +08:00
; RUN: llc -march=amdgcn -mtriple=amdgcn-- -mcpu=verde -amdgpu-use-divergent-register-indexing -verify-machineinstrs < %s | FileCheck --check-prefixes=GCN,SI,SIVI,MUBUF %s
; RUN: llc -march=amdgcn -mtriple=amdgcn-- -mcpu=gfx803 -mattr=-flat-for-global -amdgpu-use-divergent-register-indexing -verify-machineinstrs < %s | FileCheck --check-prefixes=GCN,VI,SIVI,MUBUF %s
; RUN: llc -march=amdgcn -mtriple=amdgcn-- -mcpu=gfx900 -mattr=-flat-for-global -amdgpu-use-divergent-register-indexing -verify-machineinstrs < %s | FileCheck --check-prefixes=GCN,GFX9,GFX9_10,MUBUF,GFX9-MUBUF,GFX9_10-MUBUF %s
[llvm-readobj] Update tests because of changes in llvm-readobj behavior This patch updates tests using llvm-readobj and llvm-readelf, because soon reading from stdin will be achievable only via a '-' as described here: https://bugs.llvm.org/show_bug.cgi?id=46400. Patch with changes to llvm-readobj behavior is here: https://reviews.llvm.org/D83704 Differential Revision: https://reviews.llvm.org/D83912 Reviewed by: jhenderson, MaskRay, grimar
2020-07-20 15:39:14 +08:00
; RUN: llc -march=amdgcn -mtriple=amdgcn-- -mcpu=gfx900 -filetype=obj -amdgpu-use-divergent-register-indexing < %s | llvm-readobj -r - | FileCheck --check-prefix=RELS %s
[AMDGPU] Use flat scratch instructions where available The support is disabled by default. So far there is instruction selection, spilling, and frame elimination. It also changes SP from unswizzled to swizzled as used by flat scratch instructions, so it cannot be mixed with MUBUF stack access. At the very least missing: - GlobalISel; - Some optimizations in frame elimination in between vector and scalar ALU; - It shall finally allow to always materialize frame index as an SGPR, but that is not implemented and frame elimination cannot handle it yet; - Unaligned and/or multidword flat scratch shall work, but it is legalized now for MUBUF; - Operand folding cannot optimize FI like with MUBUF yet; - It will need scaling the value of the SP/FP in the DWARF expression to recover the unswizzled scratch address; Differential Revision: https://reviews.llvm.org/D89170
2020-10-22 05:27:03 +08:00
; RUN: llc -march=amdgcn -mtriple=amdgcn-- -mcpu=gfx1010 -mattr=-flat-for-global -amdgpu-use-divergent-register-indexing -verify-machineinstrs < %s | FileCheck --check-prefixes=GCN,GFX10_W32,GFX9_10,MUBUF,GFX10_W32-MUBUF,GFX9_10-MUBUF %s
; RUN: llc -march=amdgcn -mtriple=amdgcn-- -mcpu=gfx1010 -mattr=-flat-for-global,+wavefrontsize64 -amdgpu-use-divergent-register-indexing -verify-machineinstrs < %s | FileCheck --check-prefixes=GCN,GFX10_W64,GFX9_10,MUBUF,GFX10_W64-MUBUF,GFX9_10-MUBUF %s
; RUN: llc -march=amdgcn -mtriple=amdgcn-- -mcpu=gfx900 -mattr=-flat-for-global -amdgpu-use-divergent-register-indexing -amdgpu-enable-flat-scratch -verify-machineinstrs < %s | FileCheck --check-prefixes=GCN,GFX9,GFX9_10,FLATSCR,GFX9-FLATSCR %s
; RUN: llc -march=amdgcn -mtriple=amdgcn-- -mcpu=gfx1030 -mattr=-flat-for-global -amdgpu-use-divergent-register-indexing -amdgpu-enable-flat-scratch -verify-machineinstrs < %s | FileCheck --check-prefixes=GCN,GFX10_W32,GFX9_10,FLATSCR,GFX10-FLATSCR,GFX9_10-FLATSCR %s
AMDGPU: Use an ABS32_LO relocation for SCRATCH_RSRC_DWORD1 Summary: Using HI here makes no logical sense, since the dword is only 32 bits to begin with. Current Mesa master does not look at the relocation type at all, so this change is fine. Future Mesa will rely on this, however. Change-Id: I91085707834c4ac0370926602b93c94b90e44cb1 Reviewers: arsenm, rampitec, mareko Subscribers: kzhuravl, jvesely, wdng, yaxunl, dstuttard, tpr, t-tye, llvm-commits Differential Revision: https://reviews.llvm.org/D55369 llvm-svn: 349620
2018-12-19 19:55:03 +08:00
; RELS: R_AMDGPU_ABS32_LO SCRATCH_RSRC_DWORD0 0x0
; RELS: R_AMDGPU_ABS32_LO SCRATCH_RSRC_DWORD1 0x0
AMDGPU: GFX9 GS and HS shaders always have the scratch wave offset in SGPR5 Reviewers: arsenm, nhaehnle Subscribers: kzhuravl, wdng, yaxunl, dstuttard, tpr, t-tye, llvm-commits Differential Revision: https://reviews.llvm.org/D32645 llvm-svn: 302200
2017-05-05 06:25:20 +08:00
; This used to fail due to a v_add_i32 instruction with an illegal immediate
; operand that was created during Local Stack Slot Allocation. Test case derived
; from https://bugs.freedesktop.org/show_bug.cgi?id=96602
;
; GCN-LABEL: {{^}}ps_main:
[AMDGPU] Use flat scratch instructions where available The support is disabled by default. So far there is instruction selection, spilling, and frame elimination. It also changes SP from unswizzled to swizzled as used by flat scratch instructions, so it cannot be mixed with MUBUF stack access. At the very least missing: - GlobalISel; - Some optimizations in frame elimination in between vector and scalar ALU; - It shall finally allow to always materialize frame index as an SGPR, but that is not implemented and frame elimination cannot handle it yet; - Unaligned and/or multidword flat scratch shall work, but it is legalized now for MUBUF; - Operand folding cannot optimize FI like with MUBUF yet; - It will need scaling the value of the SP/FP in the DWARF expression to recover the unswizzled scratch address; Differential Revision: https://reviews.llvm.org/D89170
2020-10-22 05:27:03 +08:00
; GFX9-FLATSCR: s_add_u32 flat_scratch_lo, s0, s2
; GFX9-FLATSCR: s_addc_u32 flat_scratch_hi, s1, 0
; GFX10-FLATSCR: s_add_u32 s0, s0, s2
; GFX10-FLATSCR: s_addc_u32 s1, s1, 0
; GFX10-FLATSCR: s_setreg_b32 hwreg(HW_REG_FLAT_SCR_LO), s0
; GFX10-FLATSCR: s_setreg_b32 hwreg(HW_REG_FLAT_SCR_HI), s1
; MUBUF-DAG: s_mov_b32 s0, SCRATCH_RSRC_DWORD0
; MUBUF-DAG: s_mov_b32 s1, SCRATCH_RSRC_DWORD1
; MUBUF-DAG: s_mov_b32 s2, -1
For PAL, make sure Scratch Buffer Descriptor do not clobber GIT pointer Since SRSRC has alignment requirements, first find non GIT pointer clobbered registers for SRSRC and then if those registers clobber preloaded Scratch Wave Offset register, copy the Scratch Wave Offset register to a free SGPR.
2020-05-01 02:25:24 +08:00
; SI-DAG: s_mov_b32 s3, 0xe8f000
; VI-DAG: s_mov_b32 s3, 0xe80000
[AMDGPU] Use flat scratch instructions where available The support is disabled by default. So far there is instruction selection, spilling, and frame elimination. It also changes SP from unswizzled to swizzled as used by flat scratch instructions, so it cannot be mixed with MUBUF stack access. At the very least missing: - GlobalISel; - Some optimizations in frame elimination in between vector and scalar ALU; - It shall finally allow to always materialize frame index as an SGPR, but that is not implemented and frame elimination cannot handle it yet; - Unaligned and/or multidword flat scratch shall work, but it is legalized now for MUBUF; - Operand folding cannot optimize FI like with MUBUF yet; - It will need scaling the value of the SP/FP in the DWARF expression to recover the unswizzled scratch address; Differential Revision: https://reviews.llvm.org/D89170
2020-10-22 05:27:03 +08:00
; GFX9-MUBUF-DAG: s_mov_b32 s3, 0xe00000
; GFX10_W32-MUBUF-DAG: s_mov_b32 s3, 0x31c16000
; GFX10_W64-MUBUF-DAG: s_mov_b32 s3, 0x31e16000
; FLATSCR-NOT: SCRATCH_RSRC_DWORD
; GFX9-FLATSCR: s_mov_b32 [[SP:[^,]+]], 0
; GFX9-FLATSCR: scratch_store_dword off, v2, [[SP]] offset:
; GFX9-FLATSCR: s_mov_b32 [[SP:[^,]+]], 0
; GFX9-FLATSCR: scratch_store_dword off, v2, [[SP]] offset:
; GFX10-FLATSCR: scratch_store_dword off, v2, off offset:
; GFX10-FLATSCR: scratch_store_dword off, v2, off offset:
AMDGPU: GFX9 GS and HS shaders always have the scratch wave offset in SGPR5 Reviewers: arsenm, nhaehnle Subscribers: kzhuravl, wdng, yaxunl, dstuttard, tpr, t-tye, llvm-commits Differential Revision: https://reviews.llvm.org/D32645 llvm-svn: 302200
2017-05-05 06:25:20 +08:00
; GCN-DAG: v_lshlrev_b32_e32 [[BYTES:v[0-9]+]], 2, v0
; GCN-DAG: v_and_b32_e32 [[CLAMP_IDX:v[0-9]+]], 0x1fc, [[BYTES]]
[AMDGPU] Fix bundle scheduling Bundles coming to scheduler considered free, i.e. zero latency. Fixed. Differential Revision: https://reviews.llvm.org/D72487
2020-01-10 06:28:49 +08:00
; GCN-NOT: s_mov_b32 s0
AMDGPU: GFX9 GS and HS shaders always have the scratch wave offset in SGPR5 Reviewers: arsenm, nhaehnle Subscribers: kzhuravl, wdng, yaxunl, dstuttard, tpr, t-tye, llvm-commits Differential Revision: https://reviews.llvm.org/D32645 llvm-svn: 302200
2017-05-05 06:25:20 +08:00
[CodeGen] Ensure callers of CreateStackTemporary use sensible alignments In two instances of CreateStackTemporary we are sometimes promoting alignments beyond the stack alignment. I have introduced a new function called getReducedAlign that will return the alignment for the broken down parts of illegal vector types. For example, on NEON a <32 x i8> type is made up of two <16 x i8> types - in this case the sensible alignment is 16 bytes, not 32. In the legalization code wherever we create stack temporaries I have started using the reduced alignments instead for illegal vector types. I added a test to CodeGen/AArch64/build-one-lane.ll that tries to insert an element into an illegal fixed vector type that involves creating a temporary stack object. Differential Revision: https://reviews.llvm.org/D80370
2020-06-02 18:16:23 +08:00
; GCN-DAG: v_add{{_|_nc_}}{{i|u}}32_e32 [[HI_OFF:v[0-9]+]],{{.*}} 0x280, [[CLAMP_IDX]]
; GCN-DAG: v_add{{_|_nc_}}{{i|u}}32_e32 [[LO_OFF:v[0-9]+]],{{.*}} {{v2|0x80}}, [[CLAMP_IDX]]
AMDGPU: GFX9 GS and HS shaders always have the scratch wave offset in SGPR5 Reviewers: arsenm, nhaehnle Subscribers: kzhuravl, wdng, yaxunl, dstuttard, tpr, t-tye, llvm-commits Differential Revision: https://reviews.llvm.org/D32645 llvm-svn: 302200
2017-05-05 06:25:20 +08:00
[AMDGPU] Use flat scratch instructions where available The support is disabled by default. So far there is instruction selection, spilling, and frame elimination. It also changes SP from unswizzled to swizzled as used by flat scratch instructions, so it cannot be mixed with MUBUF stack access. At the very least missing: - GlobalISel; - Some optimizations in frame elimination in between vector and scalar ALU; - It shall finally allow to always materialize frame index as an SGPR, but that is not implemented and frame elimination cannot handle it yet; - Unaligned and/or multidword flat scratch shall work, but it is legalized now for MUBUF; - Operand folding cannot optimize FI like with MUBUF yet; - It will need scaling the value of the SP/FP in the DWARF expression to recover the unswizzled scratch address; Differential Revision: https://reviews.llvm.org/D89170
2020-10-22 05:27:03 +08:00
; MUBUF: buffer_load_dword {{v[0-9]+}}, [[LO_OFF]], {{s\[[0-9]+:[0-9]+\]}}, 0 offen
; MUBUF: buffer_load_dword {{v[0-9]+}}, [[HI_OFF]], {{s\[[0-9]+:[0-9]+\]}}, 0 offen
; FLATSCR: scratch_load_dword {{v[0-9]+}}, [[LO_OFF]], off
; FLATSCR: scratch_load_dword {{v[0-9]+}}, [[HI_OFF]], off
AMDGPU: GFX9 GS and HS shaders always have the scratch wave offset in SGPR5 Reviewers: arsenm, nhaehnle Subscribers: kzhuravl, wdng, yaxunl, dstuttard, tpr, t-tye, llvm-commits Differential Revision: https://reviews.llvm.org/D32645 llvm-svn: 302200
2017-05-05 06:25:20 +08:00
define amdgpu_ps float @ps_main(i32 %idx) {
%v1 = extractelement <81 x float> <float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float 0x3FE41CFEA0000000, float 0xBFE7A693C0000000, float 0xBFEA477C60000000, float 0xBFEBE5DC60000000, float 0xBFEC71C720000000, float 0xBFEBE5DC60000000, float 0xBFEA477C60000000, float 0xBFE7A693C0000000, float 0xBFE41CFEA0000000, float 0x3FDF9B13E0000000, float 0x3FDF9B1380000000, float 0x3FD5C53B80000000, float 0x3FD5C53B00000000, float 0x3FC6326AC0000000, float 0x3FC63269E0000000, float 0xBEE05CEB00000000, float 0xBEE086A320000000, float 0xBFC63269E0000000, float 0xBFC6326AC0000000, float 0xBFD5C53B80000000, float 0xBFD5C53B80000000, float 0xBFDF9B13E0000000, float 0xBFDF9B1460000000, float 0xBFE41CFE80000000, float 0x3FE7A693C0000000, float 0x3FEA477C20000000, float 0x3FEBE5DC40000000, float 0x3FEC71C6E0000000, float 0x3FEBE5DC40000000, float 0x3FEA477C20000000, float 0x3FE7A693C0000000, float 0xBFE41CFE80000000>, i32 %idx
%v2 = extractelement <81 x float> <float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float 0xBFE41CFEA0000000, float 0xBFDF9B13E0000000, float 0xBFD5C53B80000000, float 0xBFC6326AC0000000, float 0x3EE0789320000000, float 0x3FC6326AC0000000, float 0x3FD5C53B80000000, float 0x3FDF9B13E0000000, float 0x3FE41CFEA0000000, float 0xBFE7A693C0000000, float 0x3FE7A693C0000000, float 0xBFEA477C20000000, float 0x3FEA477C20000000, float 0xBFEBE5DC40000000, float 0x3FEBE5DC40000000, float 0xBFEC71C720000000, float 0x3FEC71C6E0000000, float 0xBFEBE5DC60000000, float 0x3FEBE5DC40000000, float 0xBFEA477C20000000, float 0x3FEA477C20000000, float 0xBFE7A693C0000000, float 0x3FE7A69380000000, float 0xBFE41CFEA0000000, float 0xBFDF9B13E0000000, float 0xBFD5C53B80000000, float 0xBFC6326AC0000000, float 0x3EE0789320000000, float 0x3FC6326AC0000000, float 0x3FD5C53B80000000, float 0x3FDF9B13E0000000, float 0x3FE41CFE80000000>, i32 %idx
%r = fadd float %v1, %v2
ret float %r
}
; GCN-LABEL: {{^}}vs_main:
[AMDGPU] Use flat scratch instructions where available The support is disabled by default. So far there is instruction selection, spilling, and frame elimination. It also changes SP from unswizzled to swizzled as used by flat scratch instructions, so it cannot be mixed with MUBUF stack access. At the very least missing: - GlobalISel; - Some optimizations in frame elimination in between vector and scalar ALU; - It shall finally allow to always materialize frame index as an SGPR, but that is not implemented and frame elimination cannot handle it yet; - Unaligned and/or multidword flat scratch shall work, but it is legalized now for MUBUF; - Operand folding cannot optimize FI like with MUBUF yet; - It will need scaling the value of the SP/FP in the DWARF expression to recover the unswizzled scratch address; Differential Revision: https://reviews.llvm.org/D89170
2020-10-22 05:27:03 +08:00
; GFX9-FLATSCR: s_add_u32 flat_scratch_lo, s0, s2
; GFX9-FLATSCR: s_addc_u32 flat_scratch_hi, s1, 0
; GFX10-FLATSCR: s_add_u32 s0, s0, s2
; GFX10-FLATSCR: s_addc_u32 s1, s1, 0
; GFX10-FLATSCR: s_setreg_b32 hwreg(HW_REG_FLAT_SCR_LO), s0
; GFX10-FLATSCR: s_setreg_b32 hwreg(HW_REG_FLAT_SCR_HI), s1
; MUBUF-DAG: s_mov_b32 s0, SCRATCH_RSRC_DWORD0
AMDGPU: Figure out private memory regs after lowering Introduce pseudo-registers for registers needed for stack access, which are replaced during finalizeLowering. Note these pseudo-registers are currently only used for the used register location, and not for determining their input argument register. This is better because it avoids the need to try to predict whether a call will be emitted from the IR, and also detects stack objects introduced by legalization. Test changes are from the HasStackObjects check being more accurate since stack objects introduced during legalization are now known. llvm-svn: 308325
2017-07-19 00:44:56 +08:00
; GCN-NOT: s_mov_b32 s0
[AMDGPU] Use flat scratch instructions where available The support is disabled by default. So far there is instruction selection, spilling, and frame elimination. It also changes SP from unswizzled to swizzled as used by flat scratch instructions, so it cannot be mixed with MUBUF stack access. At the very least missing: - GlobalISel; - Some optimizations in frame elimination in between vector and scalar ALU; - It shall finally allow to always materialize frame index as an SGPR, but that is not implemented and frame elimination cannot handle it yet; - Unaligned and/or multidword flat scratch shall work, but it is legalized now for MUBUF; - Operand folding cannot optimize FI like with MUBUF yet; - It will need scaling the value of the SP/FP in the DWARF expression to recover the unswizzled scratch address; Differential Revision: https://reviews.llvm.org/D89170
2020-10-22 05:27:03 +08:00
; FLATSCR-NOT: SCRATCH_RSRC_DWORD
; MUBUF: buffer_load_dword {{v[0-9]+}}, {{v[0-9]+}}, {{s\[[0-9]+:[0-9]+\]}}, 0 offen
; MUBUF: buffer_load_dword {{v[0-9]+}}, {{v[0-9]+}}, {{s\[[0-9]+:[0-9]+\]}}, 0 offen
; GFX9-FLATSCR: s_mov_b32 [[SP:[^,]+]], 0
; GFX9-FLATSCR: scratch_store_dword off, v2, [[SP]] offset:
; GFX9-FLATSCR: s_mov_b32 [[SP:[^,]+]], 0
; GFX9-FLATSCR: scratch_store_dword off, v2, [[SP]] offset:
; FLATSCR: scratch_load_dword {{v[0-9]+}}, {{v[0-9]+}}, off
; FLATSCR: scratch_load_dword {{v[0-9]+}}, {{v[0-9]+}}, off
AMDGPU: GFX9 GS and HS shaders always have the scratch wave offset in SGPR5 Reviewers: arsenm, nhaehnle Subscribers: kzhuravl, wdng, yaxunl, dstuttard, tpr, t-tye, llvm-commits Differential Revision: https://reviews.llvm.org/D32645 llvm-svn: 302200
2017-05-05 06:25:20 +08:00
define amdgpu_vs float @vs_main(i32 %idx) {
%v1 = extractelement <81 x float> <float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float 0x3FE41CFEA0000000, float 0xBFE7A693C0000000, float 0xBFEA477C60000000, float 0xBFEBE5DC60000000, float 0xBFEC71C720000000, float 0xBFEBE5DC60000000, float 0xBFEA477C60000000, float 0xBFE7A693C0000000, float 0xBFE41CFEA0000000, float 0x3FDF9B13E0000000, float 0x3FDF9B1380000000, float 0x3FD5C53B80000000, float 0x3FD5C53B00000000, float 0x3FC6326AC0000000, float 0x3FC63269E0000000, float 0xBEE05CEB00000000, float 0xBEE086A320000000, float 0xBFC63269E0000000, float 0xBFC6326AC0000000, float 0xBFD5C53B80000000, float 0xBFD5C53B80000000, float 0xBFDF9B13E0000000, float 0xBFDF9B1460000000, float 0xBFE41CFE80000000, float 0x3FE7A693C0000000, float 0x3FEA477C20000000, float 0x3FEBE5DC40000000, float 0x3FEC71C6E0000000, float 0x3FEBE5DC40000000, float 0x3FEA477C20000000, float 0x3FE7A693C0000000, float 0xBFE41CFE80000000>, i32 %idx
%v2 = extractelement <81 x float> <float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float 0xBFE41CFEA0000000, float 0xBFDF9B13E0000000, float 0xBFD5C53B80000000, float 0xBFC6326AC0000000, float 0x3EE0789320000000, float 0x3FC6326AC0000000, float 0x3FD5C53B80000000, float 0x3FDF9B13E0000000, float 0x3FE41CFEA0000000, float 0xBFE7A693C0000000, float 0x3FE7A693C0000000, float 0xBFEA477C20000000, float 0x3FEA477C20000000, float 0xBFEBE5DC40000000, float 0x3FEBE5DC40000000, float 0xBFEC71C720000000, float 0x3FEC71C6E0000000, float 0xBFEBE5DC60000000, float 0x3FEBE5DC40000000, float 0xBFEA477C20000000, float 0x3FEA477C20000000, float 0xBFE7A693C0000000, float 0x3FE7A69380000000, float 0xBFE41CFEA0000000, float 0xBFDF9B13E0000000, float 0xBFD5C53B80000000, float 0xBFC6326AC0000000, float 0x3EE0789320000000, float 0x3FC6326AC0000000, float 0x3FD5C53B80000000, float 0x3FDF9B13E0000000, float 0x3FE41CFE80000000>, i32 %idx
%r = fadd float %v1, %v2
ret float %r
}
; GCN-LABEL: {{^}}cs_main:
[AMDGPU] Use flat scratch instructions where available The support is disabled by default. So far there is instruction selection, spilling, and frame elimination. It also changes SP from unswizzled to swizzled as used by flat scratch instructions, so it cannot be mixed with MUBUF stack access. At the very least missing: - GlobalISel; - Some optimizations in frame elimination in between vector and scalar ALU; - It shall finally allow to always materialize frame index as an SGPR, but that is not implemented and frame elimination cannot handle it yet; - Unaligned and/or multidword flat scratch shall work, but it is legalized now for MUBUF; - Operand folding cannot optimize FI like with MUBUF yet; - It will need scaling the value of the SP/FP in the DWARF expression to recover the unswizzled scratch address; Differential Revision: https://reviews.llvm.org/D89170
2020-10-22 05:27:03 +08:00
; GFX9-FLATSCR: s_add_u32 flat_scratch_lo, s0, s2
; GFX9-FLATSCR: s_addc_u32 flat_scratch_hi, s1, 0
; GFX10-FLATSCR: s_add_u32 s0, s0, s2
; GFX10-FLATSCR: s_addc_u32 s1, s1, 0
; GFX10-FLATSCR: s_setreg_b32 hwreg(HW_REG_FLAT_SCR_LO), s0
; GFX10-FLATSCR: s_setreg_b32 hwreg(HW_REG_FLAT_SCR_HI), s1
; MUBUF-DAG: s_mov_b32 s0, SCRATCH_RSRC_DWORD0
; FLATSCR-NOT: SCRATCH_RSRC_DWORD
; MUBUF: buffer_load_dword {{v[0-9]+}}, {{v[0-9]+}}, {{s\[[0-9]+:[0-9]+\]}}, 0 offen
; MUBUF: buffer_load_dword {{v[0-9]+}}, {{v[0-9]+}}, {{s\[[0-9]+:[0-9]+\]}}, 0 offen
; FLATSCR: scratch_load_dword {{v[0-9]+}}, {{v[0-9]+}}, off
; FLATSCR: scratch_load_dword {{v[0-9]+}}, {{v[0-9]+}}, off
AMDGPU: GFX9 GS and HS shaders always have the scratch wave offset in SGPR5 Reviewers: arsenm, nhaehnle Subscribers: kzhuravl, wdng, yaxunl, dstuttard, tpr, t-tye, llvm-commits Differential Revision: https://reviews.llvm.org/D32645 llvm-svn: 302200
2017-05-05 06:25:20 +08:00
define amdgpu_cs float @cs_main(i32 %idx) {
%v1 = extractelement <81 x float> <float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float 0x3FE41CFEA0000000, float 0xBFE7A693C0000000, float 0xBFEA477C60000000, float 0xBFEBE5DC60000000, float 0xBFEC71C720000000, float 0xBFEBE5DC60000000, float 0xBFEA477C60000000, float 0xBFE7A693C0000000, float 0xBFE41CFEA0000000, float 0x3FDF9B13E0000000, float 0x3FDF9B1380000000, float 0x3FD5C53B80000000, float 0x3FD5C53B00000000, float 0x3FC6326AC0000000, float 0x3FC63269E0000000, float 0xBEE05CEB00000000, float 0xBEE086A320000000, float 0xBFC63269E0000000, float 0xBFC6326AC0000000, float 0xBFD5C53B80000000, float 0xBFD5C53B80000000, float 0xBFDF9B13E0000000, float 0xBFDF9B1460000000, float 0xBFE41CFE80000000, float 0x3FE7A693C0000000, float 0x3FEA477C20000000, float 0x3FEBE5DC40000000, float 0x3FEC71C6E0000000, float 0x3FEBE5DC40000000, float 0x3FEA477C20000000, float 0x3FE7A693C0000000, float 0xBFE41CFE80000000>, i32 %idx
%v2 = extractelement <81 x float> <float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float 0xBFE41CFEA0000000, float 0xBFDF9B13E0000000, float 0xBFD5C53B80000000, float 0xBFC6326AC0000000, float 0x3EE0789320000000, float 0x3FC6326AC0000000, float 0x3FD5C53B80000000, float 0x3FDF9B13E0000000, float 0x3FE41CFEA0000000, float 0xBFE7A693C0000000, float 0x3FE7A693C0000000, float 0xBFEA477C20000000, float 0x3FEA477C20000000, float 0xBFEBE5DC40000000, float 0x3FEBE5DC40000000, float 0xBFEC71C720000000, float 0x3FEC71C6E0000000, float 0xBFEBE5DC60000000, float 0x3FEBE5DC40000000, float 0xBFEA477C20000000, float 0x3FEA477C20000000, float 0xBFE7A693C0000000, float 0x3FE7A69380000000, float 0xBFE41CFEA0000000, float 0xBFDF9B13E0000000, float 0xBFD5C53B80000000, float 0xBFC6326AC0000000, float 0x3EE0789320000000, float 0x3FC6326AC0000000, float 0x3FD5C53B80000000, float 0x3FDF9B13E0000000, float 0x3FE41CFE80000000>, i32 %idx
%r = fadd float %v1, %v2
ret float %r
}
; GCN-LABEL: {{^}}hs_main:
[AMDGPU] Use flat scratch instructions where available The support is disabled by default. So far there is instruction selection, spilling, and frame elimination. It also changes SP from unswizzled to swizzled as used by flat scratch instructions, so it cannot be mixed with MUBUF stack access. At the very least missing: - GlobalISel; - Some optimizations in frame elimination in between vector and scalar ALU; - It shall finally allow to always materialize frame index as an SGPR, but that is not implemented and frame elimination cannot handle it yet; - Unaligned and/or multidword flat scratch shall work, but it is legalized now for MUBUF; - Operand folding cannot optimize FI like with MUBUF yet; - It will need scaling the value of the SP/FP in the DWARF expression to recover the unswizzled scratch address; Differential Revision: https://reviews.llvm.org/D89170
2020-10-22 05:27:03 +08:00
; GFX9-FLATSCR: s_add_u32 flat_scratch_lo, s0, s5
; GFX9-FLATSCR: s_addc_u32 flat_scratch_hi, s1, 0
; GFX10-FLATSCR: s_add_u32 s0, s0, s5
; GFX10-FLATSCR: s_addc_u32 s1, s1, 0
; GFX10-FLATSCR: s_setreg_b32 hwreg(HW_REG_FLAT_SCR_LO), s0
; GFX10-FLATSCR: s_setreg_b32 hwreg(HW_REG_FLAT_SCR_HI), s1
For PAL, make sure Scratch Buffer Descriptor do not clobber GIT pointer Since SRSRC has alignment requirements, first find non GIT pointer clobbered registers for SRSRC and then if those registers clobber preloaded Scratch Wave Offset register, copy the Scratch Wave Offset register to a free SGPR.
2020-05-01 02:25:24 +08:00
; SIVI: s_mov_b32 s0, SCRATCH_RSRC_DWORD0
AMDGPU/GFX10: fix scratch resource descriptor Summary: The stride should depend on the wave size, not the hardware generation. Also, the 32_FLOAT format is 0x16, not 16; though that shouldn't be relevant. Change-Id: I088f93bf6708974d085d1c50967f119061da6dc6 Reviewers: arsenm, rampitec, mareko Subscribers: kzhuravl, jvesely, wdng, yaxunl, dstuttard, tpr, t-tye, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D63808 llvm-svn: 364788
2019-07-01 23:43:00 +08:00
; SIVI-NOT: s_mov_b32 s0
[AMDGPU] Add Scratch Wave Offset to Scratch Buffer Descriptor in entry functions Add the scratch wave offset to the scratch buffer descriptor (SRSrc) in the entry function prologue. This allows us to removes the scratch wave offset register from the calling convention ABI. As part of this change, allow the use of an inline constant zero for the SOffset of MUBUF instructions accessing the stack in entry functions when a frame pointer is not requested/required. Entry functions with calls still need to set up the calling convention ABI stack pointer register, and reference it in order to address arguments of called functions. The ABI stack pointer register remains unswizzled, but is now wave-relative instead of queue-relative. Non-entry functions also use an inline constant zero SOffset for wave-relative scratch access, but continue to use the stack and frame pointers as before. When the stack or frame pointer is converted to a swizzled offset it is now scaled directly, as the scratch wave offset no longer needs to be subtracted first. Update llvm/docs/AMDGPUUsage.rst to reflect these changes to the calling convention. Tags: #llvm Differential Revision: https://reviews.llvm.org/D75138
2020-01-22 06:27:57 +08:00
; SIVI: buffer_load_dword {{v[0-9]+}}, {{v[0-9]+}}, {{s\[[0-9]+:[0-9]+\]}}, 0 offen
; SIVI: buffer_load_dword {{v[0-9]+}}, {{v[0-9]+}}, {{s\[[0-9]+:[0-9]+\]}}, 0 offen
AMDGPU/GFX10: fix scratch resource descriptor Summary: The stride should depend on the wave size, not the hardware generation. Also, the 32_FLOAT format is 0x16, not 16; though that shouldn't be relevant. Change-Id: I088f93bf6708974d085d1c50967f119061da6dc6 Reviewers: arsenm, rampitec, mareko Subscribers: kzhuravl, jvesely, wdng, yaxunl, dstuttard, tpr, t-tye, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D63808 llvm-svn: 364788
2019-07-01 23:43:00 +08:00
[AMDGPU] Use flat scratch instructions where available The support is disabled by default. So far there is instruction selection, spilling, and frame elimination. It also changes SP from unswizzled to swizzled as used by flat scratch instructions, so it cannot be mixed with MUBUF stack access. At the very least missing: - GlobalISel; - Some optimizations in frame elimination in between vector and scalar ALU; - It shall finally allow to always materialize frame index as an SGPR, but that is not implemented and frame elimination cannot handle it yet; - Unaligned and/or multidword flat scratch shall work, but it is legalized now for MUBUF; - Operand folding cannot optimize FI like with MUBUF yet; - It will need scaling the value of the SP/FP in the DWARF expression to recover the unswizzled scratch address; Differential Revision: https://reviews.llvm.org/D89170
2020-10-22 05:27:03 +08:00
; GFX9_10-MUBUF: s_mov_b32 s0, SCRATCH_RSRC_DWORD0
; GFX9_10-NOT: s_mov_b32 s5
; GFX9_10-MUBUF: buffer_load_dword {{v[0-9]+}}, {{v[0-9]+}}, {{s\[[0-9]+:[0-9]+\]}}, 0 offen
; GFX9_10-MUBUF: buffer_load_dword {{v[0-9]+}}, {{v[0-9]+}}, {{s\[[0-9]+:[0-9]+\]}}, 0 offen
; FLATSCR-NOT: SCRATCH_RSRC_DWORD
; FLATSCR: scratch_load_dword {{v[0-9]+}}, {{v[0-9]+}}, off
; FLATSCR: scratch_load_dword {{v[0-9]+}}, {{v[0-9]+}}, off
AMDGPU: GFX9 GS and HS shaders always have the scratch wave offset in SGPR5 Reviewers: arsenm, nhaehnle Subscribers: kzhuravl, wdng, yaxunl, dstuttard, tpr, t-tye, llvm-commits Differential Revision: https://reviews.llvm.org/D32645 llvm-svn: 302200
2017-05-05 06:25:20 +08:00
define amdgpu_hs float @hs_main(i32 %idx) {
%v1 = extractelement <81 x float> <float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float 0x3FE41CFEA0000000, float 0xBFE7A693C0000000, float 0xBFEA477C60000000, float 0xBFEBE5DC60000000, float 0xBFEC71C720000000, float 0xBFEBE5DC60000000, float 0xBFEA477C60000000, float 0xBFE7A693C0000000, float 0xBFE41CFEA0000000, float 0x3FDF9B13E0000000, float 0x3FDF9B1380000000, float 0x3FD5C53B80000000, float 0x3FD5C53B00000000, float 0x3FC6326AC0000000, float 0x3FC63269E0000000, float 0xBEE05CEB00000000, float 0xBEE086A320000000, float 0xBFC63269E0000000, float 0xBFC6326AC0000000, float 0xBFD5C53B80000000, float 0xBFD5C53B80000000, float 0xBFDF9B13E0000000, float 0xBFDF9B1460000000, float 0xBFE41CFE80000000, float 0x3FE7A693C0000000, float 0x3FEA477C20000000, float 0x3FEBE5DC40000000, float 0x3FEC71C6E0000000, float 0x3FEBE5DC40000000, float 0x3FEA477C20000000, float 0x3FE7A693C0000000, float 0xBFE41CFE80000000>, i32 %idx
%v2 = extractelement <81 x float> <float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float 0xBFE41CFEA0000000, float 0xBFDF9B13E0000000, float 0xBFD5C53B80000000, float 0xBFC6326AC0000000, float 0x3EE0789320000000, float 0x3FC6326AC0000000, float 0x3FD5C53B80000000, float 0x3FDF9B13E0000000, float 0x3FE41CFEA0000000, float 0xBFE7A693C0000000, float 0x3FE7A693C0000000, float 0xBFEA477C20000000, float 0x3FEA477C20000000, float 0xBFEBE5DC40000000, float 0x3FEBE5DC40000000, float 0xBFEC71C720000000, float 0x3FEC71C6E0000000, float 0xBFEBE5DC60000000, float 0x3FEBE5DC40000000, float 0xBFEA477C20000000, float 0x3FEA477C20000000, float 0xBFE7A693C0000000, float 0x3FE7A69380000000, float 0xBFE41CFEA0000000, float 0xBFDF9B13E0000000, float 0xBFD5C53B80000000, float 0xBFC6326AC0000000, float 0x3EE0789320000000, float 0x3FC6326AC0000000, float 0x3FD5C53B80000000, float 0x3FDF9B13E0000000, float 0x3FE41CFE80000000>, i32 %idx
%r = fadd float %v1, %v2
ret float %r
}
; GCN-LABEL: {{^}}gs_main:
[AMDGPU] Use flat scratch instructions where available The support is disabled by default. So far there is instruction selection, spilling, and frame elimination. It also changes SP from unswizzled to swizzled as used by flat scratch instructions, so it cannot be mixed with MUBUF stack access. At the very least missing: - GlobalISel; - Some optimizations in frame elimination in between vector and scalar ALU; - It shall finally allow to always materialize frame index as an SGPR, but that is not implemented and frame elimination cannot handle it yet; - Unaligned and/or multidword flat scratch shall work, but it is legalized now for MUBUF; - Operand folding cannot optimize FI like with MUBUF yet; - It will need scaling the value of the SP/FP in the DWARF expression to recover the unswizzled scratch address; Differential Revision: https://reviews.llvm.org/D89170
2020-10-22 05:27:03 +08:00
; GFX9-FLATSCR: s_add_u32 flat_scratch_lo, s0, s5
; GFX9-FLATSCR: s_addc_u32 flat_scratch_hi, s1, 0
; GFX10-FLATSCR: s_add_u32 s0, s0, s5
; GFX10-FLATSCR: s_addc_u32 s1, s1, 0
; GFX10-FLATSCR: s_setreg_b32 hwreg(HW_REG_FLAT_SCR_LO), s0
; GFX10-FLATSCR: s_setreg_b32 hwreg(HW_REG_FLAT_SCR_HI), s1
For PAL, make sure Scratch Buffer Descriptor do not clobber GIT pointer Since SRSRC has alignment requirements, first find non GIT pointer clobbered registers for SRSRC and then if those registers clobber preloaded Scratch Wave Offset register, copy the Scratch Wave Offset register to a free SGPR.
2020-05-01 02:25:24 +08:00
; SIVI: s_mov_b32 s0, SCRATCH_RSRC_DWORD0
[AMDGPU] Add Scratch Wave Offset to Scratch Buffer Descriptor in entry functions Add the scratch wave offset to the scratch buffer descriptor (SRSrc) in the entry function prologue. This allows us to removes the scratch wave offset register from the calling convention ABI. As part of this change, allow the use of an inline constant zero for the SOffset of MUBUF instructions accessing the stack in entry functions when a frame pointer is not requested/required. Entry functions with calls still need to set up the calling convention ABI stack pointer register, and reference it in order to address arguments of called functions. The ABI stack pointer register remains unswizzled, but is now wave-relative instead of queue-relative. Non-entry functions also use an inline constant zero SOffset for wave-relative scratch access, but continue to use the stack and frame pointers as before. When the stack or frame pointer is converted to a swizzled offset it is now scaled directly, as the scratch wave offset no longer needs to be subtracted first. Update llvm/docs/AMDGPUUsage.rst to reflect these changes to the calling convention. Tags: #llvm Differential Revision: https://reviews.llvm.org/D75138
2020-01-22 06:27:57 +08:00
; SIVI: buffer_load_dword {{v[0-9]+}}, {{v[0-9]+}}, {{s\[[0-9]+:[0-9]+\]}}, 0 offen
; SIVI: buffer_load_dword {{v[0-9]+}}, {{v[0-9]+}}, {{s\[[0-9]+:[0-9]+\]}}, 0 offen
AMDGPU: Figure out private memory regs after lowering Introduce pseudo-registers for registers needed for stack access, which are replaced during finalizeLowering. Note these pseudo-registers are currently only used for the used register location, and not for determining their input argument register. This is better because it avoids the need to try to predict whether a call will be emitted from the IR, and also detects stack objects introduced by legalization. Test changes are from the HasStackObjects check being more accurate since stack objects introduced during legalization are now known. llvm-svn: 308325
2017-07-19 00:44:56 +08:00
[AMDGPU] Use flat scratch instructions where available The support is disabled by default. So far there is instruction selection, spilling, and frame elimination. It also changes SP from unswizzled to swizzled as used by flat scratch instructions, so it cannot be mixed with MUBUF stack access. At the very least missing: - GlobalISel; - Some optimizations in frame elimination in between vector and scalar ALU; - It shall finally allow to always materialize frame index as an SGPR, but that is not implemented and frame elimination cannot handle it yet; - Unaligned and/or multidword flat scratch shall work, but it is legalized now for MUBUF; - Operand folding cannot optimize FI like with MUBUF yet; - It will need scaling the value of the SP/FP in the DWARF expression to recover the unswizzled scratch address; Differential Revision: https://reviews.llvm.org/D89170
2020-10-22 05:27:03 +08:00
; GFX9_10-MUBUF: s_mov_b32 s0, SCRATCH_RSRC_DWORD0
; GFX9_10-MUBUF: buffer_load_dword {{v[0-9]+}}, {{v[0-9]+}}, {{s\[[0-9]+:[0-9]+\]}}, 0 offen
; GFX9_10-MUBUF: buffer_load_dword {{v[0-9]+}}, {{v[0-9]+}}, {{s\[[0-9]+:[0-9]+\]}}, 0 offen
; FLATSCR-NOT: SCRATCH_RSRC_DWORD
; FLATSCR: scratch_load_dword {{v[0-9]+}}, {{v[0-9]+}}, off
; FLATSCR: scratch_load_dword {{v[0-9]+}}, {{v[0-9]+}}, off
AMDGPU: GFX9 GS and HS shaders always have the scratch wave offset in SGPR5 Reviewers: arsenm, nhaehnle Subscribers: kzhuravl, wdng, yaxunl, dstuttard, tpr, t-tye, llvm-commits Differential Revision: https://reviews.llvm.org/D32645 llvm-svn: 302200
2017-05-05 06:25:20 +08:00
define amdgpu_gs float @gs_main(i32 %idx) {
%v1 = extractelement <81 x float> <float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float 0x3FE41CFEA0000000, float 0xBFE7A693C0000000, float 0xBFEA477C60000000, float 0xBFEBE5DC60000000, float 0xBFEC71C720000000, float 0xBFEBE5DC60000000, float 0xBFEA477C60000000, float 0xBFE7A693C0000000, float 0xBFE41CFEA0000000, float 0x3FDF9B13E0000000, float 0x3FDF9B1380000000, float 0x3FD5C53B80000000, float 0x3FD5C53B00000000, float 0x3FC6326AC0000000, float 0x3FC63269E0000000, float 0xBEE05CEB00000000, float 0xBEE086A320000000, float 0xBFC63269E0000000, float 0xBFC6326AC0000000, float 0xBFD5C53B80000000, float 0xBFD5C53B80000000, float 0xBFDF9B13E0000000, float 0xBFDF9B1460000000, float 0xBFE41CFE80000000, float 0x3FE7A693C0000000, float 0x3FEA477C20000000, float 0x3FEBE5DC40000000, float 0x3FEC71C6E0000000, float 0x3FEBE5DC40000000, float 0x3FEA477C20000000, float 0x3FE7A693C0000000, float 0xBFE41CFE80000000>, i32 %idx
%v2 = extractelement <81 x float> <float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float 0xBFE41CFEA0000000, float 0xBFDF9B13E0000000, float 0xBFD5C53B80000000, float 0xBFC6326AC0000000, float 0x3EE0789320000000, float 0x3FC6326AC0000000, float 0x3FD5C53B80000000, float 0x3FDF9B13E0000000, float 0x3FE41CFEA0000000, float 0xBFE7A693C0000000, float 0x3FE7A693C0000000, float 0xBFEA477C20000000, float 0x3FEA477C20000000, float 0xBFEBE5DC40000000, float 0x3FEBE5DC40000000, float 0xBFEC71C720000000, float 0x3FEC71C6E0000000, float 0xBFEBE5DC60000000, float 0x3FEBE5DC40000000, float 0xBFEA477C20000000, float 0x3FEA477C20000000, float 0xBFE7A693C0000000, float 0x3FE7A69380000000, float 0xBFE41CFEA0000000, float 0xBFDF9B13E0000000, float 0xBFD5C53B80000000, float 0xBFC6326AC0000000, float 0x3EE0789320000000, float 0x3FC6326AC0000000, float 0x3FD5C53B80000000, float 0x3FDF9B13E0000000, float 0x3FE41CFE80000000>, i32 %idx
%r = fadd float %v1, %v2
ret float %r
}
[AMDGPU] Add Scratch Wave Offset to Scratch Buffer Descriptor in entry functions Add the scratch wave offset to the scratch buffer descriptor (SRSrc) in the entry function prologue. This allows us to removes the scratch wave offset register from the calling convention ABI. As part of this change, allow the use of an inline constant zero for the SOffset of MUBUF instructions accessing the stack in entry functions when a frame pointer is not requested/required. Entry functions with calls still need to set up the calling convention ABI stack pointer register, and reference it in order to address arguments of called functions. The ABI stack pointer register remains unswizzled, but is now wave-relative instead of queue-relative. Non-entry functions also use an inline constant zero SOffset for wave-relative scratch access, but continue to use the stack and frame pointers as before. When the stack or frame pointer is converted to a swizzled offset it is now scaled directly, as the scratch wave offset no longer needs to be subtracted first. Update llvm/docs/AMDGPUUsage.rst to reflect these changes to the calling convention. Tags: #llvm Differential Revision: https://reviews.llvm.org/D75138
2020-01-22 06:27:57 +08:00
; Mesa GS and HS shaders have the preloaded scratch wave offset SGPR fixed at
; SGPR5, and the inreg implementation is used to reference it in the IR. The
; following tests confirm the shader and anything inserted after the return
; (i.e. SI_RETURN_TO_EPILOG) can access the scratch wave offset.
AMDGPU: GFX9 GS and HS shaders always have the scratch wave offset in SGPR5 Reviewers: arsenm, nhaehnle Subscribers: kzhuravl, wdng, yaxunl, dstuttard, tpr, t-tye, llvm-commits Differential Revision: https://reviews.llvm.org/D32645 llvm-svn: 302200
2017-05-05 06:25:20 +08:00
; GCN-LABEL: {{^}}hs_ir_uses_scratch_offset:
[AMDGPU] Use flat scratch instructions where available The support is disabled by default. So far there is instruction selection, spilling, and frame elimination. It also changes SP from unswizzled to swizzled as used by flat scratch instructions, so it cannot be mixed with MUBUF stack access. At the very least missing: - GlobalISel; - Some optimizations in frame elimination in between vector and scalar ALU; - It shall finally allow to always materialize frame index as an SGPR, but that is not implemented and frame elimination cannot handle it yet; - Unaligned and/or multidword flat scratch shall work, but it is legalized now for MUBUF; - Operand folding cannot optimize FI like with MUBUF yet; - It will need scaling the value of the SP/FP in the DWARF expression to recover the unswizzled scratch address; Differential Revision: https://reviews.llvm.org/D89170
2020-10-22 05:27:03 +08:00
; GFX9-FLATSCR: s_add_u32 flat_scratch_lo, s0, s5
; GFX9-FLATSCR: s_addc_u32 flat_scratch_hi, s1, 0
; GFX10-FLATSCR: s_add_u32 s0, s0, s5
; GFX10-FLATSCR: s_addc_u32 s1, s1, 0
; GFX10-FLATSCR: s_setreg_b32 hwreg(HW_REG_FLAT_SCR_LO), s0
; GFX10-FLATSCR: s_setreg_b32 hwreg(HW_REG_FLAT_SCR_HI), s1
; MUBUF: s_mov_b32 s8, SCRATCH_RSRC_DWORD0
; FLATSCR-NOT: SCRATCH_RSRC_DWORD
AMDGPU: Figure out private memory regs after lowering Introduce pseudo-registers for registers needed for stack access, which are replaced during finalizeLowering. Note these pseudo-registers are currently only used for the used register location, and not for determining their input argument register. This is better because it avoids the need to try to predict whether a call will be emitted from the IR, and also detects stack objects introduced by legalization. Test changes are from the HasStackObjects check being more accurate since stack objects introduced during legalization are now known. llvm-svn: 308325
2017-07-19 00:44:56 +08:00
AMDGPU/GFX10: fix scratch resource descriptor Summary: The stride should depend on the wave size, not the hardware generation. Also, the 32_FLOAT format is 0x16, not 16; though that shouldn't be relevant. Change-Id: I088f93bf6708974d085d1c50967f119061da6dc6 Reviewers: arsenm, rampitec, mareko Subscribers: kzhuravl, jvesely, wdng, yaxunl, dstuttard, tpr, t-tye, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D63808 llvm-svn: 364788
2019-07-01 23:43:00 +08:00
; SIVI-NOT: s_mov_b32 s6
[AMDGPU] Add Scratch Wave Offset to Scratch Buffer Descriptor in entry functions Add the scratch wave offset to the scratch buffer descriptor (SRSrc) in the entry function prologue. This allows us to removes the scratch wave offset register from the calling convention ABI. As part of this change, allow the use of an inline constant zero for the SOffset of MUBUF instructions accessing the stack in entry functions when a frame pointer is not requested/required. Entry functions with calls still need to set up the calling convention ABI stack pointer register, and reference it in order to address arguments of called functions. The ABI stack pointer register remains unswizzled, but is now wave-relative instead of queue-relative. Non-entry functions also use an inline constant zero SOffset for wave-relative scratch access, but continue to use the stack and frame pointers as before. When the stack or frame pointer is converted to a swizzled offset it is now scaled directly, as the scratch wave offset no longer needs to be subtracted first. Update llvm/docs/AMDGPUUsage.rst to reflect these changes to the calling convention. Tags: #llvm Differential Revision: https://reviews.llvm.org/D75138
2020-01-22 06:27:57 +08:00
; SIVI: buffer_load_dword {{v[0-9]+}}, {{v[0-9]+}}, {{s\[[0-9]+:[0-9]+\]}}, 0 offen
; SIVI: buffer_load_dword {{v[0-9]+}}, {{v[0-9]+}}, {{s\[[0-9]+:[0-9]+\]}}, 0 offen
AMDGPU: Figure out private memory regs after lowering Introduce pseudo-registers for registers needed for stack access, which are replaced during finalizeLowering. Note these pseudo-registers are currently only used for the used register location, and not for determining their input argument register. This is better because it avoids the need to try to predict whether a call will be emitted from the IR, and also detects stack objects introduced by legalization. Test changes are from the HasStackObjects check being more accurate since stack objects introduced during legalization are now known. llvm-svn: 308325
2017-07-19 00:44:56 +08:00
AMDGPU/GFX10: fix scratch resource descriptor Summary: The stride should depend on the wave size, not the hardware generation. Also, the 32_FLOAT format is 0x16, not 16; though that shouldn't be relevant. Change-Id: I088f93bf6708974d085d1c50967f119061da6dc6 Reviewers: arsenm, rampitec, mareko Subscribers: kzhuravl, jvesely, wdng, yaxunl, dstuttard, tpr, t-tye, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D63808 llvm-svn: 364788
2019-07-01 23:43:00 +08:00
; GFX9_10-NOT: s_mov_b32 s5
[AMDGPU] Use flat scratch instructions where available The support is disabled by default. So far there is instruction selection, spilling, and frame elimination. It also changes SP from unswizzled to swizzled as used by flat scratch instructions, so it cannot be mixed with MUBUF stack access. At the very least missing: - GlobalISel; - Some optimizations in frame elimination in between vector and scalar ALU; - It shall finally allow to always materialize frame index as an SGPR, but that is not implemented and frame elimination cannot handle it yet; - Unaligned and/or multidword flat scratch shall work, but it is legalized now for MUBUF; - Operand folding cannot optimize FI like with MUBUF yet; - It will need scaling the value of the SP/FP in the DWARF expression to recover the unswizzled scratch address; Differential Revision: https://reviews.llvm.org/D89170
2020-10-22 05:27:03 +08:00
; GFX9_10-MUBUF-DAG: buffer_load_dword {{v[0-9]+}}, {{v[0-9]+}}, {{s\[[0-9]+:[0-9]+\]}}, 0 offen
; GFX9_10-MUBUF-DAG: buffer_load_dword {{v[0-9]+}}, {{v[0-9]+}}, {{s\[[0-9]+:[0-9]+\]}}, 0 offen
; MUBUF-DAG: s_mov_b32 s2, s5
AMDGPU: Figure out private memory regs after lowering Introduce pseudo-registers for registers needed for stack access, which are replaced during finalizeLowering. Note these pseudo-registers are currently only used for the used register location, and not for determining their input argument register. This is better because it avoids the need to try to predict whether a call will be emitted from the IR, and also detects stack objects introduced by legalization. Test changes are from the HasStackObjects check being more accurate since stack objects introduced during legalization are now known. llvm-svn: 308325
2017-07-19 00:44:56 +08:00
[AMDGPU] Use flat scratch instructions where available The support is disabled by default. So far there is instruction selection, spilling, and frame elimination. It also changes SP from unswizzled to swizzled as used by flat scratch instructions, so it cannot be mixed with MUBUF stack access. At the very least missing: - GlobalISel; - Some optimizations in frame elimination in between vector and scalar ALU; - It shall finally allow to always materialize frame index as an SGPR, but that is not implemented and frame elimination cannot handle it yet; - Unaligned and/or multidword flat scratch shall work, but it is legalized now for MUBUF; - Operand folding cannot optimize FI like with MUBUF yet; - It will need scaling the value of the SP/FP in the DWARF expression to recover the unswizzled scratch address; Differential Revision: https://reviews.llvm.org/D89170
2020-10-22 05:27:03 +08:00
; FLATSCR-DAG: scratch_load_dword {{v[0-9]+}}, {{v[0-9]+}}, off
; FLATSCR-DAG: scratch_load_dword {{v[0-9]+}}, {{v[0-9]+}}, off
AMDGPU: GFX9 GS and HS shaders always have the scratch wave offset in SGPR5 Reviewers: arsenm, nhaehnle Subscribers: kzhuravl, wdng, yaxunl, dstuttard, tpr, t-tye, llvm-commits Differential Revision: https://reviews.llvm.org/D32645 llvm-svn: 302200
2017-05-05 06:25:20 +08:00
define amdgpu_hs <{i32, i32, i32, float}> @hs_ir_uses_scratch_offset(i32 inreg, i32 inreg, i32 inreg, i32 inreg, i32 inreg, i32 inreg %swo, i32 %idx) {
%v1 = extractelement <81 x float> <float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float 0x3FE41CFEA0000000, float 0xBFE7A693C0000000, float 0xBFEA477C60000000, float 0xBFEBE5DC60000000, float 0xBFEC71C720000000, float 0xBFEBE5DC60000000, float 0xBFEA477C60000000, float 0xBFE7A693C0000000, float 0xBFE41CFEA0000000, float 0x3FDF9B13E0000000, float 0x3FDF9B1380000000, float 0x3FD5C53B80000000, float 0x3FD5C53B00000000, float 0x3FC6326AC0000000, float 0x3FC63269E0000000, float 0xBEE05CEB00000000, float 0xBEE086A320000000, float 0xBFC63269E0000000, float 0xBFC6326AC0000000, float 0xBFD5C53B80000000, float 0xBFD5C53B80000000, float 0xBFDF9B13E0000000, float 0xBFDF9B1460000000, float 0xBFE41CFE80000000, float 0x3FE7A693C0000000, float 0x3FEA477C20000000, float 0x3FEBE5DC40000000, float 0x3FEC71C6E0000000, float 0x3FEBE5DC40000000, float 0x3FEA477C20000000, float 0x3FE7A693C0000000, float 0xBFE41CFE80000000>, i32 %idx
%v2 = extractelement <81 x float> <float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float 0xBFE41CFEA0000000, float 0xBFDF9B13E0000000, float 0xBFD5C53B80000000, float 0xBFC6326AC0000000, float 0x3EE0789320000000, float 0x3FC6326AC0000000, float 0x3FD5C53B80000000, float 0x3FDF9B13E0000000, float 0x3FE41CFEA0000000, float 0xBFE7A693C0000000, float 0x3FE7A693C0000000, float 0xBFEA477C20000000, float 0x3FEA477C20000000, float 0xBFEBE5DC40000000, float 0x3FEBE5DC40000000, float 0xBFEC71C720000000, float 0x3FEC71C6E0000000, float 0xBFEBE5DC60000000, float 0x3FEBE5DC40000000, float 0xBFEA477C20000000, float 0x3FEA477C20000000, float 0xBFE7A693C0000000, float 0x3FE7A69380000000, float 0xBFE41CFEA0000000, float 0xBFDF9B13E0000000, float 0xBFD5C53B80000000, float 0xBFC6326AC0000000, float 0x3EE0789320000000, float 0x3FC6326AC0000000, float 0x3FD5C53B80000000, float 0x3FDF9B13E0000000, float 0x3FE41CFE80000000>, i32 %idx
%f = fadd float %v1, %v2
%r1 = insertvalue <{i32, i32, i32, float}> undef, i32 %swo, 2
%r2 = insertvalue <{i32, i32, i32, float}> %r1, float %f, 3
ret <{i32, i32, i32, float}> %r2
}
; GCN-LABEL: {{^}}gs_ir_uses_scratch_offset:
[AMDGPU] Use flat scratch instructions where available The support is disabled by default. So far there is instruction selection, spilling, and frame elimination. It also changes SP from unswizzled to swizzled as used by flat scratch instructions, so it cannot be mixed with MUBUF stack access. At the very least missing: - GlobalISel; - Some optimizations in frame elimination in between vector and scalar ALU; - It shall finally allow to always materialize frame index as an SGPR, but that is not implemented and frame elimination cannot handle it yet; - Unaligned and/or multidword flat scratch shall work, but it is legalized now for MUBUF; - Operand folding cannot optimize FI like with MUBUF yet; - It will need scaling the value of the SP/FP in the DWARF expression to recover the unswizzled scratch address; Differential Revision: https://reviews.llvm.org/D89170
2020-10-22 05:27:03 +08:00
; GFX9-FLATSCR: s_add_u32 flat_scratch_lo, s0, s5
; GFX9-FLATSCR: s_addc_u32 flat_scratch_hi, s1, 0
; GFX10-FLATSCR: s_add_u32 s0, s0, s5
; GFX10-FLATSCR: s_addc_u32 s1, s1, 0
; GFX10-FLATSCR: s_setreg_b32 hwreg(HW_REG_FLAT_SCR_LO), s0
; GFX10-FLATSCR: s_setreg_b32 hwreg(HW_REG_FLAT_SCR_HI), s1
; MUBUF: s_mov_b32 s8, SCRATCH_RSRC_DWORD0
; FLATSCR-NOT: SCRATCH_RSRC_DWORD
AMDGPU: Figure out private memory regs after lowering Introduce pseudo-registers for registers needed for stack access, which are replaced during finalizeLowering. Note these pseudo-registers are currently only used for the used register location, and not for determining their input argument register. This is better because it avoids the need to try to predict whether a call will be emitted from the IR, and also detects stack objects introduced by legalization. Test changes are from the HasStackObjects check being more accurate since stack objects introduced during legalization are now known. llvm-svn: 308325
2017-07-19 00:44:56 +08:00
[AMDGPU] Add Scratch Wave Offset to Scratch Buffer Descriptor in entry functions Add the scratch wave offset to the scratch buffer descriptor (SRSrc) in the entry function prologue. This allows us to removes the scratch wave offset register from the calling convention ABI. As part of this change, allow the use of an inline constant zero for the SOffset of MUBUF instructions accessing the stack in entry functions when a frame pointer is not requested/required. Entry functions with calls still need to set up the calling convention ABI stack pointer register, and reference it in order to address arguments of called functions. The ABI stack pointer register remains unswizzled, but is now wave-relative instead of queue-relative. Non-entry functions also use an inline constant zero SOffset for wave-relative scratch access, but continue to use the stack and frame pointers as before. When the stack or frame pointer is converted to a swizzled offset it is now scaled directly, as the scratch wave offset no longer needs to be subtracted first. Update llvm/docs/AMDGPUUsage.rst to reflect these changes to the calling convention. Tags: #llvm Differential Revision: https://reviews.llvm.org/D75138
2020-01-22 06:27:57 +08:00
; SIVI: buffer_load_dword {{v[0-9]+}}, {{v[0-9]+}}, {{s\[[0-9]+:[0-9]+\]}}, 0 offen
; SIVI: buffer_load_dword {{v[0-9]+}}, {{v[0-9]+}}, {{s\[[0-9]+:[0-9]+\]}}, 0 offen
AMDGPU: Figure out private memory regs after lowering Introduce pseudo-registers for registers needed for stack access, which are replaced during finalizeLowering. Note these pseudo-registers are currently only used for the used register location, and not for determining their input argument register. This is better because it avoids the need to try to predict whether a call will be emitted from the IR, and also detects stack objects introduced by legalization. Test changes are from the HasStackObjects check being more accurate since stack objects introduced during legalization are now known. llvm-svn: 308325
2017-07-19 00:44:56 +08:00
[AMDGPU] Use flat scratch instructions where available The support is disabled by default. So far there is instruction selection, spilling, and frame elimination. It also changes SP from unswizzled to swizzled as used by flat scratch instructions, so it cannot be mixed with MUBUF stack access. At the very least missing: - GlobalISel; - Some optimizations in frame elimination in between vector and scalar ALU; - It shall finally allow to always materialize frame index as an SGPR, but that is not implemented and frame elimination cannot handle it yet; - Unaligned and/or multidword flat scratch shall work, but it is legalized now for MUBUF; - Operand folding cannot optimize FI like with MUBUF yet; - It will need scaling the value of the SP/FP in the DWARF expression to recover the unswizzled scratch address; Differential Revision: https://reviews.llvm.org/D89170
2020-10-22 05:27:03 +08:00
; GFX9_10-MUBUF-DAG: buffer_load_dword {{v[0-9]+}}, {{v[0-9]+}}, {{s\[[0-9]+:[0-9]+\]}}, 0 offen
; GFX9_10-MUBUF-DAG: buffer_load_dword {{v[0-9]+}}, {{v[0-9]+}}, {{s\[[0-9]+:[0-9]+\]}}, 0 offen
; MUBUF-DAG: s_mov_b32 s2, s5
AMDGPU: Figure out private memory regs after lowering Introduce pseudo-registers for registers needed for stack access, which are replaced during finalizeLowering. Note these pseudo-registers are currently only used for the used register location, and not for determining their input argument register. This is better because it avoids the need to try to predict whether a call will be emitted from the IR, and also detects stack objects introduced by legalization. Test changes are from the HasStackObjects check being more accurate since stack objects introduced during legalization are now known. llvm-svn: 308325
2017-07-19 00:44:56 +08:00
[AMDGPU] Use flat scratch instructions where available The support is disabled by default. So far there is instruction selection, spilling, and frame elimination. It also changes SP from unswizzled to swizzled as used by flat scratch instructions, so it cannot be mixed with MUBUF stack access. At the very least missing: - GlobalISel; - Some optimizations in frame elimination in between vector and scalar ALU; - It shall finally allow to always materialize frame index as an SGPR, but that is not implemented and frame elimination cannot handle it yet; - Unaligned and/or multidword flat scratch shall work, but it is legalized now for MUBUF; - Operand folding cannot optimize FI like with MUBUF yet; - It will need scaling the value of the SP/FP in the DWARF expression to recover the unswizzled scratch address; Differential Revision: https://reviews.llvm.org/D89170
2020-10-22 05:27:03 +08:00
; FLATSCR-DAG: scratch_load_dword {{v[0-9]+}}, {{v[0-9]+}}, off
; FLATSCR-DAG: scratch_load_dword {{v[0-9]+}}, {{v[0-9]+}}, off
AMDGPU: GFX9 GS and HS shaders always have the scratch wave offset in SGPR5 Reviewers: arsenm, nhaehnle Subscribers: kzhuravl, wdng, yaxunl, dstuttard, tpr, t-tye, llvm-commits Differential Revision: https://reviews.llvm.org/D32645 llvm-svn: 302200
2017-05-05 06:25:20 +08:00
define amdgpu_gs <{i32, i32, i32, float}> @gs_ir_uses_scratch_offset(i32 inreg, i32 inreg, i32 inreg, i32 inreg, i32 inreg, i32 inreg %swo, i32 %idx) {
%v1 = extractelement <81 x float> <float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float 0x3FE41CFEA0000000, float 0xBFE7A693C0000000, float 0xBFEA477C60000000, float 0xBFEBE5DC60000000, float 0xBFEC71C720000000, float 0xBFEBE5DC60000000, float 0xBFEA477C60000000, float 0xBFE7A693C0000000, float 0xBFE41CFEA0000000, float 0x3FDF9B13E0000000, float 0x3FDF9B1380000000, float 0x3FD5C53B80000000, float 0x3FD5C53B00000000, float 0x3FC6326AC0000000, float 0x3FC63269E0000000, float 0xBEE05CEB00000000, float 0xBEE086A320000000, float 0xBFC63269E0000000, float 0xBFC6326AC0000000, float 0xBFD5C53B80000000, float 0xBFD5C53B80000000, float 0xBFDF9B13E0000000, float 0xBFDF9B1460000000, float 0xBFE41CFE80000000, float 0x3FE7A693C0000000, float 0x3FEA477C20000000, float 0x3FEBE5DC40000000, float 0x3FEC71C6E0000000, float 0x3FEBE5DC40000000, float 0x3FEA477C20000000, float 0x3FE7A693C0000000, float 0xBFE41CFE80000000>, i32 %idx
%v2 = extractelement <81 x float> <float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float undef, float 0xBFE41CFEA0000000, float 0xBFDF9B13E0000000, float 0xBFD5C53B80000000, float 0xBFC6326AC0000000, float 0x3EE0789320000000, float 0x3FC6326AC0000000, float 0x3FD5C53B80000000, float 0x3FDF9B13E0000000, float 0x3FE41CFEA0000000, float 0xBFE7A693C0000000, float 0x3FE7A693C0000000, float 0xBFEA477C20000000, float 0x3FEA477C20000000, float 0xBFEBE5DC40000000, float 0x3FEBE5DC40000000, float 0xBFEC71C720000000, float 0x3FEC71C6E0000000, float 0xBFEBE5DC60000000, float 0x3FEBE5DC40000000, float 0xBFEA477C20000000, float 0x3FEA477C20000000, float 0xBFE7A693C0000000, float 0x3FE7A69380000000, float 0xBFE41CFEA0000000, float 0xBFDF9B13E0000000, float 0xBFD5C53B80000000, float 0xBFC6326AC0000000, float 0x3EE0789320000000, float 0x3FC6326AC0000000, float 0x3FD5C53B80000000, float 0x3FDF9B13E0000000, float 0x3FE41CFE80000000>, i32 %idx
%f = fadd float %v1, %v2
%r1 = insertvalue <{i32, i32, i32, float}> undef, i32 %swo, 2
%r2 = insertvalue <{i32, i32, i32, float}> %r1, float %f, 3
ret <{i32, i32, i32, float}> %r2
}