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/spill-wide-sgpr.ll

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

303 lines
10 KiB
LLVM
Raw Normal View History

[NFC][AMDGPU] Clean up some lit test prefixes Replace some instances of "ALL" with "GCN" where it applies. Committed as obvious.
2020-11-10 02:39:55 +08:00
; RUN: llc -O0 -march=amdgcn -mcpu=fiji -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VGPR %s
; RUN: llc -O0 -march=amdgcn -mcpu=fiji -amdgpu-spill-sgpr-to-vgpr=0 -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=VMEM %s
AMDGPU: Use wider scalar spills for SGPR spilling Since the spill is for the whole wave, these don't have the swizzling problems that vector stores do and a single 4-byte allocation is enough to spill a 64 element register. This should reduce the number of spill instructions and put all the spills for a register in the same cacheline. This should save allocated private size, but for now it doesn't. The extra slots are allocated for each component, but never used because the frame layout is essentially finalized before frame indices are replaced. For always using the scalar store path, this should probably be moved into processFunctionBeforeFrameFinalized. llvm-svn: 288445
2016-12-02 08:54:45 +08:00
[NFC][AMDGPU] Clean up some lit test prefixes Replace some instances of "ALL" with "GCN" where it applies. Committed as obvious.
2020-11-10 02:39:55 +08:00
; GCN-LABEL: {{^}}spill_sgpr_x2:
AMDGPU: Use wider scalar spills for SGPR spilling Since the spill is for the whole wave, these don't have the swizzling problems that vector stores do and a single 4-byte allocation is enough to spill a 64 element register. This should reduce the number of spill instructions and put all the spills for a register in the same cacheline. This should save allocated private size, but for now it doesn't. The extra slots are allocated for each component, but never used because the frame layout is essentially finalized before frame indices are replaced. For always using the scalar store path, this should probably be moved into processFunctionBeforeFrameFinalized. llvm-svn: 288445
2016-12-02 08:54:45 +08:00
; VGPR: v_writelane_b32 v{{[0-9]+}}, s{{[0-9]+}}, 0
; VGPR: v_writelane_b32 v{{[0-9]+}}, s{{[0-9]+}}, 1
; VGPR: s_cbranch_scc1
; VGPR: v_readlane_b32 s{{[0-9]+}}, v{{[0-9]+}}, 0
; VGPR: v_readlane_b32 s{{[0-9]+}}, v{{[0-9]+}}, 1
[AMDGPU] Remove -amdgpu-spill-sgpr-to-smem. Summary: The implementation was never completed and never used except in tests. Reviewers: arsenm, mareko Subscribers: qcolombet, kzhuravl, jvesely, wdng, nhaehnle, yaxunl, dstuttard, tpr, t-tye, hiraditya, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D69163 llvm-svn: 375293
2019-10-19 05:48:22 +08:00
AMDGPU: Use wider scalar spills for SGPR spilling Since the spill is for the whole wave, these don't have the swizzling problems that vector stores do and a single 4-byte allocation is enough to spill a 64 element register. This should reduce the number of spill instructions and put all the spills for a register in the same cacheline. This should save allocated private size, but for now it doesn't. The extra slots are allocated for each component, but never used because the frame layout is essentially finalized before frame indices are replaced. For always using the scalar store path, this should probably be moved into processFunctionBeforeFrameFinalized. llvm-svn: 288445
2016-12-02 08:54:45 +08:00
; VMEM: buffer_store_dword
; VMEM: s_cbranch_scc1
; VMEM: buffer_load_dword
define amdgpu_kernel void @spill_sgpr_x2(i32 addrspace(1)* %out, i32 %in) #0 {
%wide.sgpr = call <2 x i32> asm sideeffect "; def $0", "=s" () #0
%cmp = icmp eq i32 %in, 0
br i1 %cmp, label %bb0, label %ret
bb0:
call void asm sideeffect "; use $0", "s"(<2 x i32> %wide.sgpr) #0
br label %ret
ret:
ret void
}
[NFC][AMDGPU] Clean up some lit test prefixes Replace some instances of "ALL" with "GCN" where it applies. Committed as obvious.
2020-11-10 02:39:55 +08:00
; GCN-LABEL: {{^}}spill_sgpr_x3:
[AMDGPU] Support for v3i32/v3f32 Added support for dwordx3 for most load/store types, but not DS, and not intrinsics yet. SI (gfx6) does not have dwordx3 instructions, so they are not enabled there. Some of this patch is from Matt Arsenault, also of AMD. Differential Revision: https://reviews.llvm.org/D58902 Change-Id: I913ef54f1433a7149da8d72f4af54dbb13436bd9 llvm-svn: 356659
2019-03-21 20:01:21 +08:00
; VGPR: v_writelane_b32 v{{[0-9]+}}, s{{[0-9]+}}, 0
; VGPR: v_writelane_b32 v{{[0-9]+}}, s{{[0-9]+}}, 1
; VGPR: v_writelane_b32 v{{[0-9]+}}, s{{[0-9]+}}, 2
; VGPR: s_cbranch_scc1
; VGPR: v_readlane_b32 s{{[0-9]+}}, v{{[0-9]+}}, 0
; VGPR: v_readlane_b32 s{{[0-9]+}}, v{{[0-9]+}}, 1
; VGPR: v_readlane_b32 s{{[0-9]+}}, v{{[0-9]+}}, 2
; VMEM: buffer_store_dword
; VMEM: s_cbranch_scc1
; VMEM: buffer_load_dword
define amdgpu_kernel void @spill_sgpr_x3(i32 addrspace(1)* %out, i32 %in) #0 {
%wide.sgpr = call <3 x i32> asm sideeffect "; def $0", "=s" () #0
%cmp = icmp eq i32 %in, 0
br i1 %cmp, label %bb0, label %ret
bb0:
call void asm sideeffect "; use $0", "s"(<3 x i32> %wide.sgpr) #0
br label %ret
ret:
ret void
}
[NFC][AMDGPU] Clean up some lit test prefixes Replace some instances of "ALL" with "GCN" where it applies. Committed as obvious.
2020-11-10 02:39:55 +08:00
; GCN-LABEL: {{^}}spill_sgpr_x4:
AMDGPU: Use wider scalar spills for SGPR spilling Since the spill is for the whole wave, these don't have the swizzling problems that vector stores do and a single 4-byte allocation is enough to spill a 64 element register. This should reduce the number of spill instructions and put all the spills for a register in the same cacheline. This should save allocated private size, but for now it doesn't. The extra slots are allocated for each component, but never used because the frame layout is essentially finalized before frame indices are replaced. For always using the scalar store path, this should probably be moved into processFunctionBeforeFrameFinalized. llvm-svn: 288445
2016-12-02 08:54:45 +08:00
; VGPR: v_writelane_b32 v{{[0-9]+}}, s{{[0-9]+}}, 0
; VGPR: v_writelane_b32 v{{[0-9]+}}, s{{[0-9]+}}, 1
; VGPR: v_writelane_b32 v{{[0-9]+}}, s{{[0-9]+}}, 2
; VGPR: v_writelane_b32 v{{[0-9]+}}, s{{[0-9]+}}, 3
; VGPR: s_cbranch_scc1
; VGPR: v_readlane_b32 s{{[0-9]+}}, v{{[0-9]+}}, 0
; VGPR: v_readlane_b32 s{{[0-9]+}}, v{{[0-9]+}}, 1
; VGPR: v_readlane_b32 s{{[0-9]+}}, v{{[0-9]+}}, 2
; VGPR: v_readlane_b32 s{{[0-9]+}}, v{{[0-9]+}}, 3
; VMEM: buffer_store_dword
; VMEM: s_cbranch_scc1
; VMEM: buffer_load_dword
define amdgpu_kernel void @spill_sgpr_x4(i32 addrspace(1)* %out, i32 %in) #0 {
%wide.sgpr = call <4 x i32> asm sideeffect "; def $0", "=s" () #0
%cmp = icmp eq i32 %in, 0
br i1 %cmp, label %bb0, label %ret
bb0:
call void asm sideeffect "; use $0", "s"(<4 x i32> %wide.sgpr) #0
br label %ret
ret:
ret void
}
[NFC][AMDGPU] Clean up some lit test prefixes Replace some instances of "ALL" with "GCN" where it applies. Committed as obvious.
2020-11-10 02:39:55 +08:00
; GCN-LABEL: {{^}}spill_sgpr_x5:
[AMDGPU] Added v5i32 and v5f32 register classes They are not used by anything yet, but a subsequent commit will start using them for image ops that return 5 dwords. Differential Revision: https://reviews.llvm.org/D58903 Change-Id: I63e1904081e39a6d66e4eb96d51df25ad399d271 llvm-svn: 356735
2019-03-22 18:11:21 +08:00
; VGPR: v_writelane_b32 v{{[0-9]+}}, s{{[0-9]+}}, 0
; VGPR: v_writelane_b32 v{{[0-9]+}}, s{{[0-9]+}}, 1
; VGPR: v_writelane_b32 v{{[0-9]+}}, s{{[0-9]+}}, 2
; VGPR: v_writelane_b32 v{{[0-9]+}}, s{{[0-9]+}}, 3
; VGPR: v_writelane_b32 v{{[0-9]+}}, s{{[0-9]+}}, 4
; VGPR: s_cbranch_scc1
; VGPR: v_readlane_b32 s{{[0-9]+}}, v{{[0-9]+}}, 0
; VGPR: v_readlane_b32 s{{[0-9]+}}, v{{[0-9]+}}, 1
; VGPR: v_readlane_b32 s{{[0-9]+}}, v{{[0-9]+}}, 2
; VGPR: v_readlane_b32 s{{[0-9]+}}, v{{[0-9]+}}, 3
; VGPR: v_readlane_b32 s{{[0-9]+}}, v{{[0-9]+}}, 4
; VMEM: buffer_store_dword
; VMEM: s_cbranch_scc1
; VMEM: buffer_load_dword
define amdgpu_kernel void @spill_sgpr_x5(i32 addrspace(1)* %out, i32 %in) #0 {
%wide.sgpr = call <5 x i32> asm sideeffect "; def $0", "=s" () #0
%cmp = icmp eq i32 %in, 0
br i1 %cmp, label %bb0, label %ret
bb0:
call void asm sideeffect "; use $0", "s"(<5 x i32> %wide.sgpr) #0
br label %ret
ret:
ret void
}
[NFC][AMDGPU] Clean up some lit test prefixes Replace some instances of "ALL" with "GCN" where it applies. Committed as obvious.
2020-11-10 02:39:55 +08:00
; GCN-LABEL: {{^}}spill_sgpr_x8:
AMDGPU: Use wider scalar spills for SGPR spilling Since the spill is for the whole wave, these don't have the swizzling problems that vector stores do and a single 4-byte allocation is enough to spill a 64 element register. This should reduce the number of spill instructions and put all the spills for a register in the same cacheline. This should save allocated private size, but for now it doesn't. The extra slots are allocated for each component, but never used because the frame layout is essentially finalized before frame indices are replaced. For always using the scalar store path, this should probably be moved into processFunctionBeforeFrameFinalized. llvm-svn: 288445
2016-12-02 08:54:45 +08:00
; VGPR: v_writelane_b32 v{{[0-9]+}}, s{{[0-9]+}}, 0
; VGPR: v_writelane_b32 v{{[0-9]+}}, s{{[0-9]+}}, 1
; VGPR: v_writelane_b32 v{{[0-9]+}}, s{{[0-9]+}}, 2
; VGPR: v_writelane_b32 v{{[0-9]+}}, s{{[0-9]+}}, 3
; VGPR: v_writelane_b32 v{{[0-9]+}}, s{{[0-9]+}}, 4
; VGPR: v_writelane_b32 v{{[0-9]+}}, s{{[0-9]+}}, 5
; VGPR: v_writelane_b32 v{{[0-9]+}}, s{{[0-9]+}}, 6
; VGPR: v_writelane_b32 v{{[0-9]+}}, s{{[0-9]+}}, 7
; VGPR: s_cbranch_scc1
; VGPR: v_readlane_b32 s{{[0-9]+}}, v{{[0-9]+}}, 0
; VGPR: v_readlane_b32 s{{[0-9]+}}, v{{[0-9]+}}, 1
; VGPR: v_readlane_b32 s{{[0-9]+}}, v{{[0-9]+}}, 2
; VGPR: v_readlane_b32 s{{[0-9]+}}, v{{[0-9]+}}, 3
; VGPR: v_readlane_b32 s{{[0-9]+}}, v{{[0-9]+}}, 4
; VGPR: v_readlane_b32 s{{[0-9]+}}, v{{[0-9]+}}, 5
; VGPR: v_readlane_b32 s{{[0-9]+}}, v{{[0-9]+}}, 6
; VGPR: v_readlane_b32 s{{[0-9]+}}, v{{[0-9]+}}, 7
; VMEM: buffer_store_dword
; VMEM: s_cbranch_scc1
; VMEM: buffer_load_dword
define amdgpu_kernel void @spill_sgpr_x8(i32 addrspace(1)* %out, i32 %in) #0 {
%wide.sgpr = call <8 x i32> asm sideeffect "; def $0", "=s" () #0
%cmp = icmp eq i32 %in, 0
br i1 %cmp, label %bb0, label %ret
bb0:
call void asm sideeffect "; use $0", "s"(<8 x i32> %wide.sgpr) #0
br label %ret
ret:
ret void
}
[NFC][AMDGPU] Clean up some lit test prefixes Replace some instances of "ALL" with "GCN" where it applies. Committed as obvious.
2020-11-10 02:39:55 +08:00
; GCN-LABEL: {{^}}spill_sgpr_x16:
[AMDGPU] Remove assertion on S1024 SGPR to VGPR spill Summary: Replace an assertion that blocks S1024 SGPR to VGPR spill. The assertion pre-dates S1024 and is not wave size dependent. Reviewers: arsenm, sameerds, rampitec Reviewed By: arsenm Subscribers: qcolombet, kzhuravl, jvesely, wdng, nhaehnle, yaxunl, dstuttard, tpr, t-tye, hiraditya, kerbowa, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D80783
2020-05-30 10:15:39 +08:00
; VGPR: v_writelane_b32 v{{[0-9]+}}, s{{[0-9]+}}, 0
; VGPR: v_writelane_b32 v{{[0-9]+}}, s{{[0-9]+}}, 1
; VGPR: v_writelane_b32 v{{[0-9]+}}, s{{[0-9]+}}, 2
; VGPR: v_writelane_b32 v{{[0-9]+}}, s{{[0-9]+}}, 3
; VGPR: v_writelane_b32 v{{[0-9]+}}, s{{[0-9]+}}, 4
; VGPR: v_writelane_b32 v{{[0-9]+}}, s{{[0-9]+}}, 5
; VGPR: v_writelane_b32 v{{[0-9]+}}, s{{[0-9]+}}, 6
; VGPR: v_writelane_b32 v{{[0-9]+}}, s{{[0-9]+}}, 7
; VGPR: v_writelane_b32 v{{[0-9]+}}, s{{[0-9]+}}, 8
; VGPR: v_writelane_b32 v{{[0-9]+}}, s{{[0-9]+}}, 9
; VGPR: v_writelane_b32 v{{[0-9]+}}, s{{[0-9]+}}, 10
; VGPR: v_writelane_b32 v{{[0-9]+}}, s{{[0-9]+}}, 11
; VGPR: v_writelane_b32 v{{[0-9]+}}, s{{[0-9]+}}, 12
; VGPR: v_writelane_b32 v{{[0-9]+}}, s{{[0-9]+}}, 13
; VGPR: v_writelane_b32 v{{[0-9]+}}, s{{[0-9]+}}, 14
; VGPR: v_writelane_b32 v{{[0-9]+}}, s{{[0-9]+}}, 15
; VGPR: s_cbranch_scc1
; VGPR: v_readlane_b32 s{{[0-9]+}}, v{{[0-9]+}}, 0
; VGPR: v_readlane_b32 s{{[0-9]+}}, v{{[0-9]+}}, 1
; VGPR: v_readlane_b32 s{{[0-9]+}}, v{{[0-9]+}}, 2
; VGPR: v_readlane_b32 s{{[0-9]+}}, v{{[0-9]+}}, 3
; VGPR: v_readlane_b32 s{{[0-9]+}}, v{{[0-9]+}}, 4
; VGPR: v_readlane_b32 s{{[0-9]+}}, v{{[0-9]+}}, 5
; VGPR: v_readlane_b32 s{{[0-9]+}}, v{{[0-9]+}}, 6
; VGPR: v_readlane_b32 s{{[0-9]+}}, v{{[0-9]+}}, 7
; VGPR: v_readlane_b32 s{{[0-9]+}}, v{{[0-9]+}}, 8
; VGPR: v_readlane_b32 s{{[0-9]+}}, v{{[0-9]+}}, 9
; VGPR: v_readlane_b32 s{{[0-9]+}}, v{{[0-9]+}}, 10
; VGPR: v_readlane_b32 s{{[0-9]+}}, v{{[0-9]+}}, 11
; VGPR: v_readlane_b32 s{{[0-9]+}}, v{{[0-9]+}}, 12
; VGPR: v_readlane_b32 s{{[0-9]+}}, v{{[0-9]+}}, 13
; VGPR: v_readlane_b32 s{{[0-9]+}}, v{{[0-9]+}}, 14
; VGPR: v_readlane_b32 s{{[0-9]+}}, v{{[0-9]+}}, 15
; VMEM: buffer_store_dword
; VMEM: s_cbranch_scc1
; VMEM: buffer_load_dword
define amdgpu_kernel void @spill_sgpr_x16(i32 addrspace(1)* %out, i32 %in) #0 {
%wide.sgpr = call <16 x i32> asm sideeffect "; def $0", "=s" () #0
%cmp = icmp eq i32 %in, 0
br i1 %cmp, label %bb0, label %ret
bb0:
call void asm sideeffect "; use $0", "s"(<16 x i32> %wide.sgpr) #0
br label %ret
ret:
ret void
}
[NFC][AMDGPU] Clean up some lit test prefixes Replace some instances of "ALL" with "GCN" where it applies. Committed as obvious.
2020-11-10 02:39:55 +08:00
; GCN-LABEL: {{^}}spill_sgpr_x32:
[AMDGPU] Remove assertion on S1024 SGPR to VGPR spill Summary: Replace an assertion that blocks S1024 SGPR to VGPR spill. The assertion pre-dates S1024 and is not wave size dependent. Reviewers: arsenm, sameerds, rampitec Reviewed By: arsenm Subscribers: qcolombet, kzhuravl, jvesely, wdng, nhaehnle, yaxunl, dstuttard, tpr, t-tye, hiraditya, kerbowa, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D80783
2020-05-30 10:15:39 +08:00
; VGPR: v_writelane_b32 v{{[0-9]+}}, s{{[0-9]+}}, 0
; VGPR: v_writelane_b32 v{{[0-9]+}}, s{{[0-9]+}}, 1
; VGPR: v_writelane_b32 v{{[0-9]+}}, s{{[0-9]+}}, 2
; VGPR: v_writelane_b32 v{{[0-9]+}}, s{{[0-9]+}}, 3
; VGPR: v_writelane_b32 v{{[0-9]+}}, s{{[0-9]+}}, 4
; VGPR: v_writelane_b32 v{{[0-9]+}}, s{{[0-9]+}}, 5
; VGPR: v_writelane_b32 v{{[0-9]+}}, s{{[0-9]+}}, 6
; VGPR: v_writelane_b32 v{{[0-9]+}}, s{{[0-9]+}}, 7
; VGPR: v_writelane_b32 v{{[0-9]+}}, s{{[0-9]+}}, 8
; VGPR: v_writelane_b32 v{{[0-9]+}}, s{{[0-9]+}}, 9
; VGPR: v_writelane_b32 v{{[0-9]+}}, s{{[0-9]+}}, 10
; VGPR: v_writelane_b32 v{{[0-9]+}}, s{{[0-9]+}}, 11
; VGPR: v_writelane_b32 v{{[0-9]+}}, s{{[0-9]+}}, 12
; VGPR: v_writelane_b32 v{{[0-9]+}}, s{{[0-9]+}}, 13
; VGPR: v_writelane_b32 v{{[0-9]+}}, s{{[0-9]+}}, 14
; VGPR: v_writelane_b32 v{{[0-9]+}}, s{{[0-9]+}}, 15
; VGPR: v_writelane_b32 v{{[0-9]+}}, s{{[0-9]+}}, 16
; VGPR: v_writelane_b32 v{{[0-9]+}}, s{{[0-9]+}}, 17
; VGPR: v_writelane_b32 v{{[0-9]+}}, s{{[0-9]+}}, 18
; VGPR: v_writelane_b32 v{{[0-9]+}}, s{{[0-9]+}}, 19
; VGPR: v_writelane_b32 v{{[0-9]+}}, s{{[0-9]+}}, 20
; VGPR: v_writelane_b32 v{{[0-9]+}}, s{{[0-9]+}}, 21
; VGPR: v_writelane_b32 v{{[0-9]+}}, s{{[0-9]+}}, 22
; VGPR: v_writelane_b32 v{{[0-9]+}}, s{{[0-9]+}}, 23
; VGPR: v_writelane_b32 v{{[0-9]+}}, s{{[0-9]+}}, 24
; VGPR: v_writelane_b32 v{{[0-9]+}}, s{{[0-9]+}}, 25
; VGPR: v_writelane_b32 v{{[0-9]+}}, s{{[0-9]+}}, 26
; VGPR: v_writelane_b32 v{{[0-9]+}}, s{{[0-9]+}}, 27
; VGPR: v_writelane_b32 v{{[0-9]+}}, s{{[0-9]+}}, 28
; VGPR: v_writelane_b32 v{{[0-9]+}}, s{{[0-9]+}}, 29
; VGPR: v_writelane_b32 v{{[0-9]+}}, s{{[0-9]+}}, 30
; VGPR: v_writelane_b32 v{{[0-9]+}}, s{{[0-9]+}}, 31
; VGPR: s_cbranch_scc1
; VGPR: v_readlane_b32 s{{[0-9]+}}, v{{[0-9]+}}, 0
; VGPR: v_readlane_b32 s{{[0-9]+}}, v{{[0-9]+}}, 1
; VGPR: v_readlane_b32 s{{[0-9]+}}, v{{[0-9]+}}, 2
; VGPR: v_readlane_b32 s{{[0-9]+}}, v{{[0-9]+}}, 3
; VGPR: v_readlane_b32 s{{[0-9]+}}, v{{[0-9]+}}, 4
; VGPR: v_readlane_b32 s{{[0-9]+}}, v{{[0-9]+}}, 5
; VGPR: v_readlane_b32 s{{[0-9]+}}, v{{[0-9]+}}, 6
; VGPR: v_readlane_b32 s{{[0-9]+}}, v{{[0-9]+}}, 7
; VGPR: v_readlane_b32 s{{[0-9]+}}, v{{[0-9]+}}, 8
; VGPR: v_readlane_b32 s{{[0-9]+}}, v{{[0-9]+}}, 9
; VGPR: v_readlane_b32 s{{[0-9]+}}, v{{[0-9]+}}, 10
; VGPR: v_readlane_b32 s{{[0-9]+}}, v{{[0-9]+}}, 11
; VGPR: v_readlane_b32 s{{[0-9]+}}, v{{[0-9]+}}, 12
; VGPR: v_readlane_b32 s{{[0-9]+}}, v{{[0-9]+}}, 13
; VGPR: v_readlane_b32 s{{[0-9]+}}, v{{[0-9]+}}, 14
; VGPR: v_readlane_b32 s{{[0-9]+}}, v{{[0-9]+}}, 15
; VGPR: v_readlane_b32 s{{[0-9]+}}, v{{[0-9]+}}, 16
; VGPR: v_readlane_b32 s{{[0-9]+}}, v{{[0-9]+}}, 17
; VGPR: v_readlane_b32 s{{[0-9]+}}, v{{[0-9]+}}, 18
; VGPR: v_readlane_b32 s{{[0-9]+}}, v{{[0-9]+}}, 19
; VGPR: v_readlane_b32 s{{[0-9]+}}, v{{[0-9]+}}, 20
; VGPR: v_readlane_b32 s{{[0-9]+}}, v{{[0-9]+}}, 21
; VGPR: v_readlane_b32 s{{[0-9]+}}, v{{[0-9]+}}, 22
; VGPR: v_readlane_b32 s{{[0-9]+}}, v{{[0-9]+}}, 23
; VGPR: v_readlane_b32 s{{[0-9]+}}, v{{[0-9]+}}, 24
; VGPR: v_readlane_b32 s{{[0-9]+}}, v{{[0-9]+}}, 25
; VGPR: v_readlane_b32 s{{[0-9]+}}, v{{[0-9]+}}, 26
; VGPR: v_readlane_b32 s{{[0-9]+}}, v{{[0-9]+}}, 27
; VGPR: v_readlane_b32 s{{[0-9]+}}, v{{[0-9]+}}, 28
; VGPR: v_readlane_b32 s{{[0-9]+}}, v{{[0-9]+}}, 29
; VGPR: v_readlane_b32 s{{[0-9]+}}, v{{[0-9]+}}, 30
; VGPR: v_readlane_b32 s{{[0-9]+}}, v{{[0-9]+}}, 31
; VMEM: buffer_store_dword
; VMEM: s_cbranch_scc1
; VMEM: buffer_load_dword
define amdgpu_kernel void @spill_sgpr_x32(i32 addrspace(1)* %out, i32 %in) #0 {
%wide.sgpr = call <32 x i32> asm sideeffect "; def $0", "=s" () #0
%cmp = icmp eq i32 %in, 0
br i1 %cmp, label %bb0, label %ret
bb0:
call void asm sideeffect "; use $0", "s"(<32 x i32> %wide.sgpr) #0
br label %ret
ret:
ret void
}
AMDGPU: Use wider scalar spills for SGPR spilling Since the spill is for the whole wave, these don't have the swizzling problems that vector stores do and a single 4-byte allocation is enough to spill a 64 element register. This should reduce the number of spill instructions and put all the spills for a register in the same cacheline. This should save allocated private size, but for now it doesn't. The extra slots are allocated for each component, but never used because the frame layout is essentially finalized before frame indices are replaced. For always using the scalar store path, this should probably be moved into processFunctionBeforeFrameFinalized. llvm-svn: 288445
2016-12-02 08:54:45 +08:00
attributes #0 = { nounwind }