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/waitcnt-looptest.ll

147 lines
13 KiB
LLVM
Raw Normal View History

[AMDGPU] Increased vector length for global/constant loads. Summary: GCN ISA supports instructions that can read 16 consecutive dwords from memory through the scalar data cache; loadstoreVectorizer should take advantage of the wider vector length and pack 16/8 elements of dwords/quadwords. Author: FarhanaAleen Reviewed By: rampitec Subscribers: llvm-commits, AMDGPU Differential Revision: https://reviews.llvm.org/D44179 llvm-svn: 326910
2018-03-08 01:09:18 +08:00
; RUN: llc < %s -mtriple=amdgcn--amdhsa -mcpu=fiji -mattr=-flat-for-global -amdgpu-load-store-vectorizer=0 | FileCheck --check-prefix=GCN %s
[AMDGPU] In the new waitcnt insertion pass, use getHeader instead of getTopBlock to find the loop header. Differential Revision: https://reviews.llvm.org/D32831 llvm-svn: 302290
2017-05-06 05:10:17 +08:00
; Check that the waitcnt insertion algorithm correctly propagates wait counts
; from before a loop to the loop header.
; GCN-LABEL: {{^}}testKernel
; GCN: BB0_1:
; GCN: s_waitcnt vmcnt(0) lgkmcnt(0)
AMDGPU: Rewrite SILowerI1Copies to always stay on SALU Summary: Instead of writing boolean values temporarily into 32-bit VGPRs if they are involved in PHIs or are observed from outside a loop, we use bitwise masking operations to combine lane masks in a way that is consistent with wave control flow. Move SIFixSGPRCopies to before this pass, since that pass incorrectly attempts to move SGPR phis to VGPRs. This should recover most of the code quality that was lost with the bug fix in "AMDGPU: Remove PHI loop condition optimization". There are still some relevant cases where code quality could be improved, in particular: - We often introduce redundant masks with EXEC. Ideally, we'd have a generic computeKnownBits-like analysis to determine whether masks are already masked by EXEC, so we can avoid this masking both here and when lowering uniform control flow. - The criterion we use to determine whether a def is observed from outside a loop is conservative: it doesn't check whether (loop) branch conditions are uniform. Change-Id: Ibabdb373a7510e426b90deef00f5e16c5d56e64b Reviewers: arsenm, rampitec, tpr Subscribers: kzhuravl, jvesely, wdng, mgorny, yaxunl, dstuttard, t-tye, eraman, llvm-commits Differential Revision: https://reviews.llvm.org/D53496 llvm-svn: 345719
2018-10-31 21:27:08 +08:00
; GCN-NEXT: v_cmp_eq_f32_e32
[AMDGPU] In the new waitcnt insertion pass, use getHeader instead of getTopBlock to find the loop header. Differential Revision: https://reviews.llvm.org/D32831 llvm-svn: 302290
2017-05-06 05:10:17 +08:00
; GCN: s_waitcnt vmcnt(0) lgkmcnt(0)
; GCN-NEXT: v_cmp_eq_f32_e32
; GCN: s_waitcnt vmcnt(0) lgkmcnt(0)
; GCN-NEXT: v_cmp_eq_f32_e32
@data_generic = addrspace(1) global [100 x float] [float 0.000000e+00, float 0x3FB99999A0000000, float 0x3FC99999A0000000, float 0x3FD3333340000000, float 0x3FD99999A0000000, float 5.000000e-01, float 0x3FE3333340000000, float 0x3FE6666660000000, float 0x3FE99999A0000000, float 0x3FECCCCCC0000000, float 1.000000e+00, float 0x3FF19999A0000000, float 0x3FF3333340000000, float 0x3FF4CCCCC0000000, float 0x3FF6666660000000, float 1.500000e+00, float 0x3FF99999A0000000, float 0x3FFB333340000000, float 0x3FFCCCCCC0000000, float 0x3FFE666660000000, float 2.000000e+00, float 0x4000CCCCC0000000, float 0x40019999A0000000, float 0x4002666660000000, float 0x4003333340000000, float 2.500000e+00, float 0x4004CCCCC0000000, float 0x40059999A0000000, float 0x4006666660000000, float 0x4007333340000000, float 3.000000e+00, float 0x4008CCCCC0000000, float 0x40099999A0000000, float 0x400A666660000000, float 0x400B333340000000, float 3.500000e+00, float 0x400CCCCCC0000000, float 0x400D9999A0000000, float 0x400E666660000000, float 0x400F333340000000, float 4.000000e+00, float 0x4010666660000000, float 0x4010CCCCC0000000, float 0x4011333340000000, float 0x40119999A0000000, float 4.500000e+00, float 0x4012666660000000, float 0x4012CCCCC0000000, float 0x4013333340000000, float 0x40139999A0000000, float 5.000000e+00, float 0x4014666660000000, float 0x4014CCCCC0000000, float 0x4015333340000000, float 0x40159999A0000000, float 5.500000e+00, float 0x4016666660000000, float 0x4016CCCCC0000000, float 0x4017333340000000, float 0x40179999A0000000, float 6.000000e+00, float 0x4018666660000000, float 0x4018CCCCC0000000, float 0x4019333340000000, float 0x40199999A0000000, float 6.500000e+00, float 0x401A666660000000, float 0x401ACCCCC0000000, float 0x401B333340000000, float 0x401B9999A0000000, float 7.000000e+00, float 0x401C666660000000, float 0x401CCCCCC0000000, float 0x401D333340000000, float 0x401D9999A0000000, float 7.500000e+00, float 0x401E666660000000, float 0x401ECCCCC0000000, float 0x401F333340000000, float 0x401F9999A0000000, float 8.000000e+00, float 0x4020333340000000, float 0x4020666660000000, float 0x40209999A0000000, float 0x4020CCCCC0000000, float 8.500000e+00, float 0x4021333340000000, float 0x4021666660000000, float 0x40219999A0000000, float 0x4021CCCCC0000000, float 9.000000e+00, float 0x4022333340000000, float 0x4022666660000000, float 0x40229999A0000000, float 0x4022CCCCC0000000, float 9.500000e+00, float 0x4023333340000000, float 0x4023666660000000, float 0x40239999A0000000, float 0x4023CCCCC0000000], align 4
@data_reference = addrspace(1) global [100 x float] [float 0.000000e+00, float 0x3FB99999A0000000, float 0x3FC99999A0000000, float 0x3FD3333340000000, float 0x3FD99999A0000000, float 5.000000e-01, float 0x3FE3333340000000, float 0x3FE6666660000000, float 0x3FE99999A0000000, float 0x3FECCCCCC0000000, float 1.000000e+00, float 0x3FF19999A0000000, float 0x3FF3333340000000, float 0x3FF4CCCCC0000000, float 0x3FF6666660000000, float 1.500000e+00, float 0x3FF99999A0000000, float 0x3FFB333340000000, float 0x3FFCCCCCC0000000, float 0x3FFE666660000000, float 2.000000e+00, float 0x4000CCCCC0000000, float 0x40019999A0000000, float 0x4002666660000000, float 0x4003333340000000, float 2.500000e+00, float 0x4004CCCCC0000000, float 0x40059999A0000000, float 0x4006666660000000, float 0x4007333340000000, float 3.000000e+00, float 0x4008CCCCC0000000, float 0x40099999A0000000, float 0x400A666660000000, float 0x400B333340000000, float 3.500000e+00, float 0x400CCCCCC0000000, float 0x400D9999A0000000, float 0x400E666660000000, float 0x400F333340000000, float 4.000000e+00, float 0x4010666660000000, float 0x4010CCCCC0000000, float 0x4011333340000000, float 0x40119999A0000000, float 4.500000e+00, float 0x4012666660000000, float 0x4012CCCCC0000000, float 0x4013333340000000, float 0x40139999A0000000, float 5.000000e+00, float 0x4014666660000000, float 0x4014CCCCC0000000, float 0x4015333340000000, float 0x40159999A0000000, float 5.500000e+00, float 0x4016666660000000, float 0x4016CCCCC0000000, float 0x4017333340000000, float 0x40179999A0000000, float 6.000000e+00, float 0x4018666660000000, float 0x4018CCCCC0000000, float 0x4019333340000000, float 0x40199999A0000000, float 6.500000e+00, float 0x401A666660000000, float 0x401ACCCCC0000000, float 0x401B333340000000, float 0x401B9999A0000000, float 7.000000e+00, float 0x401C666660000000, float 0x401CCCCCC0000000, float 0x401D333340000000, float 0x401D9999A0000000, float 7.500000e+00, float 0x401E666660000000, float 0x401ECCCCC0000000, float 0x401F333340000000, float 0x401F9999A0000000, float 8.000000e+00, float 0x4020333340000000, float 0x4020666660000000, float 0x40209999A0000000, float 0x4020CCCCC0000000, float 8.500000e+00, float 0x4021333340000000, float 0x4021666660000000, float 0x40219999A0000000, float 0x4021CCCCC0000000, float 9.000000e+00, float 0x4022333340000000, float 0x4022666660000000, float 0x40229999A0000000, float 0x4022CCCCC0000000, float 9.500000e+00, float 0x4023333340000000, float 0x4023666660000000, float 0x40239999A0000000, float 0x4023CCCCC0000000], align 4
define amdgpu_kernel void @testKernel(i32 addrspace(1)* nocapture %arg) local_unnamed_addr #0 {
bb:
[AMDGPU] Switch to the new addr space mapping by default This requires corresponding clang change. Differential Revision: https://reviews.llvm.org/D40955 llvm-svn: 324101
2018-02-03 00:07:16 +08:00
store <2 x float> <float 1.000000e+00, float 1.000000e+00>, <2 x float>* bitcast (float* getelementptr ([100 x float], [100 x float]* addrspacecast ([100 x float] addrspace(1)* @data_generic to [100 x float]*), i64 0, i64 4) to <2 x float>*), align 4
store <2 x float> <float 1.000000e+00, float 1.000000e+00>, <2 x float>* bitcast (float* getelementptr ([100 x float], [100 x float]* addrspacecast ([100 x float] addrspace(1)* @data_reference to [100 x float]*), i64 0, i64 4) to <2 x float>*), align 4
[AMDGPU] In the new waitcnt insertion pass, use getHeader instead of getTopBlock to find the loop header. Differential Revision: https://reviews.llvm.org/D32831 llvm-svn: 302290
2017-05-06 05:10:17 +08:00
br label %bb18
bb1: ; preds = %bb18
[AMDGPU] Change constant addr space to 4 Differential Revision: https://reviews.llvm.org/D43170 llvm-svn: 325030
2018-02-14 02:00:25 +08:00
%tmp = tail call i8 addrspace(4)* @llvm.amdgcn.dispatch.ptr()
[AMDGPU] In the new waitcnt insertion pass, use getHeader instead of getTopBlock to find the loop header. Differential Revision: https://reviews.llvm.org/D32831 llvm-svn: 302290
2017-05-06 05:10:17 +08:00
%tmp2 = tail call i32 @llvm.amdgcn.workitem.id.x()
%tmp3 = tail call i32 @llvm.amdgcn.workgroup.id.x()
[AMDGPU] Change constant addr space to 4 Differential Revision: https://reviews.llvm.org/D43170 llvm-svn: 325030
2018-02-14 02:00:25 +08:00
%tmp4 = getelementptr inbounds i8, i8 addrspace(4)* %tmp, i64 4
%tmp5 = bitcast i8 addrspace(4)* %tmp4 to i16 addrspace(4)*
%tmp6 = load i16, i16 addrspace(4)* %tmp5, align 4
[AMDGPU] In the new waitcnt insertion pass, use getHeader instead of getTopBlock to find the loop header. Differential Revision: https://reviews.llvm.org/D32831 llvm-svn: 302290
2017-05-06 05:10:17 +08:00
%tmp7 = zext i16 %tmp6 to i32
%tmp8 = mul i32 %tmp3, %tmp7
%tmp9 = add i32 %tmp8, %tmp2
[AMDGPU] Change constant addr space to 4 Differential Revision: https://reviews.llvm.org/D43170 llvm-svn: 325030
2018-02-14 02:00:25 +08:00
%tmp10 = tail call i8 addrspace(4)* @llvm.amdgcn.implicitarg.ptr()
[AMDGPU] In the new waitcnt insertion pass, use getHeader instead of getTopBlock to find the loop header. Differential Revision: https://reviews.llvm.org/D32831 llvm-svn: 302290
2017-05-06 05:10:17 +08:00
%tmp11 = zext i32 %tmp9 to i64
[AMDGPU] Change constant addr space to 4 Differential Revision: https://reviews.llvm.org/D43170 llvm-svn: 325030
2018-02-14 02:00:25 +08:00
%tmp12 = bitcast i8 addrspace(4)* %tmp10 to i64 addrspace(4)*
%tmp13 = load i64, i64 addrspace(4)* %tmp12, align 8
[AMDGPU] In the new waitcnt insertion pass, use getHeader instead of getTopBlock to find the loop header. Differential Revision: https://reviews.llvm.org/D32831 llvm-svn: 302290
2017-05-06 05:10:17 +08:00
%tmp14 = add i64 %tmp13, %tmp11
%tmp15 = zext i1 %tmp99 to i32
%tmp16 = and i64 %tmp14, 4294967295
%tmp17 = getelementptr inbounds i32, i32 addrspace(1)* %arg, i64 %tmp16
store i32 %tmp15, i32 addrspace(1)* %tmp17, align 4
ret void
bb18: ; preds = %bb18, %bb
%tmp19 = phi i64 [ 0, %bb ], [ %tmp102, %bb18 ]
%tmp20 = phi i32 [ 0, %bb ], [ %tmp100, %bb18 ]
%tmp21 = phi i1 [ true, %bb ], [ %tmp99, %bb18 ]
%tmp22 = getelementptr inbounds [100 x float], [100 x float] addrspace(1)* @data_generic, i64 0, i64 %tmp19
%tmp23 = load float, float addrspace(1)* %tmp22, align 4
%tmp24 = getelementptr inbounds [100 x float], [100 x float] addrspace(1)* @data_reference, i64 0, i64 %tmp19
%tmp25 = load float, float addrspace(1)* %tmp24, align 4
%tmp26 = fcmp oeq float %tmp23, %tmp25
%tmp27 = and i1 %tmp21, %tmp26
%tmp28 = or i32 %tmp20, 1
%tmp29 = sext i32 %tmp28 to i64
%tmp30 = getelementptr inbounds [100 x float], [100 x float] addrspace(1)* @data_generic, i64 0, i64 %tmp29
%tmp31 = load float, float addrspace(1)* %tmp30, align 4
%tmp32 = getelementptr inbounds [100 x float], [100 x float] addrspace(1)* @data_reference, i64 0, i64 %tmp29
%tmp33 = load float, float addrspace(1)* %tmp32, align 4
%tmp34 = fcmp oeq float %tmp31, %tmp33
%tmp35 = and i1 %tmp27, %tmp34
%tmp36 = add nuw nsw i32 %tmp20, 2
%tmp37 = sext i32 %tmp36 to i64
%tmp38 = getelementptr inbounds [100 x float], [100 x float] addrspace(1)* @data_generic, i64 0, i64 %tmp37
%tmp39 = load float, float addrspace(1)* %tmp38, align 4
%tmp40 = getelementptr inbounds [100 x float], [100 x float] addrspace(1)* @data_reference, i64 0, i64 %tmp37
%tmp41 = load float, float addrspace(1)* %tmp40, align 4
%tmp42 = fcmp oeq float %tmp39, %tmp41
%tmp43 = and i1 %tmp35, %tmp42
%tmp44 = add nuw nsw i32 %tmp20, 3
%tmp45 = sext i32 %tmp44 to i64
%tmp46 = getelementptr inbounds [100 x float], [100 x float] addrspace(1)* @data_generic, i64 0, i64 %tmp45
%tmp47 = load float, float addrspace(1)* %tmp46, align 4
%tmp48 = getelementptr inbounds [100 x float], [100 x float] addrspace(1)* @data_reference, i64 0, i64 %tmp45
%tmp49 = load float, float addrspace(1)* %tmp48, align 4
%tmp50 = fcmp oeq float %tmp47, %tmp49
%tmp51 = and i1 %tmp43, %tmp50
%tmp52 = add nuw nsw i32 %tmp20, 4
%tmp53 = sext i32 %tmp52 to i64
%tmp54 = getelementptr inbounds [100 x float], [100 x float] addrspace(1)* @data_generic, i64 0, i64 %tmp53
%tmp55 = load float, float addrspace(1)* %tmp54, align 4
%tmp56 = getelementptr inbounds [100 x float], [100 x float] addrspace(1)* @data_reference, i64 0, i64 %tmp53
%tmp57 = load float, float addrspace(1)* %tmp56, align 4
%tmp58 = fcmp oeq float %tmp55, %tmp57
%tmp59 = and i1 %tmp51, %tmp58
%tmp60 = add nuw nsw i32 %tmp20, 5
%tmp61 = sext i32 %tmp60 to i64
%tmp62 = getelementptr inbounds [100 x float], [100 x float] addrspace(1)* @data_generic, i64 0, i64 %tmp61
%tmp63 = load float, float addrspace(1)* %tmp62, align 4
%tmp64 = getelementptr inbounds [100 x float], [100 x float] addrspace(1)* @data_reference, i64 0, i64 %tmp61
%tmp65 = load float, float addrspace(1)* %tmp64, align 4
%tmp66 = fcmp oeq float %tmp63, %tmp65
%tmp67 = and i1 %tmp59, %tmp66
%tmp68 = add nuw nsw i32 %tmp20, 6
%tmp69 = sext i32 %tmp68 to i64
%tmp70 = getelementptr inbounds [100 x float], [100 x float] addrspace(1)* @data_generic, i64 0, i64 %tmp69
%tmp71 = load float, float addrspace(1)* %tmp70, align 4
%tmp72 = getelementptr inbounds [100 x float], [100 x float] addrspace(1)* @data_reference, i64 0, i64 %tmp69
%tmp73 = load float, float addrspace(1)* %tmp72, align 4
%tmp74 = fcmp oeq float %tmp71, %tmp73
%tmp75 = and i1 %tmp67, %tmp74
%tmp76 = add nuw nsw i32 %tmp20, 7
%tmp77 = sext i32 %tmp76 to i64
%tmp78 = getelementptr inbounds [100 x float], [100 x float] addrspace(1)* @data_generic, i64 0, i64 %tmp77
%tmp79 = load float, float addrspace(1)* %tmp78, align 4
%tmp80 = getelementptr inbounds [100 x float], [100 x float] addrspace(1)* @data_reference, i64 0, i64 %tmp77
%tmp81 = load float, float addrspace(1)* %tmp80, align 4
%tmp82 = fcmp oeq float %tmp79, %tmp81
%tmp83 = and i1 %tmp75, %tmp82
%tmp84 = add nuw nsw i32 %tmp20, 8
%tmp85 = sext i32 %tmp84 to i64
%tmp86 = getelementptr inbounds [100 x float], [100 x float] addrspace(1)* @data_generic, i64 0, i64 %tmp85
%tmp87 = load float, float addrspace(1)* %tmp86, align 4
%tmp88 = getelementptr inbounds [100 x float], [100 x float] addrspace(1)* @data_reference, i64 0, i64 %tmp85
%tmp89 = load float, float addrspace(1)* %tmp88, align 4
%tmp90 = fcmp oeq float %tmp87, %tmp89
%tmp91 = and i1 %tmp83, %tmp90
%tmp92 = add nuw nsw i32 %tmp20, 9
%tmp93 = sext i32 %tmp92 to i64
%tmp94 = getelementptr inbounds [100 x float], [100 x float] addrspace(1)* @data_generic, i64 0, i64 %tmp93
%tmp95 = load float, float addrspace(1)* %tmp94, align 4
%tmp96 = getelementptr inbounds [100 x float], [100 x float] addrspace(1)* @data_reference, i64 0, i64 %tmp93
%tmp97 = load float, float addrspace(1)* %tmp96, align 4
%tmp98 = fcmp oeq float %tmp95, %tmp97
%tmp99 = and i1 %tmp91, %tmp98
%tmp100 = add nuw nsw i32 %tmp20, 10
%tmp101 = icmp eq i32 %tmp100, 100
%tmp102 = sext i32 %tmp100 to i64
br i1 %tmp101, label %bb1, label %bb18
}
; Function Attrs: nounwind readnone speculatable
[AMDGPU] Change constant addr space to 4 Differential Revision: https://reviews.llvm.org/D43170 llvm-svn: 325030
2018-02-14 02:00:25 +08:00
declare i8 addrspace(4)* @llvm.amdgcn.dispatch.ptr() #1
[AMDGPU] In the new waitcnt insertion pass, use getHeader instead of getTopBlock to find the loop header. Differential Revision: https://reviews.llvm.org/D32831 llvm-svn: 302290
2017-05-06 05:10:17 +08:00
; Function Attrs: nounwind readnone speculatable
declare i32 @llvm.amdgcn.workitem.id.x() #1
; Function Attrs: nounwind readnone speculatable
declare i32 @llvm.amdgcn.workgroup.id.x() #1
; Function Attrs: nounwind readnone speculatable
[AMDGPU] Change constant addr space to 4 Differential Revision: https://reviews.llvm.org/D43170 llvm-svn: 325030
2018-02-14 02:00:25 +08:00
declare i8 addrspace(4)* @llvm.amdgcn.implicitarg.ptr() #1
[AMDGPU] In the new waitcnt insertion pass, use getHeader instead of getTopBlock to find the loop header. Differential Revision: https://reviews.llvm.org/D32831 llvm-svn: 302290
2017-05-06 05:10:17 +08:00
attributes #0 = { "target-cpu"="fiji" "target-features"="-flat-for-global" }
attributes #1 = { nounwind readnone speculatable }