From 918e81c3d4191fd415d604d7048dd7cb5732378c Mon Sep 17 00:00:00 2001 From: Matt Arsenault Date: Tue, 26 Jul 2016 23:15:59 +0000 Subject: [PATCH] AMDGPU: Add more tests for LDS size with occupancy llvm-svn: 276821 --- .../AMDGPU/large-work-group-promote-alloca.ll | 151 ++++++++++++++++++++- 1 file changed, 149 insertions(+), 2 deletions(-) diff --git a/llvm/test/CodeGen/AMDGPU/large-work-group-promote-alloca.ll b/llvm/test/CodeGen/AMDGPU/large-work-group-promote-alloca.ll index f661939..79cdaf5 100644 --- a/llvm/test/CodeGen/AMDGPU/large-work-group-promote-alloca.ll +++ b/llvm/test/CodeGen/AMDGPU/large-work-group-promote-alloca.ll @@ -66,7 +66,7 @@ entry: ret void } -; CHECK: @occupancy_0( +; CHECK-LABEL: @occupancy_0( ; CHECK: alloca [5 x i32] define void @occupancy_0(i32 addrspace(1)* nocapture %out, i32 addrspace(1)* nocapture %in) #3 { entry: @@ -88,7 +88,7 @@ entry: ret void } -; CHECK: @occupancy_max( +; CHECK-LABEL: @occupancy_max( ; CHECK: alloca [5 x i32] define void @occupancy_max(i32 addrspace(1)* nocapture %out, i32 addrspace(1)* nocapture %in) #4 { entry: @@ -110,8 +110,155 @@ entry: ret void } +; CHECK-LABEL: @occupancy_6( +; CHECK-NOT: alloca +define void @occupancy_6(i8 addrspace(1)* nocapture %out, i8 addrspace(1)* nocapture %in) #5 { +entry: + %stack = alloca [42 x i8], align 4 + %tmp = load i8, i8 addrspace(1)* %in, align 1 + %tmp4 = sext i8 %tmp to i64 + %arrayidx1 = getelementptr inbounds [42 x i8], [42 x i8]* %stack, i64 0, i64 %tmp4 + store i8 4, i8* %arrayidx1, align 1 + %arrayidx2 = getelementptr inbounds i8, i8 addrspace(1)* %in, i64 1 + %tmp1 = load i8, i8 addrspace(1)* %arrayidx2, align 1 + %tmp5 = sext i8 %tmp1 to i64 + %arrayidx3 = getelementptr inbounds [42 x i8], [42 x i8]* %stack, i64 0, i64 %tmp5 + store i8 5, i8* %arrayidx3, align 1 + %arrayidx10 = getelementptr inbounds [42 x i8], [42 x i8]* %stack, i64 0, i64 0 + %tmp2 = load i8, i8* %arrayidx10, align 1 + store i8 %tmp2, i8 addrspace(1)* %out, align 1 + %arrayidx12 = getelementptr inbounds [42 x i8], [42 x i8]* %stack, i64 0, i64 1 + %tmp3 = load i8, i8* %arrayidx12, align 1 + %arrayidx13 = getelementptr inbounds i8, i8 addrspace(1)* %out, i64 1 + store i8 %tmp3, i8 addrspace(1)* %arrayidx13, align 1 + ret void +} + +; CHECK-LABEL: @occupancy_6_over( +; CHECK: alloca [43 x i8] +define void @occupancy_6_over(i8 addrspace(1)* nocapture %out, i8 addrspace(1)* nocapture %in) #5 { +entry: + %stack = alloca [43 x i8], align 4 + %tmp = load i8, i8 addrspace(1)* %in, align 1 + %tmp4 = sext i8 %tmp to i64 + %arrayidx1 = getelementptr inbounds [43 x i8], [43 x i8]* %stack, i64 0, i64 %tmp4 + store i8 4, i8* %arrayidx1, align 1 + %arrayidx2 = getelementptr inbounds i8, i8 addrspace(1)* %in, i64 1 + %tmp1 = load i8, i8 addrspace(1)* %arrayidx2, align 1 + %tmp5 = sext i8 %tmp1 to i64 + %arrayidx3 = getelementptr inbounds [43 x i8], [43 x i8]* %stack, i64 0, i64 %tmp5 + store i8 5, i8* %arrayidx3, align 1 + %arrayidx10 = getelementptr inbounds [43 x i8], [43 x i8]* %stack, i64 0, i64 0 + %tmp2 = load i8, i8* %arrayidx10, align 1 + store i8 %tmp2, i8 addrspace(1)* %out, align 1 + %arrayidx12 = getelementptr inbounds [43 x i8], [43 x i8]* %stack, i64 0, i64 1 + %tmp3 = load i8, i8* %arrayidx12, align 1 + %arrayidx13 = getelementptr inbounds i8, i8 addrspace(1)* %out, i64 1 + store i8 %tmp3, i8 addrspace(1)* %arrayidx13, align 1 + ret void +} + +; CHECK-LABEL: @occupancy_8( +; CHECK-NOT: alloca +define void @occupancy_8(i8 addrspace(1)* nocapture %out, i8 addrspace(1)* nocapture %in) #6 { +entry: + %stack = alloca [32 x i8], align 4 + %tmp = load i8, i8 addrspace(1)* %in, align 1 + %tmp4 = sext i8 %tmp to i64 + %arrayidx1 = getelementptr inbounds [32 x i8], [32 x i8]* %stack, i64 0, i64 %tmp4 + store i8 4, i8* %arrayidx1, align 1 + %arrayidx2 = getelementptr inbounds i8, i8 addrspace(1)* %in, i64 1 + %tmp1 = load i8, i8 addrspace(1)* %arrayidx2, align 1 + %tmp5 = sext i8 %tmp1 to i64 + %arrayidx3 = getelementptr inbounds [32 x i8], [32 x i8]* %stack, i64 0, i64 %tmp5 + store i8 5, i8* %arrayidx3, align 1 + %arrayidx10 = getelementptr inbounds [32 x i8], [32 x i8]* %stack, i64 0, i64 0 + %tmp2 = load i8, i8* %arrayidx10, align 1 + store i8 %tmp2, i8 addrspace(1)* %out, align 1 + %arrayidx12 = getelementptr inbounds [32 x i8], [32 x i8]* %stack, i64 0, i64 1 + %tmp3 = load i8, i8* %arrayidx12, align 1 + %arrayidx13 = getelementptr inbounds i8, i8 addrspace(1)* %out, i64 1 + store i8 %tmp3, i8 addrspace(1)* %arrayidx13, align 1 + ret void +} + +; CHECK-LABEL: @occupancy_8_over( +; CHECK: alloca [33 x i8] +define void @occupancy_8_over(i8 addrspace(1)* nocapture %out, i8 addrspace(1)* nocapture %in) #6 { +entry: + %stack = alloca [33 x i8], align 4 + %tmp = load i8, i8 addrspace(1)* %in, align 1 + %tmp4 = sext i8 %tmp to i64 + %arrayidx1 = getelementptr inbounds [33 x i8], [33 x i8]* %stack, i64 0, i64 %tmp4 + store i8 4, i8* %arrayidx1, align 1 + %arrayidx2 = getelementptr inbounds i8, i8 addrspace(1)* %in, i64 1 + %tmp1 = load i8, i8 addrspace(1)* %arrayidx2, align 1 + %tmp5 = sext i8 %tmp1 to i64 + %arrayidx3 = getelementptr inbounds [33 x i8], [33 x i8]* %stack, i64 0, i64 %tmp5 + store i8 5, i8* %arrayidx3, align 1 + %arrayidx10 = getelementptr inbounds [33 x i8], [33 x i8]* %stack, i64 0, i64 0 + %tmp2 = load i8, i8* %arrayidx10, align 1 + store i8 %tmp2, i8 addrspace(1)* %out, align 1 + %arrayidx12 = getelementptr inbounds [33 x i8], [33 x i8]* %stack, i64 0, i64 1 + %tmp3 = load i8, i8* %arrayidx12, align 1 + %arrayidx13 = getelementptr inbounds i8, i8 addrspace(1)* %out, i64 1 + store i8 %tmp3, i8 addrspace(1)* %arrayidx13, align 1 + ret void +} + +; CHECK-LABEL: @occupancy_9( +; CHECK-NOT: alloca +define void @occupancy_9(i8 addrspace(1)* nocapture %out, i8 addrspace(1)* nocapture %in) #7 { +entry: + %stack = alloca [28 x i8], align 4 + %tmp = load i8, i8 addrspace(1)* %in, align 1 + %tmp4 = sext i8 %tmp to i64 + %arrayidx1 = getelementptr inbounds [28 x i8], [28 x i8]* %stack, i64 0, i64 %tmp4 + store i8 4, i8* %arrayidx1, align 1 + %arrayidx2 = getelementptr inbounds i8, i8 addrspace(1)* %in, i64 1 + %tmp1 = load i8, i8 addrspace(1)* %arrayidx2, align 1 + %tmp5 = sext i8 %tmp1 to i64 + %arrayidx3 = getelementptr inbounds [28 x i8], [28 x i8]* %stack, i64 0, i64 %tmp5 + store i8 5, i8* %arrayidx3, align 1 + %arrayidx10 = getelementptr inbounds [28 x i8], [28 x i8]* %stack, i64 0, i64 0 + %tmp2 = load i8, i8* %arrayidx10, align 1 + store i8 %tmp2, i8 addrspace(1)* %out, align 1 + %arrayidx12 = getelementptr inbounds [28 x i8], [28 x i8]* %stack, i64 0, i64 1 + %tmp3 = load i8, i8* %arrayidx12, align 1 + %arrayidx13 = getelementptr inbounds i8, i8 addrspace(1)* %out, i64 1 + store i8 %tmp3, i8 addrspace(1)* %arrayidx13, align 1 + ret void +} + +; CHECK-LABEL: @occupancy_9_over( +; CHECK: alloca [29 x i8] +define void @occupancy_9_over(i8 addrspace(1)* nocapture %out, i8 addrspace(1)* nocapture %in) #7 { +entry: + %stack = alloca [29 x i8], align 4 + %tmp = load i8, i8 addrspace(1)* %in, align 1 + %tmp4 = sext i8 %tmp to i64 + %arrayidx1 = getelementptr inbounds [29 x i8], [29 x i8]* %stack, i64 0, i64 %tmp4 + store i8 4, i8* %arrayidx1, align 1 + %arrayidx2 = getelementptr inbounds i8, i8 addrspace(1)* %in, i64 1 + %tmp1 = load i8, i8 addrspace(1)* %arrayidx2, align 1 + %tmp5 = sext i8 %tmp1 to i64 + %arrayidx3 = getelementptr inbounds [29 x i8], [29 x i8]* %stack, i64 0, i64 %tmp5 + store i8 5, i8* %arrayidx3, align 1 + %arrayidx10 = getelementptr inbounds [29 x i8], [29 x i8]* %stack, i64 0, i64 0 + %tmp2 = load i8, i8* %arrayidx10, align 1 + store i8 %tmp2, i8 addrspace(1)* %out, align 1 + %arrayidx12 = getelementptr inbounds [29 x i8], [29 x i8]* %stack, i64 0, i64 1 + %tmp3 = load i8, i8* %arrayidx12, align 1 + %arrayidx13 = getelementptr inbounds i8, i8 addrspace(1)* %out, i64 1 + store i8 %tmp3, i8 addrspace(1)* %arrayidx13, align 1 + ret void +} + attributes #0 = { nounwind "amdgpu-max-work-group-size"="63" } attributes #1 = { nounwind "amdgpu-max-waves-per-eu"="3" "amdgpu-max-work-group-size"="256" } attributes #2 = { nounwind "amdgpu-max-waves-per-eu"="1" "amdgpu-max-work-group-size"="1600" } attributes #3 = { nounwind "amdgpu-max-waves-per-eu"="0" } attributes #4 = { nounwind "amdgpu-max-waves-per-eu"="-1" } +attributes #5 = { nounwind "amdgpu-max-waves-per-eu"="6" "amdgpu-max-work-group-size"="64" } +attributes #6 = { nounwind "amdgpu-max-waves-per-eu"="8" "amdgpu-max-work-group-size"="64" } +attributes #7 = { nounwind "amdgpu-max-waves-per-eu"="9" "amdgpu-max-work-group-size"="64" } -- 2.7.4