1; RUN: opt -S -mtriple=amdgcn-- -amdgpu-lower-module-lds --amdgpu-lower-module-lds-strategy=module < %s | FileCheck %s 2; RUN: opt -S -mtriple=amdgcn-- -passes=amdgpu-lower-module-lds --amdgpu-lower-module-lds-strategy=module < %s | FileCheck %s 3 4; CHECK: %llvm.amdgcn.module.lds.t = type { float, float } 5; CHECK: %llvm.amdgcn.kernel.timestwo.lds.t = type { float, float } 6 7@a_func = addrspace(3) global float poison, align 4 8 9@kern = addrspace(3) global float poison, align 4 10 11; @a_func is only used from a non-kernel function so is rewritten 12; CHECK-NOT: @a_func 13; @b_both is used from a non-kernel function so is rewritten 14; CHECK-NOT: @b_both 15; sorted both < func, so @b_both at null and @a_func at 4 16@b_both = addrspace(3) global float poison, align 4 17 18; CHECK: @llvm.amdgcn.module.lds = internal addrspace(3) global %llvm.amdgcn.module.lds.t poison, align 4 19; CHECK: @llvm.amdgcn.kernel.timestwo.lds = internal addrspace(3) global %llvm.amdgcn.kernel.timestwo.lds.t poison, align 4 20 21; CHECK-LABEL: @get_func() 22; CHECK: %0 = addrspacecast ptr addrspace(3) @llvm.amdgcn.module.lds to ptr 23; CHECK: %1 = ptrtoint ptr %0 to i64 24; CHECK: %2 = addrspacecast ptr addrspace(3) @llvm.amdgcn.module.lds to ptr 25; CHECK: %3 = ptrtoint ptr %2 to i64 26; CHECK: %4 = add i64 %1, %3 27; CHECK: %5 = inttoptr i64 %4 to ptr 28; CHECK: %6 = load i32, ptr %5, align 4 29; CHECK: ret i32 %6 30define i32 @get_func() local_unnamed_addr #0 { 31entry: 32 %0 = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr addrspacecast (ptr addrspace(3) @a_func to ptr) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(3) @a_func to ptr) to i64)) to ptr), align 4 33 ret i32 %0 34} 35 36; CHECK-LABEL: @set_func(i32 %x) 37; CHECK: %0 = addrspacecast ptr addrspace(3) getelementptr inbounds (%llvm.amdgcn.module.lds.t, ptr addrspace(3) @llvm.amdgcn.module.lds, i32 0, i32 1) to ptr 38; CHECK: %1 = ptrtoint ptr %0 to i64 39; CHECK: %2 = addrspacecast ptr addrspace(3) getelementptr inbounds (%llvm.amdgcn.module.lds.t, ptr addrspace(3) @llvm.amdgcn.module.lds, i32 0, i32 1) to ptr 40; CHECK: %3 = ptrtoint ptr %2 to i64 41; CHECK: %4 = add i64 %1, %3 42; CHECK: %5 = inttoptr i64 %4 to ptr 43; CHECK: store i32 %x, ptr %5, align 4 44; CHECK: ret void 45define void @set_func(i32 %x) { 46entry: 47 store i32 %x, ptr inttoptr (i64 add (i64 ptrtoint (ptr addrspacecast (ptr addrspace(3) @b_both to ptr) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(3) @b_both to ptr) to i64)) to ptr), align 4 48 ret void 49} 50 51; CHECK-LABEL: @timestwo() #0 52; CHECK-NOT: call void @llvm.donothing() 53 54; CHECK: %1 = addrspacecast ptr addrspace(3) @llvm.amdgcn.kernel.timestwo.lds to ptr 55; CHECK: %2 = ptrtoint ptr %1 to i64 56; CHECK: %3 = addrspacecast ptr addrspace(3) getelementptr inbounds (%llvm.amdgcn.kernel.timestwo.lds.t, ptr addrspace(3) @llvm.amdgcn.kernel.timestwo.lds, i32 0, i32 1) to ptr 57; CHECK: %4 = ptrtoint ptr %3 to i64 58; CHECK: %5 = add i64 %2, %4 59; CHECK: %6 = inttoptr i64 %5 to ptr 60; CHECK: %ld = load i32, ptr %6, align 4 61; CHECK: %mul = mul i32 %ld, 2 62; CHECK: %7 = addrspacecast ptr addrspace(3) getelementptr inbounds (%llvm.amdgcn.kernel.timestwo.lds.t, ptr addrspace(3) @llvm.amdgcn.kernel.timestwo.lds, i32 0, i32 1) to ptr 63; CHECK: %8 = ptrtoint ptr %7 to i64 64; CHECK: %9 = addrspacecast ptr addrspace(3) @llvm.amdgcn.kernel.timestwo.lds to ptr 65; CHECK: %10 = ptrtoint ptr %9 to i64 66; CHECK: %11 = add i64 %8, %10 67; CHECK: %12 = inttoptr i64 %11 to ptr 68; CHECK: store i32 %mul, ptr %12, align 4 69; CHECK: ret void 70define amdgpu_kernel void @timestwo() { 71 %ld = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr addrspacecast (ptr addrspace(3) @b_both to ptr) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(3) @kern to ptr) to i64)) to ptr), align 4 72 %mul = mul i32 %ld, 2 73 store i32 %mul, ptr inttoptr (i64 add (i64 ptrtoint (ptr addrspacecast (ptr addrspace(3) @kern to ptr) to i64), i64 ptrtoint (ptr addrspacecast (ptr addrspace(3) @b_both to ptr) to i64)) to ptr), align 4 74 ret void 75} 76 77; CHECK-LABEL: @through_functions() #0 78define amdgpu_kernel void @through_functions() { 79 %ld = call i32 @get_func() 80 %mul = mul i32 %ld, 4 81 call void @set_func(i32 %mul) 82 ret void 83} 84 85; CHECK: attributes #0 = { "amdgpu-lds-size"="8" } 86