diff options
Diffstat (limited to 'test/CodeGen/R600/private-memory.ll')
-rw-r--r-- | test/CodeGen/R600/private-memory.ll | 119 |
1 files changed, 94 insertions, 25 deletions
diff --git a/test/CodeGen/R600/private-memory.ll b/test/CodeGen/R600/private-memory.ll index d3453f2..89122be 100644 --- a/test/CodeGen/R600/private-memory.ll +++ b/test/CodeGen/R600/private-memory.ll @@ -1,24 +1,19 @@ ; RUN: llc -march=r600 -mcpu=redwood < %s | FileCheck %s --check-prefix=R600-CHECK --check-prefix=FUNC ; RUN: llc -verify-machineinstrs -march=r600 -mcpu=SI < %s | FileCheck %s --check-prefix=SI-CHECK --check-prefix=FUNC -; This test checks that uses and defs of the AR register happen in the same -; instruction clause. +declare i32 @llvm.r600.read.tidig.x() nounwind readnone ; FUNC-LABEL: @mova_same_clause -; R600-CHECK: MOVA_INT -; R600-CHECK-NOT: ALU clause -; R600-CHECK: 0 + AR.x -; R600-CHECK: MOVA_INT -; R600-CHECK-NOT: ALU clause -; R600-CHECK: 0 + AR.x - -; SI-CHECK: V_READFIRSTLANE_B32 vcc_lo -; SI-CHECK: V_MOVRELD -; SI-CHECK: S_CBRANCH -; SI-CHECK: V_READFIRSTLANE_B32 vcc_lo -; SI-CHECK: V_MOVRELD -; SI-CHECK: S_CBRANCH +; R600-CHECK: LDS_WRITE +; R600-CHECK: LDS_WRITE +; R600-CHECK: LDS_READ +; R600-CHECK: LDS_READ + +; SI-CHECK: DS_WRITE_B32 +; SI-CHECK: DS_WRITE_B32 +; SI-CHECK: DS_READ_B32 +; SI-CHECK: DS_READ_B32 define void @mova_same_clause(i32 addrspace(1)* nocapture %out, i32 addrspace(1)* nocapture %in) { entry: %stack = alloca [5 x i32], align 4 @@ -114,12 +109,8 @@ for.end: ; FUNC-LABEL: @short_array -; R600-CHECK: MOV {{\** *}}T{{[0-9]\.[XYZW]}}, literal -; R600-CHECK: 65536 -; R600-CHECK: * ; R600-CHECK: MOVA_INT -; SI-CHECK: V_MOV_B32_e32 v{{[0-9]}}, 0x10000 ; SI-CHECK: V_MOVRELS_B32_e32 define void @short_array(i32 addrspace(1)* %out, i32 %index) { entry: @@ -137,10 +128,7 @@ entry: ; FUNC-LABEL: @char_array -; R600-CHECK: OR_INT {{\** *}}T{{[0-9]\.[XYZW]}}, {{[PVT0-9]+\.[XYZW]}}, literal -; R600-CHECK: 256 -; R600-CHECK: * -; R600-CHECK-NEXT: MOVA_INT +; R600-CHECK: MOVA_INT ; SI-CHECK: V_OR_B32_e32 v{{[0-9]}}, 0x100 ; SI-CHECK: V_MOVRELS_B32_e32 @@ -185,7 +173,9 @@ entry: ; Test that two stack objects are not stored in the same register ; The second stack object should be in T3.X ; FUNC-LABEL: @no_overlap -; R600-CHECK: MOV {{\** *}}T3.X +; R600_CHECK: MOV +; R600_CHECK: [[CHAN:[XYZW]]]+ +; R600-CHECK-NOT: [[CHAN]]+ ; SI-CHECK: V_MOV_B32_e32 v3 define void @no_overlap(i32 addrspace(1)* %out, i32 %in) { entry: @@ -211,6 +201,85 @@ entry: ret void } +define void @char_array_array(i32 addrspace(1)* %out, i32 %index) { +entry: + %alloca = alloca [2 x [2 x i8]] + %gep0 = getelementptr [2 x [2 x i8]]* %alloca, i32 0, i32 0, i32 0 + %gep1 = getelementptr [2 x [2 x i8]]* %alloca, i32 0, i32 0, i32 1 + store i8 0, i8* %gep0 + store i8 1, i8* %gep1 + %gep2 = getelementptr [2 x [2 x i8]]* %alloca, i32 0, i32 0, i32 %index + %load = load i8* %gep2 + %sext = sext i8 %load to i32 + store i32 %sext, i32 addrspace(1)* %out + ret void +} +define void @i32_array_array(i32 addrspace(1)* %out, i32 %index) { +entry: + %alloca = alloca [2 x [2 x i32]] + %gep0 = getelementptr [2 x [2 x i32]]* %alloca, i32 0, i32 0, i32 0 + %gep1 = getelementptr [2 x [2 x i32]]* %alloca, i32 0, i32 0, i32 1 + store i32 0, i32* %gep0 + store i32 1, i32* %gep1 + %gep2 = getelementptr [2 x [2 x i32]]* %alloca, i32 0, i32 0, i32 %index + %load = load i32* %gep2 + store i32 %load, i32 addrspace(1)* %out + ret void +} + +define void @i64_array_array(i64 addrspace(1)* %out, i32 %index) { +entry: + %alloca = alloca [2 x [2 x i64]] + %gep0 = getelementptr [2 x [2 x i64]]* %alloca, i32 0, i32 0, i32 0 + %gep1 = getelementptr [2 x [2 x i64]]* %alloca, i32 0, i32 0, i32 1 + store i64 0, i64* %gep0 + store i64 1, i64* %gep1 + %gep2 = getelementptr [2 x [2 x i64]]* %alloca, i32 0, i32 0, i32 %index + %load = load i64* %gep2 + store i64 %load, i64 addrspace(1)* %out + ret void +} + +%struct.pair32 = type { i32, i32 } + +define void @struct_array_array(i32 addrspace(1)* %out, i32 %index) { +entry: + %alloca = alloca [2 x [2 x %struct.pair32]] + %gep0 = getelementptr [2 x [2 x %struct.pair32]]* %alloca, i32 0, i32 0, i32 0, i32 1 + %gep1 = getelementptr [2 x [2 x %struct.pair32]]* %alloca, i32 0, i32 0, i32 1, i32 1 + store i32 0, i32* %gep0 + store i32 1, i32* %gep1 + %gep2 = getelementptr [2 x [2 x %struct.pair32]]* %alloca, i32 0, i32 0, i32 %index, i32 0 + %load = load i32* %gep2 + store i32 %load, i32 addrspace(1)* %out + ret void +} + +define void @struct_pair32_array(i32 addrspace(1)* %out, i32 %index) { +entry: + %alloca = alloca [2 x %struct.pair32] + %gep0 = getelementptr [2 x %struct.pair32]* %alloca, i32 0, i32 0, i32 1 + %gep1 = getelementptr [2 x %struct.pair32]* %alloca, i32 0, i32 1, i32 0 + store i32 0, i32* %gep0 + store i32 1, i32* %gep1 + %gep2 = getelementptr [2 x %struct.pair32]* %alloca, i32 0, i32 %index, i32 0 + %load = load i32* %gep2 + store i32 %load, i32 addrspace(1)* %out + ret void +} + +define void @select_private(i32 addrspace(1)* %out, i32 %in) nounwind { +entry: + %tmp = alloca [2 x i32] + %tmp1 = getelementptr [2 x i32]* %tmp, i32 0, i32 0 + %tmp2 = getelementptr [2 x i32]* %tmp, i32 0, i32 1 + store i32 0, i32* %tmp1 + store i32 1, i32* %tmp2 + %cmp = icmp eq i32 %in, 0 + %sel = select i1 %cmp, i32* %tmp1, i32* %tmp2 + %load = load i32* %sel + store i32 %load, i32 addrspace(1)* %out + ret void +} -declare i32 @llvm.r600.read.tidig.x() nounwind readnone |