diff options
Diffstat (limited to 'llvm/test/Transforms/LoopVectorize/AMDGPU')
4 files changed, 93 insertions, 0 deletions
diff --git a/llvm/test/Transforms/LoopVectorize/AMDGPU/divergent-runtime-check.ll b/llvm/test/Transforms/LoopVectorize/AMDGPU/divergent-runtime-check.ll new file mode 100644 index 00000000000..91a916798c5 --- /dev/null +++ b/llvm/test/Transforms/LoopVectorize/AMDGPU/divergent-runtime-check.ll @@ -0,0 +1,29 @@ +; RUN: opt -S -mtriple=amdgcn-unknown-amdhsa -mcpu=gfx900 -loop-vectorize -simplifycfg < %s | FileCheck -check-prefixes=GCN,GFX9 %s +; RUN: opt -S -mtriple=amdgcn-unknown-amdhsa -mcpu=gfx900 -loop-vectorize -pass-remarks-analysis='loop-vectorize' < %s 2>&1 | FileCheck -check-prefixes=REMARK %s + +; GCN-LABEL: @runtime_check_divergent_target( +; GCN-NOT: load <2 x half> +; GCN-NOT: store <2 x half> + +; REMARK: remark: <unknown>:0:0: loop not vectorized: runtime pointer checks needed. Not enabled for divergent target +define amdgpu_kernel void @runtime_check_divergent_target(half addrspace(1)* nocapture %a, half addrspace(1)* nocapture %b) #0 { +entry: + br label %for.body + +for.body: ; preds = %entry, %for.body + %indvars.iv = phi i64 [ %indvars.iv.next, %for.body ], [ 0, %entry ] + %arrayidx = getelementptr inbounds half, half addrspace(1)* %b, i64 %indvars.iv + %load = load half, half addrspace(1)* %arrayidx, align 4 + %mul = fmul half %load, 3.0 + %arrayidx2 = getelementptr inbounds half, half addrspace(1)* %a, i64 %indvars.iv + store half %mul, half addrspace(1)* %arrayidx2, align 4 + %indvars.iv.next = add i64 %indvars.iv, 1 + %lftr.wideiv = trunc i64 %indvars.iv.next to i32 + %exitcond = icmp eq i32 %lftr.wideiv, 1024 + br i1 %exitcond, label %for.end, label %for.body + +for.end: ; preds = %for.body, %entry + ret void +} + +attributes #0 = { nounwind } diff --git a/llvm/test/Transforms/LoopVectorize/AMDGPU/lit.local.cfg b/llvm/test/Transforms/LoopVectorize/AMDGPU/lit.local.cfg new file mode 100644 index 00000000000..2a665f06be7 --- /dev/null +++ b/llvm/test/Transforms/LoopVectorize/AMDGPU/lit.local.cfg @@ -0,0 +1,2 @@ +if not 'AMDGPU' in config.root.targets: + config.unsupported = True diff --git a/llvm/test/Transforms/LoopVectorize/AMDGPU/packed-math.ll b/llvm/test/Transforms/LoopVectorize/AMDGPU/packed-math.ll new file mode 100644 index 00000000000..832843983eb --- /dev/null +++ b/llvm/test/Transforms/LoopVectorize/AMDGPU/packed-math.ll @@ -0,0 +1,34 @@ +; RUN: opt -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 < %s -loop-vectorize -dce -instcombine -S | FileCheck -check-prefix=GFX9 -check-prefix=GCN %s +; RUN: opt -mtriple=amdgcn-amd-amdhsa -mcpu=fiji < %s -loop-vectorize -dce -instcombine -S | FileCheck -check-prefix=CIVI -check-prefix=GCN %s +; RUN: opt -mtriple=amdgcn-amd-amdhsa -mcpu=hawaii < %s -loop-vectorize -dce -instcombine -S | FileCheck -check-prefix=CIVI -check-prefix=GCN %s + +; GCN-LABEL: @vectorize_v2f16_loop( +; GFX9: vector.body: +; GFX9: phi <2 x half> +; GFX9: load <2 x half> +; GFX9: fadd fast <2 x half> + +; GFX9: middle.block: +; GFX9: fadd fast <2 x half> + +; VI: phi half +; VI: phi load half +; VI: fadd fast half +define half @vectorize_v2f16_loop(half addrspace(1)* noalias %s) { +entry: + br label %for.body + +for.body: + %indvars.iv = phi i64 [ 0, %entry ], [ %indvars.iv.next, %for.body ] + %q.04 = phi half [ 0.0, %entry ], [ %add, %for.body ] + %arrayidx = getelementptr inbounds half, half addrspace(1)* %s, i64 %indvars.iv + %0 = load half, half addrspace(1)* %arrayidx, align 2 + %add = fadd fast half %q.04, %0 + %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1 + %exitcond = icmp eq i64 %indvars.iv.next, 256 + br i1 %exitcond, label %for.end, label %for.body + +for.end: + %add.lcssa = phi half [ %add, %for.body ] + ret half %add.lcssa +} diff --git a/llvm/test/Transforms/LoopVectorize/AMDGPU/unroll-in-loop-vectorizer.ll b/llvm/test/Transforms/LoopVectorize/AMDGPU/unroll-in-loop-vectorizer.ll new file mode 100644 index 00000000000..f303ed5377e --- /dev/null +++ b/llvm/test/Transforms/LoopVectorize/AMDGPU/unroll-in-loop-vectorizer.ll @@ -0,0 +1,28 @@ +; RUN: opt -S -mtriple=amdgcn-unknown-amdhsa -mcpu=fiji -loop-vectorize < %s | FileCheck %s + + +; For AMDGPU, loop unroll in loop vectorizer is disabled when VF==1. +; +; CHECK-LABEL: @small_loop( +; CHECK: store i32 +; CHECK-NOT: store i32 +; CHECK: ret +define amdgpu_kernel void @small_loop(i32* nocapture %inArray, i32 %size) nounwind { +entry: + %0 = icmp sgt i32 %size, 0 + br i1 %0, label %loop, label %exit + +loop: ; preds = %entry, %loop + %iv = phi i32 [ %iv1, %loop ], [ 0, %entry ] + %1 = getelementptr inbounds i32, i32* %inArray, i32 %iv + %2 = load i32, i32* %1, align 4 + %3 = add nsw i32 %2, 6 + store i32 %3, i32* %1, align 4 + %iv1 = add i32 %iv, 1 +; %lftr.wideiv = trunc i64 %indvars.iv.next to i32 + %cond = icmp eq i32 %iv1, %size + br i1 %cond, label %exit, label %loop + +exit: ; preds = %loop, %entry + ret void +} |
