From c6faf15459ad276ed8e44341783d4d663d8a4d2a Mon Sep 17 00:00:00 2001 From: Hiroshi Inoue Date: Mon, 18 Dec 2017 06:47:37 +0000 Subject: [PATCH] [SROA] Disable non-whole-alloca splits by default This patch introduce a switch to control splitting of non-whole-alloca slices with default off. The switch will be default on again after fixing an issue reported in PR35657. llvm-svn: 320958 --- llvm/lib/Transforms/Scalar/SROA.cpp | 7 +++++- llvm/test/DebugInfo/X86/sroasplit-2.ll | 3 +-- llvm/test/Transforms/SROA/basictest.ll | 33 ++++----------------------- llvm/test/Transforms/SROA/big-endian.ll | 40 ++++++++++----------------------- 4 files changed, 23 insertions(+), 60 deletions(-) diff --git a/llvm/lib/Transforms/Scalar/SROA.cpp b/llvm/lib/Transforms/Scalar/SROA.cpp index 2e502339..bfe3754 100644 --- a/llvm/lib/Transforms/Scalar/SROA.cpp +++ b/llvm/lib/Transforms/Scalar/SROA.cpp @@ -124,6 +124,11 @@ static cl::opt SROARandomShuffleSlices("sroa-random-shuffle-slices", static cl::opt SROAStrictInbounds("sroa-strict-inbounds", cl::init(false), cl::Hidden); +/// Hidden option to allow more aggressive splitting. +static cl::opt +SROASplitNonWholeAllocaSlices("sroa-split-nonwhole-alloca-slices", + cl::init(false), cl::Hidden); + namespace { /// \brief A custom IRBuilder inserter which prefixes all names, but only in @@ -4058,7 +4063,7 @@ bool SROA::splitAlloca(AllocaInst &AI, AllocaSlices &AS) { uint64_t AllocaSize = DL.getTypeAllocSize(AI.getAllocatedType()); const uint64_t MaxBitVectorSize = 1024; - if (AllocaSize <= MaxBitVectorSize) { + if (SROASplitNonWholeAllocaSlices && AllocaSize <= MaxBitVectorSize) { // If a byte boundary is included in any load or store, a slice starting or // ending at the boundary is not splittable. SmallBitVector SplittableOffset(AllocaSize + 1, true); diff --git a/llvm/test/DebugInfo/X86/sroasplit-2.ll b/llvm/test/DebugInfo/X86/sroasplit-2.ll index 3e99ec1..b2bec7ced 100644 --- a/llvm/test/DebugInfo/X86/sroasplit-2.ll +++ b/llvm/test/DebugInfo/X86/sroasplit-2.ll @@ -21,8 +21,7 @@ ; Verify that SROA creates a variable piece when splitting i1. ; CHECK: call void @llvm.dbg.value(metadata i64 %outer.coerce0, metadata ![[O:[0-9]+]], metadata !DIExpression(DW_OP_LLVM_fragment, 0, 64)), -; CHECK: call void @llvm.dbg.value(metadata i32 {{.*}}, metadata ![[O]], metadata !DIExpression(DW_OP_LLVM_fragment, 64, 32)), -; CHECK: call void @llvm.dbg.value(metadata i32 {{.*}}, metadata ![[O]], metadata !DIExpression(DW_OP_LLVM_fragment, 96, 32)), +; CHECK: call void @llvm.dbg.value(metadata i64 %outer.coerce1, metadata ![[O]], metadata !DIExpression(DW_OP_LLVM_fragment, 64, 64)), ; CHECK: call void @llvm.dbg.value({{.*}}, metadata ![[I1:[0-9]+]], metadata !DIExpression(DW_OP_LLVM_fragment, 0, 32)), ; CHECK-DAG: ![[O]] = !DILocalVariable(name: "outer",{{.*}} line: 10 ; CHECK-DAG: ![[I1]] = !DILocalVariable(name: "i1",{{.*}} line: 11 diff --git a/llvm/test/Transforms/SROA/basictest.ll b/llvm/test/Transforms/SROA/basictest.ll index 9cf2191..aa00e89 100644 --- a/llvm/test/Transforms/SROA/basictest.ll +++ b/llvm/test/Transforms/SROA/basictest.ll @@ -1615,13 +1615,13 @@ define i16 @PR24463() { ; Ensure we can handle a very interesting case where there is an integer-based ; rewrite of the uses of the alloca, but where one of the integers in that is ; a sub-integer that requires extraction *and* extends past the end of the -; alloca. SROA can split the alloca to avoid shift or trunc. +; alloca. In this case, we should extract the i8 and then zext it to i16. ; ; CHECK-LABEL: @PR24463( ; CHECK-NOT: alloca -; CHECK-NOT: trunc -; CHECK-NOT: lshr -; CHECK: %[[ZEXT:.*]] = zext i8 {{.*}} to i16 +; CHECK: %[[SHIFT:.*]] = lshr i16 0, 8 +; CHECK: %[[TRUNC:.*]] = trunc i16 %[[SHIFT]] to i8 +; CHECK: %[[ZEXT:.*]] = zext i8 %[[TRUNC]] to i16 ; CHECK: ret i16 %[[ZEXT]] entry: %alloca = alloca [3 x i8] @@ -1695,28 +1695,3 @@ bb1: call void @llvm.lifetime.end.p0i8(i64 2, i8* %0) ret void } - -define void @test28(i64 %v) #0 { -; SROA should split the first i64 store to avoid additional and/or instructions -; when storing into i32 fields - -; CHECK-LABEL: @test28( -; CHECK-NOT: alloca -; CHECK-NOT: and -; CHECK-NOT: or -; CHECK: %[[shift:.*]] = lshr i64 %v, 32 -; CHECK-NEXT: %{{.*}} = trunc i64 %[[shift]] to i32 -; CHECK-NEXT: ret void - -entry: - %t = alloca { i64, i32, i32 } - - %b = getelementptr { i64, i32, i32 }, { i64, i32, i32 }* %t, i32 0, i32 1 - %0 = bitcast i32* %b to i64* - store i64 %v, i64* %0 - - %1 = load i32, i32* %b - %c = getelementptr { i64, i32, i32 }, { i64, i32, i32 }* %t, i32 0, i32 2 - store i32 %1, i32* %c - ret void -} diff --git a/llvm/test/Transforms/SROA/big-endian.ll b/llvm/test/Transforms/SROA/big-endian.ll index fc4b8b2..ea41a20 100644 --- a/llvm/test/Transforms/SROA/big-endian.ll +++ b/llvm/test/Transforms/SROA/big-endian.ll @@ -83,34 +83,19 @@ entry: store i16 1, i16* %a0i16ptr store i8 1, i8* %a2ptr +; CHECK: %[[mask1:.*]] = and i40 undef, 4294967295 +; CHECK-NEXT: %[[insert1:.*]] = or i40 %[[mask1]], 4294967296 %a3i24ptr = bitcast i8* %a3ptr to i24* store i24 1, i24* %a3i24ptr +; CHECK-NEXT: %[[mask2:.*]] = and i40 %[[insert1]], -4294967041 +; CHECK-NEXT: %[[insert2:.*]] = or i40 %[[mask2]], 256 %a2i40ptr = bitcast i8* %a2ptr to i40* store i40 1, i40* %a2i40ptr - -; the alloca is splitted into multiple slices -; Here, i8 1 is for %a[6] -; CHECK: %[[ext1:.*]] = zext i8 1 to i40 -; CHECK-NEXT: %[[mask1:.*]] = and i40 undef, -256 -; CHECK-NEXT: %[[insert1:.*]] = or i40 %[[mask1]], %[[ext1]] - -; Here, i24 0 is for %a[3] to %a[5] -; CHECK-NEXT: %[[ext2:.*]] = zext i24 0 to i40 -; CHECK-NEXT: %[[shift2:.*]] = shl i40 %[[ext2]], 8 -; CHECK-NEXT: %[[mask2:.*]] = and i40 %[[insert1]], -4294967041 -; CHECK-NEXT: %[[insert2:.*]] = or i40 %[[mask2]], %[[shift2]] - -; Here, i8 0 is for %a[2] -; CHECK-NEXT: %[[ext3:.*]] = zext i8 0 to i40 -; CHECK-NEXT: %[[shift3:.*]] = shl i40 %[[ext3]], 32 -; CHECK-NEXT: %[[mask3:.*]] = and i40 %[[insert2]], 4294967295 -; CHECK-NEXT: %[[insert3:.*]] = or i40 %[[mask3]], %[[shift3]] - -; CHECK-NEXT: %[[ext4:.*]] = zext i40 %[[insert3]] to i56 -; CHECK-NEXT: %[[mask4:.*]] = and i56 undef, -1099511627776 -; CHECK-NEXT: %[[insert4:.*]] = or i56 %[[mask4]], %[[ext4]] +; CHECK-NEXT: %[[ext3:.*]] = zext i40 1 to i56 +; CHECK-NEXT: %[[mask3:.*]] = and i56 undef, -1099511627776 +; CHECK-NEXT: %[[insert3:.*]] = or i56 %[[mask3]], %[[ext3]] ; CHECK-NOT: store ; CHECK-NOT: load @@ -119,12 +104,11 @@ entry: %ai = load i56, i56* %aiptr %ret = zext i56 %ai to i64 ret i64 %ret -; Here, i16 1 is for %a[0] to %a[1] -; CHECK-NEXT: %[[ext5:.*]] = zext i16 1 to i56 -; CHECK-NEXT: %[[shift5:.*]] = shl i56 %[[ext5]], 40 -; CHECK-NEXT: %[[mask5:.*]] = and i56 %[[insert4]], 1099511627775 -; CHECK-NEXT: %[[insert5:.*]] = or i56 %[[mask5]], %[[shift5]] -; CHECK-NEXT: %[[ret:.*]] = zext i56 %[[insert5]] to i64 +; CHECK-NEXT: %[[ext4:.*]] = zext i16 1 to i56 +; CHECK-NEXT: %[[shift4:.*]] = shl i56 %[[ext4]], 40 +; CHECK-NEXT: %[[mask4:.*]] = and i56 %[[insert3]], 1099511627775 +; CHECK-NEXT: %[[insert4:.*]] = or i56 %[[mask4]], %[[shift4]] +; CHECK-NEXT: %[[ret:.*]] = zext i56 %[[insert4]] to i64 ; CHECK-NEXT: ret i64 %[[ret]] } -- 2.7.4