diff options
| author | Nadav Rotem <nrotem@apple.com> | 2012-10-04 22:35:15 +0000 |
|---|---|---|
| committer | Nadav Rotem <nrotem@apple.com> | 2012-10-04 22:35:15 +0000 |
| commit | ea2c50c0416555a91cf963618f07c90a4c791708 (patch) | |
| tree | ba795472b86cd373ae15558a033453e979586ad3 /test/CodeGen | |
| parent | e18c2ae7b2c725ccac041973f3df9ce20bfab26e (diff) | |
| download | llvm-ea2c50c0416555a91cf963618f07c90a4c791708.tar.gz llvm-ea2c50c0416555a91cf963618f07c90a4c791708.tar.bz2 llvm-ea2c50c0416555a91cf963618f07c90a4c791708.tar.xz | |
When merging connsecutive stores, use vectors to store the constant zero.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@165267 91177308-0d34-0410-b5e6-96231b3b80d8
Diffstat (limited to 'test/CodeGen')
| -rw-r--r-- | test/CodeGen/X86/MergeConsecutiveStores.ll | 38 |
1 files changed, 35 insertions, 3 deletions
diff --git a/test/CodeGen/X86/MergeConsecutiveStores.ll b/test/CodeGen/X86/MergeConsecutiveStores.ll index 79f8ee54a2..64825bac97 100644 --- a/test/CodeGen/X86/MergeConsecutiveStores.ll +++ b/test/CodeGen/X86/MergeConsecutiveStores.ll @@ -1,4 +1,4 @@ -; RUN: llc -march=x86-64 -mcpu=corei7 < %s | FileCheck %s +; RUN: llc -march=x86-64 -mcpu=corei7 -mattr=+avx < %s | FileCheck %s target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128" target triple = "x86_64-apple-macosx10.8.0" @@ -6,7 +6,6 @@ target triple = "x86_64-apple-macosx10.8.0" %struct.A = type { i8, i8, i8, i8, i8, i8, i8, i8 } %struct.B = type { i32, i32, i32, i32, i32, i32, i32, i32 } -; Move all of the constants using a single vector store. ; CHECK: merge_const_store ; save 1,2,3 ... as one big integer. ; CHECK: movabsq $578437695752307201 @@ -41,6 +40,40 @@ define void @merge_const_store(i32 %count, %struct.A* nocapture %p) nounwind uwt ret void } +; Move the constants using a single vector store. +; CHECK: merge_const_store_vec +; CHECK: vmovups %ymm0, (%rsi) +; CHECK: ret +define void @merge_const_store_vec(i32 %count, %struct.B* nocapture %p) nounwind uwtable noinline ssp { + %1 = icmp sgt i32 %count, 0 + br i1 %1, label %.lr.ph, label %._crit_edge +.lr.ph: + %i.02 = phi i32 [ %10, %.lr.ph ], [ 0, %0 ] + %.01 = phi %struct.B* [ %11, %.lr.ph ], [ %p, %0 ] + %2 = getelementptr inbounds %struct.B* %.01, i64 0, i32 0 + store i32 0, i32* %2, align 4 + %3 = getelementptr inbounds %struct.B* %.01, i64 0, i32 1 + store i32 0, i32* %3, align 4 + %4 = getelementptr inbounds %struct.B* %.01, i64 0, i32 2 + store i32 0, i32* %4, align 4 + %5 = getelementptr inbounds %struct.B* %.01, i64 0, i32 3 + store i32 0, i32* %5, align 4 + %6 = getelementptr inbounds %struct.B* %.01, i64 0, i32 4 + store i32 0, i32* %6, align 4 + %7 = getelementptr inbounds %struct.B* %.01, i64 0, i32 5 + store i32 0, i32* %7, align 4 + %8 = getelementptr inbounds %struct.B* %.01, i64 0, i32 6 + store i32 0, i32* %8, align 4 + %9 = getelementptr inbounds %struct.B* %.01, i64 0, i32 7 + store i32 0, i32* %9, align 4 + %10 = add nsw i32 %i.02, 1 + %11 = getelementptr inbounds %struct.B* %.01, i64 1 + %exitcond = icmp eq i32 %10, %count + br i1 %exitcond, label %._crit_edge, label %.lr.ph +._crit_edge: + ret void +} + ; Move the first 4 constants as a single vector. Move the rest as scalars. ; CHECK: merge_nonconst_store ; CHECK: movl $67305985 @@ -223,7 +256,6 @@ block4: ; preds = %4, %.lr.ph ret void } - ;CHECK: merge_loads_no_align ; load: ;CHECK: movl |