When merging connsecutive stores, use vectors to store the constant zero.
llvm-svn: 165267
diff --git a/llvm/test/CodeGen/X86/MergeConsecutiveStores.ll b/llvm/test/CodeGen/X86/MergeConsecutiveStores.ll
index 79f8ee5..64825ba 100644
--- a/llvm/test/CodeGen/X86/MergeConsecutiveStores.ll
+++ b/llvm/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 @@
%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 @@
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 @@
ret void
}
-
;CHECK: merge_loads_no_align
; load:
;CHECK: movl