1; RUN: llc < %s -mattr=+sse2 -mtriple=i686-apple-darwin -mcpu=core2 | FileCheck %s -check-prefix=SSE2-Darwin 2; RUN: llc < %s -mattr=+sse2 -mtriple=i686-pc-mingw32 -mcpu=core2 | FileCheck %s -check-prefix=SSE2-Mingw32 3; RUN: llc < %s -mattr=+sse,-sse2 -mtriple=i686-apple-darwin -mcpu=core2 | FileCheck %s -check-prefix=SSE1 4; RUN: llc < %s -mattr=-sse -mtriple=i686-apple-darwin -mcpu=core2 | FileCheck %s -check-prefix=NOSSE 5; RUN: llc < %s -mtriple=x86_64-apple-darwin -mcpu=core2 | FileCheck %s -check-prefix=X86-64 6; RUN: llc < %s -mtriple=x86_64-apple-darwin -mcpu=nehalem | FileCheck %s -check-prefix=NHM_64 7 8 9@.str = internal constant [25 x i8] c"image\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00\00" 10@.str2 = internal constant [30 x i8] c"xxxxxxxxxxxxxxxxxxxxxxxxxxxxx\00", align 4 11 12define void @t1(i32 %argc, i8** %argv) nounwind { 13entry: 14; SSE2-Darwin-LABEL: t1: 15; SSE2-Darwin: movsd _.str+16, %xmm0 16; SSE2-Darwin: movsd %xmm0, 16(%esp) 17; SSE2-Darwin: movaps _.str, %xmm0 18; SSE2-Darwin: movaps %xmm0 19; SSE2-Darwin: movb $0, 24(%esp) 20 21; SSE2-Mingw32-LABEL: t1: 22; SSE2-Mingw32: movsd _.str+16, %xmm0 23; SSE2-Mingw32: movsd %xmm0, 16(%esp) 24; SSE2-Mingw32: movaps _.str, %xmm0 25; SSE2-Mingw32: movups %xmm0 26; SSE2-Mingw32: movb $0, 24(%esp) 27 28; SSE1-LABEL: t1: 29; SSE1: movaps _.str, %xmm0 30; SSE1: movaps %xmm0 31; SSE1: movb $0, 24(%esp) 32; SSE1: movl $0, 20(%esp) 33; SSE1: movl $0, 16(%esp) 34 35; NOSSE-LABEL: t1: 36; NOSSE: movb $0 37; NOSSE: movl $0 38; NOSSE: movl $0 39; NOSSE: movl $0 40; NOSSE: movl $0 41; NOSSE: movl $101 42; NOSSE: movl $1734438249 43 44; X86-64-LABEL: t1: 45; X86-64: movaps _.str(%rip), %xmm0 46; X86-64: movaps %xmm0 47; X86-64: movb $0 48; X86-64: movq $0 49 %tmp1 = alloca [25 x i8] 50 %tmp2 = bitcast [25 x i8]* %tmp1 to i8* 51 call void @llvm.memcpy.p0i8.p0i8.i32(i8* %tmp2, i8* getelementptr inbounds ([25 x i8], [25 x i8]* @.str, i32 0, i32 0), i32 25, i32 1, i1 false) 52 unreachable 53} 54 55;rdar://7774704 56%struct.s0 = type { [2 x double] } 57 58define void @t2(%struct.s0* nocapture %a, %struct.s0* nocapture %b) nounwind ssp { 59entry: 60; SSE2-Darwin-LABEL: t2: 61; SSE2-Darwin: movaps (%ecx), %xmm0 62; SSE2-Darwin: movaps %xmm0, (%eax) 63 64; SSE2-Mingw32-LABEL: t2: 65; SSE2-Mingw32: movaps (%ecx), %xmm0 66; SSE2-Mingw32: movaps %xmm0, (%eax) 67 68; SSE1-LABEL: t2: 69; SSE1: movaps (%ecx), %xmm0 70; SSE1: movaps %xmm0, (%eax) 71 72; NOSSE-LABEL: t2: 73; NOSSE: movl 74; NOSSE: movl 75; NOSSE: movl 76; NOSSE: movl 77; NOSSE: movl 78; NOSSE: movl 79; NOSSE: movl 80; NOSSE: movl 81; NOSSE: movl 82; NOSSE: movl 83 84; X86-64-LABEL: t2: 85; X86-64: movaps (%rsi), %xmm0 86; X86-64: movaps %xmm0, (%rdi) 87 %tmp2 = bitcast %struct.s0* %a to i8* ; <i8*> [#uses=1] 88 %tmp3 = bitcast %struct.s0* %b to i8* ; <i8*> [#uses=1] 89 tail call void @llvm.memcpy.p0i8.p0i8.i32(i8* %tmp2, i8* %tmp3, i32 16, i32 16, i1 false) 90 ret void 91} 92 93define void @t3(%struct.s0* nocapture %a, %struct.s0* nocapture %b) nounwind ssp { 94entry: 95; SSE2-Darwin-LABEL: t3: 96; SSE2-Darwin: movsd (%ecx), %xmm0 97; SSE2-Darwin: movsd 8(%ecx), %xmm1 98; SSE2-Darwin: movsd %xmm1, 8(%eax) 99; SSE2-Darwin: movsd %xmm0, (%eax) 100 101; SSE2-Mingw32-LABEL: t3: 102; SSE2-Mingw32: movsd (%ecx), %xmm0 103; SSE2-Mingw32: movsd 8(%ecx), %xmm1 104; SSE2-Mingw32: movsd %xmm1, 8(%eax) 105; SSE2-Mingw32: movsd %xmm0, (%eax) 106 107; SSE1-LABEL: t3: 108; SSE1: movl 109; SSE1: movl 110; SSE1: movl 111; SSE1: movl 112; SSE1: movl 113; SSE1: movl 114; SSE1: movl 115; SSE1: movl 116; SSE1: movl 117; SSE1: movl 118 119; NOSSE-LABEL: t3: 120; NOSSE: movl 121; NOSSE: movl 122; NOSSE: movl 123; NOSSE: movl 124; NOSSE: movl 125; NOSSE: movl 126; NOSSE: movl 127; NOSSE: movl 128; NOSSE: movl 129; NOSSE: movl 130 131; X86-64-LABEL: t3: 132; X86-64: movq (%rsi), %rax 133; X86-64: movq 8(%rsi), %rcx 134; X86-64: movq %rcx, 8(%rdi) 135; X86-64: movq %rax, (%rdi) 136 %tmp2 = bitcast %struct.s0* %a to i8* ; <i8*> [#uses=1] 137 %tmp3 = bitcast %struct.s0* %b to i8* ; <i8*> [#uses=1] 138 tail call void @llvm.memcpy.p0i8.p0i8.i32(i8* %tmp2, i8* %tmp3, i32 16, i32 8, i1 false) 139 ret void 140} 141 142define void @t4() nounwind { 143entry: 144; SSE2-Darwin-LABEL: t4: 145; SSE2-Darwin: movw $120 146; SSE2-Darwin: movl $2021161080 147; SSE2-Darwin: movl $2021161080 148; SSE2-Darwin: movl $2021161080 149; SSE2-Darwin: movl $2021161080 150; SSE2-Darwin: movl $2021161080 151; SSE2-Darwin: movl $2021161080 152; SSE2-Darwin: movl $2021161080 153 154; SSE2-Mingw32-LABEL: t4: 155; SSE2-Mingw32: movw $120 156; SSE2-Mingw32: movl $2021161080 157; SSE2-Mingw32: movl $2021161080 158; SSE2-Mingw32: movl $2021161080 159; SSE2-Mingw32: movl $2021161080 160; SSE2-Mingw32: movl $2021161080 161; SSE2-Mingw32: movl $2021161080 162; SSE2-Mingw32: movl $2021161080 163 164; SSE1-LABEL: t4: 165; SSE1: movw $120 166; SSE1: movl $2021161080 167; SSE1: movl $2021161080 168; SSE1: movl $2021161080 169; SSE1: movl $2021161080 170; SSE1: movl $2021161080 171; SSE1: movl $2021161080 172; SSE1: movl $2021161080 173 174; NOSSE-LABEL: t4: 175; NOSSE: movw $120 176; NOSSE: movl $2021161080 177; NOSSE: movl $2021161080 178; NOSSE: movl $2021161080 179; NOSSE: movl $2021161080 180; NOSSE: movl $2021161080 181; NOSSE: movl $2021161080 182; NOSSE: movl $2021161080 183 184;;; TODO: (1) Some of the loads and stores are certainly unaligned and (2) the first load and first 185;;; store overlap with the second load and second store respectively. 186;;; 187;;; Is either of the sequences ideal? 188 189; X86-64-LABEL: t4: 190; X86-64: movabsq $33909456017848440, %rax ## imm = 0x78787878787878 191; X86-64: movq %rax, -10(%rsp) 192; X86-64: movabsq $8680820740569200760, %rax ## imm = 0x7878787878787878 193; X86-64: movq %rax, -16(%rsp) 194; X86-64: movq %rax, -24(%rsp) 195; X86-64: movq %rax, -32(%rsp) 196 197; NHM_64-LABEL: t4: 198; NHM_64: movups _.str2+14(%rip), %xmm0 199; NHM_64: movups %xmm0, -26(%rsp) 200; NHM_64: movups _.str2(%rip), %xmm0 201; NHM_64: movaps %xmm0, -40(%rsp) 202 203 %tmp1 = alloca [30 x i8] 204 %tmp2 = bitcast [30 x i8]* %tmp1 to i8* 205 call void @llvm.memcpy.p0i8.p0i8.i32(i8* %tmp2, i8* getelementptr inbounds ([30 x i8], [30 x i8]* @.str2, i32 0, i32 0), i32 30, i32 1, i1 false) 206 unreachable 207} 208 209declare void @llvm.memcpy.p0i8.p0i8.i32(i8* nocapture, i8* nocapture, i32, i32, i1) nounwind 210