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
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
12 define void @t1(i32 %argc, i8** %argv) nounwind {
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)
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)
29 ; SSE1: movaps _.str, %xmm0
31 ; SSE1: movb $0, 24(%esp)
32 ; SSE1: movl $0, 20(%esp)
33 ; SSE1: movl $0, 16(%esp)
42 ; NOSSE: movl $1734438249
45 ; X86-64: movaps _.str(%rip), %xmm0
46 ; X86-64: movaps %xmm0
49 %tmp1 = alloca [25 x i8]
50 %tmp2 = bitcast [25 x i8]* %tmp1 to i8*
51 call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 1 %tmp2, i8* align 1 getelementptr inbounds ([25 x i8], [25 x i8]* @.str, i32 0, i32 0), i32 25, i1 false)
56 %struct.s0 = type { [2 x double] }
58 define void @t2(%struct.s0* nocapture %a, %struct.s0* nocapture %b) nounwind ssp {
60 ; SSE2-Darwin-LABEL: t2:
61 ; SSE2-Darwin: movaps (%ecx), %xmm0
62 ; SSE2-Darwin: movaps %xmm0, (%eax)
64 ; SSE2-Mingw32-LABEL: t2:
65 ; SSE2-Mingw32: movaps (%ecx), %xmm0
66 ; SSE2-Mingw32: movaps %xmm0, (%eax)
69 ; SSE1: movaps (%ecx), %xmm0
70 ; SSE1: movaps %xmm0, (%eax)
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* align 16 %tmp2, i8* align 16 %tmp3, i32 16, i1 false)
93 define void @t3(%struct.s0* nocapture %a, %struct.s0* nocapture %b) nounwind ssp {
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)
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)
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* align 8 %tmp2, i8* align 8 %tmp3, i32 16, i1 false)
142 define void @t4() nounwind {
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
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
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
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
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.
187 ;;; Is either of the sequences ideal?
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)
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)
203 %tmp1 = alloca [30 x i8]
204 %tmp2 = bitcast [30 x i8]* %tmp1 to i8*
205 call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 1 %tmp2, i8* align 1 getelementptr inbounds ([30 x i8], [30 x i8]* @.str2, i32 0, i32 0), i32 30, i1 false)
209 declare void @llvm.memcpy.p0i8.p0i8.i32(i8* nocapture, i8* nocapture, i32, i1) nounwind