1 ; RUN: llc < %s -mtriple=i686-apple-darwin9 -mattr=sse41 | FileCheck %s -check-prefix=X32
2 ; RUN: llc < %s -mtriple=x86_64-apple-darwin9 -mattr=sse41 | FileCheck %s -check-prefix=X64
4 @g16 = external global i16
6 define <4 x i32> @pinsrd_1(i32 %s, <4 x i32> %tmp) nounwind {
7 %tmp1 = insertelement <4 x i32> %tmp, i32 %s, i32 1
10 ; X32: pinsrd $1, 4(%esp), %xmm0
13 ; X64: pinsrd $1, %edi, %xmm0
16 define <16 x i8> @pinsrb_1(i8 %s, <16 x i8> %tmp) nounwind {
17 %tmp1 = insertelement <16 x i8> %tmp, i8 %s, i32 1
20 ; X32: pinsrb $1, 4(%esp), %xmm0
23 ; X64: pinsrb $1, %edi, %xmm0
27 define <2 x i64> @pmovsxbd_1(i32* %p) nounwind {
29 %0 = load i32* %p, align 4
30 %1 = insertelement <4 x i32> undef, i32 %0, i32 0
31 %2 = insertelement <4 x i32> %1, i32 0, i32 1
32 %3 = insertelement <4 x i32> %2, i32 0, i32 2
33 %4 = insertelement <4 x i32> %3, i32 0, i32 3
34 %5 = bitcast <4 x i32> %4 to <16 x i8>
35 %6 = tail call <4 x i32> @llvm.x86.sse41.pmovsxbd(<16 x i8> %5) nounwind readnone
36 %7 = bitcast <4 x i32> %6 to <2 x i64>
40 ; X32: movl 4(%esp), %eax
41 ; X32: pmovsxbd (%eax), %xmm0
44 ; X64: pmovsxbd (%rdi), %xmm0
47 define <2 x i64> @pmovsxwd_1(i64* %p) nounwind readonly {
49 %0 = load i64* %p ; <i64> [#uses=1]
50 %tmp2 = insertelement <2 x i64> zeroinitializer, i64 %0, i32 0 ; <<2 x i64>> [#uses=1]
51 %1 = bitcast <2 x i64> %tmp2 to <8 x i16> ; <<8 x i16>> [#uses=1]
52 %2 = tail call <4 x i32> @llvm.x86.sse41.pmovsxwd(<8 x i16> %1) nounwind readnone ; <<4 x i32>> [#uses=1]
53 %3 = bitcast <4 x i32> %2 to <2 x i64> ; <<2 x i64>> [#uses=1]
57 ; X32: movl 4(%esp), %eax
58 ; X32: pmovsxwd (%eax), %xmm0
61 ; X64: pmovsxwd (%rdi), %xmm0
67 define <2 x i64> @pmovzxbq_1() nounwind {
69 %0 = load i16* @g16, align 2 ; <i16> [#uses=1]
70 %1 = insertelement <8 x i16> undef, i16 %0, i32 0 ; <<8 x i16>> [#uses=1]
71 %2 = bitcast <8 x i16> %1 to <16 x i8> ; <<16 x i8>> [#uses=1]
72 %3 = tail call <2 x i64> @llvm.x86.sse41.pmovzxbq(<16 x i8> %2) nounwind readnone ; <<2 x i64>> [#uses=1]
76 ; X32: movl L_g16$non_lazy_ptr, %eax
77 ; X32: pmovzxbq (%eax), %xmm0
80 ; X64: movq _g16@GOTPCREL(%rip), %rax
81 ; X64: pmovzxbq (%rax), %xmm0
84 declare <4 x i32> @llvm.x86.sse41.pmovsxbd(<16 x i8>) nounwind readnone
85 declare <4 x i32> @llvm.x86.sse41.pmovsxwd(<8 x i16>) nounwind readnone
86 declare <2 x i64> @llvm.x86.sse41.pmovzxbq(<16 x i8>) nounwind readnone
91 define i32 @extractps_1(<4 x float> %v) nounwind {
92 %s = extractelement <4 x float> %v, i32 3
93 %i = bitcast float %s to i32
97 ; X32: extractps $3, %xmm0, %eax
100 ; X64: extractps $3, %xmm0, %eax
102 define i32 @extractps_2(<4 x float> %v) nounwind {
103 %t = bitcast <4 x float> %v to <4 x i32>
104 %s = extractelement <4 x i32> %t, i32 3
108 ; X32: extractps $3, %xmm0, %eax
111 ; X64: extractps $3, %xmm0, %eax
115 ; The non-store form of extractps puts its result into a GPR.
116 ; This makes it suitable for an extract from a <4 x float> that
117 ; is bitcasted to i32, but unsuitable for much of anything else.
119 define float @ext_1(<4 x float> %v) nounwind {
120 %s = extractelement <4 x float> %v, i32 3
121 %t = fadd float %s, 1.0
125 ; X32: pshufd $3, %xmm0, %xmm0
126 ; X32: addss LCPI7_0, %xmm0
129 ; X64: pshufd $3, %xmm0, %xmm0
130 ; X64: addss LCPI7_0(%rip), %xmm0
132 define float @ext_2(<4 x float> %v) nounwind {
133 %s = extractelement <4 x float> %v, i32 3
137 ; X32: pshufd $3, %xmm0, %xmm0
140 ; X64: pshufd $3, %xmm0, %xmm0
142 define i32 @ext_3(<4 x i32> %v) nounwind {
143 %i = extractelement <4 x i32> %v, i32 3
147 ; X32: pextrd $3, %xmm0, %eax
150 ; X64: pextrd $3, %xmm0, %eax
153 define <4 x float> @insertps_1(<4 x float> %t1, <4 x float> %t2) nounwind {
154 %tmp1 = call <4 x float> @llvm.x86.sse41.insertps(<4 x float> %t1, <4 x float> %t2, i32 1) nounwind readnone
155 ret <4 x float> %tmp1
157 ; X32: insertps $1, %xmm1, %xmm0
160 ; X64: insertps $1, %xmm1, %xmm0
163 declare <4 x float> @llvm.x86.sse41.insertps(<4 x float>, <4 x float>, i32) nounwind readnone
165 define <4 x float> @insertps_2(<4 x float> %t1, float %t2) nounwind {
166 %tmp1 = insertelement <4 x float> %t1, float %t2, i32 0
167 ret <4 x float> %tmp1
169 ; X32: insertps $0, 4(%esp), %xmm0
172 ; X64: insertps $0, %xmm1, %xmm0
175 define <4 x float> @insertps_3(<4 x float> %t1, <4 x float> %t2) nounwind {
176 %tmp2 = extractelement <4 x float> %t2, i32 0
177 %tmp1 = insertelement <4 x float> %t1, float %tmp2, i32 0
178 ret <4 x float> %tmp1
180 ; X32: insertps $0, %xmm1, %xmm0
183 ; X64: insertps $0, %xmm1, %xmm0
186 define i32 @ptestz_1(<4 x float> %t1, <4 x float> %t2) nounwind {
187 %tmp1 = call i32 @llvm.x86.sse41.ptestz(<4 x float> %t1, <4 x float> %t2) nounwind readnone
190 ; X32: ptest %xmm1, %xmm0
194 ; X64: ptest %xmm1, %xmm0
198 define i32 @ptestz_2(<4 x float> %t1, <4 x float> %t2) nounwind {
199 %tmp1 = call i32 @llvm.x86.sse41.ptestc(<4 x float> %t1, <4 x float> %t2) nounwind readnone
202 ; X32: ptest %xmm1, %xmm0
206 ; X64: ptest %xmm1, %xmm0
210 define i32 @ptestz_3(<4 x float> %t1, <4 x float> %t2) nounwind {
211 %tmp1 = call i32 @llvm.x86.sse41.ptestnzc(<4 x float> %t1, <4 x float> %t2) nounwind readnone
214 ; X32: ptest %xmm1, %xmm0
218 ; X64: ptest %xmm1, %xmm0
223 declare i32 @llvm.x86.sse41.ptestz(<4 x float>, <4 x float>) nounwind readnone
224 declare i32 @llvm.x86.sse41.ptestc(<4 x float>, <4 x float>) nounwind readnone
225 declare i32 @llvm.x86.sse41.ptestnzc(<4 x float>, <4 x float>) nounwind readnone
227 ; This used to compile to insertps $0 + insertps $16. insertps $0 is always
229 define <2 x float> @buildvector(<2 x float> %A, <2 x float> %B) nounwind {
231 %tmp7 = extractelement <2 x float> %A, i32 0
232 %tmp5 = extractelement <2 x float> %A, i32 1
233 %tmp3 = extractelement <2 x float> %B, i32 0
234 %tmp1 = extractelement <2 x float> %B, i32 1
235 %add.r = fadd float %tmp7, %tmp3
236 %add.i = fadd float %tmp5, %tmp1
237 %tmp11 = insertelement <2 x float> undef, float %add.r, i32 0
238 %tmp9 = insertelement <2 x float> %tmp11, float %add.i, i32 1
239 ret <2 x float> %tmp9
241 ; X32-NOT: insertps $0
243 ; X32-NOT: insertps $0
246 ; X64-NOT: insertps $0
248 ; X64-NOT: insertps $0