define <2 x double> @shuffle_v2f64_32(<2 x double> %a, <2 x double> %b) {
; SSE-LABEL: shuffle_v2f64_32:
; SSE: # BB#0:
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm1[2,3,0,1]
+; SSE-NEXT: shufpd {{.*#+}} xmm1 = xmm1[1,0]
+; SSE-NEXT: movapd %xmm1, %xmm0
; SSE-NEXT: retq
;
; AVX-LABEL: shuffle_v2f64_32:
define <2 x double> @shuffle_v2f64_03(<2 x double> %a, <2 x double> %b) {
; SSE2-LABEL: shuffle_v2f64_03:
; SSE2: # BB#0:
-; SSE2-NEXT: shufpd {{.*#+}} xmm0 = xmm0[0],xmm1[1]
+; SSE2-NEXT: movsd %xmm0, %xmm1
+; SSE2-NEXT: movaps %xmm1, %xmm0
; SSE2-NEXT: retq
;
; SSE3-LABEL: shuffle_v2f64_03:
; SSE3: # BB#0:
-; SSE3-NEXT: shufpd {{.*#+}} xmm0 = xmm0[0],xmm1[1]
+; SSE3-NEXT: movsd %xmm0, %xmm1
+; SSE3-NEXT: movaps %xmm1, %xmm0
; SSE3-NEXT: retq
;
; SSSE3-LABEL: shuffle_v2f64_03:
; SSSE3: # BB#0:
-; SSSE3-NEXT: shufpd {{.*#+}} xmm0 = xmm0[0],xmm1[1]
+; SSSE3-NEXT: movsd %xmm0, %xmm1
+; SSSE3-NEXT: movaps %xmm1, %xmm0
; SSSE3-NEXT: retq
;
; SSE41-LABEL: shuffle_v2f64_03:
define <2 x double> @shuffle_v2f64_21(<2 x double> %a, <2 x double> %b) {
; SSE2-LABEL: shuffle_v2f64_21:
; SSE2: # BB#0:
-; SSE2-NEXT: shufpd {{.*#+}} xmm1 = xmm1[0],xmm0[1]
-; SSE2-NEXT: movapd %xmm1, %xmm0
+; SSE2-NEXT: movsd %xmm1, %xmm0
; SSE2-NEXT: retq
;
; SSE3-LABEL: shuffle_v2f64_21:
; SSE3: # BB#0:
-; SSE3-NEXT: shufpd {{.*#+}} xmm1 = xmm1[0],xmm0[1]
-; SSE3-NEXT: movapd %xmm1, %xmm0
+; SSE3-NEXT: movsd %xmm1, %xmm0
; SSE3-NEXT: retq
;
; SSSE3-LABEL: shuffle_v2f64_21:
; SSSE3: # BB#0:
-; SSSE3-NEXT: shufpd {{.*#+}} xmm1 = xmm1[0],xmm0[1]
-; SSSE3-NEXT: movapd %xmm1, %xmm0
+; SSSE3-NEXT: movsd %xmm1, %xmm0
; SSSE3-NEXT: retq
;
; SSE41-LABEL: shuffle_v2f64_21:
; SSE41: # BB#0:
-; SSE41-NEXT: blendpd {{.*#+}} xmm1 = xmm1[0],xmm0[1]
-; SSE41-NEXT: movapd %xmm1, %xmm0
+; SSE41-NEXT: blendpd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
; SSE41-NEXT: retq
;
; AVX-LABEL: shuffle_v2f64_21:
define <2 x i64> @shuffle_v2i64_03(<2 x i64> %a, <2 x i64> %b) {
; SSE2-LABEL: shuffle_v2i64_03:
; SSE2: # BB#0:
-; SSE2-NEXT: shufpd {{.*#+}} xmm0 = xmm0[0],xmm1[1]
+; SSE2-NEXT: movsd %xmm0, %xmm1
+; SSE2-NEXT: movaps %xmm1, %xmm0
; SSE2-NEXT: retq
;
; SSE3-LABEL: shuffle_v2i64_03:
; SSE3: # BB#0:
-; SSE3-NEXT: shufpd {{.*#+}} xmm0 = xmm0[0],xmm1[1]
+; SSE3-NEXT: movsd %xmm0, %xmm1
+; SSE3-NEXT: movaps %xmm1, %xmm0
; SSE3-NEXT: retq
;
; SSSE3-LABEL: shuffle_v2i64_03:
; SSSE3: # BB#0:
-; SSSE3-NEXT: shufpd {{.*#+}} xmm0 = xmm0[0],xmm1[1]
+; SSSE3-NEXT: movsd %xmm0, %xmm1
+; SSSE3-NEXT: movaps %xmm1, %xmm0
; SSSE3-NEXT: retq
;
; SSE41-LABEL: shuffle_v2i64_03:
define <2 x i64> @shuffle_v2i64_03_copy(<2 x i64> %nonce, <2 x i64> %a, <2 x i64> %b) {
; SSE2-LABEL: shuffle_v2i64_03_copy:
; SSE2: # BB#0:
-; SSE2-NEXT: shufpd {{.*#+}} xmm1 = xmm1[0],xmm2[1]
-; SSE2-NEXT: movapd %xmm1, %xmm0
+; SSE2-NEXT: movsd %xmm1, %xmm2
+; SSE2-NEXT: movaps %xmm2, %xmm0
; SSE2-NEXT: retq
;
; SSE3-LABEL: shuffle_v2i64_03_copy:
; SSE3: # BB#0:
-; SSE3-NEXT: shufpd {{.*#+}} xmm1 = xmm1[0],xmm2[1]
-; SSE3-NEXT: movapd %xmm1, %xmm0
+; SSE3-NEXT: movsd %xmm1, %xmm2
+; SSE3-NEXT: movaps %xmm2, %xmm0
; SSE3-NEXT: retq
;
; SSSE3-LABEL: shuffle_v2i64_03_copy:
; SSSE3: # BB#0:
-; SSSE3-NEXT: shufpd {{.*#+}} xmm1 = xmm1[0],xmm2[1]
-; SSSE3-NEXT: movapd %xmm1, %xmm0
+; SSSE3-NEXT: movsd %xmm1, %xmm2
+; SSSE3-NEXT: movaps %xmm2, %xmm0
; SSSE3-NEXT: retq
;
; SSE41-LABEL: shuffle_v2i64_03_copy:
define <2 x i64> @shuffle_v2i64_21(<2 x i64> %a, <2 x i64> %b) {
; SSE2-LABEL: shuffle_v2i64_21:
; SSE2: # BB#0:
-; SSE2-NEXT: shufpd {{.*#+}} xmm1 = xmm1[0],xmm0[1]
-; SSE2-NEXT: movapd %xmm1, %xmm0
+; SSE2-NEXT: movsd %xmm1, %xmm0
; SSE2-NEXT: retq
;
; SSE3-LABEL: shuffle_v2i64_21:
; SSE3: # BB#0:
-; SSE3-NEXT: shufpd {{.*#+}} xmm1 = xmm1[0],xmm0[1]
-; SSE3-NEXT: movapd %xmm1, %xmm0
+; SSE3-NEXT: movsd %xmm1, %xmm0
; SSE3-NEXT: retq
;
; SSSE3-LABEL: shuffle_v2i64_21:
; SSSE3: # BB#0:
-; SSSE3-NEXT: shufpd {{.*#+}} xmm1 = xmm1[0],xmm0[1]
-; SSSE3-NEXT: movapd %xmm1, %xmm0
+; SSSE3-NEXT: movsd %xmm1, %xmm0
; SSSE3-NEXT: retq
;
; SSE41-LABEL: shuffle_v2i64_21:
; SSE41: # BB#0:
-; SSE41-NEXT: pblendw {{.*#+}} xmm1 = xmm1[0,1,2,3],xmm0[4,5,6,7]
-; SSE41-NEXT: movdqa %xmm1, %xmm0
+; SSE41-NEXT: pblendw {{.*#+}} xmm0 = xmm1[0,1,2,3],xmm0[4,5,6,7]
; SSE41-NEXT: retq
;
; AVX1-LABEL: shuffle_v2i64_21:
define <2 x i64> @shuffle_v2i64_21_copy(<2 x i64> %nonce, <2 x i64> %a, <2 x i64> %b) {
; SSE2-LABEL: shuffle_v2i64_21_copy:
; SSE2: # BB#0:
-; SSE2-NEXT: shufpd {{.*#+}} xmm2 = xmm2[0],xmm1[1]
-; SSE2-NEXT: movapd %xmm2, %xmm0
+; SSE2-NEXT: movsd %xmm2, %xmm1
+; SSE2-NEXT: movaps %xmm1, %xmm0
; SSE2-NEXT: retq
;
; SSE3-LABEL: shuffle_v2i64_21_copy:
; SSE3: # BB#0:
-; SSE3-NEXT: shufpd {{.*#+}} xmm2 = xmm2[0],xmm1[1]
-; SSE3-NEXT: movapd %xmm2, %xmm0
+; SSE3-NEXT: movsd %xmm2, %xmm1
+; SSE3-NEXT: movaps %xmm1, %xmm0
; SSE3-NEXT: retq
;
; SSSE3-LABEL: shuffle_v2i64_21_copy:
; SSSE3: # BB#0:
-; SSSE3-NEXT: shufpd {{.*#+}} xmm2 = xmm2[0],xmm1[1]
-; SSSE3-NEXT: movapd %xmm2, %xmm0
+; SSSE3-NEXT: movsd %xmm2, %xmm1
+; SSSE3-NEXT: movaps %xmm1, %xmm0
; SSSE3-NEXT: retq
;
; SSE41-LABEL: shuffle_v2i64_21_copy:
; SSE41: # BB#0:
-; SSE41-NEXT: pblendw {{.*#+}} xmm2 = xmm2[0,1,2,3],xmm1[4,5,6,7]
-; SSE41-NEXT: movdqa %xmm2, %xmm0
+; SSE41-NEXT: pblendw {{.*#+}} xmm1 = xmm2[0,1,2,3],xmm1[4,5,6,7]
+; SSE41-NEXT: movdqa %xmm1, %xmm0
; SSE41-NEXT: retq
;
; AVX1-LABEL: shuffle_v2i64_21_copy:
}
define <2 x i64> @shuffle_v2i64_0z(<2 x i64> %a) {
-; SSE2-LABEL: shuffle_v2i64_0z:
-; SSE2: # BB#0:
-; SSE2-NEXT: xorpd %xmm1, %xmm1
-; SSE2-NEXT: shufpd {{.*#+}} xmm0 = xmm0[0],xmm1[1]
-; SSE2-NEXT: retq
-;
-; SSE3-LABEL: shuffle_v2i64_0z:
-; SSE3: # BB#0:
-; SSE3-NEXT: xorpd %xmm1, %xmm1
-; SSE3-NEXT: shufpd {{.*#+}} xmm0 = xmm0[0],xmm1[1]
-; SSE3-NEXT: retq
-;
-; SSSE3-LABEL: shuffle_v2i64_0z:
-; SSSE3: # BB#0:
-; SSSE3-NEXT: xorpd %xmm1, %xmm1
-; SSSE3-NEXT: shufpd {{.*#+}} xmm0 = xmm0[0],xmm1[1]
-; SSSE3-NEXT: retq
-;
-; SSE41-LABEL: shuffle_v2i64_0z:
-; SSE41: # BB#0:
-; SSE41-NEXT: pxor %xmm1, %xmm1
-; SSE41-NEXT: pblendw {{.*#+}} xmm0 = xmm0[0,1,2,3],xmm1[4,5,6,7]
-; SSE41-NEXT: retq
-;
-; AVX1-LABEL: shuffle_v2i64_0z:
-; AVX1: # BB#0:
-; AVX1-NEXT: vpxor %xmm1, %xmm1, %xmm1
-; AVX1-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3],xmm1[4,5,6,7]
-; AVX1-NEXT: retq
+; SSE-LABEL: shuffle_v2i64_0z:
+; SSE: # BB#0:
+; SSE-NEXT: movq %xmm0, %xmm0
+; SSE-NEXT: retq
;
-; AVX2-LABEL: shuffle_v2i64_0z:
-; AVX2: # BB#0:
-; AVX2-NEXT: vpxor %xmm1, %xmm1, %xmm1
-; AVX2-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
-; AVX2-NEXT: retq
+; AVX-LABEL: shuffle_v2i64_0z:
+; AVX: # BB#0:
+; AVX-NEXT: vmovq %xmm0, %xmm0
+; AVX-NEXT: retq
%shuffle = shufflevector <2 x i64> %a, <2 x i64> zeroinitializer, <2 x i32> <i32 0, i32 3>
ret <2 x i64> %shuffle
}
define <2 x i64> @shuffle_v2i64_1z(<2 x i64> %a) {
; SSE-LABEL: shuffle_v2i64_1z:
; SSE: # BB#0:
-; SSE-NEXT: pxor %xmm1, %xmm1
-; SSE-NEXT: punpckhqdq {{.*#+}} xmm0 = xmm0[1],xmm1[1]
+; SSE-NEXT: psrldq {{.*#+}} xmm0 = xmm0[8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero
; SSE-NEXT: retq
;
; AVX-LABEL: shuffle_v2i64_1z:
; AVX: # BB#0:
-; AVX-NEXT: vpxor %xmm1, %xmm1, %xmm1
-; AVX-NEXT: vpunpckhqdq {{.*#+}} xmm0 = xmm0[1],xmm1[1]
+; AVX-NEXT: vpsrldq {{.*#+}} xmm0 = xmm0[8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero
; AVX-NEXT: retq
%shuffle = shufflevector <2 x i64> %a, <2 x i64> zeroinitializer, <2 x i32> <i32 1, i32 3>
ret <2 x i64> %shuffle
define <2 x i64> @shuffle_v2i64_z0(<2 x i64> %a) {
; SSE-LABEL: shuffle_v2i64_z0:
; SSE: # BB#0:
-; SSE-NEXT: pxor %xmm1, %xmm1
-; SSE-NEXT: punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm0[0]
-; SSE-NEXT: movdqa %xmm1, %xmm0
+; SSE-NEXT: pslldq {{.*#+}} xmm0 = zero,zero,zero,zero,zero,zero,zero,zero,xmm0[0,1,2,3,4,5,6,7]
; SSE-NEXT: retq
;
; AVX-LABEL: shuffle_v2i64_z0:
; AVX: # BB#0:
-; AVX-NEXT: vpxor %xmm1, %xmm1, %xmm1
-; AVX-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm1[0],xmm0[0]
+; AVX-NEXT: vpslldq {{.*#+}} xmm0 = zero,zero,zero,zero,zero,zero,zero,zero,xmm0[0,1,2,3,4,5,6,7]
; AVX-NEXT: retq
%shuffle = shufflevector <2 x i64> %a, <2 x i64> zeroinitializer, <2 x i32> <i32 2, i32 0>
ret <2 x i64> %shuffle
define <2 x i64> @shuffle_v2i64_z1(<2 x i64> %a) {
; SSE2-LABEL: shuffle_v2i64_z1:
; SSE2: # BB#0:
-; SSE2-NEXT: xorpd %xmm1, %xmm1
-; SSE2-NEXT: shufpd {{.*#+}} xmm1 = xmm1[0],xmm0[1]
-; SSE2-NEXT: movapd %xmm1, %xmm0
+; SSE2-NEXT: xorps %xmm1, %xmm1
+; SSE2-NEXT: movsd %xmm1, %xmm0
; SSE2-NEXT: retq
;
; SSE3-LABEL: shuffle_v2i64_z1:
; SSE3: # BB#0:
-; SSE3-NEXT: xorpd %xmm1, %xmm1
-; SSE3-NEXT: shufpd {{.*#+}} xmm1 = xmm1[0],xmm0[1]
-; SSE3-NEXT: movapd %xmm1, %xmm0
+; SSE3-NEXT: xorps %xmm1, %xmm1
+; SSE3-NEXT: movsd %xmm1, %xmm0
; SSE3-NEXT: retq
;
; SSSE3-LABEL: shuffle_v2i64_z1:
; SSSE3: # BB#0:
-; SSSE3-NEXT: xorpd %xmm1, %xmm1
-; SSSE3-NEXT: shufpd {{.*#+}} xmm1 = xmm1[0],xmm0[1]
-; SSSE3-NEXT: movapd %xmm1, %xmm0
+; SSSE3-NEXT: xorps %xmm1, %xmm1
+; SSSE3-NEXT: movsd %xmm1, %xmm0
; SSSE3-NEXT: retq
;
; SSE41-LABEL: shuffle_v2i64_z1:
; SSE41: # BB#0:
; SSE41-NEXT: pxor %xmm1, %xmm1
-; SSE41-NEXT: pblendw {{.*#+}} xmm1 = xmm1[0,1,2,3],xmm0[4,5,6,7]
-; SSE41-NEXT: movdqa %xmm1, %xmm0
+; SSE41-NEXT: pblendw {{.*#+}} xmm0 = xmm1[0,1,2,3],xmm0[4,5,6,7]
; SSE41-NEXT: retq
;
; AVX1-LABEL: shuffle_v2i64_z1:
}
define <2 x double> @shuffle_v2f64_0z(<2 x double> %a) {
-; SSE2-LABEL: shuffle_v2f64_0z:
-; SSE2: # BB#0:
-; SSE2-NEXT: xorpd %xmm1, %xmm1
-; SSE2-NEXT: shufpd {{.*#+}} xmm0 = xmm0[0],xmm1[1]
-; SSE2-NEXT: retq
-;
-; SSE3-LABEL: shuffle_v2f64_0z:
-; SSE3: # BB#0:
-; SSE3-NEXT: xorpd %xmm1, %xmm1
-; SSE3-NEXT: shufpd {{.*#+}} xmm0 = xmm0[0],xmm1[1]
-; SSE3-NEXT: retq
-;
-; SSSE3-LABEL: shuffle_v2f64_0z:
-; SSSE3: # BB#0:
-; SSSE3-NEXT: xorpd %xmm1, %xmm1
-; SSSE3-NEXT: shufpd {{.*#+}} xmm0 = xmm0[0],xmm1[1]
-; SSSE3-NEXT: retq
-;
-; SSE41-LABEL: shuffle_v2f64_0z:
-; SSE41: # BB#0:
-; SSE41-NEXT: xorpd %xmm1, %xmm1
-; SSE41-NEXT: blendpd {{.*#+}} xmm0 = xmm0[0],xmm1[1]
-; SSE41-NEXT: retq
+; SSE-LABEL: shuffle_v2f64_0z:
+; SSE: # BB#0:
+; SSE-NEXT: movq %xmm0, %xmm0
+; SSE-NEXT: retq
;
; AVX-LABEL: shuffle_v2f64_0z:
; AVX: # BB#0:
-; AVX-NEXT: vxorpd %xmm1, %xmm1, %xmm1
-; AVX-NEXT: vblendpd {{.*#+}} xmm0 = xmm0[0],xmm1[1]
+; AVX-NEXT: vmovq %xmm0, %xmm0
; AVX-NEXT: retq
%shuffle = shufflevector <2 x double> %a, <2 x double> zeroinitializer, <2 x i32> <i32 0, i32 3>
ret <2 x double> %shuffle
define <2 x double> @shuffle_v2f64_z1(<2 x double> %a) {
; SSE2-LABEL: shuffle_v2f64_z1:
; SSE2: # BB#0:
-; SSE2-NEXT: xorpd %xmm1, %xmm1
-; SSE2-NEXT: shufpd {{.*#+}} xmm1 = xmm1[0],xmm0[1]
-; SSE2-NEXT: movapd %xmm1, %xmm0
+; SSE2-NEXT: xorps %xmm1, %xmm1
+; SSE2-NEXT: movsd %xmm1, %xmm0
; SSE2-NEXT: retq
;
; SSE3-LABEL: shuffle_v2f64_z1:
; SSE3: # BB#0:
-; SSE3-NEXT: xorpd %xmm1, %xmm1
-; SSE3-NEXT: shufpd {{.*#+}} xmm1 = xmm1[0],xmm0[1]
-; SSE3-NEXT: movapd %xmm1, %xmm0
+; SSE3-NEXT: xorps %xmm1, %xmm1
+; SSE3-NEXT: movsd %xmm1, %xmm0
; SSE3-NEXT: retq
;
; SSSE3-LABEL: shuffle_v2f64_z1:
; SSSE3: # BB#0:
-; SSSE3-NEXT: xorpd %xmm1, %xmm1
-; SSSE3-NEXT: shufpd {{.*#+}} xmm1 = xmm1[0],xmm0[1]
-; SSSE3-NEXT: movapd %xmm1, %xmm0
+; SSSE3-NEXT: xorps %xmm1, %xmm1
+; SSSE3-NEXT: movsd %xmm1, %xmm0
; SSSE3-NEXT: retq
;
; SSE41-LABEL: shuffle_v2f64_z1:
; SSE41: # BB#0:
; SSE41-NEXT: xorpd %xmm1, %xmm1
-; SSE41-NEXT: blendpd {{.*#+}} xmm1 = xmm1[0],xmm0[1]
-; SSE41-NEXT: movapd %xmm1, %xmm0
+; SSE41-NEXT: blendpd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
; SSE41-NEXT: retq
;
; AVX-LABEL: shuffle_v2f64_z1:
; SSE41-LABEL: insert_reg_lo_v2i64:
; SSE41: # BB#0:
; SSE41-NEXT: movd %rdi, %xmm1
-; SSE41-NEXT: pblendw {{.*#+}} xmm1 = xmm1[0,1,2,3],xmm0[4,5,6,7]
-; SSE41-NEXT: movdqa %xmm1, %xmm0
+; SSE41-NEXT: pblendw {{.*#+}} xmm0 = xmm1[0,1,2,3],xmm0[4,5,6,7]
; SSE41-NEXT: retq
;
; AVX1-LABEL: insert_reg_lo_v2i64:
; SSE41-LABEL: insert_mem_lo_v2i64:
; SSE41: # BB#0:
; SSE41-NEXT: movq (%rdi), %xmm1
-; SSE41-NEXT: pblendw {{.*#+}} xmm1 = xmm1[0,1,2,3],xmm0[4,5,6,7]
-; SSE41-NEXT: movdqa %xmm1, %xmm0
+; SSE41-NEXT: pblendw {{.*#+}} xmm0 = xmm1[0,1,2,3],xmm0[4,5,6,7]
; SSE41-NEXT: retq
;
; AVX1-LABEL: insert_mem_lo_v2i64: