; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=avx | FileCheck %s --check-prefixes=CHECK,AVX,AVX12,AVX1
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=avx2 | FileCheck %s --check-prefixes=CHECK,AVX,AVX12,AVX2
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=avx512f | FileCheck %s --check-prefixes=CHECK,AVX,AVX512,AVX512F
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=avx512vl | FileCheck %s --check-prefixes=CHECK,AVX,AVX512,AVX512VL
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=xop | FileCheck %s --check-prefixes=CHECK,XOP
; The condition vector for BLENDV* only cares about the sign bit of each element.
; So in these tests, if we generate BLENDV*, we should be able to remove the redundant cmp op.
; Test 128-bit vectors for all legal element types.
define <16 x i8> @signbit_sel_v16i8(<16 x i8> %x, <16 x i8> %y, <16 x i8> %mask) {
; CHECK-LABEL: signbit_sel_v16i8:
; CHECK: # %bb.0:
; CHECK-NEXT: vpblendvb %xmm2, %xmm0, %xmm1, %xmm0
; CHECK-NEXT: retq
%tr = icmp slt <16 x i8> %mask, zeroinitializer
%z = select <16 x i1> %tr, <16 x i8> %x, <16 x i8> %y
ret <16 x i8> %z
}
; Sorry 16-bit, you're not important enough to support?
define <8 x i16> @signbit_sel_v8i16(<8 x i16> %x, <8 x i16> %y, <8 x i16> %mask) {
; AVX12-LABEL: signbit_sel_v8i16:
; AVX12: # %bb.0:
; AVX12-NEXT: vpxor %xmm3, %xmm3, %xmm3
; AVX12-NEXT: vpcmpgtw %xmm2, %xmm3, %xmm2
; AVX12-NEXT: vpblendvb %xmm2, %xmm0, %xmm1, %xmm0
; AVX12-NEXT: retq
;
; AVX512F-LABEL: signbit_sel_v8i16:
; AVX512F: # %bb.0:
; AVX512F-NEXT: vpxor %xmm3, %xmm3, %xmm3
; AVX512F-NEXT: vpcmpgtw %xmm2, %xmm3, %xmm2
; AVX512F-NEXT: vpblendvb %xmm2, %xmm0, %xmm1, %xmm0
; AVX512F-NEXT: retq
;
; AVX512VL-LABEL: signbit_sel_v8i16:
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vpxor %xmm3, %xmm3, %xmm3
; AVX512VL-NEXT: vpcmpgtw %xmm2, %xmm3, %xmm2
; AVX512VL-NEXT: vpternlogq $226, %xmm1, %xmm2, %xmm0
; AVX512VL-NEXT: retq
;
; XOP-LABEL: signbit_sel_v8i16:
; XOP: # %bb.0:
; XOP-NEXT: vpxor %xmm3, %xmm3, %xmm3
; XOP-NEXT: vpcomltw %xmm3, %xmm2, %xmm2
; XOP-NEXT: vpblendvb %xmm2, %xmm0, %xmm1, %xmm0
; XOP-NEXT: retq
%tr = icmp slt <8 x i16> %mask, zeroinitializer
%z = select <8 x i1> %tr, <8 x i16> %x, <8 x i16> %y
ret <8 x i16> %z
}
define <4 x i32> @signbit_sel_v4i32(<4 x i32> %x, <4 x i32> %y, <4 x i32> %mask) {
; AVX12-LABEL: signbit_sel_v4i32:
; AVX12: # %bb.0:
; AVX12-NEXT: vblendvps %xmm2, %xmm0, %xmm1, %xmm0
; AVX12-NEXT: retq
;
; AVX512F-LABEL: signbit_sel_v4i32:
; AVX512F: # %bb.0:
; AVX512F-NEXT: # kill: def $xmm2 killed $xmm2 def $zmm2
; AVX512F-NEXT: # kill: def $xmm1 killed $xmm1 def $zmm1
; AVX512F-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0
; AVX512F-NEXT: vpxor %xmm3, %xmm3, %xmm3
; AVX512F-NEXT: vpcmpgtd %zmm2, %zmm3, %k1
; AVX512F-NEXT: vpblendmd %zmm0, %zmm1, %zmm0 {%k1}
; AVX512F-NEXT: # kill: def $xmm0 killed $xmm0 killed $zmm0
; AVX512F-NEXT: vzeroupper
; AVX512F-NEXT: retq
;
; AVX512VL-LABEL: signbit_sel_v4i32:
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vpxor %xmm3, %xmm3, %xmm3
; AVX512VL-NEXT: vpcmpgtd %xmm2, %xmm3, %k1
; AVX512VL-NEXT: vpblendmd %xmm0, %xmm1, %xmm0 {%k1}
; AVX512VL-NEXT: retq
;
; XOP-LABEL: signbit_sel_v4i32:
; XOP: # %bb.0:
; XOP-NEXT: vblendvps %xmm2, %xmm0, %xmm1, %xmm0
; XOP-NEXT: retq
%tr = icmp slt <4 x i32> %mask, zeroinitializer
%z = select <4 x i1> %tr, <4 x i32> %x, <4 x i32> %y
ret <4 x i32> %z
}
define <2 x i64> @signbit_sel_v2i64(<2 x i64> %x, <2 x i64> %y, <2 x i64> %mask) {
; AVX12-LABEL: signbit_sel_v2i64:
; AVX12: # %bb.0:
; AVX12-NEXT: vblendvpd %xmm2, %xmm0, %xmm1, %xmm0
; AVX12-NEXT: retq
;
; AVX512F-LABEL: signbit_sel_v2i64:
; AVX512F: # %bb.0:
; AVX512F-NEXT: # kill: def $xmm2 killed $xmm2 def $zmm2
; AVX512F-NEXT: # kill: def $xmm1 killed $xmm1 def $zmm1
; AVX512F-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0
; AVX512F-NEXT: vpxor %xmm3, %xmm3, %xmm3
; AVX512F-NEXT: vpcmpgtq %zmm2, %zmm3, %k1
; AVX512F-NEXT: vpblendmq %zmm0, %zmm1, %zmm0 {%k1}
; AVX512F-NEXT: # kill: def $xmm0 killed $xmm0 killed $zmm0
; AVX512F-NEXT: vzeroupper
; AVX512F-NEXT: retq
;
; AVX512VL-LABEL: signbit_sel_v2i64:
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vpxor %xmm3, %xmm3, %xmm3
; AVX512VL-NEXT: vpcmpgtq %xmm2, %xmm3, %k1
; AVX512VL-NEXT: vpblendmq %xmm0, %xmm1, %xmm0 {%k1}
; AVX512VL-NEXT: retq
;
; XOP-LABEL: signbit_sel_v2i64:
; XOP: # %bb.0:
; XOP-NEXT: vblendvpd %xmm2, %xmm0, %xmm1, %xmm0
; XOP-NEXT: retq
%tr = icmp slt <2 x i64> %mask, zeroinitializer
%z = select <2 x i1> %tr, <2 x i64> %x, <2 x i64> %y
ret <2 x i64> %z
}
define <4 x float> @signbit_sel_v4f32(<4 x float> %x, <4 x float> %y, <4 x i32> %mask) {
; AVX12-LABEL: signbit_sel_v4f32:
; AVX12: # %bb.0:
; AVX12-NEXT: vblendvps %xmm2, %xmm0, %xmm1, %xmm0
; AVX12-NEXT: retq
;
; AVX512F-LABEL: signbit_sel_v4f32:
; AVX512F: # %bb.0:
; AVX512F-NEXT: # kill: def $xmm2 killed $xmm2 def $zmm2
; AVX512F-NEXT: # kill: def $xmm1 killed $xmm1 def $zmm1
; AVX512F-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0
; AVX512F-NEXT: vpxor %xmm3, %xmm3, %xmm3
; AVX512F-NEXT: vpcmpgtd %zmm2, %zmm3, %k1
; AVX512F-NEXT: vblendmps %zmm0, %zmm1, %zmm0 {%k1}
; AVX512F-NEXT: # kill: def $xmm0 killed $xmm0 killed $zmm0
; AVX512F-NEXT: vzeroupper
; AVX512F-NEXT: retq
;
; AVX512VL-LABEL: signbit_sel_v4f32:
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vpxor %xmm3, %xmm3, %xmm3
; AVX512VL-NEXT: vpcmpgtd %xmm2, %xmm3, %k1
; AVX512VL-NEXT: vblendmps %xmm0, %xmm1, %xmm0 {%k1}
; AVX512VL-NEXT: retq
;
; XOP-LABEL: signbit_sel_v4f32:
; XOP: # %bb.0:
; XOP-NEXT: vblendvps %xmm2, %xmm0, %xmm1, %xmm0
; XOP-NEXT: retq
%tr = icmp slt <4 x i32> %mask, zeroinitializer
%z = select <4 x i1> %tr, <4 x float> %x, <4 x float> %y
ret <4 x float> %z
}
define <2 x double> @signbit_sel_v2f64(<2 x double> %x, <2 x double> %y, <2 x i64> %mask) {
; AVX12-LABEL: signbit_sel_v2f64:
; AVX12: # %bb.0:
; AVX12-NEXT: vblendvpd %xmm2, %xmm0, %xmm1, %xmm0
; AVX12-NEXT: retq
;
; AVX512F-LABEL: signbit_sel_v2f64:
; AVX512F: # %bb.0:
; AVX512F-NEXT: # kill: def $xmm2 killed $xmm2 def $zmm2
; AVX512F-NEXT: # kill: def $xmm1 killed $xmm1 def $zmm1
; AVX512F-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0
; AVX512F-NEXT: vpxor %xmm3, %xmm3, %xmm3
; AVX512F-NEXT: vpcmpgtq %zmm2, %zmm3, %k1
; AVX512F-NEXT: vblendmpd %zmm0, %zmm1, %zmm0 {%k1}
; AVX512F-NEXT: # kill: def $xmm0 killed $xmm0 killed $zmm0
; AVX512F-NEXT: vzeroupper
; AVX512F-NEXT: retq
;
; AVX512VL-LABEL: signbit_sel_v2f64:
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vpxor %xmm3, %xmm3, %xmm3
; AVX512VL-NEXT: vpcmpgtq %xmm2, %xmm3, %k1
; AVX512VL-NEXT: vblendmpd %xmm0, %xmm1, %xmm0 {%k1}
; AVX512VL-NEXT: retq
;
; XOP-LABEL: signbit_sel_v2f64:
; XOP: # %bb.0:
; XOP-NEXT: vblendvpd %xmm2, %xmm0, %xmm1, %xmm0
; XOP-NEXT: retq
%tr = icmp slt <2 x i64> %mask, zeroinitializer
%z = select <2 x i1> %tr, <2 x double> %x, <2 x double> %y
ret <2 x double> %z
}
; Test 256-bit vectors to see differences between AVX1 and AVX2.
define <32 x i8> @signbit_sel_v32i8(<32 x i8> %x, <32 x i8> %y, <32 x i8> %mask) {
; AVX1-LABEL: signbit_sel_v32i8:
; AVX1: # %bb.0:
; AVX1-NEXT: vextractf128 $1, %ymm2, %xmm3
; AVX1-NEXT: vpxor %xmm4, %xmm4, %xmm4
; AVX1-NEXT: vpcmpgtb %xmm3, %xmm4, %xmm3
; AVX1-NEXT: vpcmpgtb %xmm2, %xmm4, %xmm2
; AVX1-NEXT: vinsertf128 $1, %xmm3, %ymm2, %ymm2
; AVX1-NEXT: vandnps %ymm1, %ymm2, %ymm1
; AVX1-NEXT: vandps %ymm2, %ymm0, %ymm0
; AVX1-NEXT: vorps %ymm1, %ymm0, %ymm0
; AVX1-NEXT: retq
;
; AVX2-LABEL: signbit_sel_v32i8:
; AVX2: # %bb.0:
; AVX2-NEXT: vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
; AVX2-NEXT: retq
;
; AVX512-LABEL: signbit_sel_v32i8:
; AVX512: # %bb.0:
; AVX512-NEXT: vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
; AVX512-NEXT: retq
;
; XOP-LABEL: signbit_sel_v32i8:
; XOP: # %bb.0:
; XOP-NEXT: vextractf128 $1, %ymm2, %xmm3
; XOP-NEXT: vpxor %xmm4, %xmm4, %xmm4
; XOP-NEXT: vpcomltb %xmm4, %xmm3, %xmm3
; XOP-NEXT: vpcomltb %xmm4, %xmm2, %xmm2
; XOP-NEXT: vinsertf128 $1, %xmm3, %ymm2, %ymm2
; XOP-NEXT: vpcmov %ymm2, %ymm1, %ymm0, %ymm0
; XOP-NEXT: retq
%tr = icmp slt <32 x i8> %mask, zeroinitializer
%z = select <32 x i1> %tr, <32 x i8> %x, <32 x i8> %y
ret <32 x i8> %z
}
; Sorry 16-bit, you'll never be important enough to support?
define <16 x i16> @signbit_sel_v16i16(<16 x i16> %x, <16 x i16> %y, <16 x i16> %mask) {
; AVX1-LABEL: signbit_sel_v16i16:
; AVX1: # %bb.0:
; AVX1-NEXT: vextractf128 $1, %ymm2, %xmm3
; AVX1-NEXT: vpxor %xmm4, %xmm4, %xmm4
; AVX1-NEXT: vpcmpgtw %xmm3, %xmm4, %xmm3
; AVX1-NEXT: vpcmpgtw %xmm2, %xmm4, %xmm2
; AVX1-NEXT: vinsertf128 $1, %xmm3, %ymm2, %ymm2
; AVX1-NEXT: vandnps %ymm1, %ymm2, %ymm1
; AVX1-NEXT: vandps %ymm2, %ymm0, %ymm0
; AVX1-NEXT: vorps %ymm1, %ymm0, %ymm0
; AVX1-NEXT: retq
;
; AVX2-LABEL: signbit_sel_v16i16:
; AVX2: # %bb.0:
; AVX2-NEXT: vpxor %xmm3, %xmm3, %xmm3
; AVX2-NEXT: vpcmpgtw %ymm2, %ymm3, %ymm2
; AVX2-NEXT: vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
; AVX2-NEXT: retq
;
; AVX512F-LABEL: signbit_sel_v16i16:
; AVX512F: # %bb.0:
; AVX512F-NEXT: vpxor %xmm3, %xmm3, %xmm3
; AVX512F-NEXT: vpcmpgtw %ymm2, %ymm3, %ymm2
; AVX512F-NEXT: vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
; AVX512F-NEXT: retq
;
; AVX512VL-LABEL: signbit_sel_v16i16:
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vpxor %xmm3, %xmm3, %xmm3
; AVX512VL-NEXT: vpcmpgtw %ymm2, %ymm3, %ymm2
; AVX512VL-NEXT: vpternlogq $226, %ymm1, %ymm2, %ymm0
; AVX512VL-NEXT: retq
;
; XOP-LABEL: signbit_sel_v16i16:
; XOP: # %bb.0:
; XOP-NEXT: vextractf128 $1, %ymm2, %xmm3
; XOP-NEXT: vpxor %xmm4, %xmm4, %xmm4
; XOP-NEXT: vpcomltw %xmm4, %xmm3, %xmm3
; XOP-NEXT: vpcomltw %xmm4, %xmm2, %xmm2
; XOP-NEXT: vinsertf128 $1, %xmm3, %ymm2, %ymm2
; XOP-NEXT: vpcmov %ymm2, %ymm1, %ymm0, %ymm0
; XOP-NEXT: retq
%tr = icmp slt <16 x i16> %mask, zeroinitializer
%z = select <16 x i1> %tr, <16 x i16> %x, <16 x i16> %y
ret <16 x i16> %z
}
define <8 x i32> @signbit_sel_v8i32(<8 x i32> %x, <8 x i32> %y, <8 x i32> %mask) {
; AVX12-LABEL: signbit_sel_v8i32:
; AVX12: # %bb.0:
; AVX12-NEXT: vblendvps %ymm2, %ymm0, %ymm1, %ymm0
; AVX12-NEXT: retq
;
; AVX512F-LABEL: signbit_sel_v8i32:
; AVX512F: # %bb.0:
; AVX512F-NEXT: # kill: def $ymm2 killed $ymm2 def $zmm2
; AVX512F-NEXT: # kill: def $ymm1 killed $ymm1 def $zmm1
; AVX512F-NEXT: # kill: def $ymm0 killed $ymm0 def $zmm0
; AVX512F-NEXT: vpxor %xmm3, %xmm3, %xmm3
; AVX512F-NEXT: vpcmpgtd %zmm2, %zmm3, %k1
; AVX512F-NEXT: vpblendmd %zmm0, %zmm1, %zmm0 {%k1}
; AVX512F-NEXT: # kill: def $ymm0 killed $ymm0 killed $zmm0
; AVX512F-NEXT: retq
;
; AVX512VL-LABEL: signbit_sel_v8i32:
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vpxor %xmm3, %xmm3, %xmm3
; AVX512VL-NEXT: vpcmpgtd %ymm2, %ymm3, %k1
; AVX512VL-NEXT: vpblendmd %ymm0, %ymm1, %ymm0 {%k1}
; AVX512VL-NEXT: retq
;
; XOP-LABEL: signbit_sel_v8i32:
; XOP: # %bb.0:
; XOP-NEXT: vblendvps %ymm2, %ymm0, %ymm1, %ymm0
; XOP-NEXT: retq
%tr = icmp slt <8 x i32> %mask, zeroinitializer
%z = select <8 x i1> %tr, <8 x i32> %x, <8 x i32> %y
ret <8 x i32> %z
}
define <4 x i64> @signbit_sel_v4i64(<4 x i64> %x, <4 x i64> %y, <4 x i64> %mask) {
; AVX12-LABEL: signbit_sel_v4i64:
; AVX12: # %bb.0:
; AVX12-NEXT: vblendvpd %ymm2, %ymm0, %ymm1, %ymm0
; AVX12-NEXT: retq
;
; AVX512F-LABEL: signbit_sel_v4i64:
; AVX512F: # %bb.0:
; AVX512F-NEXT: # kill: def $ymm2 killed $ymm2 def $zmm2
; AVX512F-NEXT: # kill: def $ymm1 killed $ymm1 def $zmm1
; AVX512F-NEXT: # kill: def $ymm0 killed $ymm0 def $zmm0
; AVX512F-NEXT: vpxor %xmm3, %xmm3, %xmm3
; AVX512F-NEXT: vpcmpgtq %zmm2, %zmm3, %k1
; AVX512F-NEXT: vpblendmq %zmm0, %zmm1, %zmm0 {%k1}
; AVX512F-NEXT: # kill: def $ymm0 killed $ymm0 killed $zmm0
; AVX512F-NEXT: retq
;
; AVX512VL-LABEL: signbit_sel_v4i64:
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vpxor %xmm3, %xmm3, %xmm3
; AVX512VL-NEXT: vpcmpgtq %ymm2, %ymm3, %k1
; AVX512VL-NEXT: vpblendmq %ymm0, %ymm1, %ymm0 {%k1}
; AVX512VL-NEXT: retq
;
; XOP-LABEL: signbit_sel_v4i64:
; XOP: # %bb.0:
; XOP-NEXT: vblendvpd %ymm2, %ymm0, %ymm1, %ymm0
; XOP-NEXT: retq
%tr = icmp slt <4 x i64> %mask, zeroinitializer
%z = select <4 x i1> %tr, <4 x i64> %x, <4 x i64> %y
ret <4 x i64> %z
}
define <4 x double> @signbit_sel_v4f64(<4 x double> %x, <4 x double> %y, <4 x i64> %mask) {
; AVX12-LABEL: signbit_sel_v4f64:
; AVX12: # %bb.0:
; AVX12-NEXT: vblendvpd %ymm2, %ymm0, %ymm1, %ymm0
; AVX12-NEXT: retq
;
; AVX512F-LABEL: signbit_sel_v4f64:
; AVX512F: # %bb.0:
; AVX512F-NEXT: # kill: def $ymm2 killed $ymm2 def $zmm2
; AVX512F-NEXT: # kill: def $ymm1 killed $ymm1 def $zmm1
; AVX512F-NEXT: # kill: def $ymm0 killed $ymm0 def $zmm0
; AVX512F-NEXT: vpxor %xmm3, %xmm3, %xmm3
; AVX512F-NEXT: vpcmpgtq %zmm2, %zmm3, %k1
; AVX512F-NEXT: vblendmpd %zmm0, %zmm1, %zmm0 {%k1}
; AVX512F-NEXT: # kill: def $ymm0 killed $ymm0 killed $zmm0
; AVX512F-NEXT: retq
;
; AVX512VL-LABEL: signbit_sel_v4f64:
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vpxor %xmm3, %xmm3, %xmm3
; AVX512VL-NEXT: vpcmpgtq %ymm2, %ymm3, %k1
; AVX512VL-NEXT: vblendmpd %ymm0, %ymm1, %ymm0 {%k1}
; AVX512VL-NEXT: retq
;
; XOP-LABEL: signbit_sel_v4f64:
; XOP: # %bb.0:
; XOP-NEXT: vblendvpd %ymm2, %ymm0, %ymm1, %ymm0
; XOP-NEXT: retq
%tr = icmp slt <4 x i64> %mask, zeroinitializer
%z = select <4 x i1> %tr, <4 x double> %x, <4 x double> %y
ret <4 x double> %z
}
; Try a condition with a different type than the select operands.
define <4 x double> @signbit_sel_v4f64_small_mask(<4 x double> %x, <4 x double> %y, <4 x i32> %mask) {
; AVX1-LABEL: signbit_sel_v4f64_small_mask:
; AVX1: # %bb.0:
; AVX1-NEXT: vpmovsxdq %xmm2, %xmm3
; AVX1-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[2,3,2,3]
; AVX1-NEXT: vpmovsxdq %xmm2, %xmm2
; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm3, %ymm2
; AVX1-NEXT: vblendvpd %ymm2, %ymm0, %ymm1, %ymm0
; AVX1-NEXT: retq
;
; AVX2-LABEL: signbit_sel_v4f64_small_mask:
; AVX2: # %bb.0:
; AVX2-NEXT: vpmovsxdq %xmm2, %ymm2
; AVX2-NEXT: vblendvpd %ymm2, %ymm0, %ymm1, %ymm0
; AVX2-NEXT: retq
;
; AVX512F-LABEL: signbit_sel_v4f64_small_mask:
; AVX512F: # %bb.0:
; AVX512F-NEXT: # kill: def $xmm2 killed $xmm2 def $zmm2
; AVX512F-NEXT: # kill: def $ymm1 killed $ymm1 def $zmm1
; AVX512F-NEXT: # kill: def $ymm0 killed $ymm0 def $zmm0
; AVX512F-NEXT: vpxor %xmm3, %xmm3, %xmm3
; AVX512F-NEXT: vpcmpgtd %zmm2, %zmm3, %k1
; AVX512F-NEXT: vblendmpd %zmm0, %zmm1, %zmm0 {%k1}
; AVX512F-NEXT: # kill: def $ymm0 killed $ymm0 killed $zmm0
; AVX512F-NEXT: retq
;
; AVX512VL-LABEL: signbit_sel_v4f64_small_mask:
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vpxor %xmm3, %xmm3, %xmm3
; AVX512VL-NEXT: vpcmpgtd %xmm2, %xmm3, %k1
; AVX512VL-NEXT: vblendmpd %ymm0, %ymm1, %ymm0 {%k1}
; AVX512VL-NEXT: retq
;
; XOP-LABEL: signbit_sel_v4f64_small_mask:
; XOP: # %bb.0:
; XOP-NEXT: vpmovsxdq %xmm2, %xmm3
; XOP-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[2,3,2,3]
; XOP-NEXT: vpmovsxdq %xmm2, %xmm2
; XOP-NEXT: vinsertf128 $1, %xmm2, %ymm3, %ymm2
; XOP-NEXT: vblendvpd %ymm2, %ymm0, %ymm1, %ymm0
; XOP-NEXT: retq
%tr = icmp slt <4 x i32> %mask, zeroinitializer
%z = select <4 x i1> %tr, <4 x double> %x, <4 x double> %y
ret <4 x double> %z
}
; Try a 512-bit vector to make sure AVX-512 is handled as expected.
define <8 x double> @signbit_sel_v8f64(<8 x double> %x, <8 x double> %y, <8 x i64> %mask) {
; AVX12-LABEL: signbit_sel_v8f64:
; AVX12: # %bb.0:
; AVX12-NEXT: vblendvpd %ymm4, %ymm0, %ymm2, %ymm0
; AVX12-NEXT: vblendvpd %ymm5, %ymm1, %ymm3, %ymm1
; AVX12-NEXT: retq
;
; AVX512-LABEL: signbit_sel_v8f64:
; AVX512: # %bb.0:
; AVX512-NEXT: vpxor %xmm3, %xmm3, %xmm3
; AVX512-NEXT: vpcmpgtq %zmm2, %zmm3, %k1
; AVX512-NEXT: vblendmpd %zmm0, %zmm1, %zmm0 {%k1}
; AVX512-NEXT: retq
;
; XOP-LABEL: signbit_sel_v8f64:
; XOP: # %bb.0:
; XOP-NEXT: vblendvpd %ymm4, %ymm0, %ymm2, %ymm0
; XOP-NEXT: vblendvpd %ymm5, %ymm1, %ymm3, %ymm1
; XOP-NEXT: retq
%tr = icmp slt <8 x i64> %mask, zeroinitializer
%z = select <8 x i1> %tr, <8 x double> %x, <8 x double> %y
ret <8 x double> %z
}
; If we have a floating-point compare:
; (1) Don't die.
; (2) FIXME: If we don't care about signed-zero (and NaN?), the compare should still get folded.
define <4 x float> @signbit_sel_v4f32_fcmp(<4 x float> %x, <4 x float> %y, <4 x float> %mask) #0 {
; AVX12-LABEL: signbit_sel_v4f32_fcmp:
; AVX12: # %bb.0:
; AVX12-NEXT: vxorps %xmm2, %xmm2, %xmm2
; AVX12-NEXT: vcmpltps %xmm2, %xmm0, %xmm2
; AVX12-NEXT: vblendvps %xmm2, %xmm0, %xmm1, %xmm0
; AVX12-NEXT: retq
;
; AVX512F-LABEL: signbit_sel_v4f32_fcmp:
; AVX512F: # %bb.0:
; AVX512F-NEXT: # kill: def $xmm1 killed $xmm1 def $zmm1
; AVX512F-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0
; AVX512F-NEXT: vxorps %xmm2, %xmm2, %xmm2
; AVX512F-NEXT: vcmpltps %zmm2, %zmm0, %k1
; AVX512F-NEXT: vblendmps %zmm0, %zmm1, %zmm0 {%k1}
; AVX512F-NEXT: # kill: def $xmm0 killed $xmm0 killed $zmm0
; AVX512F-NEXT: vzeroupper
; AVX512F-NEXT: retq
;
; AVX512VL-LABEL: signbit_sel_v4f32_fcmp:
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vxorps %xmm2, %xmm2, %xmm2
; AVX512VL-NEXT: vcmpltps %xmm2, %xmm0, %k1
; AVX512VL-NEXT: vblendmps %xmm0, %xmm1, %xmm0 {%k1}
; AVX512VL-NEXT: retq
;
; XOP-LABEL: signbit_sel_v4f32_fcmp:
; XOP: # %bb.0:
; XOP-NEXT: vxorps %xmm2, %xmm2, %xmm2
; XOP-NEXT: vcmpltps %xmm2, %xmm0, %xmm2
; XOP-NEXT: vblendvps %xmm2, %xmm0, %xmm1, %xmm0
; XOP-NEXT: retq
%cmp = fcmp olt <4 x float> %x, zeroinitializer
%sel = select <4 x i1> %cmp, <4 x float> %x, <4 x float> %y
ret <4 x float> %sel
}
define <4 x i64> @blend_splat1_mask_cond_v4i64(<4 x i64> %x, <4 x i64> %y, <4 x i64> %z) {
; AVX1-LABEL: blend_splat1_mask_cond_v4i64:
; AVX1: # %bb.0:
; AVX1-NEXT: vpsllq $63, %xmm0, %xmm3
; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm0
; AVX1-NEXT: vpsllq $63, %xmm0, %xmm0
; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm3, %ymm0
; AVX1-NEXT: vblendvpd %ymm0, %ymm2, %ymm1, %ymm0
; AVX1-NEXT: retq
;
; AVX2-LABEL: blend_splat1_mask_cond_v4i64:
; AVX2: # %bb.0:
; AVX2-NEXT: vpsllq $63, %ymm0, %ymm0
; AVX2-NEXT: vblendvpd %ymm0, %ymm2, %ymm1, %ymm0
; AVX2-NEXT: retq
;
; AVX512F-LABEL: blend_splat1_mask_cond_v4i64:
; AVX512F: # %bb.0:
; AVX512F-NEXT: # kill: def $ymm2 killed $ymm2 def $zmm2
; AVX512F-NEXT: # kill: def $ymm1 killed $ymm1 def $zmm1
; AVX512F-NEXT: # kill: def $ymm0 killed $ymm0 def $zmm0
; AVX512F-NEXT: vptestnmq {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to8}, %zmm0, %k1
; AVX512F-NEXT: vpblendmq %zmm1, %zmm2, %zmm0 {%k1}
; AVX512F-NEXT: # kill: def $ymm0 killed $ymm0 killed $zmm0
; AVX512F-NEXT: retq
;
; AVX512VL-LABEL: blend_splat1_mask_cond_v4i64:
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vptestnmq {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to4}, %ymm0, %k1
; AVX512VL-NEXT: vpblendmq %ymm1, %ymm2, %ymm0 {%k1}
; AVX512VL-NEXT: retq
;
; XOP-LABEL: blend_splat1_mask_cond_v4i64:
; XOP: # %bb.0:
; XOP-NEXT: vextractf128 $1, %ymm0, %xmm3
; XOP-NEXT: vpsllq $63, %xmm3, %xmm3
; XOP-NEXT: vpmovsxbq {{.*#+}} xmm4 = [18446744073709551553,18446744073709551553]
; XOP-NEXT: vpshaq %xmm4, %xmm3, %xmm3
; XOP-NEXT: vpsllq $63, %xmm0, %xmm0
; XOP-NEXT: vpshaq %xmm4, %xmm0, %xmm0
; XOP-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm0
; XOP-NEXT: vblendvpd %ymm0, %ymm2, %ymm1, %ymm0
; XOP-NEXT: retq
%a = and <4 x i64> %x, <i64 1, i64 1, i64 1, i64 1>
%c = icmp eq <4 x i64> %a, zeroinitializer
%r = select <4 x i1> %c, <4 x i64> %y, <4 x i64> %z
ret <4 x i64> %r
}
define <4 x i32> @blend_splat1_mask_cond_v4i32(<4 x i32> %x, <4 x i32> %y, <4 x i32> %z) {
; AVX12-LABEL: blend_splat1_mask_cond_v4i32:
; AVX12: # %bb.0:
; AVX12-NEXT: vpslld $31, %xmm0, %xmm0
; AVX12-NEXT: vblendvps %xmm0, %xmm2, %xmm1, %xmm0
; AVX12-NEXT: retq
;
; AVX512F-LABEL: blend_splat1_mask_cond_v4i32:
; AVX512F: # %bb.0:
; AVX512F-NEXT: # kill: def $xmm2 killed $xmm2 def $zmm2
; AVX512F-NEXT: # kill: def $xmm1 killed $xmm1 def $zmm1
; AVX512F-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0
; AVX512F-NEXT: vptestnmd {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to16}, %zmm0, %k1
; AVX512F-NEXT: vpblendmd %zmm1, %zmm2, %zmm0 {%k1}
; AVX512F-NEXT: # kill: def $xmm0 killed $xmm0 killed $zmm0
; AVX512F-NEXT: vzeroupper
; AVX512F-NEXT: retq
;
; AVX512VL-LABEL: blend_splat1_mask_cond_v4i32:
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vptestnmd {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to4}, %xmm0, %k1
; AVX512VL-NEXT: vpblendmd %xmm1, %xmm2, %xmm0 {%k1}
; AVX512VL-NEXT: retq
;
; XOP-LABEL: blend_splat1_mask_cond_v4i32:
; XOP: # %bb.0:
; XOP-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
; XOP-NEXT: vpxor %xmm3, %xmm3, %xmm3
; XOP-NEXT: vpcomneqd %xmm3, %xmm0, %xmm0
; XOP-NEXT: vblendvps %xmm0, %xmm2, %xmm1, %xmm0
; XOP-NEXT: retq
%a = and <4 x i32> %x, <i32 1, i32 1, i32 1, i32 1>
%c = icmp eq <4 x i32> %a, zeroinitializer
%r = select <4 x i1> %c, <4 x i32> %y, <4 x i32> %z
ret <4 x i32> %r
}
define <16 x i16> @blend_splat1_mask_cond_v16i16(<16 x i16> %x, <16 x i16> %y, <16 x i16> %z) {
; AVX1-LABEL: blend_splat1_mask_cond_v16i16:
; AVX1: # %bb.0:
; AVX1-NEXT: vpsllw $15, %xmm0, %xmm3
; AVX1-NEXT: vpsraw $15, %xmm3, %xmm3
; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm0
; AVX1-NEXT: vpsllw $15, %xmm0, %xmm0
; AVX1-NEXT: vpsraw $15, %xmm0, %xmm0
; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm3, %ymm0
; AVX1-NEXT: vandnps %ymm1, %ymm0, %ymm1
; AVX1-NEXT: vandps %ymm0, %ymm2, %ymm0
; AVX1-NEXT: vorps %ymm1, %ymm0, %ymm0
; AVX1-NEXT: retq
;
; AVX2-LABEL: blend_splat1_mask_cond_v16i16:
; AVX2: # %bb.0:
; AVX2-NEXT: vpsllw $15, %ymm0, %ymm0
; AVX2-NEXT: vpsraw $15, %ymm0, %ymm0
; AVX2-NEXT: vpblendvb %ymm0, %ymm2, %ymm1, %ymm0
; AVX2-NEXT: retq
;
; AVX512F-LABEL: blend_splat1_mask_cond_v16i16:
; AVX512F: # %bb.0:
; AVX512F-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm0
; AVX512F-NEXT: vpxor %xmm3, %xmm3, %xmm3
; AVX512F-NEXT: vpcmpeqw %ymm3, %ymm0, %ymm0
; AVX512F-NEXT: vpblendvb %ymm0, %ymm1, %ymm2, %ymm0
; AVX512F-NEXT: retq
;
; AVX512VL-LABEL: blend_splat1_mask_cond_v16i16:
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vpandd {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to8}, %ymm0, %ymm0
; AVX512VL-NEXT: vpxor %xmm3, %xmm3, %xmm3
; AVX512VL-NEXT: vpcmpeqw %ymm3, %ymm0, %ymm0
; AVX512VL-NEXT: vpternlogq $202, %ymm2, %ymm1, %ymm0
; AVX512VL-NEXT: retq
;
; XOP-LABEL: blend_splat1_mask_cond_v16i16:
; XOP: # %bb.0:
; XOP-NEXT: vpsllw $15, %xmm0, %xmm3
; XOP-NEXT: vpsraw $15, %xmm3, %xmm3
; XOP-NEXT: vextractf128 $1, %ymm0, %xmm0
; XOP-NEXT: vpsllw $15, %xmm0, %xmm0
; XOP-NEXT: vpsraw $15, %xmm0, %xmm0
; XOP-NEXT: vinsertf128 $1, %xmm0, %ymm3, %ymm0
; XOP-NEXT: vpcmov %ymm0, %ymm1, %ymm2, %ymm0
; XOP-NEXT: retq
%a = and <16 x i16> %x, <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1>
%c = icmp eq <16 x i16> %a, zeroinitializer
%r = select <16 x i1> %c, <16 x i16> %y, <16 x i16> %z
ret <16 x i16> %r
}
define <16 x i8> @blend_splat1_mask_cond_v16i8(<16 x i8> %x, <16 x i8> %y, <16 x i8> %z) {
; AVX12-LABEL: blend_splat1_mask_cond_v16i8:
; AVX12: # %bb.0:
; AVX12-NEXT: vpsllw $7, %xmm0, %xmm0
; AVX12-NEXT: vpblendvb %xmm0, %xmm2, %xmm1, %xmm0
; AVX12-NEXT: retq
;
; AVX512F-LABEL: blend_splat1_mask_cond_v16i8:
; AVX512F: # %bb.0:
; AVX512F-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
; AVX512F-NEXT: vpxor %xmm3, %xmm3, %xmm3
; AVX512F-NEXT: vpcmpeqb %xmm3, %xmm0, %xmm0
; AVX512F-NEXT: vpblendvb %xmm0, %xmm1, %xmm2, %xmm0
; AVX512F-NEXT: retq
;
; AVX512VL-LABEL: blend_splat1_mask_cond_v16i8:
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vpandd {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to4}, %xmm0, %xmm0
; AVX512VL-NEXT: vpxor %xmm3, %xmm3, %xmm3
; AVX512VL-NEXT: vpcmpeqb %xmm3, %xmm0, %xmm0
; AVX512VL-NEXT: vpternlogq $202, %xmm2, %xmm1, %xmm0
; AVX512VL-NEXT: retq
;
; XOP-LABEL: blend_splat1_mask_cond_v16i8:
; XOP: # %bb.0:
; XOP-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
; XOP-NEXT: vpxor %xmm3, %xmm3, %xmm3
; XOP-NEXT: vpcomneqb %xmm3, %xmm0, %xmm0
; XOP-NEXT: vpblendvb %xmm0, %xmm2, %xmm1, %xmm0
; XOP-NEXT: retq
%a = and <16 x i8> %x, <i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1>
%c = icmp eq <16 x i8> %a, zeroinitializer
%r = select <16 x i1> %c, <16 x i8> %y, <16 x i8> %z
ret <16 x i8> %r
}
define <2 x i64> @blend_splatmax_mask_cond_v2i64(<2 x i64> %x, <2 x i64> %y, <2 x i64> %z) {
; AVX12-LABEL: blend_splatmax_mask_cond_v2i64:
; AVX12: # %bb.0:
; AVX12-NEXT: vblendvpd %xmm0, %xmm2, %xmm1, %xmm0
; AVX12-NEXT: retq
;
; AVX512F-LABEL: blend_splatmax_mask_cond_v2i64:
; AVX512F: # %bb.0:
; AVX512F-NEXT: # kill: def $xmm2 killed $xmm2 def $zmm2
; AVX512F-NEXT: # kill: def $xmm1 killed $xmm1 def $zmm1
; AVX512F-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0
; AVX512F-NEXT: vpbroadcastq {{.*#+}} xmm3 = [9223372036854775808,9223372036854775808]
; AVX512F-NEXT: vptestnmq %zmm3, %zmm0, %k1
; AVX512F-NEXT: vpblendmq %zmm1, %zmm2, %zmm0 {%k1}
; AVX512F-NEXT: # kill: def $xmm0 killed $xmm0 killed $zmm0
; AVX512F-NEXT: vzeroupper
; AVX512F-NEXT: retq
;
; AVX512VL-LABEL: blend_splatmax_mask_cond_v2i64:
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vptestnmq {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to2}, %xmm0, %k1
; AVX512VL-NEXT: vpblendmq %xmm1, %xmm2, %xmm0 {%k1}
; AVX512VL-NEXT: retq
;
; XOP-LABEL: blend_splatmax_mask_cond_v2i64:
; XOP: # %bb.0:
; XOP-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
; XOP-NEXT: vpxor %xmm3, %xmm3, %xmm3
; XOP-NEXT: vpcomneqq %xmm3, %xmm0, %xmm0
; XOP-NEXT: vblendvpd %xmm0, %xmm2, %xmm1, %xmm0
; XOP-NEXT: retq
%a = and <2 x i64> %x, <i64 9223372036854775808, i64 9223372036854775808>
%c = icmp eq <2 x i64> %a, zeroinitializer
%r = select <2 x i1> %c, <2 x i64> %y, <2 x i64> %z
ret <2 x i64> %r
}
define <8 x i32> @blend_splatmax_mask_cond_v8i32(<8 x i32> %x, <8 x i32> %y, <8 x i32> %z) {
; AVX12-LABEL: blend_splatmax_mask_cond_v8i32:
; AVX12: # %bb.0:
; AVX12-NEXT: vblendvps %ymm0, %ymm2, %ymm1, %ymm0
; AVX12-NEXT: retq
;
; AVX512F-LABEL: blend_splatmax_mask_cond_v8i32:
; AVX512F: # %bb.0:
; AVX512F-NEXT: # kill: def $ymm2 killed $ymm2 def $zmm2
; AVX512F-NEXT: # kill: def $ymm1 killed $ymm1 def $zmm1
; AVX512F-NEXT: # kill: def $ymm0 killed $ymm0 def $zmm0
; AVX512F-NEXT: vptestnmd {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to16}, %zmm0, %k1
; AVX512F-NEXT: vpblendmd %zmm1, %zmm2, %zmm0 {%k1}
; AVX512F-NEXT: # kill: def $ymm0 killed $ymm0 killed $zmm0
; AVX512F-NEXT: retq
;
; AVX512VL-LABEL: blend_splatmax_mask_cond_v8i32:
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vptestnmd {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to8}, %ymm0, %k1
; AVX512VL-NEXT: vpblendmd %ymm1, %ymm2, %ymm0 {%k1}
; AVX512VL-NEXT: retq
;
; XOP-LABEL: blend_splatmax_mask_cond_v8i32:
; XOP: # %bb.0:
; XOP-NEXT: vblendvps %ymm0, %ymm2, %ymm1, %ymm0
; XOP-NEXT: retq
%a = and <8 x i32> %x, <i32 2147483648, i32 2147483648, i32 2147483648, i32 2147483648, i32 2147483648, i32 2147483648, i32 2147483648, i32 2147483648>
%c = icmp eq <8 x i32> %a, zeroinitializer
%r = select <8 x i1> %c, <8 x i32> %y, <8 x i32> %z
ret <8 x i32> %r
}
define <8 x i16> @blend_splatmax_mask_cond_v8i16(<8 x i16> %x, <8 x i16> %y, <8 x i16> %z) {
; AVX12-LABEL: blend_splatmax_mask_cond_v8i16:
; AVX12: # %bb.0:
; AVX12-NEXT: vpsraw $15, %xmm0, %xmm0
; AVX12-NEXT: vpblendvb %xmm0, %xmm2, %xmm1, %xmm0
; AVX12-NEXT: retq
;
; AVX512F-LABEL: blend_splatmax_mask_cond_v8i16:
; AVX512F: # %bb.0:
; AVX512F-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
; AVX512F-NEXT: vpxor %xmm3, %xmm3, %xmm3
; AVX512F-NEXT: vpcmpeqw %xmm3, %xmm0, %xmm0
; AVX512F-NEXT: vpblendvb %xmm0, %xmm1, %xmm2, %xmm0
; AVX512F-NEXT: retq
;
; AVX512VL-LABEL: blend_splatmax_mask_cond_v8i16:
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vpandd {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to4}, %xmm0, %xmm0
; AVX512VL-NEXT: vpxor %xmm3, %xmm3, %xmm3
; AVX512VL-NEXT: vpcmpeqw %xmm3, %xmm0, %xmm0
; AVX512VL-NEXT: vpternlogq $202, %xmm2, %xmm1, %xmm0
; AVX512VL-NEXT: retq
;
; XOP-LABEL: blend_splatmax_mask_cond_v8i16:
; XOP: # %bb.0:
; XOP-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
; XOP-NEXT: vpxor %xmm3, %xmm3, %xmm3
; XOP-NEXT: vpcomneqw %xmm3, %xmm0, %xmm0
; XOP-NEXT: vpblendvb %xmm0, %xmm2, %xmm1, %xmm0
; XOP-NEXT: retq
%a = and <8 x i16> %x, <i16 32768, i16 32768, i16 32768, i16 32768, i16 32768, i16 32768, i16 32768, i16 32768>
%c = icmp eq <8 x i16> %a, zeroinitializer
%r = select <8 x i1> %c, <8 x i16> %y, <8 x i16> %z
ret <8 x i16> %r
}
define <32 x i8> @blend_splatmax_mask_cond_v32i8(<32 x i8> %x, <32 x i8> %y, <32 x i8> %z) {
; AVX1-LABEL: blend_splatmax_mask_cond_v32i8:
; AVX1: # %bb.0:
; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm3
; AVX1-NEXT: vpxor %xmm4, %xmm4, %xmm4
; AVX1-NEXT: vpcmpgtb %xmm3, %xmm4, %xmm3
; AVX1-NEXT: vpcmpgtb %xmm0, %xmm4, %xmm0
; AVX1-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm0
; AVX1-NEXT: vandnps %ymm1, %ymm0, %ymm1
; AVX1-NEXT: vandps %ymm0, %ymm2, %ymm0
; AVX1-NEXT: vorps %ymm1, %ymm0, %ymm0
; AVX1-NEXT: retq
;
; AVX2-LABEL: blend_splatmax_mask_cond_v32i8:
; AVX2: # %bb.0:
; AVX2-NEXT: vpblendvb %ymm0, %ymm2, %ymm1, %ymm0
; AVX2-NEXT: retq
;
; AVX512F-LABEL: blend_splatmax_mask_cond_v32i8:
; AVX512F: # %bb.0:
; AVX512F-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm0
; AVX512F-NEXT: vpxor %xmm3, %xmm3, %xmm3
; AVX512F-NEXT: vpcmpeqb %ymm3, %ymm0, %ymm0
; AVX512F-NEXT: vpblendvb %ymm0, %ymm1, %ymm2, %ymm0
; AVX512F-NEXT: retq
;
; AVX512VL-LABEL: blend_splatmax_mask_cond_v32i8:
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vpandd {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to8}, %ymm0, %ymm0
; AVX512VL-NEXT: vpxor %xmm3, %xmm3, %xmm3
; AVX512VL-NEXT: vpcmpeqb %ymm3, %ymm0, %ymm0
; AVX512VL-NEXT: vpternlogq $202, %ymm2, %ymm1, %ymm0
; AVX512VL-NEXT: retq
;
; XOP-LABEL: blend_splatmax_mask_cond_v32i8:
; XOP: # %bb.0:
; XOP-NEXT: vextractf128 $1, %ymm0, %xmm3
; XOP-NEXT: vpxor %xmm4, %xmm4, %xmm4
; XOP-NEXT: vpcmpgtb %xmm3, %xmm4, %xmm3
; XOP-NEXT: vpcmpgtb %xmm0, %xmm4, %xmm0
; XOP-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm0
; XOP-NEXT: vpcmov %ymm0, %ymm1, %ymm2, %ymm0
; XOP-NEXT: retq
%a = and <32 x i8> %x, <i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128>
%c = icmp eq <32 x i8> %a, zeroinitializer
%r = select <32 x i1> %c, <32 x i8> %y, <32 x i8> %z
ret <32 x i8> %r
}
define <4 x i64> @blend_splat_mask_cond_v4i64(<4 x i64> %x, <4 x i64> %y, <4 x i64> %z) {
; AVX1-LABEL: blend_splat_mask_cond_v4i64:
; AVX1: # %bb.0:
; AVX1-NEXT: vpsllq $62, %xmm0, %xmm3
; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm0
; AVX1-NEXT: vpsllq $62, %xmm0, %xmm0
; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm3, %ymm0
; AVX1-NEXT: vblendvpd %ymm0, %ymm2, %ymm1, %ymm0
; AVX1-NEXT: retq
;
; AVX2-LABEL: blend_splat_mask_cond_v4i64:
; AVX2: # %bb.0:
; AVX2-NEXT: vpsllq $62, %ymm0, %ymm0
; AVX2-NEXT: vblendvpd %ymm0, %ymm2, %ymm1, %ymm0
; AVX2-NEXT: retq
;
; AVX512F-LABEL: blend_splat_mask_cond_v4i64:
; AVX512F: # %bb.0:
; AVX512F-NEXT: # kill: def $ymm2 killed $ymm2 def $zmm2
; AVX512F-NEXT: # kill: def $ymm1 killed $ymm1 def $zmm1
; AVX512F-NEXT: # kill: def $ymm0 killed $ymm0 def $zmm0
; AVX512F-NEXT: vptestnmq {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to8}, %zmm0, %k1
; AVX512F-NEXT: vpblendmq %zmm1, %zmm2, %zmm0 {%k1}
; AVX512F-NEXT: # kill: def $ymm0 killed $ymm0 killed $zmm0
; AVX512F-NEXT: retq
;
; AVX512VL-LABEL: blend_splat_mask_cond_v4i64:
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vptestnmq {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to4}, %ymm0, %k1
; AVX512VL-NEXT: vpblendmq %ymm1, %ymm2, %ymm0 {%k1}
; AVX512VL-NEXT: retq
;
; XOP-LABEL: blend_splat_mask_cond_v4i64:
; XOP: # %bb.0:
; XOP-NEXT: vextractf128 $1, %ymm0, %xmm3
; XOP-NEXT: vpsllq $62, %xmm3, %xmm3
; XOP-NEXT: vpmovsxbq {{.*#+}} xmm4 = [18446744073709551553,18446744073709551553]
; XOP-NEXT: vpshaq %xmm4, %xmm3, %xmm3
; XOP-NEXT: vpsllq $62, %xmm0, %xmm0
; XOP-NEXT: vpshaq %xmm4, %xmm0, %xmm0
; XOP-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm0
; XOP-NEXT: vblendvpd %ymm0, %ymm2, %ymm1, %ymm0
; XOP-NEXT: retq
%a = and <4 x i64> %x, <i64 2, i64 2, i64 2, i64 2>
%c = icmp eq <4 x i64> %a, zeroinitializer
%r = select <4 x i1> %c, <4 x i64> %y, <4 x i64> %z
ret <4 x i64> %r
}
define <4 x i32> @blend_splat_mask_cond_v4i32(<4 x i32> %x, <4 x i32> %y, <4 x i32> %z) {
; AVX12-LABEL: blend_splat_mask_cond_v4i32:
; AVX12: # %bb.0:
; AVX12-NEXT: vpslld $15, %xmm0, %xmm0
; AVX12-NEXT: vblendvps %xmm0, %xmm2, %xmm1, %xmm0
; AVX12-NEXT: retq
;
; AVX512F-LABEL: blend_splat_mask_cond_v4i32:
; AVX512F: # %bb.0:
; AVX512F-NEXT: # kill: def $xmm2 killed $xmm2 def $zmm2
; AVX512F-NEXT: # kill: def $xmm1 killed $xmm1 def $zmm1
; AVX512F-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0
; AVX512F-NEXT: vptestnmd {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to16}, %zmm0, %k1
; AVX512F-NEXT: vpblendmd %zmm1, %zmm2, %zmm0 {%k1}
; AVX512F-NEXT: # kill: def $xmm0 killed $xmm0 killed $zmm0
; AVX512F-NEXT: vzeroupper
; AVX512F-NEXT: retq
;
; AVX512VL-LABEL: blend_splat_mask_cond_v4i32:
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vptestnmd {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to4}, %xmm0, %k1
; AVX512VL-NEXT: vpblendmd %xmm1, %xmm2, %xmm0 {%k1}
; AVX512VL-NEXT: retq
;
; XOP-LABEL: blend_splat_mask_cond_v4i32:
; XOP: # %bb.0:
; XOP-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
; XOP-NEXT: vpxor %xmm3, %xmm3, %xmm3
; XOP-NEXT: vpcomneqd %xmm3, %xmm0, %xmm0
; XOP-NEXT: vblendvps %xmm0, %xmm2, %xmm1, %xmm0
; XOP-NEXT: retq
%a = and <4 x i32> %x, <i32 65536, i32 65536, i32 65536, i32 65536>
%c = icmp eq <4 x i32> %a, zeroinitializer
%r = select <4 x i1> %c, <4 x i32> %y, <4 x i32> %z
ret <4 x i32> %r
}
define <16 x i16> @blend_splat_mask_cond_v16i16(<16 x i16> %x, <16 x i16> %y, <16 x i16> %z) {
; AVX1-LABEL: blend_splat_mask_cond_v16i16:
; AVX1: # %bb.0:
; AVX1-NEXT: vpsllw $5, %xmm0, %xmm3
; AVX1-NEXT: vpsraw $15, %xmm3, %xmm3
; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm0
; AVX1-NEXT: vpsllw $5, %xmm0, %xmm0
; AVX1-NEXT: vpsraw $15, %xmm0, %xmm0
; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm3, %ymm0
; AVX1-NEXT: vandnps %ymm1, %ymm0, %ymm1
; AVX1-NEXT: vandps %ymm0, %ymm2, %ymm0
; AVX1-NEXT: vorps %ymm1, %ymm0, %ymm0
; AVX1-NEXT: retq
;
; AVX2-LABEL: blend_splat_mask_cond_v16i16:
; AVX2: # %bb.0:
; AVX2-NEXT: vpsllw $5, %ymm0, %ymm0
; AVX2-NEXT: vpsraw $15, %ymm0, %ymm0
; AVX2-NEXT: vpblendvb %ymm0, %ymm2, %ymm1, %ymm0
; AVX2-NEXT: retq
;
; AVX512F-LABEL: blend_splat_mask_cond_v16i16:
; AVX512F: # %bb.0:
; AVX512F-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm0
; AVX512F-NEXT: vpxor %xmm3, %xmm3, %xmm3
; AVX512F-NEXT: vpcmpeqw %ymm3, %ymm0, %ymm0
; AVX512F-NEXT: vpblendvb %ymm0, %ymm1, %ymm2, %ymm0
; AVX512F-NEXT: retq
;
; AVX512VL-LABEL: blend_splat_mask_cond_v16i16:
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vpandd {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to8}, %ymm0, %ymm0
; AVX512VL-NEXT: vpxor %xmm3, %xmm3, %xmm3
; AVX512VL-NEXT: vpcmpeqw %ymm3, %ymm0, %ymm0
; AVX512VL-NEXT: vpternlogq $202, %ymm2, %ymm1, %ymm0
; AVX512VL-NEXT: retq
;
; XOP-LABEL: blend_splat_mask_cond_v16i16:
; XOP: # %bb.0:
; XOP-NEXT: vpsllw $5, %xmm0, %xmm3
; XOP-NEXT: vpsraw $15, %xmm3, %xmm3
; XOP-NEXT: vextractf128 $1, %ymm0, %xmm0
; XOP-NEXT: vpsllw $5, %xmm0, %xmm0
; XOP-NEXT: vpsraw $15, %xmm0, %xmm0
; XOP-NEXT: vinsertf128 $1, %xmm0, %ymm3, %ymm0
; XOP-NEXT: vpcmov %ymm0, %ymm1, %ymm2, %ymm0
; XOP-NEXT: retq
%a = and <16 x i16> %x, <i16 1024, i16 1024, i16 1024, i16 1024, i16 1024, i16 1024, i16 1024, i16 1024, i16 1024, i16 1024, i16 1024, i16 1024, i16 1024, i16 1024, i16 1024, i16 1024>
%c = icmp eq <16 x i16> %a, zeroinitializer
%r = select <16 x i1> %c, <16 x i16> %y, <16 x i16> %z
ret <16 x i16> %r
}
define <16 x i8> @blend_splat_mask_cond_v16i8(<16 x i8> %x, <16 x i8> %y, <16 x i8> %z) {
; AVX12-LABEL: blend_splat_mask_cond_v16i8:
; AVX12: # %bb.0:
; AVX12-NEXT: vpsllw $5, %xmm0, %xmm0
; AVX12-NEXT: vpblendvb %xmm0, %xmm2, %xmm1, %xmm0
; AVX12-NEXT: retq
;
; AVX512F-LABEL: blend_splat_mask_cond_v16i8:
; AVX512F: # %bb.0:
; AVX512F-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
; AVX512F-NEXT: vpxor %xmm3, %xmm3, %xmm3
; AVX512F-NEXT: vpcmpeqb %xmm3, %xmm0, %xmm0
; AVX512F-NEXT: vpblendvb %xmm0, %xmm1, %xmm2, %xmm0
; AVX512F-NEXT: retq
;
; AVX512VL-LABEL: blend_splat_mask_cond_v16i8:
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vpandd {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to4}, %xmm0, %xmm0
; AVX512VL-NEXT: vpxor %xmm3, %xmm3, %xmm3
; AVX512VL-NEXT: vpcmpeqb %xmm3, %xmm0, %xmm0
; AVX512VL-NEXT: vpternlogq $202, %xmm2, %xmm1, %xmm0
; AVX512VL-NEXT: retq
;
; XOP-LABEL: blend_splat_mask_cond_v16i8:
; XOP: # %bb.0:
; XOP-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
; XOP-NEXT: vpxor %xmm3, %xmm3, %xmm3
; XOP-NEXT: vpcomneqb %xmm3, %xmm0, %xmm0
; XOP-NEXT: vpblendvb %xmm0, %xmm2, %xmm1, %xmm0
; XOP-NEXT: retq
%a = and <16 x i8> %x, <i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4>
%c = icmp eq <16 x i8> %a, zeroinitializer
%r = select <16 x i1> %c, <16 x i8> %y, <16 x i8> %z
ret <16 x i8> %r
}
define <2 x i64> @blend_mask_cond_v2i64(<2 x i64> %x, <2 x i64> %y, <2 x i64> %z) {
; AVX1-LABEL: blend_mask_cond_v2i64:
; AVX1: # %bb.0:
; AVX1-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
; AVX1-NEXT: vpxor %xmm3, %xmm3, %xmm3
; AVX1-NEXT: vpcmpeqq %xmm3, %xmm0, %xmm0
; AVX1-NEXT: vblendvpd %xmm0, %xmm1, %xmm2, %xmm0
; AVX1-NEXT: retq
;
; AVX2-LABEL: blend_mask_cond_v2i64:
; AVX2: # %bb.0:
; AVX2-NEXT: vpsllvq {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
; AVX2-NEXT: vblendvpd %xmm0, %xmm2, %xmm1, %xmm0
; AVX2-NEXT: retq
;
; AVX512F-LABEL: blend_mask_cond_v2i64:
; AVX512F: # %bb.0:
; AVX512F-NEXT: # kill: def $xmm2 killed $xmm2 def $zmm2
; AVX512F-NEXT: # kill: def $xmm1 killed $xmm1 def $zmm1
; AVX512F-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0
; AVX512F-NEXT: vpmovsxbq {{.*#+}} xmm3 = [1,4]
; AVX512F-NEXT: vptestnmq %zmm3, %zmm0, %k1
; AVX512F-NEXT: vpblendmq %zmm1, %zmm2, %zmm0 {%k1}
; AVX512F-NEXT: # kill: def $xmm0 killed $xmm0 killed $zmm0
; AVX512F-NEXT: vzeroupper
; AVX512F-NEXT: retq
;
; AVX512VL-LABEL: blend_mask_cond_v2i64:
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vptestnmq {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %k1
; AVX512VL-NEXT: vpblendmq %xmm1, %xmm2, %xmm0 {%k1}
; AVX512VL-NEXT: retq
;
; XOP-LABEL: blend_mask_cond_v2i64:
; XOP: # %bb.0:
; XOP-NEXT: vpshlq {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
; XOP-NEXT: vblendvpd %xmm0, %xmm2, %xmm1, %xmm0
; XOP-NEXT: retq
%a = and <2 x i64> %x, <i64 1, i64 4>
%c = icmp eq <2 x i64> %a, zeroinitializer
%r = select <2 x i1> %c, <2 x i64> %y, <2 x i64> %z
ret <2 x i64> %r
}
define <4 x i32> @blend_mask_cond_v4i32(<4 x i32> %x, <4 x i32> %y, <4 x i32> %z) {
; AVX1-LABEL: blend_mask_cond_v4i32:
; AVX1: # %bb.0:
; AVX1-NEXT: vpmulld {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
; AVX1-NEXT: vblendvps %xmm0, %xmm2, %xmm1, %xmm0
; AVX1-NEXT: retq
;
; AVX2-LABEL: blend_mask_cond_v4i32:
; AVX2: # %bb.0:
; AVX2-NEXT: vpsllvd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
; AVX2-NEXT: vblendvps %xmm0, %xmm2, %xmm1, %xmm0
; AVX2-NEXT: retq
;
; AVX512F-LABEL: blend_mask_cond_v4i32:
; AVX512F: # %bb.0:
; AVX512F-NEXT: # kill: def $xmm2 killed $xmm2 def $zmm2
; AVX512F-NEXT: # kill: def $xmm1 killed $xmm1 def $zmm1
; AVX512F-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0
; AVX512F-NEXT: vmovdqa {{.*#+}} xmm3 = [65536,512,2,1]
; AVX512F-NEXT: vptestnmd %zmm3, %zmm0, %k1
; AVX512F-NEXT: vpblendmd %zmm1, %zmm2, %zmm0 {%k1}
; AVX512F-NEXT: # kill: def $xmm0 killed $xmm0 killed $zmm0
; AVX512F-NEXT: vzeroupper
; AVX512F-NEXT: retq
;
; AVX512VL-LABEL: blend_mask_cond_v4i32:
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vptestnmd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %k1
; AVX512VL-NEXT: vpblendmd %xmm1, %xmm2, %xmm0 {%k1}
; AVX512VL-NEXT: retq
;
; XOP-LABEL: blend_mask_cond_v4i32:
; XOP: # %bb.0:
; XOP-NEXT: vpshld {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
; XOP-NEXT: vblendvps %xmm0, %xmm2, %xmm1, %xmm0
; XOP-NEXT: retq
%a = and <4 x i32> %x, <i32 65536, i32 512, i32 2, i32 1>
%c = icmp eq <4 x i32> %a, zeroinitializer
%r = select <4 x i1> %c, <4 x i32> %y, <4 x i32> %z
ret <4 x i32> %r
}
define <8 x i16> @blend_mask_cond_v8i16(<8 x i16> %x, <8 x i16> %y, <8 x i16> %z) {
; AVX12-LABEL: blend_mask_cond_v8i16:
; AVX12: # %bb.0:
; AVX12-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
; AVX12-NEXT: vpxor %xmm3, %xmm3, %xmm3
; AVX12-NEXT: vpcmpeqw %xmm3, %xmm0, %xmm0
; AVX12-NEXT: vpblendvb %xmm0, %xmm1, %xmm2, %xmm0
; AVX12-NEXT: retq
;
; AVX512F-LABEL: blend_mask_cond_v8i16:
; AVX512F: # %bb.0:
; AVX512F-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
; AVX512F-NEXT: vpxor %xmm3, %xmm3, %xmm3
; AVX512F-NEXT: vpcmpeqw %xmm3, %xmm0, %xmm0
; AVX512F-NEXT: vpblendvb %xmm0, %xmm1, %xmm2, %xmm0
; AVX512F-NEXT: retq
;
; AVX512VL-LABEL: blend_mask_cond_v8i16:
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
; AVX512VL-NEXT: vpxor %xmm3, %xmm3, %xmm3
; AVX512VL-NEXT: vpcmpeqw %xmm3, %xmm0, %xmm0
; AVX512VL-NEXT: vpternlogq $202, %xmm2, %xmm1, %xmm0
; AVX512VL-NEXT: retq
;
; XOP-LABEL: blend_mask_cond_v8i16:
; XOP: # %bb.0:
; XOP-NEXT: vpxor %xmm3, %xmm3, %xmm3
; XOP-NEXT: vpshlw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
; XOP-NEXT: vpcomltw %xmm3, %xmm0, %xmm0
; XOP-NEXT: vpblendvb %xmm0, %xmm2, %xmm1, %xmm0
; XOP-NEXT: retq
%a = and <8 x i16> %x, <i16 1, i16 2, i16 8, i16 4, i16 8, i16 1024, i16 2, i16 4096>
%c = icmp eq <8 x i16> %a, zeroinitializer
%r = select <8 x i1> %c, <8 x i16> %y, <8 x i16> %z
ret <8 x i16> %r
}
define <16 x i8> @blend_mask_cond_v16i8(<16 x i8> %x, <16 x i8> %y, <16 x i8> %z) {
; AVX12-LABEL: blend_mask_cond_v16i8:
; AVX12: # %bb.0:
; AVX12-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
; AVX12-NEXT: vpxor %xmm3, %xmm3, %xmm3
; AVX12-NEXT: vpcmpeqb %xmm3, %xmm0, %xmm0
; AVX12-NEXT: vpblendvb %xmm0, %xmm1, %xmm2, %xmm0
; AVX12-NEXT: retq
;
; AVX512F-LABEL: blend_mask_cond_v16i8:
; AVX512F: # %bb.0:
; AVX512F-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
; AVX512F-NEXT: vpxor %xmm3, %xmm3, %xmm3
; AVX512F-NEXT: vpcmpeqb %xmm3, %xmm0, %xmm0
; AVX512F-NEXT: vpblendvb %xmm0, %xmm1, %xmm2, %xmm0
; AVX512F-NEXT: retq
;
; AVX512VL-LABEL: blend_mask_cond_v16i8:
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
; AVX512VL-NEXT: vpxor %xmm3, %xmm3, %xmm3
; AVX512VL-NEXT: vpcmpeqb %xmm3, %xmm0, %xmm0
; AVX512VL-NEXT: vpternlogq $202, %xmm2, %xmm1, %xmm0
; AVX512VL-NEXT: retq
;
; XOP-LABEL: blend_mask_cond_v16i8:
; XOP: # %bb.0:
; XOP-NEXT: vpshlb {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
; XOP-NEXT: vpblendvb %xmm0, %xmm2, %xmm1, %xmm0
; XOP-NEXT: retq
%a = and <16 x i8> %x, <i8 1, i8 2, i8 4, i8 8, i8 16, i8 32, i8 64, i8 128, i8 4, i8 4, i8 4, i8 4, i8 2, i8 2, i8 2, i8 2>
%c = icmp eq <16 x i8> %a, zeroinitializer
%r = select <16 x i1> %c, <16 x i8> %y, <16 x i8> %z
ret <16 x i8> %r
}
define <4 x i64> @blend_mask_cond_v4i64(<4 x i64> %x, <4 x i64> %y, <4 x i64> %z) {
; AVX1-LABEL: blend_mask_cond_v4i64:
; AVX1: # %bb.0:
; AVX1-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm0
; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm3
; AVX1-NEXT: vpxor %xmm4, %xmm4, %xmm4
; AVX1-NEXT: vpcmpeqq %xmm4, %xmm3, %xmm3
; AVX1-NEXT: vpcmpeqq %xmm4, %xmm0, %xmm0
; AVX1-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm0
; AVX1-NEXT: vblendvpd %ymm0, %ymm1, %ymm2, %ymm0
; AVX1-NEXT: retq
;
; AVX2-LABEL: blend_mask_cond_v4i64:
; AVX2: # %bb.0:
; AVX2-NEXT: vpsllvq {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm0
; AVX2-NEXT: vblendvpd %ymm0, %ymm2, %ymm1, %ymm0
; AVX2-NEXT: retq
;
; AVX512F-LABEL: blend_mask_cond_v4i64:
; AVX512F: # %bb.0:
; AVX512F-NEXT: # kill: def $ymm2 killed $ymm2 def $zmm2
; AVX512F-NEXT: # kill: def $ymm1 killed $ymm1 def $zmm1
; AVX512F-NEXT: # kill: def $ymm0 killed $ymm0 def $zmm0
; AVX512F-NEXT: vpmovzxwq {{.*#+}} ymm3 = [2,4,32768,1]
; AVX512F-NEXT: vptestnmq %zmm3, %zmm0, %k1
; AVX512F-NEXT: vpblendmq %zmm1, %zmm2, %zmm0 {%k1}
; AVX512F-NEXT: # kill: def $ymm0 killed $ymm0 killed $zmm0
; AVX512F-NEXT: retq
;
; AVX512VL-LABEL: blend_mask_cond_v4i64:
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vptestnmq {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %k1
; AVX512VL-NEXT: vpblendmq %ymm1, %ymm2, %ymm0 {%k1}
; AVX512VL-NEXT: retq
;
; XOP-LABEL: blend_mask_cond_v4i64:
; XOP: # %bb.0:
; XOP-NEXT: vpshlq {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm3
; XOP-NEXT: vextractf128 $1, %ymm0, %xmm0
; XOP-NEXT: vpshlq {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
; XOP-NEXT: vinsertf128 $1, %xmm0, %ymm3, %ymm0
; XOP-NEXT: vblendvpd %ymm0, %ymm2, %ymm1, %ymm0
; XOP-NEXT: retq
%a = and <4 x i64> %x, <i64 2, i64 4, i64 32768, i64 1>
%c = icmp eq <4 x i64> %a, zeroinitializer
%r = select <4 x i1> %c, <4 x i64> %y, <4 x i64> %z
ret <4 x i64> %r
}
define <8 x i32> @blend_mask_cond_v8i32(<8 x i32> %x, <8 x i32> %y, <8 x i32> %z) {
; AVX1-LABEL: blend_mask_cond_v8i32:
; AVX1: # %bb.0:
; AVX1-NEXT: vpmulld {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm3
; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm0
; AVX1-NEXT: vpmulld {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm3, %ymm0
; AVX1-NEXT: vblendvps %ymm0, %ymm2, %ymm1, %ymm0
; AVX1-NEXT: retq
;
; AVX2-LABEL: blend_mask_cond_v8i32:
; AVX2: # %bb.0:
; AVX2-NEXT: vpsllvd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm0
; AVX2-NEXT: vblendvps %ymm0, %ymm2, %ymm1, %ymm0
; AVX2-NEXT: retq
;
; AVX512F-LABEL: blend_mask_cond_v8i32:
; AVX512F: # %bb.0:
; AVX512F-NEXT: # kill: def $ymm2 killed $ymm2 def $zmm2
; AVX512F-NEXT: # kill: def $ymm1 killed $ymm1 def $zmm1
; AVX512F-NEXT: # kill: def $ymm0 killed $ymm0 def $zmm0
; AVX512F-NEXT: vpmovsxwd {{.*#+}} ymm3 = [1,2,8,4,8,1024,2,4096]
; AVX512F-NEXT: vptestnmd %zmm3, %zmm0, %k1
; AVX512F-NEXT: vpblendmd %zmm1, %zmm2, %zmm0 {%k1}
; AVX512F-NEXT: # kill: def $ymm0 killed $ymm0 killed $zmm0
; AVX512F-NEXT: retq
;
; AVX512VL-LABEL: blend_mask_cond_v8i32:
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vptestnmd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %k1
; AVX512VL-NEXT: vpblendmd %ymm1, %ymm2, %ymm0 {%k1}
; AVX512VL-NEXT: retq
;
; XOP-LABEL: blend_mask_cond_v8i32:
; XOP: # %bb.0:
; XOP-NEXT: vpshld {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm3
; XOP-NEXT: vextractf128 $1, %ymm0, %xmm0
; XOP-NEXT: vpshld {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
; XOP-NEXT: vinsertf128 $1, %xmm0, %ymm3, %ymm0
; XOP-NEXT: vblendvps %ymm0, %ymm2, %ymm1, %ymm0
; XOP-NEXT: retq
%a = and <8 x i32> %x, <i32 1, i32 2, i32 8, i32 4, i32 8, i32 1024, i32 2, i32 4096>
%c = icmp eq <8 x i32> %a, zeroinitializer
%r = select <8 x i1> %c, <8 x i32> %y, <8 x i32> %z
ret <8 x i32> %r
}
define <16 x i16> @blend_mask_cond_v16i16(<16 x i16> %x, <16 x i16> %y, <16 x i16> %z) {
; AVX1-LABEL: blend_mask_cond_v16i16:
; AVX1: # %bb.0:
; AVX1-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm0
; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm3
; AVX1-NEXT: vpxor %xmm4, %xmm4, %xmm4
; AVX1-NEXT: vpcmpeqw %xmm4, %xmm3, %xmm3
; AVX1-NEXT: vpcmpeqw %xmm4, %xmm0, %xmm0
; AVX1-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm0
; AVX1-NEXT: vandnps %ymm2, %ymm0, %ymm2
; AVX1-NEXT: vandps %ymm0, %ymm1, %ymm0
; AVX1-NEXT: vorps %ymm2, %ymm0, %ymm0
; AVX1-NEXT: retq
;
; AVX2-LABEL: blend_mask_cond_v16i16:
; AVX2: # %bb.0:
; AVX2-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm0
; AVX2-NEXT: vpxor %xmm3, %xmm3, %xmm3
; AVX2-NEXT: vpcmpeqw %ymm3, %ymm0, %ymm0
; AVX2-NEXT: vpblendvb %ymm0, %ymm1, %ymm2, %ymm0
; AVX2-NEXT: retq
;
; AVX512F-LABEL: blend_mask_cond_v16i16:
; AVX512F: # %bb.0:
; AVX512F-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm0
; AVX512F-NEXT: vpxor %xmm3, %xmm3, %xmm3
; AVX512F-NEXT: vpcmpeqw %ymm3, %ymm0, %ymm0
; AVX512F-NEXT: vpblendvb %ymm0, %ymm1, %ymm2, %ymm0
; AVX512F-NEXT: retq
;
; AVX512VL-LABEL: blend_mask_cond_v16i16:
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm0
; AVX512VL-NEXT: vpxor %xmm3, %xmm3, %xmm3
; AVX512VL-NEXT: vpcmpeqw %ymm3, %ymm0, %ymm0
; AVX512VL-NEXT: vpternlogq $202, %ymm2, %ymm1, %ymm0
; AVX512VL-NEXT: retq
;
; XOP-LABEL: blend_mask_cond_v16i16:
; XOP: # %bb.0:
; XOP-NEXT: vextractf128 $1, %ymm0, %xmm3
; XOP-NEXT: vpshlw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm3, %xmm3
; XOP-NEXT: vpxor %xmm4, %xmm4, %xmm4
; XOP-NEXT: vpcomltw %xmm4, %xmm3, %xmm3
; XOP-NEXT: vpshlw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
; XOP-NEXT: vpcomltw %xmm4, %xmm0, %xmm0
; XOP-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm0
; XOP-NEXT: vpcmov %ymm0, %ymm1, %ymm2, %ymm0
; XOP-NEXT: retq
%a = and <16 x i16> %x, <i16 1, i16 2, i16 8, i16 4, i16 8, i16 2, i16 2, i16 2, i16 2, i16 8, i16 8, i16 64, i16 64, i16 1024, i16 4096, i16 1024>
%c = icmp eq <16 x i16> %a, zeroinitializer
%r = select <16 x i1> %c, <16 x i16> %y, <16 x i16> %z
ret <16 x i16> %r
}
define <32 x i8> @blend_mask_cond_v32i8(<32 x i8> %x, <32 x i8> %y, <32 x i8> %z) {
; AVX1-LABEL: blend_mask_cond_v32i8:
; AVX1: # %bb.0:
; AVX1-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm0
; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm3
; AVX1-NEXT: vpxor %xmm4, %xmm4, %xmm4
; AVX1-NEXT: vpcmpeqb %xmm4, %xmm3, %xmm3
; AVX1-NEXT: vpcmpeqb %xmm4, %xmm0, %xmm0
; AVX1-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm0
; AVX1-NEXT: vandnps %ymm2, %ymm0, %ymm2
; AVX1-NEXT: vandps %ymm0, %ymm1, %ymm0
; AVX1-NEXT: vorps %ymm2, %ymm0, %ymm0
; AVX1-NEXT: retq
;
; AVX2-LABEL: blend_mask_cond_v32i8:
; AVX2: # %bb.0:
; AVX2-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm0
; AVX2-NEXT: vpxor %xmm3, %xmm3, %xmm3
; AVX2-NEXT: vpcmpeqb %ymm3, %ymm0, %ymm0
; AVX2-NEXT: vpblendvb %ymm0, %ymm1, %ymm2, %ymm0
; AVX2-NEXT: retq
;
; AVX512F-LABEL: blend_mask_cond_v32i8:
; AVX512F: # %bb.0:
; AVX512F-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm0
; AVX512F-NEXT: vpxor %xmm3, %xmm3, %xmm3
; AVX512F-NEXT: vpcmpeqb %ymm3, %ymm0, %ymm0
; AVX512F-NEXT: vpblendvb %ymm0, %ymm1, %ymm2, %ymm0
; AVX512F-NEXT: retq
;
; AVX512VL-LABEL: blend_mask_cond_v32i8:
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm0
; AVX512VL-NEXT: vpxor %xmm3, %xmm3, %xmm3
; AVX512VL-NEXT: vpcmpeqb %ymm3, %ymm0, %ymm0
; AVX512VL-NEXT: vpternlogq $202, %ymm2, %ymm1, %ymm0
; AVX512VL-NEXT: retq
;
; XOP-LABEL: blend_mask_cond_v32i8:
; XOP: # %bb.0:
; XOP-NEXT: vextractf128 $1, %ymm0, %xmm3
; XOP-NEXT: vpshlb {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm3, %xmm3
; XOP-NEXT: vpxor %xmm4, %xmm4, %xmm4
; XOP-NEXT: vpcomltb %xmm4, %xmm3, %xmm3
; XOP-NEXT: vpshlb {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
; XOP-NEXT: vpcomltb %xmm4, %xmm0, %xmm0
; XOP-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm0
; XOP-NEXT: vpcmov %ymm0, %ymm1, %ymm2, %ymm0
; XOP-NEXT: retq
%a = and <32 x i8> %x, <i8 1, i8 2, i8 4, i8 8, i8 16, i8 32, i8 64, i8 128, i8 4, i8 4, i8 4, i8 4, i8 2, i8 2, i8 2, i8 2, i8 1, i8 2, i8 4, i8 8, i8 16, i8 32, i8 64, i8 128, i8 4, i8 4, i8 4, i8 4, i8 128, i8 4, i8 2, i8 16>
%c = icmp eq <32 x i8> %a, zeroinitializer
%r = select <32 x i1> %c, <32 x i8> %y, <32 x i8> %z
ret <32 x i8> %r
}
define void @store_blend_load_v4i64(ptr %a0, ptr %a1, ptr %a2) {
; AVX1-LABEL: store_blend_load_v4i64:
; AVX1: # %bb.0:
; AVX1-NEXT: vmovapd (%rsi), %ymm0
; AVX1-NEXT: vmovddup {{.*#+}} xmm1 = [9223372036854775808,9223372036854775808]
; AVX1-NEXT: # xmm1 = mem[0,0]
; AVX1-NEXT: vpxor 16(%rdi), %xmm1, %xmm2
; AVX1-NEXT: vmovddup {{.*#+}} xmm3 = [9223372036854775815,9223372036854775815]
; AVX1-NEXT: # xmm3 = mem[0,0]
; AVX1-NEXT: vpcmpgtq %xmm3, %xmm2, %xmm2
; AVX1-NEXT: vpxor (%rdi), %xmm1, %xmm1
; AVX1-NEXT: vpcmpgtq %xmm3, %xmm1, %xmm1
; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm1, %ymm1
; AVX1-NEXT: vblendvpd %ymm1, (%rdi), %ymm0, %ymm0
; AVX1-NEXT: vmovapd %ymm0, (%rdx)
; AVX1-NEXT: vzeroupper
; AVX1-NEXT: retq
;
; AVX2-LABEL: store_blend_load_v4i64:
; AVX2: # %bb.0:
; AVX2-NEXT: vmovdqa (%rdi), %ymm0
; AVX2-NEXT: vmovapd (%rsi), %ymm1
; AVX2-NEXT: vpbroadcastq {{.*#+}} ymm2 = [9223372036854775808,9223372036854775808,9223372036854775808,9223372036854775808]
; AVX2-NEXT: vpxor %ymm2, %ymm0, %ymm2
; AVX2-NEXT: vpbroadcastq {{.*#+}} ymm3 = [9223372036854775815,9223372036854775815,9223372036854775815,9223372036854775815]
; AVX2-NEXT: vpcmpgtq %ymm3, %ymm2, %ymm2
; AVX2-NEXT: vblendvpd %ymm2, %ymm0, %ymm1, %ymm0
; AVX2-NEXT: vmovapd %ymm0, (%rdx)
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
;
; AVX512F-LABEL: store_blend_load_v4i64:
; AVX512F: # %bb.0:
; AVX512F-NEXT: vmovdqa (%rdi), %ymm0
; AVX512F-NEXT: vmovdqa (%rsi), %ymm1
; AVX512F-NEXT: vpcmpnleuq {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to8}, %zmm0, %k1
; AVX512F-NEXT: vmovdqa64 %zmm0, %zmm1 {%k1}
; AVX512F-NEXT: vmovdqa %ymm1, (%rdx)
; AVX512F-NEXT: vzeroupper
; AVX512F-NEXT: retq
;
; AVX512VL-LABEL: store_blend_load_v4i64:
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vmovdqa (%rdi), %ymm0
; AVX512VL-NEXT: vmovdqa (%rsi), %ymm1
; AVX512VL-NEXT: vpcmpnleuq {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to4}, %ymm0, %k1
; AVX512VL-NEXT: vmovdqa64 %ymm0, %ymm1 {%k1}
; AVX512VL-NEXT: vmovdqa %ymm1, (%rdx)
; AVX512VL-NEXT: vzeroupper
; AVX512VL-NEXT: retq
;
; XOP-LABEL: store_blend_load_v4i64:
; XOP: # %bb.0:
; XOP-NEXT: vmovapd (%rsi), %ymm0
; XOP-NEXT: vpmovsxbq {{.*#+}} xmm1 = [7,7]
; XOP-NEXT: vpcomltuq 16(%rdi), %xmm1, %xmm2
; XOP-NEXT: vpcomltuq (%rdi), %xmm1, %xmm1
; XOP-NEXT: vinsertf128 $1, %xmm2, %ymm1, %ymm1
; XOP-NEXT: vblendvpd %ymm1, (%rdi), %ymm0, %ymm0
; XOP-NEXT: vmovapd %ymm0, (%rdx)
; XOP-NEXT: vzeroupper
; XOP-NEXT: retq
%v0 = load <4 x i64>, ptr %a0
%v1 = load <4 x i64>, ptr %a1
%cmp = icmp ugt <4 x i64> %v0, <i64 7, i64 7, i64 7, i64 7>
%res = select <4 x i1> %cmp, <4 x i64> %v0, <4 x i64> %v1
store <4 x i64> %res, ptr %a2
ret void
}
define void @store_blend_load_v8i32(ptr %a0, ptr %a1, ptr %a2) {
; AVX1-LABEL: store_blend_load_v8i32:
; AVX1: # %bb.0:
; AVX1-NEXT: vmovaps (%rsi), %ymm0
; AVX1-NEXT: vmovdqa (%rdi), %xmm1
; AVX1-NEXT: vmovdqa 16(%rdi), %xmm2
; AVX1-NEXT: vbroadcastss {{.*#+}} xmm3 = [8,8,8,8]
; AVX1-NEXT: vpmaxud %xmm3, %xmm2, %xmm4
; AVX1-NEXT: vpcmpeqd %xmm4, %xmm2, %xmm2
; AVX1-NEXT: vpmaxud %xmm3, %xmm1, %xmm3
; AVX1-NEXT: vpcmpeqd %xmm3, %xmm1, %xmm1
; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm1, %ymm1
; AVX1-NEXT: vblendvps %ymm1, (%rdi), %ymm0, %ymm0
; AVX1-NEXT: vmovaps %ymm0, (%rdx)
; AVX1-NEXT: vzeroupper
; AVX1-NEXT: retq
;
; AVX2-LABEL: store_blend_load_v8i32:
; AVX2: # %bb.0:
; AVX2-NEXT: vmovdqa (%rdi), %ymm0
; AVX2-NEXT: vmovaps (%rsi), %ymm1
; AVX2-NEXT: vpbroadcastd {{.*#+}} ymm2 = [8,8,8,8,8,8,8,8]
; AVX2-NEXT: vpmaxud %ymm2, %ymm0, %ymm2
; AVX2-NEXT: vpcmpeqd %ymm2, %ymm0, %ymm2
; AVX2-NEXT: vblendvps %ymm2, %ymm0, %ymm1, %ymm0
; AVX2-NEXT: vmovaps %ymm0, (%rdx)
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
;
; AVX512F-LABEL: store_blend_load_v8i32:
; AVX512F: # %bb.0:
; AVX512F-NEXT: vmovdqa (%rdi), %ymm0
; AVX512F-NEXT: vmovdqa (%rsi), %ymm1
; AVX512F-NEXT: vpcmpnleud {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to16}, %zmm0, %k1
; AVX512F-NEXT: vmovdqa32 %zmm0, %zmm1 {%k1}
; AVX512F-NEXT: vmovdqa %ymm1, (%rdx)
; AVX512F-NEXT: vzeroupper
; AVX512F-NEXT: retq
;
; AVX512VL-LABEL: store_blend_load_v8i32:
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vmovdqa (%rdi), %ymm0
; AVX512VL-NEXT: vmovdqa (%rsi), %ymm1
; AVX512VL-NEXT: vpcmpnleud {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to8}, %ymm0, %k1
; AVX512VL-NEXT: vmovdqa32 %ymm0, %ymm1 {%k1}
; AVX512VL-NEXT: vmovdqa %ymm1, (%rdx)
; AVX512VL-NEXT: vzeroupper
; AVX512VL-NEXT: retq
;
; XOP-LABEL: store_blend_load_v8i32:
; XOP: # %bb.0:
; XOP-NEXT: vmovaps (%rsi), %ymm0
; XOP-NEXT: vbroadcastss {{.*#+}} xmm1 = [7,7,7,7]
; XOP-NEXT: vpcomltud 16(%rdi), %xmm1, %xmm2
; XOP-NEXT: vpcomltud (%rdi), %xmm1, %xmm1
; XOP-NEXT: vinsertf128 $1, %xmm2, %ymm1, %ymm1
; XOP-NEXT: vblendvps %ymm1, (%rdi), %ymm0, %ymm0
; XOP-NEXT: vmovaps %ymm0, (%rdx)
; XOP-NEXT: vzeroupper
; XOP-NEXT: retq
%v0 = load <8 x i32>, ptr %a0
%v1 = load <8 x i32>, ptr %a1
%cmp = icmp ugt <8 x i32> %v0, <i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7>
%res = select <8 x i1> %cmp, <8 x i32> %v0, <8 x i32> %v1
store <8 x i32> %res, ptr %a2
ret void
}
define void @store_blend_load_v16i16(ptr %a0, ptr %a1, ptr %a2) {
; AVX1-LABEL: store_blend_load_v16i16:
; AVX1: # %bb.0:
; AVX1-NEXT: vmovdqa (%rdi), %xmm0
; AVX1-NEXT: vmovdqa 16(%rdi), %xmm1
; AVX1-NEXT: vbroadcastss {{.*#+}} xmm2 = [8,8,8,8,8,8,8,8]
; AVX1-NEXT: vpmaxuw %xmm2, %xmm0, %xmm3
; AVX1-NEXT: vpcmpeqw %xmm3, %xmm0, %xmm3
; AVX1-NEXT: vpmaxuw %xmm2, %xmm1, %xmm2
; AVX1-NEXT: vpcmpeqw %xmm2, %xmm1, %xmm2
; AVX1-NEXT: vmovdqa (%rsi), %xmm4
; AVX1-NEXT: vmovdqa 16(%rsi), %xmm5
; AVX1-NEXT: vpblendvb %xmm2, %xmm1, %xmm5, %xmm1
; AVX1-NEXT: vpblendvb %xmm3, %xmm0, %xmm4, %xmm0
; AVX1-NEXT: vmovdqa %xmm0, (%rdx)
; AVX1-NEXT: vmovdqa %xmm1, 16(%rdx)
; AVX1-NEXT: retq
;
; AVX2-LABEL: store_blend_load_v16i16:
; AVX2: # %bb.0:
; AVX2-NEXT: vmovdqa (%rdi), %ymm0
; AVX2-NEXT: vmovdqa (%rsi), %ymm1
; AVX2-NEXT: vpmaxuw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm2
; AVX2-NEXT: vpcmpeqw %ymm2, %ymm0, %ymm2
; AVX2-NEXT: vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
; AVX2-NEXT: vmovdqa %ymm0, (%rdx)
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
;
; AVX512F-LABEL: store_blend_load_v16i16:
; AVX512F: # %bb.0:
; AVX512F-NEXT: vmovdqa (%rdi), %ymm0
; AVX512F-NEXT: vmovdqa (%rsi), %ymm1
; AVX512F-NEXT: vpmaxuw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm2
; AVX512F-NEXT: vpcmpeqw %ymm2, %ymm0, %ymm2
; AVX512F-NEXT: vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
; AVX512F-NEXT: vmovdqa %ymm0, (%rdx)
; AVX512F-NEXT: vzeroupper
; AVX512F-NEXT: retq
;
; AVX512VL-LABEL: store_blend_load_v16i16:
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vmovdqa (%rdi), %ymm0
; AVX512VL-NEXT: vpmaxuw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm1
; AVX512VL-NEXT: vpcmpeqw %ymm1, %ymm0, %ymm1
; AVX512VL-NEXT: vpternlogq $202, (%rsi), %ymm0, %ymm1
; AVX512VL-NEXT: vmovdqa %ymm1, (%rdx)
; AVX512VL-NEXT: vzeroupper
; AVX512VL-NEXT: retq
;
; XOP-LABEL: store_blend_load_v16i16:
; XOP: # %bb.0:
; XOP-NEXT: vmovdqa (%rdi), %ymm0
; XOP-NEXT: vbroadcastss {{.*#+}} xmm1 = [7,7,7,7,7,7,7,7]
; XOP-NEXT: vpcomltuw 16(%rdi), %xmm1, %xmm2
; XOP-NEXT: vpcomltuw (%rdi), %xmm1, %xmm1
; XOP-NEXT: vinsertf128 $1, %xmm2, %ymm1, %ymm1
; XOP-NEXT: vpcmov %ymm1, (%rsi), %ymm0, %ymm0
; XOP-NEXT: vmovdqa %ymm0, (%rdx)
; XOP-NEXT: vzeroupper
; XOP-NEXT: retq
%v0 = load <16 x i16>, ptr %a0
%v1 = load <16 x i16>, ptr %a1
%cmp = icmp ugt <16 x i16> %v0, <i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7>
%res = select <16 x i1> %cmp, <16 x i16> %v0, <16 x i16> %v1
store <16 x i16> %res, ptr %a2
ret void
}
define void @store_blend_load_v32i8(ptr %a0, ptr %a1, ptr %a2) {
; AVX1-LABEL: store_blend_load_v32i8:
; AVX1: # %bb.0:
; AVX1-NEXT: vmovdqa (%rdi), %xmm0
; AVX1-NEXT: vmovdqa 16(%rdi), %xmm1
; AVX1-NEXT: vbroadcastss {{.*#+}} xmm2 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8]
; AVX1-NEXT: vpmaxub %xmm2, %xmm0, %xmm3
; AVX1-NEXT: vpcmpeqb %xmm3, %xmm0, %xmm3
; AVX1-NEXT: vpmaxub %xmm2, %xmm1, %xmm2
; AVX1-NEXT: vpcmpeqb %xmm2, %xmm1, %xmm2
; AVX1-NEXT: vmovdqa (%rsi), %xmm4
; AVX1-NEXT: vmovdqa 16(%rsi), %xmm5
; AVX1-NEXT: vpblendvb %xmm2, %xmm1, %xmm5, %xmm1
; AVX1-NEXT: vpblendvb %xmm3, %xmm0, %xmm4, %xmm0
; AVX1-NEXT: vmovdqa %xmm0, (%rdx)
; AVX1-NEXT: vmovdqa %xmm1, 16(%rdx)
; AVX1-NEXT: retq
;
; AVX2-LABEL: store_blend_load_v32i8:
; AVX2: # %bb.0:
; AVX2-NEXT: vmovdqa (%rdi), %ymm0
; AVX2-NEXT: vmovdqa (%rsi), %ymm1
; AVX2-NEXT: vpmaxub {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm2
; AVX2-NEXT: vpcmpeqb %ymm2, %ymm0, %ymm2
; AVX2-NEXT: vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
; AVX2-NEXT: vmovdqa %ymm0, (%rdx)
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
;
; AVX512F-LABEL: store_blend_load_v32i8:
; AVX512F: # %bb.0:
; AVX512F-NEXT: vmovdqa (%rdi), %ymm0
; AVX512F-NEXT: vmovdqa (%rsi), %ymm1
; AVX512F-NEXT: vpmaxub {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm2
; AVX512F-NEXT: vpcmpeqb %ymm2, %ymm0, %ymm2
; AVX512F-NEXT: vpblendvb %ymm2, %ymm0, %ymm1, %ymm0
; AVX512F-NEXT: vmovdqa %ymm0, (%rdx)
; AVX512F-NEXT: vzeroupper
; AVX512F-NEXT: retq
;
; AVX512VL-LABEL: store_blend_load_v32i8:
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vmovdqa (%rdi), %ymm0
; AVX512VL-NEXT: vpmaxub {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm1
; AVX512VL-NEXT: vpcmpeqb %ymm1, %ymm0, %ymm1
; AVX512VL-NEXT: vpternlogq $202, (%rsi), %ymm0, %ymm1
; AVX512VL-NEXT: vmovdqa %ymm1, (%rdx)
; AVX512VL-NEXT: vzeroupper
; AVX512VL-NEXT: retq
;
; XOP-LABEL: store_blend_load_v32i8:
; XOP: # %bb.0:
; XOP-NEXT: vmovdqa (%rdi), %ymm0
; XOP-NEXT: vbroadcastss {{.*#+}} xmm1 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
; XOP-NEXT: vpcomltub 16(%rdi), %xmm1, %xmm2
; XOP-NEXT: vpcomltub (%rdi), %xmm1, %xmm1
; XOP-NEXT: vinsertf128 $1, %xmm2, %ymm1, %ymm1
; XOP-NEXT: vpcmov %ymm1, (%rsi), %ymm0, %ymm0
; XOP-NEXT: vmovdqa %ymm0, (%rdx)
; XOP-NEXT: vzeroupper
; XOP-NEXT: retq
%v0 = load <32 x i8>, ptr %a0
%v1 = load <32 x i8>, ptr %a1
%cmp = icmp ugt <32 x i8> %v0, <i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7>
%res = select <32 x i1> %cmp, <32 x i8> %v0, <32 x i8> %v1
store <32 x i8> %res, ptr %a2
ret void
}
define void @PR46531(ptr %x, ptr %y, ptr %z) {
; AVX12-LABEL: PR46531:
; AVX12: # %bb.0:
; AVX12-NEXT: vmovdqu (%rsi), %xmm0
; AVX12-NEXT: vmovdqu (%rdx), %xmm1
; AVX12-NEXT: vpor %xmm0, %xmm1, %xmm2
; AVX12-NEXT: vpxor %xmm0, %xmm1, %xmm0
; AVX12-NEXT: vpslld $31, %xmm1, %xmm1
; AVX12-NEXT: vblendvps %xmm1, %xmm0, %xmm2, %xmm0
; AVX12-NEXT: vmovups %xmm0, (%rdi)
; AVX12-NEXT: retq
;
; AVX512F-LABEL: PR46531:
; AVX512F: # %bb.0:
; AVX512F-NEXT: vmovdqu (%rsi), %xmm0
; AVX512F-NEXT: vmovdqu (%rdx), %xmm1
; AVX512F-NEXT: vpor %xmm0, %xmm1, %xmm2
; AVX512F-NEXT: vptestnmd {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to16}, %zmm1, %k1
; AVX512F-NEXT: vpxor %xmm0, %xmm1, %xmm0
; AVX512F-NEXT: vmovdqa32 %zmm2, %zmm0 {%k1}
; AVX512F-NEXT: vmovdqu %xmm0, (%rdi)
; AVX512F-NEXT: vzeroupper
; AVX512F-NEXT: retq
;
; AVX512VL-LABEL: PR46531:
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vmovdqu (%rsi), %xmm0
; AVX512VL-NEXT: vmovdqu (%rdx), %xmm1
; AVX512VL-NEXT: vptestnmd {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to4}, %xmm1, %k1
; AVX512VL-NEXT: vpxor %xmm0, %xmm1, %xmm2
; AVX512VL-NEXT: vpord %xmm0, %xmm1, %xmm2 {%k1}
; AVX512VL-NEXT: vmovdqu %xmm2, (%rdi)
; AVX512VL-NEXT: retq
;
; XOP-LABEL: PR46531:
; XOP: # %bb.0:
; XOP-NEXT: vmovdqu (%rsi), %xmm0
; XOP-NEXT: vmovdqu (%rdx), %xmm1
; XOP-NEXT: vpor %xmm0, %xmm1, %xmm2
; XOP-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm3
; XOP-NEXT: vpxor %xmm4, %xmm4, %xmm4
; XOP-NEXT: vpcomneqd %xmm4, %xmm3, %xmm3
; XOP-NEXT: vpxor %xmm0, %xmm1, %xmm0
; XOP-NEXT: vblendvps %xmm3, %xmm0, %xmm2, %xmm0
; XOP-NEXT: vmovups %xmm0, (%rdi)
; XOP-NEXT: retq
%a = load <4 x i32>, ptr %y, align 4
%b = load <4 x i32>, ptr %z, align 4
%or = or <4 x i32> %b, %a
%and = and <4 x i32> %b, <i32 1, i32 1, i32 1, i32 1>
%cmp = icmp eq <4 x i32> %and, zeroinitializer
%xor = xor <4 x i32> %b, %a
%sel = select <4 x i1> %cmp, <4 x i32> %or, <4 x i32> %xor
store <4 x i32> %sel, ptr %x, align 4
ret void
}
attributes #0 = { "no-nans-fp-math"="true" }
;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
; AVX: {{.*}}
Codebase Browser
llvm