; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=+sse2 | FileCheck %s --check-prefixes=SSE,SSE2,X86-SSE2
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+sse2 | FileCheck %s --check-prefixes=SSE,SSE2,X64-SSE,X64-SSE2
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+sse4.1 | FileCheck %s --check-prefixes=SSE,X64-SSE,SSE41
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx | FileCheck %s --check-prefixes=AVX,AVX1OR2,AVX1
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2 | FileCheck %s --check-prefixes=AVX,AVX1OR2,AVX2
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512f | FileCheck %s --check-prefixes=AVX,AVX512,AVX512F
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512f,+avx512bw | FileCheck %s --check-prefixes=AVX,AVX512,AVX512BW
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512f,+avx512bw,+avx512vl | FileCheck %s --check-prefixes=AVX,AVX512,AVX512VL
;
; Truncate
;
define i1 @trunc_v2i64_v2i1(<2 x i64>) nounwind {
; SSE2-LABEL: trunc_v2i64_v2i1:
; SSE2: # %bb.0:
; SSE2-NEXT: psllq $63, %xmm0
; SSE2-NEXT: movmskpd %xmm0, %eax
; SSE2-NEXT: testl %eax, %eax
; SSE2-NEXT: setne %al
; SSE2-NEXT: ret{{[l|q]}}
;
; SSE41-LABEL: trunc_v2i64_v2i1:
; SSE41: # %bb.0:
; SSE41-NEXT: ptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; SSE41-NEXT: setne %al
; SSE41-NEXT: retq
;
; AVX1OR2-LABEL: trunc_v2i64_v2i1:
; AVX1OR2: # %bb.0:
; AVX1OR2-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; AVX1OR2-NEXT: setne %al
; AVX1OR2-NEXT: retq
;
; AVX512F-LABEL: trunc_v2i64_v2i1:
; AVX512F: # %bb.0:
; AVX512F-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; AVX512F-NEXT: setne %al
; AVX512F-NEXT: retq
;
; AVX512BW-LABEL: trunc_v2i64_v2i1:
; AVX512BW: # %bb.0:
; AVX512BW-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; AVX512BW-NEXT: setne %al
; AVX512BW-NEXT: retq
;
; AVX512VL-LABEL: trunc_v2i64_v2i1:
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vpbroadcastq {{.*#+}} xmm1 = [1,1]
; AVX512VL-NEXT: vptest %xmm1, %xmm0
; AVX512VL-NEXT: setne %al
; AVX512VL-NEXT: retq
%a = trunc <2 x i64> %0 to <2 x i1>
%b = call i1 @llvm.vector.reduce.or.v2i1(<2 x i1> %a)
ret i1 %b
}
define i1 @trunc_v4i32_v4i1(<4 x i32>) nounwind {
; SSE2-LABEL: trunc_v4i32_v4i1:
; SSE2: # %bb.0:
; SSE2-NEXT: pslld $31, %xmm0
; SSE2-NEXT: movmskps %xmm0, %eax
; SSE2-NEXT: testl %eax, %eax
; SSE2-NEXT: setne %al
; SSE2-NEXT: ret{{[l|q]}}
;
; SSE41-LABEL: trunc_v4i32_v4i1:
; SSE41: # %bb.0:
; SSE41-NEXT: ptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; SSE41-NEXT: setne %al
; SSE41-NEXT: retq
;
; AVX1OR2-LABEL: trunc_v4i32_v4i1:
; AVX1OR2: # %bb.0:
; AVX1OR2-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; AVX1OR2-NEXT: setne %al
; AVX1OR2-NEXT: retq
;
; AVX512F-LABEL: trunc_v4i32_v4i1:
; AVX512F: # %bb.0:
; AVX512F-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; AVX512F-NEXT: setne %al
; AVX512F-NEXT: retq
;
; AVX512BW-LABEL: trunc_v4i32_v4i1:
; AVX512BW: # %bb.0:
; AVX512BW-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; AVX512BW-NEXT: setne %al
; AVX512BW-NEXT: retq
;
; AVX512VL-LABEL: trunc_v4i32_v4i1:
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vpbroadcastq {{.*#+}} xmm1 = [4294967297,4294967297]
; AVX512VL-NEXT: vptest %xmm1, %xmm0
; AVX512VL-NEXT: setne %al
; AVX512VL-NEXT: retq
%a = trunc <4 x i32> %0 to <4 x i1>
%b = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> %a)
ret i1 %b
}
define i1 @trunc_v8i16_v8i1(<8 x i16>) nounwind {
; SSE2-LABEL: trunc_v8i16_v8i1:
; SSE2: # %bb.0:
; SSE2-NEXT: psllw $7, %xmm0
; SSE2-NEXT: pmovmskb %xmm0, %eax
; SSE2-NEXT: testl $21845, %eax # imm = 0x5555
; SSE2-NEXT: setne %al
; SSE2-NEXT: ret{{[l|q]}}
;
; SSE41-LABEL: trunc_v8i16_v8i1:
; SSE41: # %bb.0:
; SSE41-NEXT: ptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; SSE41-NEXT: setne %al
; SSE41-NEXT: retq
;
; AVX1OR2-LABEL: trunc_v8i16_v8i1:
; AVX1OR2: # %bb.0:
; AVX1OR2-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; AVX1OR2-NEXT: setne %al
; AVX1OR2-NEXT: retq
;
; AVX512F-LABEL: trunc_v8i16_v8i1:
; AVX512F: # %bb.0:
; AVX512F-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; AVX512F-NEXT: setne %al
; AVX512F-NEXT: retq
;
; AVX512BW-LABEL: trunc_v8i16_v8i1:
; AVX512BW: # %bb.0:
; AVX512BW-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; AVX512BW-NEXT: setne %al
; AVX512BW-NEXT: retq
;
; AVX512VL-LABEL: trunc_v8i16_v8i1:
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vpbroadcastq {{.*#+}} xmm1 = [281479271743489,281479271743489]
; AVX512VL-NEXT: vptest %xmm1, %xmm0
; AVX512VL-NEXT: setne %al
; AVX512VL-NEXT: retq
%a = trunc <8 x i16> %0 to <8 x i1>
%b = call i1 @llvm.vector.reduce.or.v8i1(<8 x i1> %a)
ret i1 %b
}
define i1 @trunc_v16i8_v16i1(<16 x i8>) nounwind {
; SSE2-LABEL: trunc_v16i8_v16i1:
; SSE2: # %bb.0:
; SSE2-NEXT: psllw $7, %xmm0
; SSE2-NEXT: pmovmskb %xmm0, %eax
; SSE2-NEXT: testl %eax, %eax
; SSE2-NEXT: setne %al
; SSE2-NEXT: ret{{[l|q]}}
;
; SSE41-LABEL: trunc_v16i8_v16i1:
; SSE41: # %bb.0:
; SSE41-NEXT: ptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; SSE41-NEXT: setne %al
; SSE41-NEXT: retq
;
; AVX1OR2-LABEL: trunc_v16i8_v16i1:
; AVX1OR2: # %bb.0:
; AVX1OR2-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; AVX1OR2-NEXT: setne %al
; AVX1OR2-NEXT: retq
;
; AVX512F-LABEL: trunc_v16i8_v16i1:
; AVX512F: # %bb.0:
; AVX512F-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; AVX512F-NEXT: setne %al
; AVX512F-NEXT: retq
;
; AVX512BW-LABEL: trunc_v16i8_v16i1:
; AVX512BW: # %bb.0:
; AVX512BW-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; AVX512BW-NEXT: setne %al
; AVX512BW-NEXT: retq
;
; AVX512VL-LABEL: trunc_v16i8_v16i1:
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vpbroadcastq {{.*#+}} xmm1 = [72340172838076673,72340172838076673]
; AVX512VL-NEXT: vptest %xmm1, %xmm0
; AVX512VL-NEXT: setne %al
; AVX512VL-NEXT: retq
%a = trunc <16 x i8> %0 to <16 x i1>
%b = call i1 @llvm.vector.reduce.or.v16i1(<16 x i1> %a)
ret i1 %b
}
define i1 @trunc_v4i64_v4i1(<4 x i64>) nounwind {
; SSE2-LABEL: trunc_v4i64_v4i1:
; SSE2: # %bb.0:
; SSE2-NEXT: shufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
; SSE2-NEXT: pslld $31, %xmm0
; SSE2-NEXT: movmskps %xmm0, %eax
; SSE2-NEXT: testl %eax, %eax
; SSE2-NEXT: setne %al
; SSE2-NEXT: ret{{[l|q]}}
;
; SSE41-LABEL: trunc_v4i64_v4i1:
; SSE41: # %bb.0:
; SSE41-NEXT: por %xmm1, %xmm0
; SSE41-NEXT: ptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; SSE41-NEXT: setne %al
; SSE41-NEXT: retq
;
; AVX1-LABEL: trunc_v4i64_v4i1:
; AVX1: # %bb.0:
; AVX1-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0
; AVX1-NEXT: setne %al
; AVX1-NEXT: vzeroupper
; AVX1-NEXT: retq
;
; AVX2-LABEL: trunc_v4i64_v4i1:
; AVX2: # %bb.0:
; AVX2-NEXT: vpbroadcastq {{.*#+}} ymm1 = [1,1,1,1]
; AVX2-NEXT: vptest %ymm1, %ymm0
; AVX2-NEXT: setne %al
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
;
; AVX512-LABEL: trunc_v4i64_v4i1:
; AVX512: # %bb.0:
; AVX512-NEXT: vpbroadcastq {{.*#+}} ymm1 = [1,1,1,1]
; AVX512-NEXT: vptest %ymm1, %ymm0
; AVX512-NEXT: setne %al
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
%a = trunc <4 x i64> %0 to <4 x i1>
%b = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> %a)
ret i1 %b
}
define i1 @trunc_v8i32_v8i1(<8 x i32>) nounwind {
; SSE2-LABEL: trunc_v8i32_v8i1:
; SSE2: # %bb.0:
; SSE2-NEXT: por %xmm1, %xmm0
; SSE2-NEXT: pslld $31, %xmm0
; SSE2-NEXT: movmskps %xmm0, %eax
; SSE2-NEXT: testl %eax, %eax
; SSE2-NEXT: setne %al
; SSE2-NEXT: ret{{[l|q]}}
;
; SSE41-LABEL: trunc_v8i32_v8i1:
; SSE41: # %bb.0:
; SSE41-NEXT: por %xmm1, %xmm0
; SSE41-NEXT: ptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; SSE41-NEXT: setne %al
; SSE41-NEXT: retq
;
; AVX1-LABEL: trunc_v8i32_v8i1:
; AVX1: # %bb.0:
; AVX1-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0
; AVX1-NEXT: setne %al
; AVX1-NEXT: vzeroupper
; AVX1-NEXT: retq
;
; AVX2-LABEL: trunc_v8i32_v8i1:
; AVX2: # %bb.0:
; AVX2-NEXT: vpbroadcastq {{.*#+}} ymm1 = [4294967297,4294967297,4294967297,4294967297]
; AVX2-NEXT: vptest %ymm1, %ymm0
; AVX2-NEXT: setne %al
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
;
; AVX512-LABEL: trunc_v8i32_v8i1:
; AVX512: # %bb.0:
; AVX512-NEXT: vpbroadcastq {{.*#+}} ymm1 = [4294967297,4294967297,4294967297,4294967297]
; AVX512-NEXT: vptest %ymm1, %ymm0
; AVX512-NEXT: setne %al
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
%a = trunc <8 x i32> %0 to <8 x i1>
%b = call i1 @llvm.vector.reduce.or.v8i1(<8 x i1> %a)
ret i1 %b
}
define i1 @trunc_v16i16_v16i1(<16 x i16>) nounwind {
; SSE2-LABEL: trunc_v16i16_v16i1:
; SSE2: # %bb.0:
; SSE2-NEXT: por %xmm1, %xmm0
; SSE2-NEXT: psllw $7, %xmm0
; SSE2-NEXT: pmovmskb %xmm0, %eax
; SSE2-NEXT: testl $21845, %eax # imm = 0x5555
; SSE2-NEXT: setne %al
; SSE2-NEXT: ret{{[l|q]}}
;
; SSE41-LABEL: trunc_v16i16_v16i1:
; SSE41: # %bb.0:
; SSE41-NEXT: por %xmm1, %xmm0
; SSE41-NEXT: ptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; SSE41-NEXT: setne %al
; SSE41-NEXT: retq
;
; AVX1-LABEL: trunc_v16i16_v16i1:
; AVX1: # %bb.0:
; AVX1-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0
; AVX1-NEXT: setne %al
; AVX1-NEXT: vzeroupper
; AVX1-NEXT: retq
;
; AVX2-LABEL: trunc_v16i16_v16i1:
; AVX2: # %bb.0:
; AVX2-NEXT: vpbroadcastq {{.*#+}} ymm1 = [281479271743489,281479271743489,281479271743489,281479271743489]
; AVX2-NEXT: vptest %ymm1, %ymm0
; AVX2-NEXT: setne %al
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
;
; AVX512-LABEL: trunc_v16i16_v16i1:
; AVX512: # %bb.0:
; AVX512-NEXT: vpbroadcastq {{.*#+}} ymm1 = [281479271743489,281479271743489,281479271743489,281479271743489]
; AVX512-NEXT: vptest %ymm1, %ymm0
; AVX512-NEXT: setne %al
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
%a = trunc <16 x i16> %0 to <16 x i1>
%b = call i1 @llvm.vector.reduce.or.v16i1(<16 x i1> %a)
ret i1 %b
}
define i1 @trunc_v32i8_v32i1(<32 x i8>) nounwind {
; SSE2-LABEL: trunc_v32i8_v32i1:
; SSE2: # %bb.0:
; SSE2-NEXT: por %xmm1, %xmm0
; SSE2-NEXT: psllw $7, %xmm0
; SSE2-NEXT: pmovmskb %xmm0, %eax
; SSE2-NEXT: testl %eax, %eax
; SSE2-NEXT: setne %al
; SSE2-NEXT: ret{{[l|q]}}
;
; SSE41-LABEL: trunc_v32i8_v32i1:
; SSE41: # %bb.0:
; SSE41-NEXT: por %xmm1, %xmm0
; SSE41-NEXT: ptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; SSE41-NEXT: setne %al
; SSE41-NEXT: retq
;
; AVX1-LABEL: trunc_v32i8_v32i1:
; AVX1: # %bb.0:
; AVX1-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0
; AVX1-NEXT: setne %al
; AVX1-NEXT: vzeroupper
; AVX1-NEXT: retq
;
; AVX2-LABEL: trunc_v32i8_v32i1:
; AVX2: # %bb.0:
; AVX2-NEXT: vpbroadcastq {{.*#+}} ymm1 = [72340172838076673,72340172838076673,72340172838076673,72340172838076673]
; AVX2-NEXT: vptest %ymm1, %ymm0
; AVX2-NEXT: setne %al
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
;
; AVX512-LABEL: trunc_v32i8_v32i1:
; AVX512: # %bb.0:
; AVX512-NEXT: vpbroadcastq {{.*#+}} ymm1 = [72340172838076673,72340172838076673,72340172838076673,72340172838076673]
; AVX512-NEXT: vptest %ymm1, %ymm0
; AVX512-NEXT: setne %al
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
%a = trunc <32 x i8> %0 to <32 x i1>
%b = call i1 @llvm.vector.reduce.or.v32i1(<32 x i1> %a)
ret i1 %b
}
define i1 @trunc_v8i64_v8i1(<8 x i64>) nounwind {
; X86-SSE2-LABEL: trunc_v8i64_v8i1:
; X86-SSE2: # %bb.0:
; X86-SSE2-NEXT: pushl %ebp
; X86-SSE2-NEXT: movl %esp, %ebp
; X86-SSE2-NEXT: andl $-16, %esp
; X86-SSE2-NEXT: subl $16, %esp
; X86-SSE2-NEXT: shufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
; X86-SSE2-NEXT: pslld $16, %xmm0
; X86-SSE2-NEXT: psrad $16, %xmm0
; X86-SSE2-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,2],mem[0,2]
; X86-SSE2-NEXT: pslld $16, %xmm2
; X86-SSE2-NEXT: psrad $16, %xmm2
; X86-SSE2-NEXT: packssdw %xmm2, %xmm0
; X86-SSE2-NEXT: psllw $15, %xmm0
; X86-SSE2-NEXT: pmovmskb %xmm0, %eax
; X86-SSE2-NEXT: testl $43690, %eax # imm = 0xAAAA
; X86-SSE2-NEXT: setne %al
; X86-SSE2-NEXT: movl %ebp, %esp
; X86-SSE2-NEXT: popl %ebp
; X86-SSE2-NEXT: retl
;
; X64-SSE2-LABEL: trunc_v8i64_v8i1:
; X64-SSE2: # %bb.0:
; X64-SSE2-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,2],xmm3[0,2]
; X64-SSE2-NEXT: pslld $16, %xmm2
; X64-SSE2-NEXT: psrad $16, %xmm2
; X64-SSE2-NEXT: shufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
; X64-SSE2-NEXT: pslld $16, %xmm0
; X64-SSE2-NEXT: psrad $16, %xmm0
; X64-SSE2-NEXT: packssdw %xmm2, %xmm0
; X64-SSE2-NEXT: psllw $15, %xmm0
; X64-SSE2-NEXT: pmovmskb %xmm0, %eax
; X64-SSE2-NEXT: testl $43690, %eax # imm = 0xAAAA
; X64-SSE2-NEXT: setne %al
; X64-SSE2-NEXT: retq
;
; SSE41-LABEL: trunc_v8i64_v8i1:
; SSE41: # %bb.0:
; SSE41-NEXT: por %xmm3, %xmm1
; SSE41-NEXT: por %xmm2, %xmm0
; SSE41-NEXT: por %xmm1, %xmm0
; SSE41-NEXT: ptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; SSE41-NEXT: setne %al
; SSE41-NEXT: retq
;
; AVX1-LABEL: trunc_v8i64_v8i1:
; AVX1: # %bb.0:
; AVX1-NEXT: vorps %ymm1, %ymm0, %ymm0
; AVX1-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0
; AVX1-NEXT: setne %al
; AVX1-NEXT: vzeroupper
; AVX1-NEXT: retq
;
; AVX2-LABEL: trunc_v8i64_v8i1:
; AVX2: # %bb.0:
; AVX2-NEXT: vpor %ymm1, %ymm0, %ymm0
; AVX2-NEXT: vpbroadcastq {{.*#+}} ymm1 = [1,1,1,1]
; AVX2-NEXT: vptest %ymm1, %ymm0
; AVX2-NEXT: setne %al
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
;
; AVX512-LABEL: trunc_v8i64_v8i1:
; AVX512: # %bb.0:
; AVX512-NEXT: vpbroadcastq {{.*#+}} zmm1 = [1,1,1,1,1,1,1,1]
; AVX512-NEXT: vptestmd %zmm1, %zmm0, %k0
; AVX512-NEXT: kortestw %k0, %k0
; AVX512-NEXT: setne %al
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
%a = trunc <8 x i64> %0 to <8 x i1>
%b = call i1 @llvm.vector.reduce.or.v8i1(<8 x i1> %a)
ret i1 %b
}
define i1 @trunc_v16i32_v16i1(<16 x i32>) nounwind {
; X86-SSE2-LABEL: trunc_v16i32_v16i1:
; X86-SSE2: # %bb.0:
; X86-SSE2-NEXT: pushl %ebp
; X86-SSE2-NEXT: movl %esp, %ebp
; X86-SSE2-NEXT: andl $-16, %esp
; X86-SSE2-NEXT: subl $16, %esp
; X86-SSE2-NEXT: por %xmm2, %xmm0
; X86-SSE2-NEXT: por 8(%ebp), %xmm1
; X86-SSE2-NEXT: por %xmm0, %xmm1
; X86-SSE2-NEXT: pslld $31, %xmm1
; X86-SSE2-NEXT: movmskps %xmm1, %eax
; X86-SSE2-NEXT: testl %eax, %eax
; X86-SSE2-NEXT: setne %al
; X86-SSE2-NEXT: movl %ebp, %esp
; X86-SSE2-NEXT: popl %ebp
; X86-SSE2-NEXT: retl
;
; X64-SSE2-LABEL: trunc_v16i32_v16i1:
; X64-SSE2: # %bb.0:
; X64-SSE2-NEXT: por %xmm3, %xmm1
; X64-SSE2-NEXT: por %xmm2, %xmm0
; X64-SSE2-NEXT: por %xmm1, %xmm0
; X64-SSE2-NEXT: pslld $31, %xmm0
; X64-SSE2-NEXT: movmskps %xmm0, %eax
; X64-SSE2-NEXT: testl %eax, %eax
; X64-SSE2-NEXT: setne %al
; X64-SSE2-NEXT: retq
;
; SSE41-LABEL: trunc_v16i32_v16i1:
; SSE41: # %bb.0:
; SSE41-NEXT: por %xmm3, %xmm1
; SSE41-NEXT: por %xmm2, %xmm0
; SSE41-NEXT: por %xmm1, %xmm0
; SSE41-NEXT: ptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; SSE41-NEXT: setne %al
; SSE41-NEXT: retq
;
; AVX1-LABEL: trunc_v16i32_v16i1:
; AVX1: # %bb.0:
; AVX1-NEXT: vorps %ymm1, %ymm0, %ymm0
; AVX1-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0
; AVX1-NEXT: setne %al
; AVX1-NEXT: vzeroupper
; AVX1-NEXT: retq
;
; AVX2-LABEL: trunc_v16i32_v16i1:
; AVX2: # %bb.0:
; AVX2-NEXT: vpor %ymm1, %ymm0, %ymm0
; AVX2-NEXT: vpbroadcastq {{.*#+}} ymm1 = [4294967297,4294967297,4294967297,4294967297]
; AVX2-NEXT: vptest %ymm1, %ymm0
; AVX2-NEXT: setne %al
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
;
; AVX512-LABEL: trunc_v16i32_v16i1:
; AVX512: # %bb.0:
; AVX512-NEXT: vptestmd {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to16}, %zmm0, %k0
; AVX512-NEXT: kortestw %k0, %k0
; AVX512-NEXT: setne %al
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
%a = trunc <16 x i32> %0 to <16 x i1>
%b = call i1 @llvm.vector.reduce.or.v16i1(<16 x i1> %a)
ret i1 %b
}
define i1 @trunc_v32i16_v32i1(<32 x i16>) nounwind {
; X86-SSE2-LABEL: trunc_v32i16_v32i1:
; X86-SSE2: # %bb.0:
; X86-SSE2-NEXT: pushl %ebp
; X86-SSE2-NEXT: movl %esp, %ebp
; X86-SSE2-NEXT: andl $-16, %esp
; X86-SSE2-NEXT: subl $16, %esp
; X86-SSE2-NEXT: por %xmm2, %xmm0
; X86-SSE2-NEXT: por 8(%ebp), %xmm1
; X86-SSE2-NEXT: por %xmm0, %xmm1
; X86-SSE2-NEXT: psllw $7, %xmm1
; X86-SSE2-NEXT: pmovmskb %xmm1, %eax
; X86-SSE2-NEXT: testl $21845, %eax # imm = 0x5555
; X86-SSE2-NEXT: setne %al
; X86-SSE2-NEXT: movl %ebp, %esp
; X86-SSE2-NEXT: popl %ebp
; X86-SSE2-NEXT: retl
;
; X64-SSE2-LABEL: trunc_v32i16_v32i1:
; X64-SSE2: # %bb.0:
; X64-SSE2-NEXT: por %xmm3, %xmm1
; X64-SSE2-NEXT: por %xmm2, %xmm0
; X64-SSE2-NEXT: por %xmm1, %xmm0
; X64-SSE2-NEXT: psllw $7, %xmm0
; X64-SSE2-NEXT: pmovmskb %xmm0, %eax
; X64-SSE2-NEXT: testl $21845, %eax # imm = 0x5555
; X64-SSE2-NEXT: setne %al
; X64-SSE2-NEXT: retq
;
; SSE41-LABEL: trunc_v32i16_v32i1:
; SSE41: # %bb.0:
; SSE41-NEXT: por %xmm3, %xmm1
; SSE41-NEXT: por %xmm2, %xmm0
; SSE41-NEXT: por %xmm1, %xmm0
; SSE41-NEXT: ptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; SSE41-NEXT: setne %al
; SSE41-NEXT: retq
;
; AVX1-LABEL: trunc_v32i16_v32i1:
; AVX1: # %bb.0:
; AVX1-NEXT: vorps %ymm1, %ymm0, %ymm0
; AVX1-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0
; AVX1-NEXT: setne %al
; AVX1-NEXT: vzeroupper
; AVX1-NEXT: retq
;
; AVX2-LABEL: trunc_v32i16_v32i1:
; AVX2: # %bb.0:
; AVX2-NEXT: vpor %ymm1, %ymm0, %ymm0
; AVX2-NEXT: vpbroadcastq {{.*#+}} ymm1 = [281479271743489,281479271743489,281479271743489,281479271743489]
; AVX2-NEXT: vptest %ymm1, %ymm0
; AVX2-NEXT: setne %al
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
;
; AVX512-LABEL: trunc_v32i16_v32i1:
; AVX512: # %bb.0:
; AVX512-NEXT: vptestmd {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to16}, %zmm0, %k0
; AVX512-NEXT: kortestw %k0, %k0
; AVX512-NEXT: setne %al
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
%a = trunc <32 x i16> %0 to <32 x i1>
%b = call i1 @llvm.vector.reduce.or.v32i1(<32 x i1> %a)
ret i1 %b
}
define i1 @trunc_v64i8_v64i1(<64 x i8>) nounwind {
; X86-SSE2-LABEL: trunc_v64i8_v64i1:
; X86-SSE2: # %bb.0:
; X86-SSE2-NEXT: pushl %ebp
; X86-SSE2-NEXT: movl %esp, %ebp
; X86-SSE2-NEXT: andl $-16, %esp
; X86-SSE2-NEXT: subl $16, %esp
; X86-SSE2-NEXT: por %xmm2, %xmm0
; X86-SSE2-NEXT: por 8(%ebp), %xmm1
; X86-SSE2-NEXT: por %xmm0, %xmm1
; X86-SSE2-NEXT: psllw $7, %xmm1
; X86-SSE2-NEXT: pmovmskb %xmm1, %eax
; X86-SSE2-NEXT: testl %eax, %eax
; X86-SSE2-NEXT: setne %al
; X86-SSE2-NEXT: movl %ebp, %esp
; X86-SSE2-NEXT: popl %ebp
; X86-SSE2-NEXT: retl
;
; X64-SSE2-LABEL: trunc_v64i8_v64i1:
; X64-SSE2: # %bb.0:
; X64-SSE2-NEXT: por %xmm3, %xmm1
; X64-SSE2-NEXT: por %xmm2, %xmm0
; X64-SSE2-NEXT: por %xmm1, %xmm0
; X64-SSE2-NEXT: psllw $7, %xmm0
; X64-SSE2-NEXT: pmovmskb %xmm0, %eax
; X64-SSE2-NEXT: testl %eax, %eax
; X64-SSE2-NEXT: setne %al
; X64-SSE2-NEXT: retq
;
; SSE41-LABEL: trunc_v64i8_v64i1:
; SSE41: # %bb.0:
; SSE41-NEXT: por %xmm3, %xmm1
; SSE41-NEXT: por %xmm2, %xmm0
; SSE41-NEXT: por %xmm1, %xmm0
; SSE41-NEXT: ptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; SSE41-NEXT: setne %al
; SSE41-NEXT: retq
;
; AVX1-LABEL: trunc_v64i8_v64i1:
; AVX1: # %bb.0:
; AVX1-NEXT: vorps %ymm1, %ymm0, %ymm0
; AVX1-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0
; AVX1-NEXT: setne %al
; AVX1-NEXT: vzeroupper
; AVX1-NEXT: retq
;
; AVX2-LABEL: trunc_v64i8_v64i1:
; AVX2: # %bb.0:
; AVX2-NEXT: vpor %ymm1, %ymm0, %ymm0
; AVX2-NEXT: vpbroadcastq {{.*#+}} ymm1 = [72340172838076673,72340172838076673,72340172838076673,72340172838076673]
; AVX2-NEXT: vptest %ymm1, %ymm0
; AVX2-NEXT: setne %al
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
;
; AVX512-LABEL: trunc_v64i8_v64i1:
; AVX512: # %bb.0:
; AVX512-NEXT: vptestmd {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to16}, %zmm0, %k0
; AVX512-NEXT: kortestw %k0, %k0
; AVX512-NEXT: setne %al
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
%a = trunc <64 x i8> %0 to <64 x i1>
%b = call i1 @llvm.vector.reduce.or.v64i1(<64 x i1> %a)
ret i1 %b
}
;
; Comparison With Zero
;
define i1 @icmp0_v2i64_v2i1(<2 x i64>) nounwind {
; SSE2-LABEL: icmp0_v2i64_v2i1:
; SSE2: # %bb.0:
; SSE2-NEXT: pxor %xmm1, %xmm1
; SSE2-NEXT: pcmpeqd %xmm0, %xmm1
; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm1[1,0,3,2]
; SSE2-NEXT: pand %xmm1, %xmm0
; SSE2-NEXT: movmskpd %xmm0, %eax
; SSE2-NEXT: testl %eax, %eax
; SSE2-NEXT: setne %al
; SSE2-NEXT: ret{{[l|q]}}
;
; SSE41-LABEL: icmp0_v2i64_v2i1:
; SSE41: # %bb.0:
; SSE41-NEXT: pxor %xmm1, %xmm1
; SSE41-NEXT: pcmpeqq %xmm0, %xmm1
; SSE41-NEXT: movmskpd %xmm1, %eax
; SSE41-NEXT: testl %eax, %eax
; SSE41-NEXT: setne %al
; SSE41-NEXT: retq
;
; AVX1OR2-LABEL: icmp0_v2i64_v2i1:
; AVX1OR2: # %bb.0:
; AVX1OR2-NEXT: vpxor %xmm1, %xmm1, %xmm1
; AVX1OR2-NEXT: vpcmpeqq %xmm1, %xmm0, %xmm0
; AVX1OR2-NEXT: vtestpd %xmm0, %xmm0
; AVX1OR2-NEXT: setne %al
; AVX1OR2-NEXT: retq
;
; AVX512F-LABEL: icmp0_v2i64_v2i1:
; AVX512F: # %bb.0:
; AVX512F-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0
; AVX512F-NEXT: vptestnmq %zmm0, %zmm0, %k0
; AVX512F-NEXT: kmovw %k0, %eax
; AVX512F-NEXT: testb $3, %al
; AVX512F-NEXT: setne %al
; AVX512F-NEXT: vzeroupper
; AVX512F-NEXT: retq
;
; AVX512BW-LABEL: icmp0_v2i64_v2i1:
; AVX512BW: # %bb.0:
; AVX512BW-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0
; AVX512BW-NEXT: vptestnmq %zmm0, %zmm0, %k0
; AVX512BW-NEXT: kmovd %k0, %eax
; AVX512BW-NEXT: testb $3, %al
; AVX512BW-NEXT: setne %al
; AVX512BW-NEXT: vzeroupper
; AVX512BW-NEXT: retq
;
; AVX512VL-LABEL: icmp0_v2i64_v2i1:
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vptestnmq %xmm0, %xmm0, %k0
; AVX512VL-NEXT: kmovd %k0, %eax
; AVX512VL-NEXT: testb %al, %al
; AVX512VL-NEXT: setne %al
; AVX512VL-NEXT: retq
%a = icmp eq <2 x i64> %0, zeroinitializer
%b = call i1 @llvm.vector.reduce.or.v2i1(<2 x i1> %a)
ret i1 %b
}
define i1 @icmp0_v4i32_v4i1(<4 x i32>) nounwind {
; SSE-LABEL: icmp0_v4i32_v4i1:
; SSE: # %bb.0:
; SSE-NEXT: pxor %xmm1, %xmm1
; SSE-NEXT: pcmpeqd %xmm0, %xmm1
; SSE-NEXT: movmskps %xmm1, %eax
; SSE-NEXT: testl %eax, %eax
; SSE-NEXT: setne %al
; SSE-NEXT: ret{{[l|q]}}
;
; AVX1OR2-LABEL: icmp0_v4i32_v4i1:
; AVX1OR2: # %bb.0:
; AVX1OR2-NEXT: vpxor %xmm1, %xmm1, %xmm1
; AVX1OR2-NEXT: vpcmpeqd %xmm1, %xmm0, %xmm0
; AVX1OR2-NEXT: vtestps %xmm0, %xmm0
; AVX1OR2-NEXT: setne %al
; AVX1OR2-NEXT: retq
;
; AVX512F-LABEL: icmp0_v4i32_v4i1:
; AVX512F: # %bb.0:
; AVX512F-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0
; AVX512F-NEXT: vptestnmd %zmm0, %zmm0, %k0
; AVX512F-NEXT: kmovw %k0, %eax
; AVX512F-NEXT: testb $15, %al
; AVX512F-NEXT: setne %al
; AVX512F-NEXT: vzeroupper
; AVX512F-NEXT: retq
;
; AVX512BW-LABEL: icmp0_v4i32_v4i1:
; AVX512BW: # %bb.0:
; AVX512BW-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0
; AVX512BW-NEXT: vptestnmd %zmm0, %zmm0, %k0
; AVX512BW-NEXT: kmovd %k0, %eax
; AVX512BW-NEXT: testb $15, %al
; AVX512BW-NEXT: setne %al
; AVX512BW-NEXT: vzeroupper
; AVX512BW-NEXT: retq
;
; AVX512VL-LABEL: icmp0_v4i32_v4i1:
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vptestnmd %xmm0, %xmm0, %k0
; AVX512VL-NEXT: kmovd %k0, %eax
; AVX512VL-NEXT: testb %al, %al
; AVX512VL-NEXT: setne %al
; AVX512VL-NEXT: retq
%a = icmp eq <4 x i32> %0, zeroinitializer
%b = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> %a)
ret i1 %b
}
define i1 @icmp0_v8i16_v8i1(<8 x i16>) nounwind {
; SSE-LABEL: icmp0_v8i16_v8i1:
; SSE: # %bb.0:
; SSE-NEXT: pxor %xmm1, %xmm1
; SSE-NEXT: pcmpeqw %xmm0, %xmm1
; SSE-NEXT: pmovmskb %xmm1, %eax
; SSE-NEXT: testl %eax, %eax
; SSE-NEXT: setne %al
; SSE-NEXT: ret{{[l|q]}}
;
; AVX1OR2-LABEL: icmp0_v8i16_v8i1:
; AVX1OR2: # %bb.0:
; AVX1OR2-NEXT: vpxor %xmm1, %xmm1, %xmm1
; AVX1OR2-NEXT: vpcmpeqw %xmm1, %xmm0, %xmm0
; AVX1OR2-NEXT: vpmovmskb %xmm0, %eax
; AVX1OR2-NEXT: testl %eax, %eax
; AVX1OR2-NEXT: setne %al
; AVX1OR2-NEXT: retq
;
; AVX512F-LABEL: icmp0_v8i16_v8i1:
; AVX512F: # %bb.0:
; AVX512F-NEXT: vpxor %xmm1, %xmm1, %xmm1
; AVX512F-NEXT: vpcmpeqw %xmm1, %xmm0, %xmm0
; AVX512F-NEXT: vpmovsxwq %xmm0, %zmm0
; AVX512F-NEXT: vptestmd %zmm0, %zmm0, %k0
; AVX512F-NEXT: kortestw %k0, %k0
; AVX512F-NEXT: setne %al
; AVX512F-NEXT: vzeroupper
; AVX512F-NEXT: retq
;
; AVX512BW-LABEL: icmp0_v8i16_v8i1:
; AVX512BW: # %bb.0:
; AVX512BW-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0
; AVX512BW-NEXT: vptestnmw %zmm0, %zmm0, %k0
; AVX512BW-NEXT: kmovd %k0, %eax
; AVX512BW-NEXT: testb %al, %al
; AVX512BW-NEXT: setne %al
; AVX512BW-NEXT: vzeroupper
; AVX512BW-NEXT: retq
;
; AVX512VL-LABEL: icmp0_v8i16_v8i1:
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vptestnmw %xmm0, %xmm0, %k0
; AVX512VL-NEXT: kmovd %k0, %eax
; AVX512VL-NEXT: testb %al, %al
; AVX512VL-NEXT: setne %al
; AVX512VL-NEXT: retq
%a = icmp eq <8 x i16> %0, zeroinitializer
%b = call i1 @llvm.vector.reduce.or.v8i1(<8 x i1> %a)
ret i1 %b
}
define i1 @icmp0_v16i8_v16i1(<16 x i8>) nounwind {
; SSE-LABEL: icmp0_v16i8_v16i1:
; SSE: # %bb.0:
; SSE-NEXT: pxor %xmm1, %xmm1
; SSE-NEXT: pcmpeqb %xmm0, %xmm1
; SSE-NEXT: pmovmskb %xmm1, %eax
; SSE-NEXT: testl %eax, %eax
; SSE-NEXT: setne %al
; SSE-NEXT: ret{{[l|q]}}
;
; AVX1OR2-LABEL: icmp0_v16i8_v16i1:
; AVX1OR2: # %bb.0:
; AVX1OR2-NEXT: vpxor %xmm1, %xmm1, %xmm1
; AVX1OR2-NEXT: vpcmpeqb %xmm1, %xmm0, %xmm0
; AVX1OR2-NEXT: vpmovmskb %xmm0, %eax
; AVX1OR2-NEXT: testl %eax, %eax
; AVX1OR2-NEXT: setne %al
; AVX1OR2-NEXT: retq
;
; AVX512F-LABEL: icmp0_v16i8_v16i1:
; AVX512F: # %bb.0:
; AVX512F-NEXT: vpxor %xmm1, %xmm1, %xmm1
; AVX512F-NEXT: vpcmpeqb %xmm1, %xmm0, %xmm0
; AVX512F-NEXT: vpmovmskb %xmm0, %eax
; AVX512F-NEXT: testl %eax, %eax
; AVX512F-NEXT: setne %al
; AVX512F-NEXT: retq
;
; AVX512BW-LABEL: icmp0_v16i8_v16i1:
; AVX512BW: # %bb.0:
; AVX512BW-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0
; AVX512BW-NEXT: vptestnmb %zmm0, %zmm0, %k0
; AVX512BW-NEXT: kortestw %k0, %k0
; AVX512BW-NEXT: setne %al
; AVX512BW-NEXT: vzeroupper
; AVX512BW-NEXT: retq
;
; AVX512VL-LABEL: icmp0_v16i8_v16i1:
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vptestnmb %xmm0, %xmm0, %k0
; AVX512VL-NEXT: kortestw %k0, %k0
; AVX512VL-NEXT: setne %al
; AVX512VL-NEXT: retq
%a = icmp eq <16 x i8> %0, zeroinitializer
%b = call i1 @llvm.vector.reduce.or.v16i1(<16 x i1> %a)
ret i1 %b
}
define i1 @icmp0_v4i64_v4i1(<4 x i64>) nounwind {
; SSE2-LABEL: icmp0_v4i64_v4i1:
; SSE2: # %bb.0:
; SSE2-NEXT: pxor %xmm2, %xmm2
; SSE2-NEXT: pcmpeqd %xmm2, %xmm1
; SSE2-NEXT: pcmpeqd %xmm2, %xmm0
; SSE2-NEXT: movdqa %xmm0, %xmm2
; SSE2-NEXT: shufps {{.*#+}} xmm2 = xmm2[1,3],xmm1[1,3]
; SSE2-NEXT: shufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
; SSE2-NEXT: andps %xmm2, %xmm0
; SSE2-NEXT: movmskps %xmm0, %eax
; SSE2-NEXT: testl %eax, %eax
; SSE2-NEXT: setne %al
; SSE2-NEXT: ret{{[l|q]}}
;
; SSE41-LABEL: icmp0_v4i64_v4i1:
; SSE41: # %bb.0:
; SSE41-NEXT: pxor %xmm2, %xmm2
; SSE41-NEXT: pcmpeqq %xmm2, %xmm1
; SSE41-NEXT: pcmpeqq %xmm2, %xmm0
; SSE41-NEXT: packssdw %xmm1, %xmm0
; SSE41-NEXT: movmskps %xmm0, %eax
; SSE41-NEXT: testl %eax, %eax
; SSE41-NEXT: setne %al
; SSE41-NEXT: retq
;
; AVX1-LABEL: icmp0_v4i64_v4i1:
; AVX1: # %bb.0:
; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX1-NEXT: vpxor %xmm2, %xmm2, %xmm2
; AVX1-NEXT: vpcmpeqq %xmm2, %xmm1, %xmm1
; AVX1-NEXT: vpcmpeqq %xmm2, %xmm0, %xmm0
; AVX1-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vtestpd %xmm0, %xmm0
; AVX1-NEXT: setne %al
; AVX1-NEXT: vzeroupper
; AVX1-NEXT: retq
;
; AVX2-LABEL: icmp0_v4i64_v4i1:
; AVX2: # %bb.0:
; AVX2-NEXT: vpxor %xmm1, %xmm1, %xmm1
; AVX2-NEXT: vpcmpeqq %ymm1, %ymm0, %ymm0
; AVX2-NEXT: vtestpd %ymm0, %ymm0
; AVX2-NEXT: setne %al
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
;
; AVX512F-LABEL: icmp0_v4i64_v4i1:
; AVX512F: # %bb.0:
; AVX512F-NEXT: # kill: def $ymm0 killed $ymm0 def $zmm0
; AVX512F-NEXT: vptestnmq %zmm0, %zmm0, %k0
; AVX512F-NEXT: kmovw %k0, %eax
; AVX512F-NEXT: testb $15, %al
; AVX512F-NEXT: setne %al
; AVX512F-NEXT: vzeroupper
; AVX512F-NEXT: retq
;
; AVX512BW-LABEL: icmp0_v4i64_v4i1:
; AVX512BW: # %bb.0:
; AVX512BW-NEXT: # kill: def $ymm0 killed $ymm0 def $zmm0
; AVX512BW-NEXT: vptestnmq %zmm0, %zmm0, %k0
; AVX512BW-NEXT: kmovd %k0, %eax
; AVX512BW-NEXT: testb $15, %al
; AVX512BW-NEXT: setne %al
; AVX512BW-NEXT: vzeroupper
; AVX512BW-NEXT: retq
;
; AVX512VL-LABEL: icmp0_v4i64_v4i1:
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vptestnmq %ymm0, %ymm0, %k0
; AVX512VL-NEXT: kmovd %k0, %eax
; AVX512VL-NEXT: testb %al, %al
; AVX512VL-NEXT: setne %al
; AVX512VL-NEXT: vzeroupper
; AVX512VL-NEXT: retq
%a = icmp eq <4 x i64> %0, zeroinitializer
%b = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> %a)
ret i1 %b
}
define i1 @icmp0_v8i32_v8i1(<8 x i32>) nounwind {
; SSE-LABEL: icmp0_v8i32_v8i1:
; SSE: # %bb.0:
; SSE-NEXT: pxor %xmm2, %xmm2
; SSE-NEXT: pcmpeqd %xmm2, %xmm1
; SSE-NEXT: pcmpeqd %xmm2, %xmm0
; SSE-NEXT: packssdw %xmm1, %xmm0
; SSE-NEXT: pmovmskb %xmm0, %eax
; SSE-NEXT: testl %eax, %eax
; SSE-NEXT: setne %al
; SSE-NEXT: ret{{[l|q]}}
;
; AVX1-LABEL: icmp0_v8i32_v8i1:
; AVX1: # %bb.0:
; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX1-NEXT: vpxor %xmm2, %xmm2, %xmm2
; AVX1-NEXT: vpcmpeqd %xmm2, %xmm1, %xmm1
; AVX1-NEXT: vpcmpeqd %xmm2, %xmm0, %xmm0
; AVX1-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vtestps %xmm0, %xmm0
; AVX1-NEXT: setne %al
; AVX1-NEXT: vzeroupper
; AVX1-NEXT: retq
;
; AVX2-LABEL: icmp0_v8i32_v8i1:
; AVX2: # %bb.0:
; AVX2-NEXT: vpxor %xmm1, %xmm1, %xmm1
; AVX2-NEXT: vpcmpeqd %ymm1, %ymm0, %ymm0
; AVX2-NEXT: vtestps %ymm0, %ymm0
; AVX2-NEXT: setne %al
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
;
; AVX512F-LABEL: icmp0_v8i32_v8i1:
; AVX512F: # %bb.0:
; AVX512F-NEXT: # kill: def $ymm0 killed $ymm0 def $zmm0
; AVX512F-NEXT: vptestnmd %zmm0, %zmm0, %k0
; AVX512F-NEXT: kmovw %k0, %eax
; AVX512F-NEXT: testb %al, %al
; AVX512F-NEXT: setne %al
; AVX512F-NEXT: vzeroupper
; AVX512F-NEXT: retq
;
; AVX512BW-LABEL: icmp0_v8i32_v8i1:
; AVX512BW: # %bb.0:
; AVX512BW-NEXT: # kill: def $ymm0 killed $ymm0 def $zmm0
; AVX512BW-NEXT: vptestnmd %zmm0, %zmm0, %k0
; AVX512BW-NEXT: kmovd %k0, %eax
; AVX512BW-NEXT: testb %al, %al
; AVX512BW-NEXT: setne %al
; AVX512BW-NEXT: vzeroupper
; AVX512BW-NEXT: retq
;
; AVX512VL-LABEL: icmp0_v8i32_v8i1:
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vptestnmd %ymm0, %ymm0, %k0
; AVX512VL-NEXT: kmovd %k0, %eax
; AVX512VL-NEXT: testb %al, %al
; AVX512VL-NEXT: setne %al
; AVX512VL-NEXT: vzeroupper
; AVX512VL-NEXT: retq
%a = icmp eq <8 x i32> %0, zeroinitializer
%b = call i1 @llvm.vector.reduce.or.v8i1(<8 x i1> %a)
ret i1 %b
}
define i1 @icmp0_v16i16_v16i1(<16 x i16>) nounwind {
; SSE-LABEL: icmp0_v16i16_v16i1:
; SSE: # %bb.0:
; SSE-NEXT: pxor %xmm2, %xmm2
; SSE-NEXT: pcmpeqw %xmm2, %xmm1
; SSE-NEXT: pcmpeqw %xmm2, %xmm0
; SSE-NEXT: packsswb %xmm1, %xmm0
; SSE-NEXT: pmovmskb %xmm0, %eax
; SSE-NEXT: testl %eax, %eax
; SSE-NEXT: setne %al
; SSE-NEXT: ret{{[l|q]}}
;
; AVX1-LABEL: icmp0_v16i16_v16i1:
; AVX1: # %bb.0:
; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX1-NEXT: vpxor %xmm2, %xmm2, %xmm2
; AVX1-NEXT: vpcmpeqw %xmm2, %xmm1, %xmm1
; AVX1-NEXT: vpcmpeqw %xmm2, %xmm0, %xmm0
; AVX1-NEXT: vpacksswb %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vpmovmskb %xmm0, %eax
; AVX1-NEXT: testl %eax, %eax
; AVX1-NEXT: setne %al
; AVX1-NEXT: vzeroupper
; AVX1-NEXT: retq
;
; AVX2-LABEL: icmp0_v16i16_v16i1:
; AVX2: # %bb.0:
; AVX2-NEXT: vpxor %xmm1, %xmm1, %xmm1
; AVX2-NEXT: vpcmpeqw %ymm1, %ymm0, %ymm0
; AVX2-NEXT: vpmovmskb %ymm0, %eax
; AVX2-NEXT: testl %eax, %eax
; AVX2-NEXT: setne %al
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
;
; AVX512F-LABEL: icmp0_v16i16_v16i1:
; AVX512F: # %bb.0:
; AVX512F-NEXT: vpxor %xmm1, %xmm1, %xmm1
; AVX512F-NEXT: vpcmpeqw %ymm1, %ymm0, %ymm0
; AVX512F-NEXT: vpmovsxwd %ymm0, %zmm0
; AVX512F-NEXT: vptestmd %zmm0, %zmm0, %k0
; AVX512F-NEXT: kortestw %k0, %k0
; AVX512F-NEXT: setne %al
; AVX512F-NEXT: vzeroupper
; AVX512F-NEXT: retq
;
; AVX512BW-LABEL: icmp0_v16i16_v16i1:
; AVX512BW: # %bb.0:
; AVX512BW-NEXT: # kill: def $ymm0 killed $ymm0 def $zmm0
; AVX512BW-NEXT: vptestnmw %zmm0, %zmm0, %k0
; AVX512BW-NEXT: kortestw %k0, %k0
; AVX512BW-NEXT: setne %al
; AVX512BW-NEXT: vzeroupper
; AVX512BW-NEXT: retq
;
; AVX512VL-LABEL: icmp0_v16i16_v16i1:
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vptestnmw %ymm0, %ymm0, %k0
; AVX512VL-NEXT: kortestw %k0, %k0
; AVX512VL-NEXT: setne %al
; AVX512VL-NEXT: vzeroupper
; AVX512VL-NEXT: retq
%a = icmp eq <16 x i16> %0, zeroinitializer
%b = call i1 @llvm.vector.reduce.or.v16i1(<16 x i1> %a)
ret i1 %b
}
define i1 @icmp0_v32i8_v32i1(<32 x i8>) nounwind {
; SSE-LABEL: icmp0_v32i8_v32i1:
; SSE: # %bb.0:
; SSE-NEXT: pxor %xmm2, %xmm2
; SSE-NEXT: pcmpeqb %xmm2, %xmm1
; SSE-NEXT: pcmpeqb %xmm2, %xmm0
; SSE-NEXT: por %xmm1, %xmm0
; SSE-NEXT: pmovmskb %xmm0, %eax
; SSE-NEXT: testl %eax, %eax
; SSE-NEXT: setne %al
; SSE-NEXT: ret{{[l|q]}}
;
; AVX1-LABEL: icmp0_v32i8_v32i1:
; AVX1: # %bb.0:
; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX1-NEXT: vpxor %xmm2, %xmm2, %xmm2
; AVX1-NEXT: vpcmpeqb %xmm2, %xmm1, %xmm1
; AVX1-NEXT: vpcmpeqb %xmm2, %xmm0, %xmm0
; AVX1-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vpmovmskb %xmm0, %eax
; AVX1-NEXT: testl %eax, %eax
; AVX1-NEXT: setne %al
; AVX1-NEXT: vzeroupper
; AVX1-NEXT: retq
;
; AVX2-LABEL: icmp0_v32i8_v32i1:
; AVX2: # %bb.0:
; AVX2-NEXT: vpxor %xmm1, %xmm1, %xmm1
; AVX2-NEXT: vpcmpeqb %ymm1, %ymm0, %ymm0
; AVX2-NEXT: vpmovmskb %ymm0, %eax
; AVX2-NEXT: testl %eax, %eax
; AVX2-NEXT: setne %al
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
;
; AVX512F-LABEL: icmp0_v32i8_v32i1:
; AVX512F: # %bb.0:
; AVX512F-NEXT: vpxor %xmm1, %xmm1, %xmm1
; AVX512F-NEXT: vpcmpeqb %ymm1, %ymm0, %ymm0
; AVX512F-NEXT: vpmovmskb %ymm0, %eax
; AVX512F-NEXT: testl %eax, %eax
; AVX512F-NEXT: setne %al
; AVX512F-NEXT: vzeroupper
; AVX512F-NEXT: retq
;
; AVX512BW-LABEL: icmp0_v32i8_v32i1:
; AVX512BW: # %bb.0:
; AVX512BW-NEXT: # kill: def $ymm0 killed $ymm0 def $zmm0
; AVX512BW-NEXT: vptestnmb %zmm0, %zmm0, %k0
; AVX512BW-NEXT: kortestd %k0, %k0
; AVX512BW-NEXT: setne %al
; AVX512BW-NEXT: vzeroupper
; AVX512BW-NEXT: retq
;
; AVX512VL-LABEL: icmp0_v32i8_v32i1:
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vptestnmb %ymm0, %ymm0, %k0
; AVX512VL-NEXT: kortestd %k0, %k0
; AVX512VL-NEXT: setne %al
; AVX512VL-NEXT: vzeroupper
; AVX512VL-NEXT: retq
%a = icmp eq <32 x i8> %0, zeroinitializer
%b = call i1 @llvm.vector.reduce.or.v32i1(<32 x i1> %a)
ret i1 %b
}
define i1 @icmp0_v8i64_v8i1(<8 x i64>) nounwind {
; X86-SSE2-LABEL: icmp0_v8i64_v8i1:
; X86-SSE2: # %bb.0:
; X86-SSE2-NEXT: pushl %ebp
; X86-SSE2-NEXT: movl %esp, %ebp
; X86-SSE2-NEXT: andl $-16, %esp
; X86-SSE2-NEXT: subl $16, %esp
; X86-SSE2-NEXT: pxor %xmm3, %xmm3
; X86-SSE2-NEXT: pcmpeqd %xmm3, %xmm1
; X86-SSE2-NEXT: pshufd {{.*#+}} xmm4 = xmm1[1,0,3,2]
; X86-SSE2-NEXT: pand %xmm1, %xmm4
; X86-SSE2-NEXT: pcmpeqd %xmm3, %xmm0
; X86-SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm0[1,0,3,2]
; X86-SSE2-NEXT: pand %xmm0, %xmm1
; X86-SSE2-NEXT: packssdw %xmm4, %xmm1
; X86-SSE2-NEXT: pcmpeqd %xmm3, %xmm2
; X86-SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm2[1,0,3,2]
; X86-SSE2-NEXT: pand %xmm2, %xmm0
; X86-SSE2-NEXT: pcmpeqd 8(%ebp), %xmm3
; X86-SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm3[1,0,3,2]
; X86-SSE2-NEXT: pand %xmm3, %xmm2
; X86-SSE2-NEXT: packssdw %xmm2, %xmm0
; X86-SSE2-NEXT: packssdw %xmm0, %xmm1
; X86-SSE2-NEXT: pmovmskb %xmm1, %eax
; X86-SSE2-NEXT: testl $43690, %eax # imm = 0xAAAA
; X86-SSE2-NEXT: setne %al
; X86-SSE2-NEXT: movl %ebp, %esp
; X86-SSE2-NEXT: popl %ebp
; X86-SSE2-NEXT: retl
;
; X64-SSE2-LABEL: icmp0_v8i64_v8i1:
; X64-SSE2: # %bb.0:
; X64-SSE2-NEXT: pxor %xmm4, %xmm4
; X64-SSE2-NEXT: pcmpeqd %xmm4, %xmm3
; X64-SSE2-NEXT: pshufd {{.*#+}} xmm5 = xmm3[1,0,3,2]
; X64-SSE2-NEXT: pand %xmm3, %xmm5
; X64-SSE2-NEXT: pcmpeqd %xmm4, %xmm2
; X64-SSE2-NEXT: pshufd {{.*#+}} xmm3 = xmm2[1,0,3,2]
; X64-SSE2-NEXT: pand %xmm2, %xmm3
; X64-SSE2-NEXT: packssdw %xmm5, %xmm3
; X64-SSE2-NEXT: pcmpeqd %xmm4, %xmm1
; X64-SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm1[1,0,3,2]
; X64-SSE2-NEXT: pand %xmm1, %xmm2
; X64-SSE2-NEXT: pcmpeqd %xmm4, %xmm0
; X64-SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm0[1,0,3,2]
; X64-SSE2-NEXT: pand %xmm0, %xmm1
; X64-SSE2-NEXT: packssdw %xmm2, %xmm1
; X64-SSE2-NEXT: packssdw %xmm3, %xmm1
; X64-SSE2-NEXT: pmovmskb %xmm1, %eax
; X64-SSE2-NEXT: testl $43690, %eax # imm = 0xAAAA
; X64-SSE2-NEXT: setne %al
; X64-SSE2-NEXT: retq
;
; SSE41-LABEL: icmp0_v8i64_v8i1:
; SSE41: # %bb.0:
; SSE41-NEXT: pxor %xmm4, %xmm4
; SSE41-NEXT: pcmpeqq %xmm4, %xmm3
; SSE41-NEXT: pcmpeqq %xmm4, %xmm2
; SSE41-NEXT: packssdw %xmm3, %xmm2
; SSE41-NEXT: pcmpeqq %xmm4, %xmm1
; SSE41-NEXT: pcmpeqq %xmm4, %xmm0
; SSE41-NEXT: packssdw %xmm1, %xmm0
; SSE41-NEXT: packssdw %xmm2, %xmm0
; SSE41-NEXT: pmovmskb %xmm0, %eax
; SSE41-NEXT: testl %eax, %eax
; SSE41-NEXT: setne %al
; SSE41-NEXT: retq
;
; AVX1-LABEL: icmp0_v8i64_v8i1:
; AVX1: # %bb.0:
; AVX1-NEXT: vpxor %xmm2, %xmm2, %xmm2
; AVX1-NEXT: vpcmpeqq %xmm2, %xmm1, %xmm3
; AVX1-NEXT: vpcmpeqq %xmm2, %xmm0, %xmm4
; AVX1-NEXT: vpor %xmm3, %xmm4, %xmm3
; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm1
; AVX1-NEXT: vpcmpeqq %xmm2, %xmm1, %xmm1
; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm0
; AVX1-NEXT: vpcmpeqq %xmm2, %xmm0, %xmm0
; AVX1-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vpackssdw %xmm0, %xmm3, %xmm0
; AVX1-NEXT: vtestps %xmm0, %xmm0
; AVX1-NEXT: setne %al
; AVX1-NEXT: vzeroupper
; AVX1-NEXT: retq
;
; AVX2-LABEL: icmp0_v8i64_v8i1:
; AVX2: # %bb.0:
; AVX2-NEXT: vpxor %xmm2, %xmm2, %xmm2
; AVX2-NEXT: vpcmpeqq %ymm2, %ymm1, %ymm1
; AVX2-NEXT: vpcmpeqq %ymm2, %ymm0, %ymm0
; AVX2-NEXT: vpackssdw %ymm1, %ymm0, %ymm0
; AVX2-NEXT: vtestps %ymm0, %ymm0
; AVX2-NEXT: setne %al
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
;
; AVX512F-LABEL: icmp0_v8i64_v8i1:
; AVX512F: # %bb.0:
; AVX512F-NEXT: vptestnmq %zmm0, %zmm0, %k0
; AVX512F-NEXT: kmovw %k0, %eax
; AVX512F-NEXT: testb %al, %al
; AVX512F-NEXT: setne %al
; AVX512F-NEXT: vzeroupper
; AVX512F-NEXT: retq
;
; AVX512BW-LABEL: icmp0_v8i64_v8i1:
; AVX512BW: # %bb.0:
; AVX512BW-NEXT: vptestnmq %zmm0, %zmm0, %k0
; AVX512BW-NEXT: kmovd %k0, %eax
; AVX512BW-NEXT: testb %al, %al
; AVX512BW-NEXT: setne %al
; AVX512BW-NEXT: vzeroupper
; AVX512BW-NEXT: retq
;
; AVX512VL-LABEL: icmp0_v8i64_v8i1:
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vptestnmq %zmm0, %zmm0, %k0
; AVX512VL-NEXT: kmovd %k0, %eax
; AVX512VL-NEXT: testb %al, %al
; AVX512VL-NEXT: setne %al
; AVX512VL-NEXT: vzeroupper
; AVX512VL-NEXT: retq
%a = icmp eq <8 x i64> %0, zeroinitializer
%b = call i1 @llvm.vector.reduce.or.v8i1(<8 x i1> %a)
ret i1 %b
}
define i1 @icmp0_v16i32_v16i1(<16 x i32>) nounwind {
; X86-SSE2-LABEL: icmp0_v16i32_v16i1:
; X86-SSE2: # %bb.0:
; X86-SSE2-NEXT: pushl %ebp
; X86-SSE2-NEXT: movl %esp, %ebp
; X86-SSE2-NEXT: andl $-16, %esp
; X86-SSE2-NEXT: subl $16, %esp
; X86-SSE2-NEXT: pxor %xmm3, %xmm3
; X86-SSE2-NEXT: pcmpeqd %xmm3, %xmm1
; X86-SSE2-NEXT: pcmpeqd %xmm3, %xmm0
; X86-SSE2-NEXT: packssdw %xmm1, %xmm0
; X86-SSE2-NEXT: pcmpeqd %xmm3, %xmm2
; X86-SSE2-NEXT: pcmpeqd 8(%ebp), %xmm3
; X86-SSE2-NEXT: packssdw %xmm3, %xmm2
; X86-SSE2-NEXT: packsswb %xmm2, %xmm0
; X86-SSE2-NEXT: pmovmskb %xmm0, %eax
; X86-SSE2-NEXT: testl %eax, %eax
; X86-SSE2-NEXT: setne %al
; X86-SSE2-NEXT: movl %ebp, %esp
; X86-SSE2-NEXT: popl %ebp
; X86-SSE2-NEXT: retl
;
; X64-SSE-LABEL: icmp0_v16i32_v16i1:
; X64-SSE: # %bb.0:
; X64-SSE-NEXT: pxor %xmm4, %xmm4
; X64-SSE-NEXT: pcmpeqd %xmm4, %xmm3
; X64-SSE-NEXT: pcmpeqd %xmm4, %xmm2
; X64-SSE-NEXT: packssdw %xmm3, %xmm2
; X64-SSE-NEXT: pcmpeqd %xmm4, %xmm1
; X64-SSE-NEXT: pcmpeqd %xmm4, %xmm0
; X64-SSE-NEXT: packssdw %xmm1, %xmm0
; X64-SSE-NEXT: packsswb %xmm2, %xmm0
; X64-SSE-NEXT: pmovmskb %xmm0, %eax
; X64-SSE-NEXT: testl %eax, %eax
; X64-SSE-NEXT: setne %al
; X64-SSE-NEXT: retq
;
; AVX1-LABEL: icmp0_v16i32_v16i1:
; AVX1: # %bb.0:
; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm2
; AVX1-NEXT: vpxor %xmm3, %xmm3, %xmm3
; AVX1-NEXT: vpcmpeqd %xmm3, %xmm2, %xmm2
; AVX1-NEXT: vpcmpeqd %xmm3, %xmm1, %xmm1
; AVX1-NEXT: vpackssdw %xmm2, %xmm1, %xmm1
; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm2
; AVX1-NEXT: vpcmpeqd %xmm3, %xmm2, %xmm2
; AVX1-NEXT: vpcmpeqd %xmm3, %xmm0, %xmm0
; AVX1-NEXT: vpackssdw %xmm2, %xmm0, %xmm0
; AVX1-NEXT: vpacksswb %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vpmovmskb %xmm0, %eax
; AVX1-NEXT: testl %eax, %eax
; AVX1-NEXT: setne %al
; AVX1-NEXT: vzeroupper
; AVX1-NEXT: retq
;
; AVX2-LABEL: icmp0_v16i32_v16i1:
; AVX2: # %bb.0:
; AVX2-NEXT: vpxor %xmm2, %xmm2, %xmm2
; AVX2-NEXT: vpcmpeqd %ymm2, %ymm1, %ymm1
; AVX2-NEXT: vpcmpeqd %ymm2, %ymm0, %ymm0
; AVX2-NEXT: vpackssdw %ymm1, %ymm0, %ymm0
; AVX2-NEXT: vpmovmskb %ymm0, %eax
; AVX2-NEXT: testl %eax, %eax
; AVX2-NEXT: setne %al
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
;
; AVX512-LABEL: icmp0_v16i32_v16i1:
; AVX512: # %bb.0:
; AVX512-NEXT: vptestnmd %zmm0, %zmm0, %k0
; AVX512-NEXT: kortestw %k0, %k0
; AVX512-NEXT: setne %al
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
%a = icmp eq <16 x i32> %0, zeroinitializer
%b = call i1 @llvm.vector.reduce.or.v16i1(<16 x i1> %a)
ret i1 %b
}
define i1 @icmp0_v32i16_v32i1(<32 x i16>) nounwind {
; X86-SSE2-LABEL: icmp0_v32i16_v32i1:
; X86-SSE2: # %bb.0:
; X86-SSE2-NEXT: pushl %ebp
; X86-SSE2-NEXT: movl %esp, %ebp
; X86-SSE2-NEXT: andl $-16, %esp
; X86-SSE2-NEXT: subl $16, %esp
; X86-SSE2-NEXT: pxor %xmm3, %xmm3
; X86-SSE2-NEXT: pcmpeqw %xmm3, %xmm1
; X86-SSE2-NEXT: pcmpeqw %xmm3, %xmm2
; X86-SSE2-NEXT: pcmpeqw %xmm3, %xmm0
; X86-SSE2-NEXT: por %xmm2, %xmm0
; X86-SSE2-NEXT: pcmpeqw 8(%ebp), %xmm3
; X86-SSE2-NEXT: por %xmm1, %xmm3
; X86-SSE2-NEXT: packsswb %xmm3, %xmm0
; X86-SSE2-NEXT: pmovmskb %xmm0, %eax
; X86-SSE2-NEXT: testl %eax, %eax
; X86-SSE2-NEXT: setne %al
; X86-SSE2-NEXT: movl %ebp, %esp
; X86-SSE2-NEXT: popl %ebp
; X86-SSE2-NEXT: retl
;
; X64-SSE-LABEL: icmp0_v32i16_v32i1:
; X64-SSE: # %bb.0:
; X64-SSE-NEXT: pxor %xmm4, %xmm4
; X64-SSE-NEXT: pcmpeqw %xmm4, %xmm2
; X64-SSE-NEXT: pcmpeqw %xmm4, %xmm0
; X64-SSE-NEXT: por %xmm2, %xmm0
; X64-SSE-NEXT: pcmpeqw %xmm4, %xmm3
; X64-SSE-NEXT: pcmpeqw %xmm4, %xmm1
; X64-SSE-NEXT: por %xmm3, %xmm1
; X64-SSE-NEXT: packsswb %xmm1, %xmm0
; X64-SSE-NEXT: pmovmskb %xmm0, %eax
; X64-SSE-NEXT: testl %eax, %eax
; X64-SSE-NEXT: setne %al
; X64-SSE-NEXT: retq
;
; AVX1-LABEL: icmp0_v32i16_v32i1:
; AVX1: # %bb.0:
; AVX1-NEXT: vpxor %xmm2, %xmm2, %xmm2
; AVX1-NEXT: vpcmpeqw %xmm2, %xmm1, %xmm3
; AVX1-NEXT: vpcmpeqw %xmm2, %xmm0, %xmm4
; AVX1-NEXT: vpor %xmm3, %xmm4, %xmm3
; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm1
; AVX1-NEXT: vpcmpeqw %xmm2, %xmm1, %xmm1
; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm0
; AVX1-NEXT: vpcmpeqw %xmm2, %xmm0, %xmm0
; AVX1-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vpacksswb %xmm0, %xmm3, %xmm0
; AVX1-NEXT: vpmovmskb %xmm0, %eax
; AVX1-NEXT: testl %eax, %eax
; AVX1-NEXT: setne %al
; AVX1-NEXT: vzeroupper
; AVX1-NEXT: retq
;
; AVX2-LABEL: icmp0_v32i16_v32i1:
; AVX2: # %bb.0:
; AVX2-NEXT: vpxor %xmm2, %xmm2, %xmm2
; AVX2-NEXT: vpcmpeqw %ymm2, %ymm1, %ymm1
; AVX2-NEXT: vpcmpeqw %ymm2, %ymm0, %ymm0
; AVX2-NEXT: vpacksswb %ymm1, %ymm0, %ymm0
; AVX2-NEXT: vpmovmskb %ymm0, %eax
; AVX2-NEXT: testl %eax, %eax
; AVX2-NEXT: setne %al
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
;
; AVX512F-LABEL: icmp0_v32i16_v32i1:
; AVX512F: # %bb.0:
; AVX512F-NEXT: vextracti64x4 $1, %zmm0, %ymm1
; AVX512F-NEXT: vpxor %xmm2, %xmm2, %xmm2
; AVX512F-NEXT: vpcmpeqw %ymm2, %ymm1, %ymm1
; AVX512F-NEXT: vpcmpeqw %ymm2, %ymm0, %ymm0
; AVX512F-NEXT: vpor %ymm1, %ymm0, %ymm0
; AVX512F-NEXT: vpmovsxwd %ymm0, %zmm0
; AVX512F-NEXT: vptestmd %zmm0, %zmm0, %k0
; AVX512F-NEXT: kortestw %k0, %k0
; AVX512F-NEXT: setne %al
; AVX512F-NEXT: vzeroupper
; AVX512F-NEXT: retq
;
; AVX512BW-LABEL: icmp0_v32i16_v32i1:
; AVX512BW: # %bb.0:
; AVX512BW-NEXT: vptestnmw %zmm0, %zmm0, %k0
; AVX512BW-NEXT: kortestd %k0, %k0
; AVX512BW-NEXT: setne %al
; AVX512BW-NEXT: vzeroupper
; AVX512BW-NEXT: retq
;
; AVX512VL-LABEL: icmp0_v32i16_v32i1:
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vptestnmw %zmm0, %zmm0, %k0
; AVX512VL-NEXT: kortestd %k0, %k0
; AVX512VL-NEXT: setne %al
; AVX512VL-NEXT: vzeroupper
; AVX512VL-NEXT: retq
%a = icmp eq <32 x i16> %0, zeroinitializer
%b = call i1 @llvm.vector.reduce.or.v32i1(<32 x i1> %a)
ret i1 %b
}
define i1 @icmp0_v64i8_v64i1(<64 x i8>) nounwind {
; X86-SSE2-LABEL: icmp0_v64i8_v64i1:
; X86-SSE2: # %bb.0:
; X86-SSE2-NEXT: pushl %ebp
; X86-SSE2-NEXT: movl %esp, %ebp
; X86-SSE2-NEXT: andl $-16, %esp
; X86-SSE2-NEXT: subl $16, %esp
; X86-SSE2-NEXT: pxor %xmm3, %xmm3
; X86-SSE2-NEXT: pcmpeqb %xmm3, %xmm1
; X86-SSE2-NEXT: pcmpeqb %xmm3, %xmm2
; X86-SSE2-NEXT: pcmpeqb %xmm3, %xmm0
; X86-SSE2-NEXT: por %xmm2, %xmm0
; X86-SSE2-NEXT: pcmpeqb 8(%ebp), %xmm3
; X86-SSE2-NEXT: por %xmm1, %xmm3
; X86-SSE2-NEXT: por %xmm0, %xmm3
; X86-SSE2-NEXT: pmovmskb %xmm3, %eax
; X86-SSE2-NEXT: testl %eax, %eax
; X86-SSE2-NEXT: setne %al
; X86-SSE2-NEXT: movl %ebp, %esp
; X86-SSE2-NEXT: popl %ebp
; X86-SSE2-NEXT: retl
;
; X64-SSE-LABEL: icmp0_v64i8_v64i1:
; X64-SSE: # %bb.0:
; X64-SSE-NEXT: pxor %xmm4, %xmm4
; X64-SSE-NEXT: pcmpeqb %xmm4, %xmm2
; X64-SSE-NEXT: pcmpeqb %xmm4, %xmm0
; X64-SSE-NEXT: por %xmm2, %xmm0
; X64-SSE-NEXT: pcmpeqb %xmm4, %xmm3
; X64-SSE-NEXT: pcmpeqb %xmm4, %xmm1
; X64-SSE-NEXT: por %xmm3, %xmm1
; X64-SSE-NEXT: por %xmm0, %xmm1
; X64-SSE-NEXT: pmovmskb %xmm1, %eax
; X64-SSE-NEXT: testl %eax, %eax
; X64-SSE-NEXT: setne %al
; X64-SSE-NEXT: retq
;
; AVX1-LABEL: icmp0_v64i8_v64i1:
; AVX1: # %bb.0:
; AVX1-NEXT: vpxor %xmm2, %xmm2, %xmm2
; AVX1-NEXT: vpcmpeqb %xmm2, %xmm1, %xmm3
; AVX1-NEXT: vpcmpeqb %xmm2, %xmm0, %xmm4
; AVX1-NEXT: vpor %xmm3, %xmm4, %xmm3
; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm1
; AVX1-NEXT: vpcmpeqb %xmm2, %xmm1, %xmm1
; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm0
; AVX1-NEXT: vpcmpeqb %xmm2, %xmm0, %xmm0
; AVX1-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vpor %xmm0, %xmm3, %xmm0
; AVX1-NEXT: vpmovmskb %xmm0, %eax
; AVX1-NEXT: testl %eax, %eax
; AVX1-NEXT: setne %al
; AVX1-NEXT: vzeroupper
; AVX1-NEXT: retq
;
; AVX2-LABEL: icmp0_v64i8_v64i1:
; AVX2: # %bb.0:
; AVX2-NEXT: vpxor %xmm2, %xmm2, %xmm2
; AVX2-NEXT: vpcmpeqb %ymm2, %ymm1, %ymm1
; AVX2-NEXT: vpcmpeqb %ymm2, %ymm0, %ymm0
; AVX2-NEXT: vpor %ymm1, %ymm0, %ymm0
; AVX2-NEXT: vpmovmskb %ymm0, %eax
; AVX2-NEXT: testl %eax, %eax
; AVX2-NEXT: setne %al
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
;
; AVX512F-LABEL: icmp0_v64i8_v64i1:
; AVX512F: # %bb.0:
; AVX512F-NEXT: vextracti64x4 $1, %zmm0, %ymm1
; AVX512F-NEXT: vpxor %xmm2, %xmm2, %xmm2
; AVX512F-NEXT: vpcmpeqb %ymm2, %ymm1, %ymm1
; AVX512F-NEXT: vpcmpeqb %ymm2, %ymm0, %ymm0
; AVX512F-NEXT: vpor %ymm1, %ymm0, %ymm0
; AVX512F-NEXT: vpmovmskb %ymm0, %eax
; AVX512F-NEXT: testl %eax, %eax
; AVX512F-NEXT: setne %al
; AVX512F-NEXT: vzeroupper
; AVX512F-NEXT: retq
;
; AVX512BW-LABEL: icmp0_v64i8_v64i1:
; AVX512BW: # %bb.0:
; AVX512BW-NEXT: vptestnmb %zmm0, %zmm0, %k0
; AVX512BW-NEXT: kortestq %k0, %k0
; AVX512BW-NEXT: setne %al
; AVX512BW-NEXT: vzeroupper
; AVX512BW-NEXT: retq
;
; AVX512VL-LABEL: icmp0_v64i8_v64i1:
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vptestnmb %zmm0, %zmm0, %k0
; AVX512VL-NEXT: kortestq %k0, %k0
; AVX512VL-NEXT: setne %al
; AVX512VL-NEXT: vzeroupper
; AVX512VL-NEXT: retq
%a = icmp eq <64 x i8> %0, zeroinitializer
%b = call i1 @llvm.vector.reduce.or.v64i1(<64 x i1> %a)
ret i1 %b
}
; Comparison
;
define i1 @icmp_v2i64_v2i1(<2 x i64>, <2 x i64>) nounwind {
; SSE2-LABEL: icmp_v2i64_v2i1:
; SSE2: # %bb.0:
; SSE2-NEXT: pcmpeqd %xmm1, %xmm0
; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm0[1,0,3,2]
; SSE2-NEXT: pand %xmm0, %xmm1
; SSE2-NEXT: movmskpd %xmm1, %eax
; SSE2-NEXT: testl %eax, %eax
; SSE2-NEXT: setne %al
; SSE2-NEXT: ret{{[l|q]}}
;
; SSE41-LABEL: icmp_v2i64_v2i1:
; SSE41: # %bb.0:
; SSE41-NEXT: pcmpeqq %xmm1, %xmm0
; SSE41-NEXT: movmskpd %xmm0, %eax
; SSE41-NEXT: testl %eax, %eax
; SSE41-NEXT: setne %al
; SSE41-NEXT: retq
;
; AVX1OR2-LABEL: icmp_v2i64_v2i1:
; AVX1OR2: # %bb.0:
; AVX1OR2-NEXT: vpcmpeqq %xmm1, %xmm0, %xmm0
; AVX1OR2-NEXT: vtestpd %xmm0, %xmm0
; AVX1OR2-NEXT: setne %al
; AVX1OR2-NEXT: retq
;
; AVX512F-LABEL: icmp_v2i64_v2i1:
; AVX512F: # %bb.0:
; AVX512F-NEXT: # kill: def $xmm1 killed $xmm1 def $zmm1
; AVX512F-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0
; AVX512F-NEXT: vpcmpeqq %zmm1, %zmm0, %k0
; AVX512F-NEXT: kmovw %k0, %eax
; AVX512F-NEXT: testb $3, %al
; AVX512F-NEXT: setne %al
; AVX512F-NEXT: vzeroupper
; AVX512F-NEXT: retq
;
; AVX512BW-LABEL: icmp_v2i64_v2i1:
; AVX512BW: # %bb.0:
; AVX512BW-NEXT: # kill: def $xmm1 killed $xmm1 def $zmm1
; AVX512BW-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0
; AVX512BW-NEXT: vpcmpeqq %zmm1, %zmm0, %k0
; AVX512BW-NEXT: kmovd %k0, %eax
; AVX512BW-NEXT: testb $3, %al
; AVX512BW-NEXT: setne %al
; AVX512BW-NEXT: vzeroupper
; AVX512BW-NEXT: retq
;
; AVX512VL-LABEL: icmp_v2i64_v2i1:
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vpcmpeqq %xmm1, %xmm0, %k0
; AVX512VL-NEXT: kmovd %k0, %eax
; AVX512VL-NEXT: testb %al, %al
; AVX512VL-NEXT: setne %al
; AVX512VL-NEXT: retq
%a = icmp eq <2 x i64> %0, %1
%b = call i1 @llvm.vector.reduce.or.v2i1(<2 x i1> %a)
ret i1 %b
}
define i1 @icmp_v4i32_v4i1(<4 x i32>, <4 x i32>) nounwind {
; SSE-LABEL: icmp_v4i32_v4i1:
; SSE: # %bb.0:
; SSE-NEXT: pcmpeqd %xmm1, %xmm0
; SSE-NEXT: movmskps %xmm0, %eax
; SSE-NEXT: testl %eax, %eax
; SSE-NEXT: setne %al
; SSE-NEXT: ret{{[l|q]}}
;
; AVX1OR2-LABEL: icmp_v4i32_v4i1:
; AVX1OR2: # %bb.0:
; AVX1OR2-NEXT: vpcmpeqd %xmm1, %xmm0, %xmm0
; AVX1OR2-NEXT: vtestps %xmm0, %xmm0
; AVX1OR2-NEXT: setne %al
; AVX1OR2-NEXT: retq
;
; AVX512F-LABEL: icmp_v4i32_v4i1:
; AVX512F: # %bb.0:
; AVX512F-NEXT: # kill: def $xmm1 killed $xmm1 def $zmm1
; AVX512F-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0
; AVX512F-NEXT: vpcmpeqd %zmm1, %zmm0, %k0
; AVX512F-NEXT: kmovw %k0, %eax
; AVX512F-NEXT: testb $15, %al
; AVX512F-NEXT: setne %al
; AVX512F-NEXT: vzeroupper
; AVX512F-NEXT: retq
;
; AVX512BW-LABEL: icmp_v4i32_v4i1:
; AVX512BW: # %bb.0:
; AVX512BW-NEXT: # kill: def $xmm1 killed $xmm1 def $zmm1
; AVX512BW-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0
; AVX512BW-NEXT: vpcmpeqd %zmm1, %zmm0, %k0
; AVX512BW-NEXT: kmovd %k0, %eax
; AVX512BW-NEXT: testb $15, %al
; AVX512BW-NEXT: setne %al
; AVX512BW-NEXT: vzeroupper
; AVX512BW-NEXT: retq
;
; AVX512VL-LABEL: icmp_v4i32_v4i1:
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vpcmpeqd %xmm1, %xmm0, %k0
; AVX512VL-NEXT: kmovd %k0, %eax
; AVX512VL-NEXT: testb %al, %al
; AVX512VL-NEXT: setne %al
; AVX512VL-NEXT: retq
%a = icmp eq <4 x i32> %0, %1
%b = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> %a)
ret i1 %b
}
define i1 @icmp_v8i16_v8i1(<8 x i16>, <8 x i16>) nounwind {
; SSE-LABEL: icmp_v8i16_v8i1:
; SSE: # %bb.0:
; SSE-NEXT: pcmpeqw %xmm1, %xmm0
; SSE-NEXT: pmovmskb %xmm0, %eax
; SSE-NEXT: testl %eax, %eax
; SSE-NEXT: setne %al
; SSE-NEXT: ret{{[l|q]}}
;
; AVX1OR2-LABEL: icmp_v8i16_v8i1:
; AVX1OR2: # %bb.0:
; AVX1OR2-NEXT: vpcmpeqw %xmm1, %xmm0, %xmm0
; AVX1OR2-NEXT: vpmovmskb %xmm0, %eax
; AVX1OR2-NEXT: testl %eax, %eax
; AVX1OR2-NEXT: setne %al
; AVX1OR2-NEXT: retq
;
; AVX512F-LABEL: icmp_v8i16_v8i1:
; AVX512F: # %bb.0:
; AVX512F-NEXT: vpcmpeqw %xmm1, %xmm0, %xmm0
; AVX512F-NEXT: vpmovsxwq %xmm0, %zmm0
; AVX512F-NEXT: vptestmd %zmm0, %zmm0, %k0
; AVX512F-NEXT: kortestw %k0, %k0
; AVX512F-NEXT: setne %al
; AVX512F-NEXT: vzeroupper
; AVX512F-NEXT: retq
;
; AVX512BW-LABEL: icmp_v8i16_v8i1:
; AVX512BW: # %bb.0:
; AVX512BW-NEXT: # kill: def $xmm1 killed $xmm1 def $zmm1
; AVX512BW-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0
; AVX512BW-NEXT: vpcmpeqw %zmm1, %zmm0, %k0
; AVX512BW-NEXT: kmovd %k0, %eax
; AVX512BW-NEXT: testb %al, %al
; AVX512BW-NEXT: setne %al
; AVX512BW-NEXT: vzeroupper
; AVX512BW-NEXT: retq
;
; AVX512VL-LABEL: icmp_v8i16_v8i1:
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vpcmpeqw %xmm1, %xmm0, %k0
; AVX512VL-NEXT: kmovd %k0, %eax
; AVX512VL-NEXT: testb %al, %al
; AVX512VL-NEXT: setne %al
; AVX512VL-NEXT: retq
%a = icmp eq <8 x i16> %0, %1
%b = call i1 @llvm.vector.reduce.or.v8i1(<8 x i1> %a)
ret i1 %b
}
define i1 @icmp_v16i8_v16i1(<16 x i8>, <16 x i8>) nounwind {
; SSE-LABEL: icmp_v16i8_v16i1:
; SSE: # %bb.0:
; SSE-NEXT: pcmpeqb %xmm1, %xmm0
; SSE-NEXT: pmovmskb %xmm0, %eax
; SSE-NEXT: testl %eax, %eax
; SSE-NEXT: setne %al
; SSE-NEXT: ret{{[l|q]}}
;
; AVX1OR2-LABEL: icmp_v16i8_v16i1:
; AVX1OR2: # %bb.0:
; AVX1OR2-NEXT: vpcmpeqb %xmm1, %xmm0, %xmm0
; AVX1OR2-NEXT: vpmovmskb %xmm0, %eax
; AVX1OR2-NEXT: testl %eax, %eax
; AVX1OR2-NEXT: setne %al
; AVX1OR2-NEXT: retq
;
; AVX512F-LABEL: icmp_v16i8_v16i1:
; AVX512F: # %bb.0:
; AVX512F-NEXT: vpcmpeqb %xmm1, %xmm0, %xmm0
; AVX512F-NEXT: vpmovmskb %xmm0, %eax
; AVX512F-NEXT: testl %eax, %eax
; AVX512F-NEXT: setne %al
; AVX512F-NEXT: retq
;
; AVX512BW-LABEL: icmp_v16i8_v16i1:
; AVX512BW: # %bb.0:
; AVX512BW-NEXT: # kill: def $xmm1 killed $xmm1 def $zmm1
; AVX512BW-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0
; AVX512BW-NEXT: vpcmpeqb %zmm1, %zmm0, %k0
; AVX512BW-NEXT: kortestw %k0, %k0
; AVX512BW-NEXT: setne %al
; AVX512BW-NEXT: vzeroupper
; AVX512BW-NEXT: retq
;
; AVX512VL-LABEL: icmp_v16i8_v16i1:
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vpcmpeqb %xmm1, %xmm0, %k0
; AVX512VL-NEXT: kortestw %k0, %k0
; AVX512VL-NEXT: setne %al
; AVX512VL-NEXT: retq
%a = icmp eq <16 x i8> %0, %1
%b = call i1 @llvm.vector.reduce.or.v16i1(<16 x i1> %a)
ret i1 %b
}
define i1 @icmp_v4i64_v4i1(<4 x i64>, <4 x i64>) nounwind {
; X86-SSE2-LABEL: icmp_v4i64_v4i1:
; X86-SSE2: # %bb.0:
; X86-SSE2-NEXT: pushl %ebp
; X86-SSE2-NEXT: movl %esp, %ebp
; X86-SSE2-NEXT: andl $-16, %esp
; X86-SSE2-NEXT: subl $16, %esp
; X86-SSE2-NEXT: pcmpeqd %xmm2, %xmm0
; X86-SSE2-NEXT: pcmpeqd 8(%ebp), %xmm1
; X86-SSE2-NEXT: movdqa %xmm0, %xmm2
; X86-SSE2-NEXT: shufps {{.*#+}} xmm2 = xmm2[1,3],xmm1[1,3]
; X86-SSE2-NEXT: shufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
; X86-SSE2-NEXT: andps %xmm2, %xmm0
; X86-SSE2-NEXT: movmskps %xmm0, %eax
; X86-SSE2-NEXT: testl %eax, %eax
; X86-SSE2-NEXT: setne %al
; X86-SSE2-NEXT: movl %ebp, %esp
; X86-SSE2-NEXT: popl %ebp
; X86-SSE2-NEXT: retl
;
; X64-SSE2-LABEL: icmp_v4i64_v4i1:
; X64-SSE2: # %bb.0:
; X64-SSE2-NEXT: pcmpeqd %xmm3, %xmm1
; X64-SSE2-NEXT: pcmpeqd %xmm2, %xmm0
; X64-SSE2-NEXT: movdqa %xmm0, %xmm2
; X64-SSE2-NEXT: shufps {{.*#+}} xmm2 = xmm2[1,3],xmm1[1,3]
; X64-SSE2-NEXT: shufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
; X64-SSE2-NEXT: andps %xmm2, %xmm0
; X64-SSE2-NEXT: movmskps %xmm0, %eax
; X64-SSE2-NEXT: testl %eax, %eax
; X64-SSE2-NEXT: setne %al
; X64-SSE2-NEXT: retq
;
; SSE41-LABEL: icmp_v4i64_v4i1:
; SSE41: # %bb.0:
; SSE41-NEXT: pcmpeqq %xmm3, %xmm1
; SSE41-NEXT: pcmpeqq %xmm2, %xmm0
; SSE41-NEXT: packssdw %xmm1, %xmm0
; SSE41-NEXT: movmskps %xmm0, %eax
; SSE41-NEXT: testl %eax, %eax
; SSE41-NEXT: setne %al
; SSE41-NEXT: retq
;
; AVX1-LABEL: icmp_v4i64_v4i1:
; AVX1: # %bb.0:
; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm2
; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm3
; AVX1-NEXT: vpcmpeqq %xmm2, %xmm3, %xmm2
; AVX1-NEXT: vpcmpeqq %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vpor %xmm2, %xmm0, %xmm0
; AVX1-NEXT: vtestpd %xmm0, %xmm0
; AVX1-NEXT: setne %al
; AVX1-NEXT: vzeroupper
; AVX1-NEXT: retq
;
; AVX2-LABEL: icmp_v4i64_v4i1:
; AVX2: # %bb.0:
; AVX2-NEXT: vpcmpeqq %ymm1, %ymm0, %ymm0
; AVX2-NEXT: vtestpd %ymm0, %ymm0
; AVX2-NEXT: setne %al
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
;
; AVX512F-LABEL: icmp_v4i64_v4i1:
; AVX512F: # %bb.0:
; AVX512F-NEXT: # kill: def $ymm1 killed $ymm1 def $zmm1
; AVX512F-NEXT: # kill: def $ymm0 killed $ymm0 def $zmm0
; AVX512F-NEXT: vpcmpeqq %zmm1, %zmm0, %k0
; AVX512F-NEXT: kmovw %k0, %eax
; AVX512F-NEXT: testb $15, %al
; AVX512F-NEXT: setne %al
; AVX512F-NEXT: vzeroupper
; AVX512F-NEXT: retq
;
; AVX512BW-LABEL: icmp_v4i64_v4i1:
; AVX512BW: # %bb.0:
; AVX512BW-NEXT: # kill: def $ymm1 killed $ymm1 def $zmm1
; AVX512BW-NEXT: # kill: def $ymm0 killed $ymm0 def $zmm0
; AVX512BW-NEXT: vpcmpeqq %zmm1, %zmm0, %k0
; AVX512BW-NEXT: kmovd %k0, %eax
; AVX512BW-NEXT: testb $15, %al
; AVX512BW-NEXT: setne %al
; AVX512BW-NEXT: vzeroupper
; AVX512BW-NEXT: retq
;
; AVX512VL-LABEL: icmp_v4i64_v4i1:
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vpcmpeqq %ymm1, %ymm0, %k0
; AVX512VL-NEXT: kmovd %k0, %eax
; AVX512VL-NEXT: testb %al, %al
; AVX512VL-NEXT: setne %al
; AVX512VL-NEXT: vzeroupper
; AVX512VL-NEXT: retq
%a = icmp eq <4 x i64> %0, %1
%b = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> %a)
ret i1 %b
}
define i1 @icmp_v8i32_v8i1(<8 x i32>, <8 x i32>) nounwind {
; X86-SSE2-LABEL: icmp_v8i32_v8i1:
; X86-SSE2: # %bb.0:
; X86-SSE2-NEXT: pushl %ebp
; X86-SSE2-NEXT: movl %esp, %ebp
; X86-SSE2-NEXT: andl $-16, %esp
; X86-SSE2-NEXT: subl $16, %esp
; X86-SSE2-NEXT: pcmpeqd %xmm2, %xmm0
; X86-SSE2-NEXT: pcmpeqd 8(%ebp), %xmm1
; X86-SSE2-NEXT: packssdw %xmm1, %xmm0
; X86-SSE2-NEXT: pmovmskb %xmm0, %eax
; X86-SSE2-NEXT: testl %eax, %eax
; X86-SSE2-NEXT: setne %al
; X86-SSE2-NEXT: movl %ebp, %esp
; X86-SSE2-NEXT: popl %ebp
; X86-SSE2-NEXT: retl
;
; X64-SSE-LABEL: icmp_v8i32_v8i1:
; X64-SSE: # %bb.0:
; X64-SSE-NEXT: pcmpeqd %xmm3, %xmm1
; X64-SSE-NEXT: pcmpeqd %xmm2, %xmm0
; X64-SSE-NEXT: packssdw %xmm1, %xmm0
; X64-SSE-NEXT: pmovmskb %xmm0, %eax
; X64-SSE-NEXT: testl %eax, %eax
; X64-SSE-NEXT: setne %al
; X64-SSE-NEXT: retq
;
; AVX1-LABEL: icmp_v8i32_v8i1:
; AVX1: # %bb.0:
; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm2
; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm3
; AVX1-NEXT: vpcmpeqd %xmm2, %xmm3, %xmm2
; AVX1-NEXT: vpcmpeqd %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vpor %xmm2, %xmm0, %xmm0
; AVX1-NEXT: vtestps %xmm0, %xmm0
; AVX1-NEXT: setne %al
; AVX1-NEXT: vzeroupper
; AVX1-NEXT: retq
;
; AVX2-LABEL: icmp_v8i32_v8i1:
; AVX2: # %bb.0:
; AVX2-NEXT: vpcmpeqd %ymm1, %ymm0, %ymm0
; AVX2-NEXT: vtestps %ymm0, %ymm0
; AVX2-NEXT: setne %al
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
;
; AVX512F-LABEL: icmp_v8i32_v8i1:
; AVX512F: # %bb.0:
; AVX512F-NEXT: # kill: def $ymm1 killed $ymm1 def $zmm1
; AVX512F-NEXT: # kill: def $ymm0 killed $ymm0 def $zmm0
; AVX512F-NEXT: vpcmpeqd %zmm1, %zmm0, %k0
; AVX512F-NEXT: kmovw %k0, %eax
; AVX512F-NEXT: testb %al, %al
; AVX512F-NEXT: setne %al
; AVX512F-NEXT: vzeroupper
; AVX512F-NEXT: retq
;
; AVX512BW-LABEL: icmp_v8i32_v8i1:
; AVX512BW: # %bb.0:
; AVX512BW-NEXT: # kill: def $ymm1 killed $ymm1 def $zmm1
; AVX512BW-NEXT: # kill: def $ymm0 killed $ymm0 def $zmm0
; AVX512BW-NEXT: vpcmpeqd %zmm1, %zmm0, %k0
; AVX512BW-NEXT: kmovd %k0, %eax
; AVX512BW-NEXT: testb %al, %al
; AVX512BW-NEXT: setne %al
; AVX512BW-NEXT: vzeroupper
; AVX512BW-NEXT: retq
;
; AVX512VL-LABEL: icmp_v8i32_v8i1:
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vpcmpeqd %ymm1, %ymm0, %k0
; AVX512VL-NEXT: kmovd %k0, %eax
; AVX512VL-NEXT: testb %al, %al
; AVX512VL-NEXT: setne %al
; AVX512VL-NEXT: vzeroupper
; AVX512VL-NEXT: retq
%a = icmp eq <8 x i32> %0, %1
%b = call i1 @llvm.vector.reduce.or.v8i1(<8 x i1> %a)
ret i1 %b
}
define i1 @icmp_v16i16_v16i1(<16 x i16>, <16 x i16>) nounwind {
; X86-SSE2-LABEL: icmp_v16i16_v16i1:
; X86-SSE2: # %bb.0:
; X86-SSE2-NEXT: pushl %ebp
; X86-SSE2-NEXT: movl %esp, %ebp
; X86-SSE2-NEXT: andl $-16, %esp
; X86-SSE2-NEXT: subl $16, %esp
; X86-SSE2-NEXT: pcmpeqw %xmm2, %xmm0
; X86-SSE2-NEXT: pcmpeqw 8(%ebp), %xmm1
; X86-SSE2-NEXT: packsswb %xmm1, %xmm0
; X86-SSE2-NEXT: pmovmskb %xmm0, %eax
; X86-SSE2-NEXT: testl %eax, %eax
; X86-SSE2-NEXT: setne %al
; X86-SSE2-NEXT: movl %ebp, %esp
; X86-SSE2-NEXT: popl %ebp
; X86-SSE2-NEXT: retl
;
; X64-SSE-LABEL: icmp_v16i16_v16i1:
; X64-SSE: # %bb.0:
; X64-SSE-NEXT: pcmpeqw %xmm3, %xmm1
; X64-SSE-NEXT: pcmpeqw %xmm2, %xmm0
; X64-SSE-NEXT: packsswb %xmm1, %xmm0
; X64-SSE-NEXT: pmovmskb %xmm0, %eax
; X64-SSE-NEXT: testl %eax, %eax
; X64-SSE-NEXT: setne %al
; X64-SSE-NEXT: retq
;
; AVX1-LABEL: icmp_v16i16_v16i1:
; AVX1: # %bb.0:
; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm2
; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm3
; AVX1-NEXT: vpcmpeqw %xmm2, %xmm3, %xmm2
; AVX1-NEXT: vpcmpeqw %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vpacksswb %xmm2, %xmm0, %xmm0
; AVX1-NEXT: vpmovmskb %xmm0, %eax
; AVX1-NEXT: testl %eax, %eax
; AVX1-NEXT: setne %al
; AVX1-NEXT: vzeroupper
; AVX1-NEXT: retq
;
; AVX2-LABEL: icmp_v16i16_v16i1:
; AVX2: # %bb.0:
; AVX2-NEXT: vpcmpeqw %ymm1, %ymm0, %ymm0
; AVX2-NEXT: vpmovmskb %ymm0, %eax
; AVX2-NEXT: testl %eax, %eax
; AVX2-NEXT: setne %al
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
;
; AVX512F-LABEL: icmp_v16i16_v16i1:
; AVX512F: # %bb.0:
; AVX512F-NEXT: vpcmpeqw %ymm1, %ymm0, %ymm0
; AVX512F-NEXT: vpmovsxwd %ymm0, %zmm0
; AVX512F-NEXT: vptestmd %zmm0, %zmm0, %k0
; AVX512F-NEXT: kortestw %k0, %k0
; AVX512F-NEXT: setne %al
; AVX512F-NEXT: vzeroupper
; AVX512F-NEXT: retq
;
; AVX512BW-LABEL: icmp_v16i16_v16i1:
; AVX512BW: # %bb.0:
; AVX512BW-NEXT: # kill: def $ymm1 killed $ymm1 def $zmm1
; AVX512BW-NEXT: # kill: def $ymm0 killed $ymm0 def $zmm0
; AVX512BW-NEXT: vpcmpeqw %zmm1, %zmm0, %k0
; AVX512BW-NEXT: kortestw %k0, %k0
; AVX512BW-NEXT: setne %al
; AVX512BW-NEXT: vzeroupper
; AVX512BW-NEXT: retq
;
; AVX512VL-LABEL: icmp_v16i16_v16i1:
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vpcmpeqw %ymm1, %ymm0, %k0
; AVX512VL-NEXT: kortestw %k0, %k0
; AVX512VL-NEXT: setne %al
; AVX512VL-NEXT: vzeroupper
; AVX512VL-NEXT: retq
%a = icmp eq <16 x i16> %0, %1
%b = call i1 @llvm.vector.reduce.or.v16i1(<16 x i1> %a)
ret i1 %b
}
define i1 @icmp_v32i8_v32i1(<32 x i8>, <32 x i8>) nounwind {
; X86-SSE2-LABEL: icmp_v32i8_v32i1:
; X86-SSE2: # %bb.0:
; X86-SSE2-NEXT: pushl %ebp
; X86-SSE2-NEXT: movl %esp, %ebp
; X86-SSE2-NEXT: andl $-16, %esp
; X86-SSE2-NEXT: subl $16, %esp
; X86-SSE2-NEXT: pcmpeqb %xmm2, %xmm0
; X86-SSE2-NEXT: pcmpeqb 8(%ebp), %xmm1
; X86-SSE2-NEXT: por %xmm0, %xmm1
; X86-SSE2-NEXT: pmovmskb %xmm1, %eax
; X86-SSE2-NEXT: testl %eax, %eax
; X86-SSE2-NEXT: setne %al
; X86-SSE2-NEXT: movl %ebp, %esp
; X86-SSE2-NEXT: popl %ebp
; X86-SSE2-NEXT: retl
;
; X64-SSE-LABEL: icmp_v32i8_v32i1:
; X64-SSE: # %bb.0:
; X64-SSE-NEXT: pcmpeqb %xmm3, %xmm1
; X64-SSE-NEXT: pcmpeqb %xmm2, %xmm0
; X64-SSE-NEXT: por %xmm1, %xmm0
; X64-SSE-NEXT: pmovmskb %xmm0, %eax
; X64-SSE-NEXT: testl %eax, %eax
; X64-SSE-NEXT: setne %al
; X64-SSE-NEXT: retq
;
; AVX1-LABEL: icmp_v32i8_v32i1:
; AVX1: # %bb.0:
; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm2
; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm3
; AVX1-NEXT: vpcmpeqb %xmm2, %xmm3, %xmm2
; AVX1-NEXT: vpcmpeqb %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vpor %xmm2, %xmm0, %xmm0
; AVX1-NEXT: vpmovmskb %xmm0, %eax
; AVX1-NEXT: testl %eax, %eax
; AVX1-NEXT: setne %al
; AVX1-NEXT: vzeroupper
; AVX1-NEXT: retq
;
; AVX2-LABEL: icmp_v32i8_v32i1:
; AVX2: # %bb.0:
; AVX2-NEXT: vpcmpeqb %ymm1, %ymm0, %ymm0
; AVX2-NEXT: vpmovmskb %ymm0, %eax
; AVX2-NEXT: testl %eax, %eax
; AVX2-NEXT: setne %al
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
;
; AVX512F-LABEL: icmp_v32i8_v32i1:
; AVX512F: # %bb.0:
; AVX512F-NEXT: vpcmpeqb %ymm1, %ymm0, %ymm0
; AVX512F-NEXT: vpmovmskb %ymm0, %eax
; AVX512F-NEXT: testl %eax, %eax
; AVX512F-NEXT: setne %al
; AVX512F-NEXT: vzeroupper
; AVX512F-NEXT: retq
;
; AVX512BW-LABEL: icmp_v32i8_v32i1:
; AVX512BW: # %bb.0:
; AVX512BW-NEXT: # kill: def $ymm1 killed $ymm1 def $zmm1
; AVX512BW-NEXT: # kill: def $ymm0 killed $ymm0 def $zmm0
; AVX512BW-NEXT: vpcmpeqb %zmm1, %zmm0, %k0
; AVX512BW-NEXT: kortestd %k0, %k0
; AVX512BW-NEXT: setne %al
; AVX512BW-NEXT: vzeroupper
; AVX512BW-NEXT: retq
;
; AVX512VL-LABEL: icmp_v32i8_v32i1:
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vpcmpeqb %ymm1, %ymm0, %k0
; AVX512VL-NEXT: kortestd %k0, %k0
; AVX512VL-NEXT: setne %al
; AVX512VL-NEXT: vzeroupper
; AVX512VL-NEXT: retq
%a = icmp eq <32 x i8> %0, %1
%b = call i1 @llvm.vector.reduce.or.v32i1(<32 x i1> %a)
ret i1 %b
}
define i1 @icmp_v8i64_v8i1(<8 x i64>, <8 x i64>) nounwind {
; X86-SSE2-LABEL: icmp_v8i64_v8i1:
; X86-SSE2: # %bb.0:
; X86-SSE2-NEXT: pushl %ebp
; X86-SSE2-NEXT: movl %esp, %ebp
; X86-SSE2-NEXT: andl $-16, %esp
; X86-SSE2-NEXT: subl $16, %esp
; X86-SSE2-NEXT: movdqa 8(%ebp), %xmm3
; X86-SSE2-NEXT: pcmpeqd 72(%ebp), %xmm3
; X86-SSE2-NEXT: pshufd {{.*#+}} xmm4 = xmm3[1,0,3,2]
; X86-SSE2-NEXT: pand %xmm3, %xmm4
; X86-SSE2-NEXT: pcmpeqd 56(%ebp), %xmm2
; X86-SSE2-NEXT: pshufd {{.*#+}} xmm3 = xmm2[1,0,3,2]
; X86-SSE2-NEXT: pand %xmm2, %xmm3
; X86-SSE2-NEXT: packssdw %xmm4, %xmm3
; X86-SSE2-NEXT: pcmpeqd 40(%ebp), %xmm1
; X86-SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm1[1,0,3,2]
; X86-SSE2-NEXT: pand %xmm1, %xmm2
; X86-SSE2-NEXT: pcmpeqd 24(%ebp), %xmm0
; X86-SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm0[1,0,3,2]
; X86-SSE2-NEXT: pand %xmm0, %xmm1
; X86-SSE2-NEXT: packssdw %xmm2, %xmm1
; X86-SSE2-NEXT: packssdw %xmm3, %xmm1
; X86-SSE2-NEXT: pmovmskb %xmm1, %eax
; X86-SSE2-NEXT: testl $43690, %eax # imm = 0xAAAA
; X86-SSE2-NEXT: setne %al
; X86-SSE2-NEXT: movl %ebp, %esp
; X86-SSE2-NEXT: popl %ebp
; X86-SSE2-NEXT: retl
;
; X64-SSE2-LABEL: icmp_v8i64_v8i1:
; X64-SSE2: # %bb.0:
; X64-SSE2-NEXT: pcmpeqd %xmm7, %xmm3
; X64-SSE2-NEXT: pshufd {{.*#+}} xmm7 = xmm3[1,0,3,2]
; X64-SSE2-NEXT: pand %xmm3, %xmm7
; X64-SSE2-NEXT: pcmpeqd %xmm6, %xmm2
; X64-SSE2-NEXT: pshufd {{.*#+}} xmm3 = xmm2[1,0,3,2]
; X64-SSE2-NEXT: pand %xmm2, %xmm3
; X64-SSE2-NEXT: packssdw %xmm7, %xmm3
; X64-SSE2-NEXT: pcmpeqd %xmm5, %xmm1
; X64-SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm1[1,0,3,2]
; X64-SSE2-NEXT: pand %xmm1, %xmm2
; X64-SSE2-NEXT: pcmpeqd %xmm4, %xmm0
; X64-SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm0[1,0,3,2]
; X64-SSE2-NEXT: pand %xmm0, %xmm1
; X64-SSE2-NEXT: packssdw %xmm2, %xmm1
; X64-SSE2-NEXT: packssdw %xmm3, %xmm1
; X64-SSE2-NEXT: pmovmskb %xmm1, %eax
; X64-SSE2-NEXT: testl $43690, %eax # imm = 0xAAAA
; X64-SSE2-NEXT: setne %al
; X64-SSE2-NEXT: retq
;
; SSE41-LABEL: icmp_v8i64_v8i1:
; SSE41: # %bb.0:
; SSE41-NEXT: pcmpeqq %xmm7, %xmm3
; SSE41-NEXT: pcmpeqq %xmm6, %xmm2
; SSE41-NEXT: packssdw %xmm3, %xmm2
; SSE41-NEXT: pcmpeqq %xmm5, %xmm1
; SSE41-NEXT: pcmpeqq %xmm4, %xmm0
; SSE41-NEXT: packssdw %xmm1, %xmm0
; SSE41-NEXT: packssdw %xmm2, %xmm0
; SSE41-NEXT: pmovmskb %xmm0, %eax
; SSE41-NEXT: testl %eax, %eax
; SSE41-NEXT: setne %al
; SSE41-NEXT: retq
;
; AVX1-LABEL: icmp_v8i64_v8i1:
; AVX1: # %bb.0:
; AVX1-NEXT: vpcmpeqq %xmm3, %xmm1, %xmm4
; AVX1-NEXT: vpcmpeqq %xmm2, %xmm0, %xmm5
; AVX1-NEXT: vpor %xmm4, %xmm5, %xmm4
; AVX1-NEXT: vextractf128 $1, %ymm3, %xmm3
; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm1
; AVX1-NEXT: vpcmpeqq %xmm3, %xmm1, %xmm1
; AVX1-NEXT: vextractf128 $1, %ymm2, %xmm2
; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm0
; AVX1-NEXT: vpcmpeqq %xmm2, %xmm0, %xmm0
; AVX1-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vpackssdw %xmm0, %xmm4, %xmm0
; AVX1-NEXT: vtestps %xmm0, %xmm0
; AVX1-NEXT: setne %al
; AVX1-NEXT: vzeroupper
; AVX1-NEXT: retq
;
; AVX2-LABEL: icmp_v8i64_v8i1:
; AVX2: # %bb.0:
; AVX2-NEXT: vpcmpeqq %ymm3, %ymm1, %ymm1
; AVX2-NEXT: vpcmpeqq %ymm2, %ymm0, %ymm0
; AVX2-NEXT: vpackssdw %ymm1, %ymm0, %ymm0
; AVX2-NEXT: vtestps %ymm0, %ymm0
; AVX2-NEXT: setne %al
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
;
; AVX512F-LABEL: icmp_v8i64_v8i1:
; AVX512F: # %bb.0:
; AVX512F-NEXT: vpcmpeqq %zmm1, %zmm0, %k0
; AVX512F-NEXT: kmovw %k0, %eax
; AVX512F-NEXT: testb %al, %al
; AVX512F-NEXT: setne %al
; AVX512F-NEXT: vzeroupper
; AVX512F-NEXT: retq
;
; AVX512BW-LABEL: icmp_v8i64_v8i1:
; AVX512BW: # %bb.0:
; AVX512BW-NEXT: vpcmpeqq %zmm1, %zmm0, %k0
; AVX512BW-NEXT: kmovd %k0, %eax
; AVX512BW-NEXT: testb %al, %al
; AVX512BW-NEXT: setne %al
; AVX512BW-NEXT: vzeroupper
; AVX512BW-NEXT: retq
;
; AVX512VL-LABEL: icmp_v8i64_v8i1:
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vpcmpeqq %zmm1, %zmm0, %k0
; AVX512VL-NEXT: kmovd %k0, %eax
; AVX512VL-NEXT: testb %al, %al
; AVX512VL-NEXT: setne %al
; AVX512VL-NEXT: vzeroupper
; AVX512VL-NEXT: retq
%a = icmp eq <8 x i64> %0, %1
%b = call i1 @llvm.vector.reduce.or.v8i1(<8 x i1> %a)
ret i1 %b
}
define i1 @icmp_v16i32_v16i1(<16 x i32>, <16 x i32>) nounwind {
; X86-SSE2-LABEL: icmp_v16i32_v16i1:
; X86-SSE2: # %bb.0:
; X86-SSE2-NEXT: pushl %ebp
; X86-SSE2-NEXT: movl %esp, %ebp
; X86-SSE2-NEXT: andl $-16, %esp
; X86-SSE2-NEXT: subl $16, %esp
; X86-SSE2-NEXT: movdqa 8(%ebp), %xmm3
; X86-SSE2-NEXT: pcmpeqd 72(%ebp), %xmm3
; X86-SSE2-NEXT: pcmpeqd 56(%ebp), %xmm2
; X86-SSE2-NEXT: packssdw %xmm3, %xmm2
; X86-SSE2-NEXT: pcmpeqd 40(%ebp), %xmm1
; X86-SSE2-NEXT: pcmpeqd 24(%ebp), %xmm0
; X86-SSE2-NEXT: packssdw %xmm1, %xmm0
; X86-SSE2-NEXT: packsswb %xmm2, %xmm0
; X86-SSE2-NEXT: pmovmskb %xmm0, %eax
; X86-SSE2-NEXT: testl %eax, %eax
; X86-SSE2-NEXT: setne %al
; X86-SSE2-NEXT: movl %ebp, %esp
; X86-SSE2-NEXT: popl %ebp
; X86-SSE2-NEXT: retl
;
; X64-SSE-LABEL: icmp_v16i32_v16i1:
; X64-SSE: # %bb.0:
; X64-SSE-NEXT: pcmpeqd %xmm7, %xmm3
; X64-SSE-NEXT: pcmpeqd %xmm6, %xmm2
; X64-SSE-NEXT: packssdw %xmm3, %xmm2
; X64-SSE-NEXT: pcmpeqd %xmm5, %xmm1
; X64-SSE-NEXT: pcmpeqd %xmm4, %xmm0
; X64-SSE-NEXT: packssdw %xmm1, %xmm0
; X64-SSE-NEXT: packsswb %xmm2, %xmm0
; X64-SSE-NEXT: pmovmskb %xmm0, %eax
; X64-SSE-NEXT: testl %eax, %eax
; X64-SSE-NEXT: setne %al
; X64-SSE-NEXT: retq
;
; AVX1-LABEL: icmp_v16i32_v16i1:
; AVX1: # %bb.0:
; AVX1-NEXT: vextractf128 $1, %ymm3, %xmm4
; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm5
; AVX1-NEXT: vpcmpeqd %xmm4, %xmm5, %xmm4
; AVX1-NEXT: vpcmpeqd %xmm3, %xmm1, %xmm1
; AVX1-NEXT: vpackssdw %xmm4, %xmm1, %xmm1
; AVX1-NEXT: vextractf128 $1, %ymm2, %xmm3
; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm4
; AVX1-NEXT: vpcmpeqd %xmm3, %xmm4, %xmm3
; AVX1-NEXT: vpcmpeqd %xmm2, %xmm0, %xmm0
; AVX1-NEXT: vpackssdw %xmm3, %xmm0, %xmm0
; AVX1-NEXT: vpacksswb %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vpmovmskb %xmm0, %eax
; AVX1-NEXT: testl %eax, %eax
; AVX1-NEXT: setne %al
; AVX1-NEXT: vzeroupper
; AVX1-NEXT: retq
;
; AVX2-LABEL: icmp_v16i32_v16i1:
; AVX2: # %bb.0:
; AVX2-NEXT: vpcmpeqd %ymm3, %ymm1, %ymm1
; AVX2-NEXT: vpcmpeqd %ymm2, %ymm0, %ymm0
; AVX2-NEXT: vpackssdw %ymm1, %ymm0, %ymm0
; AVX2-NEXT: vpmovmskb %ymm0, %eax
; AVX2-NEXT: testl %eax, %eax
; AVX2-NEXT: setne %al
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
;
; AVX512-LABEL: icmp_v16i32_v16i1:
; AVX512: # %bb.0:
; AVX512-NEXT: vpcmpeqd %zmm1, %zmm0, %k0
; AVX512-NEXT: kortestw %k0, %k0
; AVX512-NEXT: setne %al
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
%a = icmp eq <16 x i32> %0, %1
%b = call i1 @llvm.vector.reduce.or.v16i1(<16 x i1> %a)
ret i1 %b
}
define i1 @icmp_v32i16_v32i1(<32 x i16>, <32 x i16>) nounwind {
; X86-SSE2-LABEL: icmp_v32i16_v32i1:
; X86-SSE2: # %bb.0:
; X86-SSE2-NEXT: pushl %ebp
; X86-SSE2-NEXT: movl %esp, %ebp
; X86-SSE2-NEXT: andl $-16, %esp
; X86-SSE2-NEXT: subl $16, %esp
; X86-SSE2-NEXT: movdqa 8(%ebp), %xmm3
; X86-SSE2-NEXT: pcmpeqw 56(%ebp), %xmm2
; X86-SSE2-NEXT: pcmpeqw 24(%ebp), %xmm0
; X86-SSE2-NEXT: por %xmm2, %xmm0
; X86-SSE2-NEXT: pcmpeqw 72(%ebp), %xmm3
; X86-SSE2-NEXT: pcmpeqw 40(%ebp), %xmm1
; X86-SSE2-NEXT: por %xmm3, %xmm1
; X86-SSE2-NEXT: packsswb %xmm1, %xmm0
; X86-SSE2-NEXT: pmovmskb %xmm0, %eax
; X86-SSE2-NEXT: testl %eax, %eax
; X86-SSE2-NEXT: setne %al
; X86-SSE2-NEXT: movl %ebp, %esp
; X86-SSE2-NEXT: popl %ebp
; X86-SSE2-NEXT: retl
;
; X64-SSE-LABEL: icmp_v32i16_v32i1:
; X64-SSE: # %bb.0:
; X64-SSE-NEXT: pcmpeqw %xmm6, %xmm2
; X64-SSE-NEXT: pcmpeqw %xmm4, %xmm0
; X64-SSE-NEXT: por %xmm2, %xmm0
; X64-SSE-NEXT: pcmpeqw %xmm7, %xmm3
; X64-SSE-NEXT: pcmpeqw %xmm5, %xmm1
; X64-SSE-NEXT: por %xmm3, %xmm1
; X64-SSE-NEXT: packsswb %xmm1, %xmm0
; X64-SSE-NEXT: pmovmskb %xmm0, %eax
; X64-SSE-NEXT: testl %eax, %eax
; X64-SSE-NEXT: setne %al
; X64-SSE-NEXT: retq
;
; AVX1-LABEL: icmp_v32i16_v32i1:
; AVX1: # %bb.0:
; AVX1-NEXT: vpcmpeqw %xmm3, %xmm1, %xmm4
; AVX1-NEXT: vpcmpeqw %xmm2, %xmm0, %xmm5
; AVX1-NEXT: vpor %xmm4, %xmm5, %xmm4
; AVX1-NEXT: vextractf128 $1, %ymm3, %xmm3
; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm1
; AVX1-NEXT: vpcmpeqw %xmm3, %xmm1, %xmm1
; AVX1-NEXT: vextractf128 $1, %ymm2, %xmm2
; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm0
; AVX1-NEXT: vpcmpeqw %xmm2, %xmm0, %xmm0
; AVX1-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vpacksswb %xmm0, %xmm4, %xmm0
; AVX1-NEXT: vpmovmskb %xmm0, %eax
; AVX1-NEXT: testl %eax, %eax
; AVX1-NEXT: setne %al
; AVX1-NEXT: vzeroupper
; AVX1-NEXT: retq
;
; AVX2-LABEL: icmp_v32i16_v32i1:
; AVX2: # %bb.0:
; AVX2-NEXT: vpcmpeqw %ymm3, %ymm1, %ymm1
; AVX2-NEXT: vpcmpeqw %ymm2, %ymm0, %ymm0
; AVX2-NEXT: vpacksswb %ymm1, %ymm0, %ymm0
; AVX2-NEXT: vpmovmskb %ymm0, %eax
; AVX2-NEXT: testl %eax, %eax
; AVX2-NEXT: setne %al
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
;
; AVX512F-LABEL: icmp_v32i16_v32i1:
; AVX512F: # %bb.0:
; AVX512F-NEXT: vextracti64x4 $1, %zmm1, %ymm2
; AVX512F-NEXT: vextracti64x4 $1, %zmm0, %ymm3
; AVX512F-NEXT: vpcmpeqw %ymm2, %ymm3, %ymm2
; AVX512F-NEXT: vpcmpeqw %ymm1, %ymm0, %ymm0
; AVX512F-NEXT: vpor %ymm2, %ymm0, %ymm0
; AVX512F-NEXT: vpmovsxwd %ymm0, %zmm0
; AVX512F-NEXT: vptestmd %zmm0, %zmm0, %k0
; AVX512F-NEXT: kortestw %k0, %k0
; AVX512F-NEXT: setne %al
; AVX512F-NEXT: vzeroupper
; AVX512F-NEXT: retq
;
; AVX512BW-LABEL: icmp_v32i16_v32i1:
; AVX512BW: # %bb.0:
; AVX512BW-NEXT: vpcmpeqw %zmm1, %zmm0, %k0
; AVX512BW-NEXT: kortestd %k0, %k0
; AVX512BW-NEXT: setne %al
; AVX512BW-NEXT: vzeroupper
; AVX512BW-NEXT: retq
;
; AVX512VL-LABEL: icmp_v32i16_v32i1:
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vpcmpeqw %zmm1, %zmm0, %k0
; AVX512VL-NEXT: kortestd %k0, %k0
; AVX512VL-NEXT: setne %al
; AVX512VL-NEXT: vzeroupper
; AVX512VL-NEXT: retq
%a = icmp eq <32 x i16> %0, %1
%b = call i1 @llvm.vector.reduce.or.v32i1(<32 x i1> %a)
ret i1 %b
}
define i1 @icmp_v64i8_v64i1(<64 x i8>, <64 x i8>) nounwind {
; X86-SSE2-LABEL: icmp_v64i8_v64i1:
; X86-SSE2: # %bb.0:
; X86-SSE2-NEXT: pushl %ebp
; X86-SSE2-NEXT: movl %esp, %ebp
; X86-SSE2-NEXT: andl $-16, %esp
; X86-SSE2-NEXT: subl $16, %esp
; X86-SSE2-NEXT: movdqa 8(%ebp), %xmm3
; X86-SSE2-NEXT: pcmpeqb 56(%ebp), %xmm2
; X86-SSE2-NEXT: pcmpeqb 24(%ebp), %xmm0
; X86-SSE2-NEXT: por %xmm2, %xmm0
; X86-SSE2-NEXT: pcmpeqb 72(%ebp), %xmm3
; X86-SSE2-NEXT: pcmpeqb 40(%ebp), %xmm1
; X86-SSE2-NEXT: por %xmm3, %xmm1
; X86-SSE2-NEXT: por %xmm0, %xmm1
; X86-SSE2-NEXT: pmovmskb %xmm1, %eax
; X86-SSE2-NEXT: testl %eax, %eax
; X86-SSE2-NEXT: setne %al
; X86-SSE2-NEXT: movl %ebp, %esp
; X86-SSE2-NEXT: popl %ebp
; X86-SSE2-NEXT: retl
;
; X64-SSE-LABEL: icmp_v64i8_v64i1:
; X64-SSE: # %bb.0:
; X64-SSE-NEXT: pcmpeqb %xmm6, %xmm2
; X64-SSE-NEXT: pcmpeqb %xmm4, %xmm0
; X64-SSE-NEXT: por %xmm2, %xmm0
; X64-SSE-NEXT: pcmpeqb %xmm7, %xmm3
; X64-SSE-NEXT: pcmpeqb %xmm5, %xmm1
; X64-SSE-NEXT: por %xmm3, %xmm1
; X64-SSE-NEXT: por %xmm0, %xmm1
; X64-SSE-NEXT: pmovmskb %xmm1, %eax
; X64-SSE-NEXT: testl %eax, %eax
; X64-SSE-NEXT: setne %al
; X64-SSE-NEXT: retq
;
; AVX1-LABEL: icmp_v64i8_v64i1:
; AVX1: # %bb.0:
; AVX1-NEXT: vpcmpeqb %xmm3, %xmm1, %xmm4
; AVX1-NEXT: vpcmpeqb %xmm2, %xmm0, %xmm5
; AVX1-NEXT: vpor %xmm4, %xmm5, %xmm4
; AVX1-NEXT: vextractf128 $1, %ymm3, %xmm3
; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm1
; AVX1-NEXT: vpcmpeqb %xmm3, %xmm1, %xmm1
; AVX1-NEXT: vextractf128 $1, %ymm2, %xmm2
; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm0
; AVX1-NEXT: vpcmpeqb %xmm2, %xmm0, %xmm0
; AVX1-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vpor %xmm0, %xmm4, %xmm0
; AVX1-NEXT: vpmovmskb %xmm0, %eax
; AVX1-NEXT: testl %eax, %eax
; AVX1-NEXT: setne %al
; AVX1-NEXT: vzeroupper
; AVX1-NEXT: retq
;
; AVX2-LABEL: icmp_v64i8_v64i1:
; AVX2: # %bb.0:
; AVX2-NEXT: vpcmpeqb %ymm3, %ymm1, %ymm1
; AVX2-NEXT: vpcmpeqb %ymm2, %ymm0, %ymm0
; AVX2-NEXT: vpor %ymm1, %ymm0, %ymm0
; AVX2-NEXT: vpmovmskb %ymm0, %eax
; AVX2-NEXT: testl %eax, %eax
; AVX2-NEXT: setne %al
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
;
; AVX512F-LABEL: icmp_v64i8_v64i1:
; AVX512F: # %bb.0:
; AVX512F-NEXT: vextracti64x4 $1, %zmm1, %ymm2
; AVX512F-NEXT: vextracti64x4 $1, %zmm0, %ymm3
; AVX512F-NEXT: vpcmpeqb %ymm2, %ymm3, %ymm2
; AVX512F-NEXT: vpcmpeqb %ymm1, %ymm0, %ymm0
; AVX512F-NEXT: vpor %ymm2, %ymm0, %ymm0
; AVX512F-NEXT: vpmovmskb %ymm0, %eax
; AVX512F-NEXT: testl %eax, %eax
; AVX512F-NEXT: setne %al
; AVX512F-NEXT: vzeroupper
; AVX512F-NEXT: retq
;
; AVX512BW-LABEL: icmp_v64i8_v64i1:
; AVX512BW: # %bb.0:
; AVX512BW-NEXT: vpcmpeqb %zmm1, %zmm0, %k0
; AVX512BW-NEXT: kortestq %k0, %k0
; AVX512BW-NEXT: setne %al
; AVX512BW-NEXT: vzeroupper
; AVX512BW-NEXT: retq
;
; AVX512VL-LABEL: icmp_v64i8_v64i1:
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vpcmpeqb %zmm1, %zmm0, %k0
; AVX512VL-NEXT: kortestq %k0, %k0
; AVX512VL-NEXT: setne %al
; AVX512VL-NEXT: vzeroupper
; AVX512VL-NEXT: retq
%a = icmp eq <64 x i8> %0, %1
%b = call i1 @llvm.vector.reduce.or.v64i1(<64 x i1> %a)
ret i1 %b
}
declare i1 @llvm.vector.reduce.or.v2i1(<2 x i1>)
declare i1 @llvm.vector.reduce.or.v4i1(<4 x i1>)
declare i1 @llvm.vector.reduce.or.v8i1(<8 x i1>)
declare i1 @llvm.vector.reduce.or.v16i1(<16 x i1>)
declare i1 @llvm.vector.reduce.or.v32i1(<32 x i1>)
declare i1 @llvm.vector.reduce.or.v64i1(<64 x i1>)
;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
; AVX: {{.*}}
Codebase Browser
llvm