Page MenuHomePhabricator
Differential D46179 Diff 156996 test/CodeGen/X86/sse2-intrinsics-canonical.ll

Changeset View

Standalone View

View Options

test/CodeGen/X86/sse2-intrinsics-canonical.ll

  • This file was added.
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc < %s -mtriple=i386-apple-darwin -mattr=-avx,+sse2 -show-mc-encoding | FileCheck %s --check-prefix=CHECK --check-prefix=SSE
; RUN: llc < %s -mtriple=i386-apple-darwin -mattr=+avx2 -show-mc-encoding | FileCheck %s --check-prefix=CHECK --check-prefix=VCHECK --check-prefix=AVX2
; RUN: llc < %s -mtriple=i386-apple-darwin -mcpu=skx -show-mc-encoding | FileCheck %s --check-prefix=CHECK --check-prefix=VCHECK --check-prefix=SKX
; NOTE: This should use IR equivalent to what is generated by clang/test/CodeGen/sse2-builtins.c
define <16 x i8> @test_x86_sse2_padds_b(<16 x i8> %a0, <16 x i8> %a1) {
; SSE-LABEL: test_x86_sse2_padds_b:
; SSE: ## %bb.0:
; SSE-NEXT: paddsb %xmm1, %xmm0 ## encoding: [0x66,0x0f,0xec,0xc1]
; SSE-NEXT: retl ## encoding: [0xc3]
;
; AVX2-LABEL: test_x86_sse2_padds_b:
; AVX2: ## %bb.0:
; AVX2-NEXT: vpaddsb %xmm1, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xec,0xc1]
; AVX2-NEXT: retl ## encoding: [0xc3]
;
; SKX-LABEL: test_x86_sse2_padds_b:
; SKX: ## %bb.0:
; SKX-NEXT: vpaddsb %xmm1, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xec,0xc1]
; SKX-NEXT: retl ## encoding: [0xc3]
%1 = sext <16 x i8> %a0 to <16 x i16>
%2 = sext <16 x i8> %a1 to <16 x i16>
%3 = add nsw <16 x i16> %1, %2
%4 = icmp slt <16 x i16> %3, <i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127>
%5 = select <16 x i1> %4, <16 x i16> %3, <16 x i16> <i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127>
%6 = icmp sgt <16 x i16> %5, <i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128>
%7 = select <16 x i1> %6, <16 x i16> %5, <16 x i16> <i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128>
%8 = trunc <16 x i16> %7 to <16 x i8>
ret <16 x i8> %8
}
define <8 x i16> @test_x86_sse2_padds_w(<8 x i16> %a0, <8 x i16> %a1) {
; SSE-LABEL: test_x86_sse2_padds_w:
; SSE: ## %bb.0:
; SSE-NEXT: paddsw %xmm1, %xmm0 ## encoding: [0x66,0x0f,0xed,0xc1]
; SSE-NEXT: retl ## encoding: [0xc3]
;
; AVX2-LABEL: test_x86_sse2_padds_w:
; AVX2: ## %bb.0:
; AVX2-NEXT: vpaddsw %xmm1, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xed,0xc1]
; AVX2-NEXT: retl ## encoding: [0xc3]
;
; SKX-LABEL: test_x86_sse2_padds_w:
; SKX: ## %bb.0:
; SKX-NEXT: vpaddsw %xmm1, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xed,0xc1]
; SKX-NEXT: retl ## encoding: [0xc3]
%1 = sext <8 x i16> %a0 to <8 x i32>
%2 = sext <8 x i16> %a1 to <8 x i32>
%3 = add nsw <8 x i32> %1, %2
%4 = icmp slt <8 x i32> %3, <i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767>
%5 = select <8 x i1> %4, <8 x i32> %3, <8 x i32> <i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767>
%6 = icmp sgt <8 x i32> %5, <i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768>
%7 = select <8 x i1> %6, <8 x i32> %5, <8 x i32> <i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768>
%8 = trunc <8 x i32> %7 to <8 x i16>
ret <8 x i16> %8
}
define <16 x i8> @test_x86_sse2_paddus_b(<16 x i8> %a0, <16 x i8> %a1) {
; SSE-LABEL: test_x86_sse2_paddus_b:
; SSE: ## %bb.0:
; SSE-NEXT: paddusb %xmm1, %xmm0 ## encoding: [0x66,0x0f,0xdc,0xc1]
; SSE-NEXT: retl ## encoding: [0xc3]
;
; AVX2-LABEL: test_x86_sse2_paddus_b:
; AVX2: ## %bb.0:
; AVX2-NEXT: vpaddusb %xmm1, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xdc,0xc1]
; AVX2-NEXT: retl ## encoding: [0xc3]
;
; SKX-LABEL: test_x86_sse2_paddus_b:
; SKX: ## %bb.0:
; SKX-NEXT: vpaddusb %xmm1, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xdc,0xc1]
; SKX-NEXT: retl ## encoding: [0xc3]
%1 = add <16 x i8> %a0, %a1
%2 = icmp ugt <16 x i8> %a0, %1
%3 = select <16 x i1> %2, <16 x i8> <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>, <16 x i8> %1
ret <16 x i8> %3
}
define <8 x i16> @test_x86_sse2_paddus_w(<8 x i16> %a0, <8 x i16> %a1) {
; SSE-LABEL: test_x86_sse2_paddus_w:
; SSE: ## %bb.0:
; SSE-NEXT: paddusw %xmm1, %xmm0 ## encoding: [0x66,0x0f,0xdd,0xc1]
; SSE-NEXT: retl ## encoding: [0xc3]
;
; AVX2-LABEL: test_x86_sse2_paddus_w:
; AVX2: ## %bb.0:
; AVX2-NEXT: vpaddusw %xmm1, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xdd,0xc1]
; AVX2-NEXT: retl ## encoding: [0xc3]
;
; SKX-LABEL: test_x86_sse2_paddus_w:
; SKX: ## %bb.0:
; SKX-NEXT: vpaddusw %xmm1, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xdd,0xc1]
; SKX-NEXT: retl ## encoding: [0xc3]
%1 = add <8 x i16> %a0, %a1
%2 = icmp ugt <8 x i16> %a0, %1
%3 = select <8 x i1> %2, <8 x i16> <i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1>, <8 x i16> %1
ret <8 x i16> %3
}
define <16 x i8> @test_x86_sse2_psubs_b(<16 x i8> %a0, <16 x i8> %a1) {
; SSE-LABEL: test_x86_sse2_psubs_b:
; SSE: ## %bb.0:
; SSE-NEXT: psubsb %xmm1, %xmm0 ## encoding: [0x66,0x0f,0xe8,0xc1]
; SSE-NEXT: retl ## encoding: [0xc3]
;
; AVX2-LABEL: test_x86_sse2_psubs_b:
; AVX2: ## %bb.0:
; AVX2-NEXT: vpsubsb %xmm1, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xe8,0xc1]
; AVX2-NEXT: retl ## encoding: [0xc3]
;
; SKX-LABEL: test_x86_sse2_psubs_b:
; SKX: ## %bb.0:
; SKX-NEXT: vpsubsb %xmm1, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xe8,0xc1]
; SKX-NEXT: retl ## encoding: [0xc3]
%1 = sext <16 x i8> %a0 to <16 x i16>
%2 = sext <16 x i8> %a1 to <16 x i16>
%3 = sub nsw <16 x i16> %1, %2
%4 = icmp slt <16 x i16> %3, <i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127>
%5 = select <16 x i1> %4, <16 x i16> %3, <16 x i16> <i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127>
%6 = icmp sgt <16 x i16> %5, <i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128>
%7 = select <16 x i1> %6, <16 x i16> %5, <16 x i16> <i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128>
%8 = trunc <16 x i16> %7 to <16 x i8>
ret <16 x i8> %8
}
define <8 x i16> @test_x86_sse2_psubs_w(<8 x i16> %a0, <8 x i16> %a1) {
; SSE-LABEL: test_x86_sse2_psubs_w:
; SSE: ## %bb.0:
; SSE-NEXT: psubsw %xmm1, %xmm0 ## encoding: [0x66,0x0f,0xe9,0xc1]
; SSE-NEXT: retl ## encoding: [0xc3]
;
; AVX2-LABEL: test_x86_sse2_psubs_w:
; AVX2: ## %bb.0:
; AVX2-NEXT: vpsubsw %xmm1, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xe9,0xc1]
; AVX2-NEXT: retl ## encoding: [0xc3]
;
; SKX-LABEL: test_x86_sse2_psubs_w:
; SKX: ## %bb.0:
; SKX-NEXT: vpsubsw %xmm1, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xe9,0xc1]
; SKX-NEXT: retl ## encoding: [0xc3]
%1 = sext <8 x i16> %a0 to <8 x i32>
%2 = sext <8 x i16> %a1 to <8 x i32>
%3 = sub nsw <8 x i32> %1, %2
%4 = icmp slt <8 x i32> %3, <i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767>
%5 = select <8 x i1> %4, <8 x i32> %3, <8 x i32> <i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767, i32 32767>
%6 = icmp sgt <8 x i32> %5, <i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768>
%7 = select <8 x i1> %6, <8 x i32> %5, <8 x i32> <i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768, i32 -32768>
%8 = trunc <8 x i32> %7 to <8 x i16>
ret <8 x i16> %8
}
define <16 x i8> @test_x86_sse2_psubus_b(<16 x i8> %a0, <16 x i8> %a1) {
; SSE-LABEL: test_x86_sse2_psubus_b:
; SSE: ## %bb.0:
; SSE-NEXT: psubusb %xmm1, %xmm0 ## encoding: [0x66,0x0f,0xd8,0xc1]
; SSE-NEXT: retl ## encoding: [0xc3]
;
; AVX2-LABEL: test_x86_sse2_psubus_b:
; AVX2: ## %bb.0:
; AVX2-NEXT: vpsubusb %xmm1, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xd8,0xc1]
; AVX2-NEXT: retl ## encoding: [0xc3]
;
; SKX-LABEL: test_x86_sse2_psubus_b:
; SKX: ## %bb.0:
; SKX-NEXT: vpsubusb %xmm1, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xd8,0xc1]
; SKX-NEXT: retl ## encoding: [0xc3]
%cmp = icmp ugt <16 x i8> %a0, %a1
%sel = select <16 x i1> %cmp, <16 x i8> %a0, <16 x i8> %a1
%sub = sub <16 x i8> %sel, %a1
ret <16 x i8> %sub
}
define <8 x i16> @test_x86_sse2_psubus_w(<8 x i16> %a0, <8 x i16> %a1) {
; SSE-LABEL: test_x86_sse2_psubus_w:
; SSE: ## %bb.0:
; SSE-NEXT: psubusw %xmm1, %xmm0 ## encoding: [0x66,0x0f,0xd9,0xc1]
; SSE-NEXT: retl ## encoding: [0xc3]
;
; AVX2-LABEL: test_x86_sse2_psubus_w:
; AVX2: ## %bb.0:
; AVX2-NEXT: vpsubusw %xmm1, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xd9,0xc1]
; AVX2-NEXT: retl ## encoding: [0xc3]
;
; SKX-LABEL: test_x86_sse2_psubus_w:
; SKX: ## %bb.0:
; SKX-NEXT: vpsubusw %xmm1, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xd9,0xc1]
; SKX-NEXT: retl ## encoding: [0xc3]
%cmp = icmp ugt <8 x i16> %a0, %a1
%sel = select <8 x i1> %cmp, <8 x i16> %a0, <8 x i16> %a1
%sub = sub <8 x i16> %sel, %a1
ret <8 x i16> %sub
}
define <8 x i8> @test_x86_sse2_padds_b_64(<8 x i8> %a0, <8 x i8> %a1) {
; SSE-LABEL: test_x86_sse2_padds_b_64:
; SSE: ## %bb.0:
; SSE-NEXT: psllw $8, %xmm0 ## encoding: [0x66,0x0f,0x71,0xf0,0x08]
; SSE-NEXT: psraw $8, %xmm0 ## encoding: [0x66,0x0f,0x71,0xe0,0x08]
; SSE-NEXT: psllw $8, %xmm1 ## encoding: [0x66,0x0f,0x71,0xf1,0x08]
; SSE-NEXT: psraw $8, %xmm1 ## encoding: [0x66,0x0f,0x71,0xe1,0x08]
; SSE-NEXT: paddw %xmm1, %xmm0 ## encoding: [0x66,0x0f,0xfd,0xc1]
; SSE-NEXT: pminsw LCPI8_0, %xmm0 ## encoding: [0x66,0x0f,0xea,0x05,A,A,A,A]
; SSE-NEXT: ## fixup A - offset: 4, value: LCPI8_0, kind: FK_Data_4
; SSE-NEXT: pmaxsw LCPI8_1, %xmm0 ## encoding: [0x66,0x0f,0xee,0x05,A,A,A,A]
; SSE-NEXT: ## fixup A - offset: 4, value: LCPI8_1, kind: FK_Data_4
; SSE-NEXT: retl ## encoding: [0xc3]
;
; AVX2-LABEL: test_x86_sse2_padds_b_64:
; AVX2: ## %bb.0:
; AVX2-NEXT: vpsllw $8, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0x71,0xf0,0x08]
; AVX2-NEXT: vpsraw $8, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0x71,0xe0,0x08]
; AVX2-NEXT: vpsllw $8, %xmm1, %xmm1 ## encoding: [0xc5,0xf1,0x71,0xf1,0x08]
; AVX2-NEXT: vpsraw $8, %xmm1, %xmm1 ## encoding: [0xc5,0xf1,0x71,0xe1,0x08]
; AVX2-NEXT: vpaddw %xmm1, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xfd,0xc1]
; AVX2-NEXT: vpminsw LCPI8_0, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xea,0x05,A,A,A,A]
; AVX2-NEXT: ## fixup A - offset: 4, value: LCPI8_0, kind: FK_Data_4
; AVX2-NEXT: vpmaxsw LCPI8_1, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xee,0x05,A,A,A,A]
; AVX2-NEXT: ## fixup A - offset: 4, value: LCPI8_1, kind: FK_Data_4
; AVX2-NEXT: retl ## encoding: [0xc3]
;
; SKX-LABEL: test_x86_sse2_padds_b_64:
; SKX: ## %bb.0:
; SKX-NEXT: vpsllw $8, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x71,0xf0,0x08]
; SKX-NEXT: vpsraw $8, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x71,0xe0,0x08]
; SKX-NEXT: vpsllw $8, %xmm1, %xmm1 ## EVEX TO VEX Compression encoding: [0xc5,0xf1,0x71,0xf1,0x08]
; SKX-NEXT: vpsraw $8, %xmm1, %xmm1 ## EVEX TO VEX Compression encoding: [0xc5,0xf1,0x71,0xe1,0x08]
; SKX-NEXT: vpaddw %xmm1, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc1]
; SKX-NEXT: vpminsw LCPI8_0, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xea,0x05,A,A,A,A]
; SKX-NEXT: ## fixup A - offset: 4, value: LCPI8_0, kind: FK_Data_4
; SKX-NEXT: vpmaxsw LCPI8_1, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xee,0x05,A,A,A,A]
; SKX-NEXT: ## fixup A - offset: 4, value: LCPI8_1, kind: FK_Data_4
; SKX-NEXT: retl ## encoding: [0xc3]
%1 = sext <8 x i8> %a0 to <8 x i16>
%2 = sext <8 x i8> %a1 to <8 x i16>
%3 = add nsw <8 x i16> %1, %2
%4 = icmp slt <8 x i16> %3, <i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127>
%5 = select <8 x i1> %4, <8 x i16> %3, <8 x i16> <i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127>
%6 = icmp sgt <8 x i16> %5, <i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128>
%7 = select <8 x i1> %6, <8 x i16> %5, <8 x i16> <i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128>
%8 = trunc <8 x i16> %7 to <8 x i8>
ret <8 x i8> %8
}
define <4 x i16> @test_x86_sse2_padds_w_64(<4 x i16> %a0, <4 x i16> %a1) {
; SSE-LABEL: test_x86_sse2_padds_w_64:
; SSE: ## %bb.0:
; SSE-NEXT: pslld $16, %xmm0 ## encoding: [0x66,0x0f,0x72,0xf0,0x10]
; SSE-NEXT: psrad $16, %xmm0 ## encoding: [0x66,0x0f,0x72,0xe0,0x10]
; SSE-NEXT: pslld $16, %xmm1 ## encoding: [0x66,0x0f,0x72,0xf1,0x10]
; SSE-NEXT: psrad $16, %xmm1 ## encoding: [0x66,0x0f,0x72,0xe1,0x10]
; SSE-NEXT: paddd %xmm0, %xmm1 ## encoding: [0x66,0x0f,0xfe,0xc8]
; SSE-NEXT: movdqa {{.*#+}} xmm0 = [32767,32767,32767,32767]
; SSE-NEXT: ## encoding: [0x66,0x0f,0x6f,0x05,A,A,A,A]
; SSE-NEXT: ## fixup A - offset: 4, value: LCPI9_0, kind: FK_Data_4
; SSE-NEXT: movdqa %xmm0, %xmm2 ## encoding: [0x66,0x0f,0x6f,0xd0]
; SSE-NEXT: pcmpgtd %xmm1, %xmm2 ## encoding: [0x66,0x0f,0x66,0xd1]
; SSE-NEXT: pand %xmm2, %xmm1 ## encoding: [0x66,0x0f,0xdb,0xca]
; SSE-NEXT: pandn %xmm0, %xmm2 ## encoding: [0x66,0x0f,0xdf,0xd0]
; SSE-NEXT: por %xmm1, %xmm2 ## encoding: [0x66,0x0f,0xeb,0xd1]
; SSE-NEXT: movdqa {{.*#+}} xmm1 = [4294934528,4294934528,4294934528,4294934528]
; SSE-NEXT: ## encoding: [0x66,0x0f,0x6f,0x0d,A,A,A,A]
; SSE-NEXT: ## fixup A - offset: 4, value: LCPI9_1, kind: FK_Data_4
; SSE-NEXT: movdqa %xmm2, %xmm0 ## encoding: [0x66,0x0f,0x6f,0xc2]
; SSE-NEXT: pcmpgtd %xmm1, %xmm0 ## encoding: [0x66,0x0f,0x66,0xc1]
; SSE-NEXT: pand %xmm0, %xmm2 ## encoding: [0x66,0x0f,0xdb,0xd0]
; SSE-NEXT: pandn %xmm1, %xmm0 ## encoding: [0x66,0x0f,0xdf,0xc1]
; SSE-NEXT: por %xmm2, %xmm0 ## encoding: [0x66,0x0f,0xeb,0xc2]
; SSE-NEXT: retl ## encoding: [0xc3]
;
; AVX2-LABEL: test_x86_sse2_padds_w_64:
; AVX2: ## %bb.0:
; AVX2-NEXT: vpslld $16, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0x72,0xf0,0x10]
; AVX2-NEXT: vpsrad $16, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0x72,0xe0,0x10]
; AVX2-NEXT: vpslld $16, %xmm1, %xmm1 ## encoding: [0xc5,0xf1,0x72,0xf1,0x10]
; AVX2-NEXT: vpsrad $16, %xmm1, %xmm1 ## encoding: [0xc5,0xf1,0x72,0xe1,0x10]
; AVX2-NEXT: vpaddd %xmm1, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xfe,0xc1]
; AVX2-NEXT: vpbroadcastd {{.*#+}} xmm1 = [32767,32767,32767,32767]
; AVX2-NEXT: ## encoding: [0xc4,0xe2,0x79,0x58,0x0d,A,A,A,A]
; AVX2-NEXT: ## fixup A - offset: 5, value: LCPI9_0, kind: FK_Data_4
; AVX2-NEXT: vpminsd %xmm1, %xmm0, %xmm0 ## encoding: [0xc4,0xe2,0x79,0x39,0xc1]
; AVX2-NEXT: vpbroadcastd {{.*#+}} xmm1 = [4294934528,4294934528,4294934528,4294934528]
; AVX2-NEXT: ## encoding: [0xc4,0xe2,0x79,0x58,0x0d,A,A,A,A]
; AVX2-NEXT: ## fixup A - offset: 5, value: LCPI9_1, kind: FK_Data_4
; AVX2-NEXT: vpmaxsd %xmm1, %xmm0, %xmm0 ## encoding: [0xc4,0xe2,0x79,0x3d,0xc1]
; AVX2-NEXT: retl ## encoding: [0xc3]
;
; SKX-LABEL: test_x86_sse2_padds_w_64:
; SKX: ## %bb.0:
; SKX-NEXT: vpslld $16, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x72,0xf0,0x10]
; SKX-NEXT: vpsrad $16, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x72,0xe0,0x10]
; SKX-NEXT: vpslld $16, %xmm1, %xmm1 ## EVEX TO VEX Compression encoding: [0xc5,0xf1,0x72,0xf1,0x10]
; SKX-NEXT: vpsrad $16, %xmm1, %xmm1 ## EVEX TO VEX Compression encoding: [0xc5,0xf1,0x72,0xe1,0x10]
; SKX-NEXT: vpaddd %xmm1, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfe,0xc1]
; SKX-NEXT: vpminsd LCPI9_0{1to4}, %xmm0, %xmm0 ## encoding: [0x62,0xf2,0x7d,0x18,0x39,0x05,A,A,A,A]
; SKX-NEXT: ## fixup A - offset: 6, value: LCPI9_0, kind: FK_Data_4
; SKX-NEXT: vpmaxsd LCPI9_1{1to4}, %xmm0, %xmm0 ## encoding: [0x62,0xf2,0x7d,0x18,0x3d,0x05,A,A,A,A]
; SKX-NEXT: ## fixup A - offset: 6, value: LCPI9_1, kind: FK_Data_4
; SKX-NEXT: retl ## encoding: [0xc3]
%1 = sext <4 x i16> %a0 to <4 x i32>
%2 = sext <4 x i16> %a1 to <4 x i32>
%3 = add nsw <4 x i32> %1, %2
%4 = icmp slt <4 x i32> %3, <i32 32767, i32 32767, i32 32767, i32 32767>
%5 = select <4 x i1> %4, <4 x i32> %3, <4 x i32> <i32 32767, i32 32767, i32 32767, i32 32767>
%6 = icmp sgt <4 x i32> %5, <i32 -32768, i32 -32768, i32 -32768, i32 -32768>
%7 = select <4 x i1> %6, <4 x i32> %5, <4 x i32> <i32 -32768, i32 -32768, i32 -32768, i32 -32768>
%8 = trunc <4 x i32> %7 to <4 x i16>
ret <4 x i16> %8
}
define <8 x i8> @test_x86_sse2_paddus_b_64(<8 x i8> %a0, <8 x i8> %a1) {
; SSE-LABEL: test_x86_sse2_paddus_b_64:
; SSE: ## %bb.0:
; SSE-NEXT: movdqa {{.*#+}} xmm2 = [255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0]
; SSE-NEXT: ## encoding: [0x66,0x0f,0x6f,0x15,A,A,A,A]
; SSE-NEXT: ## fixup A - offset: 4, value: LCPI10_0, kind: FK_Data_4
; SSE-NEXT: paddw %xmm0, %xmm1 ## encoding: [0x66,0x0f,0xfd,0xc8]
; SSE-NEXT: pand %xmm2, %xmm0 ## encoding: [0x66,0x0f,0xdb,0xc2]
; SSE-NEXT: pand %xmm1, %xmm2 ## encoding: [0x66,0x0f,0xdb,0xd1]
; SSE-NEXT: pcmpgtw %xmm2, %xmm0 ## encoding: [0x66,0x0f,0x65,0xc2]
; SSE-NEXT: movdqa %xmm0, %xmm2 ## encoding: [0x66,0x0f,0x6f,0xd0]
; SSE-NEXT: pandn %xmm1, %xmm2 ## encoding: [0x66,0x0f,0xdf,0xd1]
; SSE-NEXT: pand LCPI10_0, %xmm0 ## encoding: [0x66,0x0f,0xdb,0x05,A,A,A,A]
; SSE-NEXT: ## fixup A - offset: 4, value: LCPI10_0, kind: FK_Data_4
; SSE-NEXT: por %xmm2, %xmm0 ## encoding: [0x66,0x0f,0xeb,0xc2]
; SSE-NEXT: retl ## encoding: [0xc3]
;
; AVX2-LABEL: test_x86_sse2_paddus_b_64:
; AVX2: ## %bb.0:
; AVX2-NEXT: vmovdqa {{.*#+}} xmm2 = [255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0]
; AVX2-NEXT: ## encoding: [0xc5,0xf9,0x6f,0x15,A,A,A,A]
; AVX2-NEXT: ## fixup A - offset: 4, value: LCPI10_0, kind: FK_Data_4
; AVX2-NEXT: vpand %xmm2, %xmm0, %xmm3 ## encoding: [0xc5,0xf9,0xdb,0xda]
; AVX2-NEXT: vpaddw %xmm1, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xfd,0xc1]
; AVX2-NEXT: vpand %xmm2, %xmm0, %xmm1 ## encoding: [0xc5,0xf9,0xdb,0xca]
; AVX2-NEXT: vpcmpgtw %xmm1, %xmm3, %xmm1 ## encoding: [0xc5,0xe1,0x65,0xc9]
; AVX2-NEXT: vpblendvb %xmm1, %xmm2, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x4c,0xc2,0x10]
; AVX2-NEXT: retl ## encoding: [0xc3]
;
; SKX-LABEL: test_x86_sse2_paddus_b_64:
; SKX: ## %bb.0:
; SKX-NEXT: vmovdqa LCPI10_0, %xmm2 ## EVEX TO VEX Compression xmm2 = [255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0]
; SKX-NEXT: ## encoding: [0xc5,0xf9,0x6f,0x15,A,A,A,A]
; SKX-NEXT: ## fixup A - offset: 4, value: LCPI10_0, kind: FK_Data_4
; SKX-NEXT: vpand %xmm2, %xmm0, %xmm3 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xdb,0xda]
; SKX-NEXT: vpaddw %xmm1, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc1]
; SKX-NEXT: vpand %xmm2, %xmm0, %xmm1 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xdb,0xca]
; SKX-NEXT: vpcmpnleuw %xmm1, %xmm3, %k1 ## encoding: [0x62,0xf3,0xe5,0x08,0x3e,0xc9,0x06]
; SKX-NEXT: vmovdqu16 LCPI10_0, %xmm0 {%k1} ## encoding: [0x62,0xf1,0xff,0x09,0x6f,0x05,A,A,A,A]
; SKX-NEXT: ## fixup A - offset: 6, value: LCPI10_0, kind: FK_Data_4
; SKX-NEXT: retl ## encoding: [0xc3]
%1 = add <8 x i8> %a0, %a1
%2 = icmp ugt <8 x i8> %a0, %1
%3 = select <8 x i1> %2, <8 x i8> <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>, <8 x i8> %1
ret <8 x i8> %3
}
define <4 x i16> @test_x86_sse2_paddus_w_64(<4 x i16> %a0, <4 x i16> %a1) {
; SSE-LABEL: test_x86_sse2_paddus_w_64:
; SSE: ## %bb.0:
; SSE-NEXT: movdqa {{.*#+}} xmm2 = [65535,0,65535,0,65535,0,65535,0]
; SSE-NEXT: ## encoding: [0x66,0x0f,0x6f,0x15,A,A,A,A]
; SSE-NEXT: ## fixup A - offset: 4, value: LCPI11_0, kind: FK_Data_4
; SSE-NEXT: paddd %xmm0, %xmm1 ## encoding: [0x66,0x0f,0xfe,0xc8]
; SSE-NEXT: pand %xmm2, %xmm0 ## encoding: [0x66,0x0f,0xdb,0xc2]
; SSE-NEXT: pand %xmm1, %xmm2 ## encoding: [0x66,0x0f,0xdb,0xd1]
; SSE-NEXT: pcmpgtd %xmm2, %xmm0 ## encoding: [0x66,0x0f,0x66,0xc2]
; SSE-NEXT: movdqa %xmm0, %xmm2 ## encoding: [0x66,0x0f,0x6f,0xd0]
; SSE-NEXT: pandn %xmm1, %xmm2 ## encoding: [0x66,0x0f,0xdf,0xd1]
; SSE-NEXT: pand LCPI11_0, %xmm0 ## encoding: [0x66,0x0f,0xdb,0x05,A,A,A,A]
; SSE-NEXT: ## fixup A - offset: 4, value: LCPI11_0, kind: FK_Data_4
; SSE-NEXT: por %xmm2, %xmm0 ## encoding: [0x66,0x0f,0xeb,0xc2]
; SSE-NEXT: retl ## encoding: [0xc3]
;
; AVX2-LABEL: test_x86_sse2_paddus_w_64:
; AVX2: ## %bb.0:
; AVX2-NEXT: vpxor %xmm2, %xmm2, %xmm2 ## encoding: [0xc5,0xe9,0xef,0xd2]
; AVX2-NEXT: vpblendw $170, %xmm2, %xmm0, %xmm3 ## encoding: [0xc4,0xe3,0x79,0x0e,0xda,0xaa]
; AVX2-NEXT: ## xmm3 = xmm0[0],xmm2[1],xmm0[2],xmm2[3],xmm0[4],xmm2[5],xmm0[6],xmm2[7]
; AVX2-NEXT: vpaddd %xmm1, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xfe,0xc1]
; AVX2-NEXT: vpblendw $170, %xmm2, %xmm0, %xmm1 ## encoding: [0xc4,0xe3,0x79,0x0e,0xca,0xaa]
; AVX2-NEXT: ## xmm1 = xmm0[0],xmm2[1],xmm0[2],xmm2[3],xmm0[4],xmm2[5],xmm0[6],xmm2[7]
; AVX2-NEXT: vpcmpgtd %xmm1, %xmm3, %xmm1 ## encoding: [0xc5,0xe1,0x66,0xc9]
; AVX2-NEXT: vbroadcastss {{.*#+}} xmm2 = [65535,65535,65535,65535]
; AVX2-NEXT: ## encoding: [0xc4,0xe2,0x79,0x18,0x15,A,A,A,A]
; AVX2-NEXT: ## fixup A - offset: 5, value: LCPI11_0, kind: FK_Data_4
; AVX2-NEXT: vblendvps %xmm1, %xmm2, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x4a,0xc2,0x10]
; AVX2-NEXT: retl ## encoding: [0xc3]
;
; SKX-LABEL: test_x86_sse2_paddus_w_64:
; SKX: ## %bb.0:
; SKX-NEXT: vpxor %xmm2, %xmm2, %xmm2 ## EVEX TO VEX Compression encoding: [0xc5,0xe9,0xef,0xd2]
; SKX-NEXT: vpblendw $170, %xmm2, %xmm0, %xmm3 ## encoding: [0xc4,0xe3,0x79,0x0e,0xda,0xaa]
; SKX-NEXT: ## xmm3 = xmm0[0],xmm2[1],xmm0[2],xmm2[3],xmm0[4],xmm2[5],xmm0[6],xmm2[7]
; SKX-NEXT: vpaddd %xmm1, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfe,0xc1]
; SKX-NEXT: vpblendw $170, %xmm2, %xmm0, %xmm1 ## encoding: [0xc4,0xe3,0x79,0x0e,0xca,0xaa]
; SKX-NEXT: ## xmm1 = xmm0[0],xmm2[1],xmm0[2],xmm2[3],xmm0[4],xmm2[5],xmm0[6],xmm2[7]
; SKX-NEXT: vpcmpnleud %xmm1, %xmm3, %k1 ## encoding: [0x62,0xf3,0x65,0x08,0x1e,0xc9,0x06]
; SKX-NEXT: vpbroadcastd LCPI11_0, %xmm0 {%k1} ## encoding: [0x62,0xf2,0x7d,0x09,0x58,0x05,A,A,A,A]
; SKX-NEXT: ## fixup A - offset: 6, value: LCPI11_0, kind: FK_Data_4
; SKX-NEXT: retl ## encoding: [0xc3]
%1 = add <4 x i16> %a0, %a1
%2 = icmp ugt <4 x i16> %a0, %1
%3 = select <4 x i1> %2, <4 x i16> <i16 -1, i16 -1, i16 -1, i16 -1>, <4 x i16> %1
ret <4 x i16> %3
}
define <8 x i8> @test_x86_sse2_psubs_b_64(<8 x i8> %a0, <8 x i8> %a1) {
; SSE-LABEL: test_x86_sse2_psubs_b_64:
; SSE: ## %bb.0:
; SSE-NEXT: psllw $8, %xmm0 ## encoding: [0x66,0x0f,0x71,0xf0,0x08]
; SSE-NEXT: psraw $8, %xmm0 ## encoding: [0x66,0x0f,0x71,0xe0,0x08]
; SSE-NEXT: psllw $8, %xmm1 ## encoding: [0x66,0x0f,0x71,0xf1,0x08]
; SSE-NEXT: psraw $8, %xmm1 ## encoding: [0x66,0x0f,0x71,0xe1,0x08]
; SSE-NEXT: psubw %xmm1, %xmm0 ## encoding: [0x66,0x0f,0xf9,0xc1]
; SSE-NEXT: pminsw LCPI12_0, %xmm0 ## encoding: [0x66,0x0f,0xea,0x05,A,A,A,A]
; SSE-NEXT: ## fixup A - offset: 4, value: LCPI12_0, kind: FK_Data_4
; SSE-NEXT: pmaxsw LCPI12_1, %xmm0 ## encoding: [0x66,0x0f,0xee,0x05,A,A,A,A]
; SSE-NEXT: ## fixup A - offset: 4, value: LCPI12_1, kind: FK_Data_4
; SSE-NEXT: retl ## encoding: [0xc3]
;
; AVX2-LABEL: test_x86_sse2_psubs_b_64:
; AVX2: ## %bb.0:
; AVX2-NEXT: vpsllw $8, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0x71,0xf0,0x08]
; AVX2-NEXT: vpsraw $8, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0x71,0xe0,0x08]
; AVX2-NEXT: vpsllw $8, %xmm1, %xmm1 ## encoding: [0xc5,0xf1,0x71,0xf1,0x08]
; AVX2-NEXT: vpsraw $8, %xmm1, %xmm1 ## encoding: [0xc5,0xf1,0x71,0xe1,0x08]
; AVX2-NEXT: vpsubw %xmm1, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xf9,0xc1]
; AVX2-NEXT: vpminsw LCPI12_0, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xea,0x05,A,A,A,A]
; AVX2-NEXT: ## fixup A - offset: 4, value: LCPI12_0, kind: FK_Data_4
; AVX2-NEXT: vpmaxsw LCPI12_1, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xee,0x05,A,A,A,A]
; AVX2-NEXT: ## fixup A - offset: 4, value: LCPI12_1, kind: FK_Data_4
; AVX2-NEXT: retl ## encoding: [0xc3]
;
; SKX-LABEL: test_x86_sse2_psubs_b_64:
; SKX: ## %bb.0:
; SKX-NEXT: vpsllw $8, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x71,0xf0,0x08]
; SKX-NEXT: vpsraw $8, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x71,0xe0,0x08]
; SKX-NEXT: vpsllw $8, %xmm1, %xmm1 ## EVEX TO VEX Compression encoding: [0xc5,0xf1,0x71,0xf1,0x08]
; SKX-NEXT: vpsraw $8, %xmm1, %xmm1 ## EVEX TO VEX Compression encoding: [0xc5,0xf1,0x71,0xe1,0x08]
; SKX-NEXT: vpsubw %xmm1, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xf9,0xc1]
; SKX-NEXT: vpminsw LCPI12_0, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xea,0x05,A,A,A,A]
; SKX-NEXT: ## fixup A - offset: 4, value: LCPI12_0, kind: FK_Data_4
; SKX-NEXT: vpmaxsw LCPI12_1, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xee,0x05,A,A,A,A]
; SKX-NEXT: ## fixup A - offset: 4, value: LCPI12_1, kind: FK_Data_4
; SKX-NEXT: retl ## encoding: [0xc3]
%1 = sext <8 x i8> %a0 to <8 x i16>
%2 = sext <8 x i8> %a1 to <8 x i16>
%3 = sub nsw <8 x i16> %1, %2
%4 = icmp slt <8 x i16> %3, <i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127>
%5 = select <8 x i1> %4, <8 x i16> %3, <8 x i16> <i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127, i16 127>
%6 = icmp sgt <8 x i16> %5, <i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128>
%7 = select <8 x i1> %6, <8 x i16> %5, <8 x i16> <i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128, i16 -128>
%8 = trunc <8 x i16> %7 to <8 x i8>
ret <8 x i8> %8
}
define <4 x i16> @test_x86_sse2_psubs_w_64(<4 x i16> %a0, <4 x i16> %a1) {
; SSE-LABEL: test_x86_sse2_psubs_w_64:
; SSE: ## %bb.0:
; SSE-NEXT: pslld $16, %xmm0 ## encoding: [0x66,0x0f,0x72,0xf0,0x10]
; SSE-NEXT: psrad $16, %xmm0 ## encoding: [0x66,0x0f,0x72,0xe0,0x10]
; SSE-NEXT: pslld $16, %xmm1 ## encoding: [0x66,0x0f,0x72,0xf1,0x10]
; SSE-NEXT: psrad $16, %xmm1 ## encoding: [0x66,0x0f,0x72,0xe1,0x10]
; SSE-NEXT: psubd %xmm1, %xmm0 ## encoding: [0x66,0x0f,0xfa,0xc1]
; SSE-NEXT: movdqa {{.*#+}} xmm1 = [32767,32767,32767,32767]
; SSE-NEXT: ## encoding: [0x66,0x0f,0x6f,0x0d,A,A,A,A]
; SSE-NEXT: ## fixup A - offset: 4, value: LCPI13_0, kind: FK_Data_4
; SSE-NEXT: movdqa %xmm1, %xmm2 ## encoding: [0x66,0x0f,0x6f,0xd1]
; SSE-NEXT: pcmpgtd %xmm0, %xmm2 ## encoding: [0x66,0x0f,0x66,0xd0]
; SSE-NEXT: pand %xmm2, %xmm0 ## encoding: [0x66,0x0f,0xdb,0xc2]
; SSE-NEXT: pandn %xmm1, %xmm2 ## encoding: [0x66,0x0f,0xdf,0xd1]
; SSE-NEXT: por %xmm0, %xmm2 ## encoding: [0x66,0x0f,0xeb,0xd0]
; SSE-NEXT: movdqa {{.*#+}} xmm1 = [4294934528,4294934528,4294934528,4294934528]
; SSE-NEXT: ## encoding: [0x66,0x0f,0x6f,0x0d,A,A,A,A]
; SSE-NEXT: ## fixup A - offset: 4, value: LCPI13_1, kind: FK_Data_4
; SSE-NEXT: movdqa %xmm2, %xmm0 ## encoding: [0x66,0x0f,0x6f,0xc2]
; SSE-NEXT: pcmpgtd %xmm1, %xmm0 ## encoding: [0x66,0x0f,0x66,0xc1]
; SSE-NEXT: pand %xmm0, %xmm2 ## encoding: [0x66,0x0f,0xdb,0xd0]
; SSE-NEXT: pandn %xmm1, %xmm0 ## encoding: [0x66,0x0f,0xdf,0xc1]
; SSE-NEXT: por %xmm2, %xmm0 ## encoding: [0x66,0x0f,0xeb,0xc2]
; SSE-NEXT: retl ## encoding: [0xc3]
;
; AVX2-LABEL: test_x86_sse2_psubs_w_64:
; AVX2: ## %bb.0:
; AVX2-NEXT: vpslld $16, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0x72,0xf0,0x10]
; AVX2-NEXT: vpsrad $16, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0x72,0xe0,0x10]
; AVX2-NEXT: vpslld $16, %xmm1, %xmm1 ## encoding: [0xc5,0xf1,0x72,0xf1,0x10]
; AVX2-NEXT: vpsrad $16, %xmm1, %xmm1 ## encoding: [0xc5,0xf1,0x72,0xe1,0x10]
; AVX2-NEXT: vpsubd %xmm1, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xfa,0xc1]
; AVX2-NEXT: vpbroadcastd {{.*#+}} xmm1 = [32767,32767,32767,32767]
; AVX2-NEXT: ## encoding: [0xc4,0xe2,0x79,0x58,0x0d,A,A,A,A]
; AVX2-NEXT: ## fixup A - offset: 5, value: LCPI13_0, kind: FK_Data_4
; AVX2-NEXT: vpminsd %xmm1, %xmm0, %xmm0 ## encoding: [0xc4,0xe2,0x79,0x39,0xc1]
; AVX2-NEXT: vpbroadcastd {{.*#+}} xmm1 = [4294934528,4294934528,4294934528,4294934528]
; AVX2-NEXT: ## encoding: [0xc4,0xe2,0x79,0x58,0x0d,A,A,A,A]
; AVX2-NEXT: ## fixup A - offset: 5, value: LCPI13_1, kind: FK_Data_4
; AVX2-NEXT: vpmaxsd %xmm1, %xmm0, %xmm0 ## encoding: [0xc4,0xe2,0x79,0x3d,0xc1]
; AVX2-NEXT: retl ## encoding: [0xc3]
;
; SKX-LABEL: test_x86_sse2_psubs_w_64:
; SKX: ## %bb.0:
; SKX-NEXT: vpslld $16, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x72,0xf0,0x10]
; SKX-NEXT: vpsrad $16, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x72,0xe0,0x10]
; SKX-NEXT: vpslld $16, %xmm1, %xmm1 ## EVEX TO VEX Compression encoding: [0xc5,0xf1,0x72,0xf1,0x10]
; SKX-NEXT: vpsrad $16, %xmm1, %xmm1 ## EVEX TO VEX Compression encoding: [0xc5,0xf1,0x72,0xe1,0x10]
; SKX-NEXT: vpsubd %xmm1, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfa,0xc1]
; SKX-NEXT: vpminsd LCPI13_0{1to4}, %xmm0, %xmm0 ## encoding: [0x62,0xf2,0x7d,0x18,0x39,0x05,A,A,A,A]
; SKX-NEXT: ## fixup A - offset: 6, value: LCPI13_0, kind: FK_Data_4
; SKX-NEXT: vpmaxsd LCPI13_1{1to4}, %xmm0, %xmm0 ## encoding: [0x62,0xf2,0x7d,0x18,0x3d,0x05,A,A,A,A]
; SKX-NEXT: ## fixup A - offset: 6, value: LCPI13_1, kind: FK_Data_4
; SKX-NEXT: retl ## encoding: [0xc3]
%1 = sext <4 x i16> %a0 to <4 x i32>
%2 = sext <4 x i16> %a1 to <4 x i32>
%3 = sub nsw <4 x i32> %1, %2
%4 = icmp slt <4 x i32> %3, <i32 32767, i32 32767, i32 32767, i32 32767>
%5 = select <4 x i1> %4, <4 x i32> %3, <4 x i32> <i32 32767, i32 32767, i32 32767, i32 32767>
%6 = icmp sgt <4 x i32> %5, <i32 -32768, i32 -32768, i32 -32768, i32 -32768>
%7 = select <4 x i1> %6, <4 x i32> %5, <4 x i32> <i32 -32768, i32 -32768, i32 -32768, i32 -32768>
%8 = trunc <4 x i32> %7 to <4 x i16>
ret <4 x i16> %8
}
define <8 x i8> @test_x86_sse2_psubus_b_64(<8 x i8> %a0, <8 x i8> %a1) {
; SSE-LABEL: test_x86_sse2_psubus_b_64:
; SSE: ## %bb.0:
; SSE-NEXT: movdqa {{.*#+}} xmm2 = [255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0]
; SSE-NEXT: ## encoding: [0x66,0x0f,0x6f,0x15,A,A,A,A]
; SSE-NEXT: ## fixup A - offset: 4, value: LCPI14_0, kind: FK_Data_4
; SSE-NEXT: movdqa %xmm1, %xmm3 ## encoding: [0x66,0x0f,0x6f,0xd9]
; SSE-NEXT: pand %xmm2, %xmm3 ## encoding: [0x66,0x0f,0xdb,0xda]
; SSE-NEXT: pand %xmm2, %xmm0 ## encoding: [0x66,0x0f,0xdb,0xc2]
; SSE-NEXT: pmaxsw %xmm3, %xmm0 ## encoding: [0x66,0x0f,0xee,0xc3]
; SSE-NEXT: psubw %xmm1, %xmm0 ## encoding: [0x66,0x0f,0xf9,0xc1]
; SSE-NEXT: retl ## encoding: [0xc3]
;
; AVX2-LABEL: test_x86_sse2_psubus_b_64:
; AVX2: ## %bb.0:
; AVX2-NEXT: vmovdqa {{.*#+}} xmm2 = [255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0]
; AVX2-NEXT: ## encoding: [0xc5,0xf9,0x6f,0x15,A,A,A,A]
; AVX2-NEXT: ## fixup A - offset: 4, value: LCPI14_0, kind: FK_Data_4
; AVX2-NEXT: vpand %xmm2, %xmm1, %xmm3 ## encoding: [0xc5,0xf1,0xdb,0xda]
; AVX2-NEXT: vpand %xmm2, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xdb,0xc2]
; AVX2-NEXT: vpmaxuw %xmm3, %xmm0, %xmm0 ## encoding: [0xc4,0xe2,0x79,0x3e,0xc3]
; AVX2-NEXT: vpsubw %xmm1, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xf9,0xc1]
; AVX2-NEXT: retl ## encoding: [0xc3]
;
; SKX-LABEL: test_x86_sse2_psubus_b_64:
; SKX: ## %bb.0:
; SKX-NEXT: vmovdqa LCPI14_0, %xmm2 ## EVEX TO VEX Compression xmm2 = [255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0]
; SKX-NEXT: ## encoding: [0xc5,0xf9,0x6f,0x15,A,A,A,A]
; SKX-NEXT: ## fixup A - offset: 4, value: LCPI14_0, kind: FK_Data_4
; SKX-NEXT: vpand %xmm2, %xmm1, %xmm3 ## EVEX TO VEX Compression encoding: [0xc5,0xf1,0xdb,0xda]
; SKX-NEXT: vpand %xmm2, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xdb,0xc2]
; SKX-NEXT: vpmaxuw %xmm3, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x3e,0xc3]
; SKX-NEXT: vpsubw %xmm1, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xf9,0xc1]
; SKX-NEXT: retl ## encoding: [0xc3]
%cmp = icmp ugt <8 x i8> %a0, %a1
%sel = select <8 x i1> %cmp, <8 x i8> %a0, <8 x i8> %a1
%sub = sub <8 x i8> %sel, %a1
ret <8 x i8> %sub
}
define <4 x i16> @test_x86_sse2_psubus_w_64(<4 x i16> %a0, <4 x i16> %a1) {
; SSE-LABEL: test_x86_sse2_psubus_w_64:
; SSE: ## %bb.0:
; SSE-NEXT: movdqa {{.*#+}} xmm2 = [65535,0,65535,0,65535,0,65535,0]
; SSE-NEXT: ## encoding: [0x66,0x0f,0x6f,0x15,A,A,A,A]
; SSE-NEXT: ## fixup A - offset: 4, value: LCPI15_0, kind: FK_Data_4
; SSE-NEXT: movdqa %xmm1, %xmm3 ## encoding: [0x66,0x0f,0x6f,0xd9]
; SSE-NEXT: pand %xmm2, %xmm3 ## encoding: [0x66,0x0f,0xdb,0xda]
; SSE-NEXT: pand %xmm2, %xmm0 ## encoding: [0x66,0x0f,0xdb,0xc2]
; SSE-NEXT: movdqa %xmm0, %xmm2 ## encoding: [0x66,0x0f,0x6f,0xd0]
; SSE-NEXT: pcmpgtd %xmm3, %xmm2 ## encoding: [0x66,0x0f,0x66,0xd3]
; SSE-NEXT: pand %xmm2, %xmm0 ## encoding: [0x66,0x0f,0xdb,0xc2]
; SSE-NEXT: pandn %xmm3, %xmm2 ## encoding: [0x66,0x0f,0xdf,0xd3]
; SSE-NEXT: por %xmm0, %xmm2 ## encoding: [0x66,0x0f,0xeb,0xd0]
; SSE-NEXT: psubd %xmm1, %xmm2 ## encoding: [0x66,0x0f,0xfa,0xd1]
; SSE-NEXT: movdqa %xmm2, %xmm0 ## encoding: [0x66,0x0f,0x6f,0xc2]
; SSE-NEXT: retl ## encoding: [0xc3]
;
; AVX2-LABEL: test_x86_sse2_psubus_w_64:
; AVX2: ## %bb.0:
; AVX2-NEXT: vpxor %xmm2, %xmm2, %xmm2 ## encoding: [0xc5,0xe9,0xef,0xd2]
; AVX2-NEXT: vpblendw $170, %xmm2, %xmm1, %xmm3 ## encoding: [0xc4,0xe3,0x71,0x0e,0xda,0xaa]
; AVX2-NEXT: ## xmm3 = xmm1[0],xmm2[1],xmm1[2],xmm2[3],xmm1[4],xmm2[5],xmm1[6],xmm2[7]
; AVX2-NEXT: vpblendw $170, %xmm2, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x0e,0xc2,0xaa]
; AVX2-NEXT: ## xmm0 = xmm0[0],xmm2[1],xmm0[2],xmm2[3],xmm0[4],xmm2[5],xmm0[6],xmm2[7]
; AVX2-NEXT: vpmaxud %xmm3, %xmm0, %xmm0 ## encoding: [0xc4,0xe2,0x79,0x3f,0xc3]
; AVX2-NEXT: vpsubd %xmm1, %xmm0, %xmm0 ## encoding: [0xc5,0xf9,0xfa,0xc1]
; AVX2-NEXT: retl ## encoding: [0xc3]
;
; SKX-LABEL: test_x86_sse2_psubus_w_64:
; SKX: ## %bb.0:
; SKX-NEXT: vpxor %xmm2, %xmm2, %xmm2 ## EVEX TO VEX Compression encoding: [0xc5,0xe9,0xef,0xd2]
; SKX-NEXT: vpblendw $170, %xmm2, %xmm1, %xmm3 ## encoding: [0xc4,0xe3,0x71,0x0e,0xda,0xaa]
; SKX-NEXT: ## xmm3 = xmm1[0],xmm2[1],xmm1[2],xmm2[3],xmm1[4],xmm2[5],xmm1[6],xmm2[7]
; SKX-NEXT: vpblendw $170, %xmm2, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x0e,0xc2,0xaa]
; SKX-NEXT: ## xmm0 = xmm0[0],xmm2[1],xmm0[2],xmm2[3],xmm0[4],xmm2[5],xmm0[6],xmm2[7]
; SKX-NEXT: vpmaxud %xmm3, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x3f,0xc3]
; SKX-NEXT: vpsubd %xmm1, %xmm0, %xmm0 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfa,0xc1]
; SKX-NEXT: retl ## encoding: [0xc3]
%cmp = icmp ugt <4 x i16> %a0, %a1
%sel = select <4 x i1> %cmp, <4 x i16> %a0, <4 x i16> %a1
%sub = sub <4 x i16> %sel, %a1
ret <4 x i16> %sub
}