RPCS3/llvm-mirror
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-11-23 11:13:28 +01:00
Code Issues Projects Releases Wiki Activity
6a9be09c82
llvm-mirror/test/CodeGen/X86/vector-reduce-add.ll
Simon Pilgrim 21418c85a1 [X86][SSE] getV4X86ShuffleImm8 - canonicalize broadcast masks 
If the mask input to getV4X86ShuffleImm8 only refers to a single source element (+ undefs) then canonicalize to a full broadcast.

getV4X86ShuffleImm8 defaults to inline values for undefs, which can be useful for shuffle widening/narrowing but does leave SimplifyDemanded* calls thinking the shuffle depends on unnecessary elements.

I'm still investigating what we should do more generally to avoid these undemanded elements, but broadcast cases was a simpler win.
2020-07-29 11:32:44 +01:00

1392 lines
49 KiB
LLVM
Raw Blame History

; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+sse2 | FileCheck %s --check-prefixes=SSE,SSE2
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+sse4.1 | FileCheck %s --check-prefixes=SSE,SSE41
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx | FileCheck %s --check-prefixes=AVX,AVX1,AVX1-SLOW
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx,+fast-hops | FileCheck %s --check-prefixes=AVX,AVX1,AVX1-FAST
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2 | FileCheck %s --check-prefixes=AVX,AVX2
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512f,+avx512bw | FileCheck %s --check-prefixes=AVX512,AVX512BW
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512f,+avx512bw,+avx512vl | FileCheck %s --check-prefixes=AVX512,AVX512VL
;
; vXi64
;
define i64 @test_v2i64(<2 x i64> %a0) {
; SSE-LABEL: test_v2i64:
; SSE: # %bb.0:
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
; SSE-NEXT: paddq %xmm0, %xmm1
; SSE-NEXT: movq %xmm1, %rax
; SSE-NEXT: retq
;
; AVX-LABEL: test_v2i64:
; AVX: # %bb.0:
; AVX-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
; AVX-NEXT: vpaddq %xmm1, %xmm0, %xmm0
; AVX-NEXT: vmovq %xmm0, %rax
; AVX-NEXT: retq
;
; AVX512-LABEL: test_v2i64:
; AVX512: # %bb.0:
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
; AVX512-NEXT: vpaddq %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovq %xmm0, %rax
; AVX512-NEXT: retq
%1 = call i64 @llvm.experimental.vector.reduce.add.v2i64(<2 x i64> %a0)
ret i64 %1
}
define i64 @test_v4i64(<4 x i64> %a0) {
; SSE-LABEL: test_v4i64:
; SSE: # %bb.0:
; SSE-NEXT: paddq %xmm1, %xmm0
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
; SSE-NEXT: paddq %xmm0, %xmm1
; SSE-NEXT: movq %xmm1, %rax
; SSE-NEXT: retq
;
; AVX1-LABEL: test_v4i64:
; AVX1: # %bb.0:
; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX1-NEXT: vpaddq %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
; AVX1-NEXT: vpaddq %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vmovq %xmm0, %rax
; AVX1-NEXT: vzeroupper
; AVX1-NEXT: retq
;
; AVX2-LABEL: test_v4i64:
; AVX2: # %bb.0:
; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX2-NEXT: vpaddq %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
; AVX2-NEXT: vpaddq %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vmovq %xmm0, %rax
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
;
; AVX512-LABEL: test_v4i64:
; AVX512: # %bb.0:
; AVX512-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX512-NEXT: vpaddq %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
; AVX512-NEXT: vpaddq %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovq %xmm0, %rax
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
%1 = call i64 @llvm.experimental.vector.reduce.add.v4i64(<4 x i64> %a0)
ret i64 %1
}
define i64 @test_v8i64(<8 x i64> %a0) {
; SSE-LABEL: test_v8i64:
; SSE: # %bb.0:
; SSE-NEXT: paddq %xmm3, %xmm1
; SSE-NEXT: paddq %xmm2, %xmm1
; SSE-NEXT: paddq %xmm0, %xmm1
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm1[2,3,2,3]
; SSE-NEXT: paddq %xmm1, %xmm0
; SSE-NEXT: movq %xmm0, %rax
; SSE-NEXT: retq
;
; AVX1-LABEL: test_v8i64:
; AVX1: # %bb.0:
; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm2
; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm3
; AVX1-NEXT: vpaddq %xmm2, %xmm3, %xmm2
; AVX1-NEXT: vpaddq %xmm2, %xmm1, %xmm1
; AVX1-NEXT: vpaddq %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
; AVX1-NEXT: vpaddq %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vmovq %xmm0, %rax
; AVX1-NEXT: vzeroupper
; AVX1-NEXT: retq
;
; AVX2-LABEL: test_v8i64:
; AVX2: # %bb.0:
; AVX2-NEXT: vpaddq %ymm1, %ymm0, %ymm0
; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX2-NEXT: vpaddq %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
; AVX2-NEXT: vpaddq %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vmovq %xmm0, %rax
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
;
; AVX512-LABEL: test_v8i64:
; AVX512: # %bb.0:
; AVX512-NEXT: vextracti64x4 $1, %zmm0, %ymm1
; AVX512-NEXT: vpaddq %zmm1, %zmm0, %zmm0
; AVX512-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX512-NEXT: vpaddq %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
; AVX512-NEXT: vpaddq %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovq %xmm0, %rax
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
%1 = call i64 @llvm.experimental.vector.reduce.add.v8i64(<8 x i64> %a0)
ret i64 %1
}
define i64 @test_v16i64(<16 x i64> %a0) {
; SSE-LABEL: test_v16i64:
; SSE: # %bb.0:
; SSE-NEXT: paddq %xmm6, %xmm2
; SSE-NEXT: paddq %xmm7, %xmm3
; SSE-NEXT: paddq %xmm5, %xmm3
; SSE-NEXT: paddq %xmm1, %xmm3
; SSE-NEXT: paddq %xmm4, %xmm2
; SSE-NEXT: paddq %xmm3, %xmm2
; SSE-NEXT: paddq %xmm0, %xmm2
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm2[2,3,2,3]
; SSE-NEXT: paddq %xmm2, %xmm0
; SSE-NEXT: movq %xmm0, %rax
; SSE-NEXT: retq
;
; AVX1-LABEL: test_v16i64:
; AVX1: # %bb.0:
; AVX1-NEXT: vpaddq %xmm3, %xmm1, %xmm4
; AVX1-NEXT: vextractf128 $1, %ymm3, %xmm3
; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm1
; AVX1-NEXT: vpaddq %xmm3, %xmm1, %xmm1
; AVX1-NEXT: vextractf128 $1, %ymm2, %xmm3
; AVX1-NEXT: vpaddq %xmm1, %xmm3, %xmm1
; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm3
; AVX1-NEXT: vpaddq %xmm1, %xmm3, %xmm1
; AVX1-NEXT: vpaddq %xmm4, %xmm2, %xmm2
; AVX1-NEXT: vpaddq %xmm1, %xmm2, %xmm1
; AVX1-NEXT: vpaddq %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
; AVX1-NEXT: vpaddq %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vmovq %xmm0, %rax
; AVX1-NEXT: vzeroupper
; AVX1-NEXT: retq
;
; AVX2-LABEL: test_v16i64:
; AVX2: # %bb.0:
; AVX2-NEXT: vpaddq %ymm3, %ymm1, %ymm1
; AVX2-NEXT: vpaddq %ymm1, %ymm2, %ymm1
; AVX2-NEXT: vpaddq %ymm1, %ymm0, %ymm0
; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX2-NEXT: vpaddq %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
; AVX2-NEXT: vpaddq %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vmovq %xmm0, %rax
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
;
; AVX512-LABEL: test_v16i64:
; AVX512: # %bb.0:
; AVX512-NEXT: vpaddq %zmm1, %zmm0, %zmm0
; AVX512-NEXT: vextracti64x4 $1, %zmm0, %ymm1
; AVX512-NEXT: vpaddq %zmm1, %zmm0, %zmm0
; AVX512-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX512-NEXT: vpaddq %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
; AVX512-NEXT: vpaddq %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovq %xmm0, %rax
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
%1 = call i64 @llvm.experimental.vector.reduce.add.v16i64(<16 x i64> %a0)
ret i64 %1
}
;
; vXi32
;
define i32 @test_v2i32(<2 x i32> %a0) {
; SSE-LABEL: test_v2i32:
; SSE: # %bb.0:
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
; SSE-NEXT: paddd %xmm0, %xmm1
; SSE-NEXT: movd %xmm1, %eax
; SSE-NEXT: retq
;
; AVX1-SLOW-LABEL: test_v2i32:
; AVX1-SLOW: # %bb.0:
; AVX1-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
; AVX1-SLOW-NEXT: vpaddd %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT: vmovd %xmm0, %eax
; AVX1-SLOW-NEXT: retq
;
; AVX1-FAST-LABEL: test_v2i32:
; AVX1-FAST: # %bb.0:
; AVX1-FAST-NEXT: vphaddd %xmm0, %xmm0, %xmm0
; AVX1-FAST-NEXT: vmovd %xmm0, %eax
; AVX1-FAST-NEXT: retq
;
; AVX2-LABEL: test_v2i32:
; AVX2: # %bb.0:
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
; AVX2-NEXT: vpaddd %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vmovd %xmm0, %eax
; AVX2-NEXT: retq
;
; AVX512-LABEL: test_v2i32:
; AVX512: # %bb.0:
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
; AVX512-NEXT: vpaddd %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovd %xmm0, %eax
; AVX512-NEXT: retq
%1 = call i32 @llvm.experimental.vector.reduce.add.v2i32(<2 x i32> %a0)
ret i32 %1
}
define i32 @test_v4i32(<4 x i32> %a0) {
; SSE-LABEL: test_v4i32:
; SSE: # %bb.0:
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
; SSE-NEXT: paddd %xmm0, %xmm1
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm1[1,1,1,1]
; SSE-NEXT: paddd %xmm1, %xmm0
; SSE-NEXT: movd %xmm0, %eax
; SSE-NEXT: retq
;
; AVX1-SLOW-LABEL: test_v4i32:
; AVX1-SLOW: # %bb.0:
; AVX1-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
; AVX1-SLOW-NEXT: vpaddd %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
; AVX1-SLOW-NEXT: vpaddd %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT: vmovd %xmm0, %eax
; AVX1-SLOW-NEXT: retq
;
; AVX1-FAST-LABEL: test_v4i32:
; AVX1-FAST: # %bb.0:
; AVX1-FAST-NEXT: vphaddd %xmm0, %xmm0, %xmm0
; AVX1-FAST-NEXT: vphaddd %xmm0, %xmm0, %xmm0
; AVX1-FAST-NEXT: vmovd %xmm0, %eax
; AVX1-FAST-NEXT: retq
;
; AVX2-LABEL: test_v4i32:
; AVX2: # %bb.0:
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
; AVX2-NEXT: vpaddd %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
; AVX2-NEXT: vpaddd %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vmovd %xmm0, %eax
; AVX2-NEXT: retq
;
; AVX512-LABEL: test_v4i32:
; AVX512: # %bb.0:
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
; AVX512-NEXT: vpaddd %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
; AVX512-NEXT: vpaddd %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovd %xmm0, %eax
; AVX512-NEXT: retq
%1 = call i32 @llvm.experimental.vector.reduce.add.v4i32(<4 x i32> %a0)
ret i32 %1
}
define i32 @test_v8i32(<8 x i32> %a0) {
; SSE-LABEL: test_v8i32:
; SSE: # %bb.0:
; SSE-NEXT: paddd %xmm1, %xmm0
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
; SSE-NEXT: paddd %xmm0, %xmm1
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm1[1,1,1,1]
; SSE-NEXT: paddd %xmm1, %xmm0
; SSE-NEXT: movd %xmm0, %eax
; SSE-NEXT: retq
;
; AVX1-SLOW-LABEL: test_v8i32:
; AVX1-SLOW: # %bb.0:
; AVX1-SLOW-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX1-SLOW-NEXT: vpaddd %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
; AVX1-SLOW-NEXT: vpaddd %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
; AVX1-SLOW-NEXT: vpaddd %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT: vmovd %xmm0, %eax
; AVX1-SLOW-NEXT: vzeroupper
; AVX1-SLOW-NEXT: retq
;
; AVX1-FAST-LABEL: test_v8i32:
; AVX1-FAST: # %bb.0:
; AVX1-FAST-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX1-FAST-NEXT: vphaddd %xmm0, %xmm1, %xmm0
; AVX1-FAST-NEXT: vphaddd %xmm0, %xmm0, %xmm0
; AVX1-FAST-NEXT: vphaddd %xmm0, %xmm0, %xmm0
; AVX1-FAST-NEXT: vmovd %xmm0, %eax
; AVX1-FAST-NEXT: vzeroupper
; AVX1-FAST-NEXT: retq
;
; AVX2-LABEL: test_v8i32:
; AVX2: # %bb.0:
; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX2-NEXT: vpaddd %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
; AVX2-NEXT: vpaddd %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
; AVX2-NEXT: vpaddd %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vmovd %xmm0, %eax
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
;
; AVX512-LABEL: test_v8i32:
; AVX512: # %bb.0:
; AVX512-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX512-NEXT: vpaddd %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
; AVX512-NEXT: vpaddd %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
; AVX512-NEXT: vpaddd %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovd %xmm0, %eax
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
%1 = call i32 @llvm.experimental.vector.reduce.add.v8i32(<8 x i32> %a0)
ret i32 %1
}
define i32 @test_v16i32(<16 x i32> %a0) {
; SSE-LABEL: test_v16i32:
; SSE: # %bb.0:
; SSE-NEXT: paddd %xmm3, %xmm1
; SSE-NEXT: paddd %xmm2, %xmm1
; SSE-NEXT: paddd %xmm0, %xmm1
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm1[2,3,2,3]
; SSE-NEXT: paddd %xmm1, %xmm0
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
; SSE-NEXT: paddd %xmm0, %xmm1
; SSE-NEXT: movd %xmm1, %eax
; SSE-NEXT: retq
;
; AVX1-SLOW-LABEL: test_v16i32:
; AVX1-SLOW: # %bb.0:
; AVX1-SLOW-NEXT: vextractf128 $1, %ymm1, %xmm2
; AVX1-SLOW-NEXT: vextractf128 $1, %ymm0, %xmm3
; AVX1-SLOW-NEXT: vpaddd %xmm2, %xmm3, %xmm2
; AVX1-SLOW-NEXT: vpaddd %xmm2, %xmm1, %xmm1
; AVX1-SLOW-NEXT: vpaddd %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
; AVX1-SLOW-NEXT: vpaddd %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
; AVX1-SLOW-NEXT: vpaddd %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT: vmovd %xmm0, %eax
; AVX1-SLOW-NEXT: vzeroupper
; AVX1-SLOW-NEXT: retq
;
; AVX1-FAST-LABEL: test_v16i32:
; AVX1-FAST: # %bb.0:
; AVX1-FAST-NEXT: vextractf128 $1, %ymm1, %xmm2
; AVX1-FAST-NEXT: vextractf128 $1, %ymm0, %xmm3
; AVX1-FAST-NEXT: vpaddd %xmm2, %xmm3, %xmm2
; AVX1-FAST-NEXT: vpaddd %xmm2, %xmm1, %xmm1
; AVX1-FAST-NEXT: vpaddd %xmm1, %xmm0, %xmm0
; AVX1-FAST-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
; AVX1-FAST-NEXT: vpaddd %xmm1, %xmm0, %xmm0
; AVX1-FAST-NEXT: vphaddd %xmm0, %xmm0, %xmm0
; AVX1-FAST-NEXT: vmovd %xmm0, %eax
; AVX1-FAST-NEXT: vzeroupper
; AVX1-FAST-NEXT: retq
;
; AVX2-LABEL: test_v16i32:
; AVX2: # %bb.0:
; AVX2-NEXT: vpaddd %ymm1, %ymm0, %ymm0
; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX2-NEXT: vpaddd %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
; AVX2-NEXT: vpaddd %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
; AVX2-NEXT: vpaddd %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vmovd %xmm0, %eax
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
;
; AVX512-LABEL: test_v16i32:
; AVX512: # %bb.0:
; AVX512-NEXT: vextracti64x4 $1, %zmm0, %ymm1
; AVX512-NEXT: vpaddd %zmm1, %zmm0, %zmm0
; AVX512-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX512-NEXT: vpaddd %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
; AVX512-NEXT: vpaddd %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
; AVX512-NEXT: vpaddd %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovd %xmm0, %eax
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
%1 = call i32 @llvm.experimental.vector.reduce.add.v16i32(<16 x i32> %a0)
ret i32 %1
}
define i32 @test_v32i32(<32 x i32> %a0) {
; SSE-LABEL: test_v32i32:
; SSE: # %bb.0:
; SSE-NEXT: paddd %xmm6, %xmm2
; SSE-NEXT: paddd %xmm7, %xmm3
; SSE-NEXT: paddd %xmm5, %xmm3
; SSE-NEXT: paddd %xmm1, %xmm3
; SSE-NEXT: paddd %xmm4, %xmm2
; SSE-NEXT: paddd %xmm3, %xmm2
; SSE-NEXT: paddd %xmm0, %xmm2
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm2[2,3,2,3]
; SSE-NEXT: paddd %xmm2, %xmm0
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
; SSE-NEXT: paddd %xmm0, %xmm1
; SSE-NEXT: movd %xmm1, %eax
; SSE-NEXT: retq
;
; AVX1-SLOW-LABEL: test_v32i32:
; AVX1-SLOW: # %bb.0:
; AVX1-SLOW-NEXT: vpaddd %xmm3, %xmm1, %xmm4
; AVX1-SLOW-NEXT: vextractf128 $1, %ymm3, %xmm3
; AVX1-SLOW-NEXT: vextractf128 $1, %ymm1, %xmm1
; AVX1-SLOW-NEXT: vpaddd %xmm3, %xmm1, %xmm1
; AVX1-SLOW-NEXT: vextractf128 $1, %ymm2, %xmm3
; AVX1-SLOW-NEXT: vpaddd %xmm1, %xmm3, %xmm1
; AVX1-SLOW-NEXT: vextractf128 $1, %ymm0, %xmm3
; AVX1-SLOW-NEXT: vpaddd %xmm1, %xmm3, %xmm1
; AVX1-SLOW-NEXT: vpaddd %xmm4, %xmm2, %xmm2
; AVX1-SLOW-NEXT: vpaddd %xmm1, %xmm2, %xmm1
; AVX1-SLOW-NEXT: vpaddd %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
; AVX1-SLOW-NEXT: vpaddd %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
; AVX1-SLOW-NEXT: vpaddd %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT: vmovd %xmm0, %eax
; AVX1-SLOW-NEXT: vzeroupper
; AVX1-SLOW-NEXT: retq
;
; AVX1-FAST-LABEL: test_v32i32:
; AVX1-FAST: # %bb.0:
; AVX1-FAST-NEXT: vpaddd %xmm3, %xmm1, %xmm4
; AVX1-FAST-NEXT: vextractf128 $1, %ymm3, %xmm3
; AVX1-FAST-NEXT: vextractf128 $1, %ymm1, %xmm1
; AVX1-FAST-NEXT: vpaddd %xmm3, %xmm1, %xmm1
; AVX1-FAST-NEXT: vextractf128 $1, %ymm2, %xmm3
; AVX1-FAST-NEXT: vpaddd %xmm1, %xmm3, %xmm1
; AVX1-FAST-NEXT: vextractf128 $1, %ymm0, %xmm3
; AVX1-FAST-NEXT: vpaddd %xmm1, %xmm3, %xmm1
; AVX1-FAST-NEXT: vpaddd %xmm4, %xmm2, %xmm2
; AVX1-FAST-NEXT: vpaddd %xmm1, %xmm2, %xmm1
; AVX1-FAST-NEXT: vpaddd %xmm1, %xmm0, %xmm0
; AVX1-FAST-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
; AVX1-FAST-NEXT: vpaddd %xmm1, %xmm0, %xmm0
; AVX1-FAST-NEXT: vphaddd %xmm0, %xmm0, %xmm0
; AVX1-FAST-NEXT: vmovd %xmm0, %eax
; AVX1-FAST-NEXT: vzeroupper
; AVX1-FAST-NEXT: retq
;
; AVX2-LABEL: test_v32i32:
; AVX2: # %bb.0:
; AVX2-NEXT: vpaddd %ymm3, %ymm1, %ymm1
; AVX2-NEXT: vpaddd %ymm1, %ymm2, %ymm1
; AVX2-NEXT: vpaddd %ymm1, %ymm0, %ymm0
; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX2-NEXT: vpaddd %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
; AVX2-NEXT: vpaddd %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
; AVX2-NEXT: vpaddd %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vmovd %xmm0, %eax
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
;
; AVX512-LABEL: test_v32i32:
; AVX512: # %bb.0:
; AVX512-NEXT: vpaddd %zmm1, %zmm0, %zmm0
; AVX512-NEXT: vextracti64x4 $1, %zmm0, %ymm1
; AVX512-NEXT: vpaddd %zmm1, %zmm0, %zmm0
; AVX512-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX512-NEXT: vpaddd %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
; AVX512-NEXT: vpaddd %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
; AVX512-NEXT: vpaddd %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovd %xmm0, %eax
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
%1 = call i32 @llvm.experimental.vector.reduce.add.v32i32(<32 x i32> %a0)
ret i32 %1
}
;
; vXi16
;
define i16 @test_v2i16(<2 x i16> %a0) {
; SSE-LABEL: test_v2i16:
; SSE: # %bb.0:
; SSE-NEXT: movdqa %xmm0, %xmm1
; SSE-NEXT: psrld $16, %xmm1
; SSE-NEXT: paddw %xmm0, %xmm1
; SSE-NEXT: movd %xmm1, %eax
; SSE-NEXT: # kill: def $ax killed $ax killed $eax
; SSE-NEXT: retq
;
; AVX1-SLOW-LABEL: test_v2i16:
; AVX1-SLOW: # %bb.0:
; AVX1-SLOW-NEXT: vpsrld $16, %xmm0, %xmm1
; AVX1-SLOW-NEXT: vpaddw %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT: vmovd %xmm0, %eax
; AVX1-SLOW-NEXT: # kill: def $ax killed $ax killed $eax
; AVX1-SLOW-NEXT: retq
;
; AVX1-FAST-LABEL: test_v2i16:
; AVX1-FAST: # %bb.0:
; AVX1-FAST-NEXT: vphaddw %xmm0, %xmm0, %xmm0
; AVX1-FAST-NEXT: vmovd %xmm0, %eax
; AVX1-FAST-NEXT: # kill: def $ax killed $ax killed $eax
; AVX1-FAST-NEXT: retq
;
; AVX2-LABEL: test_v2i16:
; AVX2: # %bb.0:
; AVX2-NEXT: vpsrld $16, %xmm0, %xmm1
; AVX2-NEXT: vpaddw %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vmovd %xmm0, %eax
; AVX2-NEXT: # kill: def $ax killed $ax killed $eax
; AVX2-NEXT: retq
;
; AVX512-LABEL: test_v2i16:
; AVX512: # %bb.0:
; AVX512-NEXT: vpsrld $16, %xmm0, %xmm1
; AVX512-NEXT: vpaddw %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovd %xmm0, %eax
; AVX512-NEXT: # kill: def $ax killed $ax killed $eax
; AVX512-NEXT: retq
%1 = call i16 @llvm.experimental.vector.reduce.add.v2i16(<2 x i16> %a0)
ret i16 %1
}
define i16 @test_v4i16(<4 x i16> %a0) {
; SSE-LABEL: test_v4i16:
; SSE: # %bb.0:
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
; SSE-NEXT: paddw %xmm0, %xmm1
; SSE-NEXT: movdqa %xmm1, %xmm0
; SSE-NEXT: psrld $16, %xmm0
; SSE-NEXT: paddw %xmm1, %xmm0
; SSE-NEXT: movd %xmm0, %eax
; SSE-NEXT: # kill: def $ax killed $ax killed $eax
; SSE-NEXT: retq
;
; AVX1-SLOW-LABEL: test_v4i16:
; AVX1-SLOW: # %bb.0:
; AVX1-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
; AVX1-SLOW-NEXT: vpaddw %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT: vpsrld $16, %xmm0, %xmm1
; AVX1-SLOW-NEXT: vpaddw %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT: vmovd %xmm0, %eax
; AVX1-SLOW-NEXT: # kill: def $ax killed $ax killed $eax
; AVX1-SLOW-NEXT: retq
;
; AVX1-FAST-LABEL: test_v4i16:
; AVX1-FAST: # %bb.0:
; AVX1-FAST-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
; AVX1-FAST-NEXT: vpaddw %xmm1, %xmm0, %xmm0
; AVX1-FAST-NEXT: vphaddw %xmm0, %xmm0, %xmm0
; AVX1-FAST-NEXT: vmovd %xmm0, %eax
; AVX1-FAST-NEXT: # kill: def $ax killed $ax killed $eax
; AVX1-FAST-NEXT: retq
;
; AVX2-LABEL: test_v4i16:
; AVX2: # %bb.0:
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
; AVX2-NEXT: vpaddw %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vpsrld $16, %xmm0, %xmm1
; AVX2-NEXT: vpaddw %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vmovd %xmm0, %eax
; AVX2-NEXT: # kill: def $ax killed $ax killed $eax
; AVX2-NEXT: retq
;
; AVX512-LABEL: test_v4i16:
; AVX512: # %bb.0:
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
; AVX512-NEXT: vpaddw %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpsrld $16, %xmm0, %xmm1
; AVX512-NEXT: vpaddw %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovd %xmm0, %eax
; AVX512-NEXT: # kill: def $ax killed $ax killed $eax
; AVX512-NEXT: retq
%1 = call i16 @llvm.experimental.vector.reduce.add.v4i16(<4 x i16> %a0)
ret i16 %1
}
define i16 @test_v8i16(<8 x i16> %a0) {
; SSE-LABEL: test_v8i16:
; SSE: # %bb.0:
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
; SSE-NEXT: paddw %xmm0, %xmm1
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm1[1,1,1,1]
; SSE-NEXT: paddw %xmm1, %xmm0
; SSE-NEXT: movdqa %xmm0, %xmm1
; SSE-NEXT: psrld $16, %xmm1
; SSE-NEXT: paddw %xmm0, %xmm1
; SSE-NEXT: movd %xmm1, %eax
; SSE-NEXT: # kill: def $ax killed $ax killed $eax
; SSE-NEXT: retq
;
; AVX1-SLOW-LABEL: test_v8i16:
; AVX1-SLOW: # %bb.0:
; AVX1-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
; AVX1-SLOW-NEXT: vpaddw %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
; AVX1-SLOW-NEXT: vpaddw %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT: vpsrld $16, %xmm0, %xmm1
; AVX1-SLOW-NEXT: vpaddw %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT: vmovd %xmm0, %eax
; AVX1-SLOW-NEXT: # kill: def $ax killed $ax killed $eax
; AVX1-SLOW-NEXT: retq
;
; AVX1-FAST-LABEL: test_v8i16:
; AVX1-FAST: # %bb.0:
; AVX1-FAST-NEXT: vphaddw %xmm0, %xmm0, %xmm0
; AVX1-FAST-NEXT: vphaddw %xmm0, %xmm0, %xmm0
; AVX1-FAST-NEXT: vphaddw %xmm0, %xmm0, %xmm0
; AVX1-FAST-NEXT: vmovd %xmm0, %eax
; AVX1-FAST-NEXT: # kill: def $ax killed $ax killed $eax
; AVX1-FAST-NEXT: retq
;
; AVX2-LABEL: test_v8i16:
; AVX2: # %bb.0:
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
; AVX2-NEXT: vpaddw %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
; AVX2-NEXT: vpaddw %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vpsrld $16, %xmm0, %xmm1
; AVX2-NEXT: vpaddw %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vmovd %xmm0, %eax
; AVX2-NEXT: # kill: def $ax killed $ax killed $eax
; AVX2-NEXT: retq
;
; AVX512-LABEL: test_v8i16:
; AVX512: # %bb.0:
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
; AVX512-NEXT: vpaddw %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
; AVX512-NEXT: vpaddw %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpsrld $16, %xmm0, %xmm1
; AVX512-NEXT: vpaddw %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovd %xmm0, %eax
; AVX512-NEXT: # kill: def $ax killed $ax killed $eax
; AVX512-NEXT: retq
%1 = call i16 @llvm.experimental.vector.reduce.add.v8i16(<8 x i16> %a0)
ret i16 %1
}
define i16 @test_v16i16(<16 x i16> %a0) {
; SSE-LABEL: test_v16i16:
; SSE: # %bb.0:
; SSE-NEXT: paddw %xmm1, %xmm0
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
; SSE-NEXT: paddw %xmm0, %xmm1
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm1[1,1,1,1]
; SSE-NEXT: paddw %xmm1, %xmm0
; SSE-NEXT: movdqa %xmm0, %xmm1
; SSE-NEXT: psrld $16, %xmm1
; SSE-NEXT: paddw %xmm0, %xmm1
; SSE-NEXT: movd %xmm1, %eax
; SSE-NEXT: # kill: def $ax killed $ax killed $eax
; SSE-NEXT: retq
;
; AVX1-SLOW-LABEL: test_v16i16:
; AVX1-SLOW: # %bb.0:
; AVX1-SLOW-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX1-SLOW-NEXT: vpaddw %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
; AVX1-SLOW-NEXT: vpaddw %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
; AVX1-SLOW-NEXT: vpaddw %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT: vpsrld $16, %xmm0, %xmm1
; AVX1-SLOW-NEXT: vpaddw %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT: vmovd %xmm0, %eax
; AVX1-SLOW-NEXT: # kill: def $ax killed $ax killed $eax
; AVX1-SLOW-NEXT: vzeroupper
; AVX1-SLOW-NEXT: retq
;
; AVX1-FAST-LABEL: test_v16i16:
; AVX1-FAST: # %bb.0:
; AVX1-FAST-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX1-FAST-NEXT: vphaddw %xmm0, %xmm1, %xmm0
; AVX1-FAST-NEXT: vphaddw %xmm0, %xmm0, %xmm0
; AVX1-FAST-NEXT: vphaddw %xmm0, %xmm0, %xmm0
; AVX1-FAST-NEXT: vphaddw %xmm0, %xmm0, %xmm0
; AVX1-FAST-NEXT: vmovd %xmm0, %eax
; AVX1-FAST-NEXT: # kill: def $ax killed $ax killed $eax
; AVX1-FAST-NEXT: vzeroupper
; AVX1-FAST-NEXT: retq
;
; AVX2-LABEL: test_v16i16:
; AVX2: # %bb.0:
; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX2-NEXT: vpaddw %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
; AVX2-NEXT: vpaddw %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
; AVX2-NEXT: vpaddw %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vpsrld $16, %xmm0, %xmm1
; AVX2-NEXT: vpaddw %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vmovd %xmm0, %eax
; AVX2-NEXT: # kill: def $ax killed $ax killed $eax
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
;
; AVX512-LABEL: test_v16i16:
; AVX512: # %bb.0:
; AVX512-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX512-NEXT: vpaddw %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
; AVX512-NEXT: vpaddw %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
; AVX512-NEXT: vpaddw %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpsrld $16, %xmm0, %xmm1
; AVX512-NEXT: vpaddw %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovd %xmm0, %eax
; AVX512-NEXT: # kill: def $ax killed $ax killed $eax
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
%1 = call i16 @llvm.experimental.vector.reduce.add.v16i16(<16 x i16> %a0)
ret i16 %1
}
define i16 @test_v32i16(<32 x i16> %a0) {
; SSE-LABEL: test_v32i16:
; SSE: # %bb.0:
; SSE-NEXT: paddw %xmm3, %xmm1
; SSE-NEXT: paddw %xmm2, %xmm1
; SSE-NEXT: paddw %xmm0, %xmm1
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm1[2,3,2,3]
; SSE-NEXT: paddw %xmm1, %xmm0
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
; SSE-NEXT: paddw %xmm0, %xmm1
; SSE-NEXT: movdqa %xmm1, %xmm0
; SSE-NEXT: psrld $16, %xmm0
; SSE-NEXT: paddw %xmm1, %xmm0
; SSE-NEXT: movd %xmm0, %eax
; SSE-NEXT: # kill: def $ax killed $ax killed $eax
; SSE-NEXT: retq
;
; AVX1-SLOW-LABEL: test_v32i16:
; AVX1-SLOW: # %bb.0:
; AVX1-SLOW-NEXT: vextractf128 $1, %ymm1, %xmm2
; AVX1-SLOW-NEXT: vextractf128 $1, %ymm0, %xmm3
; AVX1-SLOW-NEXT: vpaddw %xmm2, %xmm3, %xmm2
; AVX1-SLOW-NEXT: vpaddw %xmm2, %xmm1, %xmm1
; AVX1-SLOW-NEXT: vpaddw %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
; AVX1-SLOW-NEXT: vpaddw %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
; AVX1-SLOW-NEXT: vpaddw %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT: vpsrld $16, %xmm0, %xmm1
; AVX1-SLOW-NEXT: vpaddw %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT: vmovd %xmm0, %eax
; AVX1-SLOW-NEXT: # kill: def $ax killed $ax killed $eax
; AVX1-SLOW-NEXT: vzeroupper
; AVX1-SLOW-NEXT: retq
;
; AVX1-FAST-LABEL: test_v32i16:
; AVX1-FAST: # %bb.0:
; AVX1-FAST-NEXT: vextractf128 $1, %ymm1, %xmm2
; AVX1-FAST-NEXT: vextractf128 $1, %ymm0, %xmm3
; AVX1-FAST-NEXT: vpaddw %xmm2, %xmm3, %xmm2
; AVX1-FAST-NEXT: vpaddw %xmm2, %xmm1, %xmm1
; AVX1-FAST-NEXT: vpaddw %xmm1, %xmm0, %xmm0
; AVX1-FAST-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
; AVX1-FAST-NEXT: vpaddw %xmm1, %xmm0, %xmm0
; AVX1-FAST-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
; AVX1-FAST-NEXT: vpaddw %xmm1, %xmm0, %xmm0
; AVX1-FAST-NEXT: vphaddw %xmm0, %xmm0, %xmm0
; AVX1-FAST-NEXT: vmovd %xmm0, %eax
; AVX1-FAST-NEXT: # kill: def $ax killed $ax killed $eax
; AVX1-FAST-NEXT: vzeroupper
; AVX1-FAST-NEXT: retq
;
; AVX2-LABEL: test_v32i16:
; AVX2: # %bb.0:
; AVX2-NEXT: vpaddw %ymm1, %ymm0, %ymm0
; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX2-NEXT: vpaddw %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
; AVX2-NEXT: vpaddw %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
; AVX2-NEXT: vpaddw %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vpsrld $16, %xmm0, %xmm1
; AVX2-NEXT: vpaddw %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vmovd %xmm0, %eax
; AVX2-NEXT: # kill: def $ax killed $ax killed $eax
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
;
; AVX512-LABEL: test_v32i16:
; AVX512: # %bb.0:
; AVX512-NEXT: vextracti64x4 $1, %zmm0, %ymm1
; AVX512-NEXT: vpaddw %zmm1, %zmm0, %zmm0
; AVX512-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX512-NEXT: vpaddw %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
; AVX512-NEXT: vpaddw %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
; AVX512-NEXT: vpaddw %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpsrld $16, %xmm0, %xmm1
; AVX512-NEXT: vpaddw %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovd %xmm0, %eax
; AVX512-NEXT: # kill: def $ax killed $ax killed $eax
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
%1 = call i16 @llvm.experimental.vector.reduce.add.v32i16(<32 x i16> %a0)
ret i16 %1
}
define i16 @test_v64i16(<64 x i16> %a0) {
; SSE-LABEL: test_v64i16:
; SSE: # %bb.0:
; SSE-NEXT: paddw %xmm6, %xmm2
; SSE-NEXT: paddw %xmm7, %xmm3
; SSE-NEXT: paddw %xmm5, %xmm3
; SSE-NEXT: paddw %xmm1, %xmm3
; SSE-NEXT: paddw %xmm4, %xmm2
; SSE-NEXT: paddw %xmm3, %xmm2
; SSE-NEXT: paddw %xmm0, %xmm2
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm2[2,3,2,3]
; SSE-NEXT: paddw %xmm2, %xmm0
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
; SSE-NEXT: paddw %xmm0, %xmm1
; SSE-NEXT: movdqa %xmm1, %xmm0
; SSE-NEXT: psrld $16, %xmm0
; SSE-NEXT: paddw %xmm1, %xmm0
; SSE-NEXT: movd %xmm0, %eax
; SSE-NEXT: # kill: def $ax killed $ax killed $eax
; SSE-NEXT: retq
;
; AVX1-SLOW-LABEL: test_v64i16:
; AVX1-SLOW: # %bb.0:
; AVX1-SLOW-NEXT: vpaddw %xmm3, %xmm1, %xmm4
; AVX1-SLOW-NEXT: vextractf128 $1, %ymm3, %xmm3
; AVX1-SLOW-NEXT: vextractf128 $1, %ymm1, %xmm1
; AVX1-SLOW-NEXT: vpaddw %xmm3, %xmm1, %xmm1
; AVX1-SLOW-NEXT: vextractf128 $1, %ymm2, %xmm3
; AVX1-SLOW-NEXT: vpaddw %xmm1, %xmm3, %xmm1
; AVX1-SLOW-NEXT: vextractf128 $1, %ymm0, %xmm3
; AVX1-SLOW-NEXT: vpaddw %xmm1, %xmm3, %xmm1
; AVX1-SLOW-NEXT: vpaddw %xmm4, %xmm2, %xmm2
; AVX1-SLOW-NEXT: vpaddw %xmm1, %xmm2, %xmm1
; AVX1-SLOW-NEXT: vpaddw %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
; AVX1-SLOW-NEXT: vpaddw %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
; AVX1-SLOW-NEXT: vpaddw %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT: vpsrld $16, %xmm0, %xmm1
; AVX1-SLOW-NEXT: vpaddw %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT: vmovd %xmm0, %eax
; AVX1-SLOW-NEXT: # kill: def $ax killed $ax killed $eax
; AVX1-SLOW-NEXT: vzeroupper
; AVX1-SLOW-NEXT: retq
;
; AVX1-FAST-LABEL: test_v64i16:
; AVX1-FAST: # %bb.0:
; AVX1-FAST-NEXT: vpaddw %xmm3, %xmm1, %xmm4
; AVX1-FAST-NEXT: vextractf128 $1, %ymm3, %xmm3
; AVX1-FAST-NEXT: vextractf128 $1, %ymm1, %xmm1
; AVX1-FAST-NEXT: vpaddw %xmm3, %xmm1, %xmm1
; AVX1-FAST-NEXT: vextractf128 $1, %ymm2, %xmm3
; AVX1-FAST-NEXT: vpaddw %xmm1, %xmm3, %xmm1
; AVX1-FAST-NEXT: vextractf128 $1, %ymm0, %xmm3
; AVX1-FAST-NEXT: vpaddw %xmm1, %xmm3, %xmm1
; AVX1-FAST-NEXT: vpaddw %xmm4, %xmm2, %xmm2
; AVX1-FAST-NEXT: vpaddw %xmm1, %xmm2, %xmm1
; AVX1-FAST-NEXT: vpaddw %xmm1, %xmm0, %xmm0
; AVX1-FAST-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
; AVX1-FAST-NEXT: vpaddw %xmm1, %xmm0, %xmm0
; AVX1-FAST-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
; AVX1-FAST-NEXT: vpaddw %xmm1, %xmm0, %xmm0
; AVX1-FAST-NEXT: vphaddw %xmm0, %xmm0, %xmm0
; AVX1-FAST-NEXT: vmovd %xmm0, %eax
; AVX1-FAST-NEXT: # kill: def $ax killed $ax killed $eax
; AVX1-FAST-NEXT: vzeroupper
; AVX1-FAST-NEXT: retq
;
; AVX2-LABEL: test_v64i16:
; AVX2: # %bb.0:
; AVX2-NEXT: vpaddw %ymm3, %ymm1, %ymm1
; AVX2-NEXT: vpaddw %ymm1, %ymm2, %ymm1
; AVX2-NEXT: vpaddw %ymm1, %ymm0, %ymm0
; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX2-NEXT: vpaddw %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
; AVX2-NEXT: vpaddw %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
; AVX2-NEXT: vpaddw %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vpsrld $16, %xmm0, %xmm1
; AVX2-NEXT: vpaddw %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vmovd %xmm0, %eax
; AVX2-NEXT: # kill: def $ax killed $ax killed $eax
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
;
; AVX512-LABEL: test_v64i16:
; AVX512: # %bb.0:
; AVX512-NEXT: vpaddw %zmm1, %zmm0, %zmm0
; AVX512-NEXT: vextracti64x4 $1, %zmm0, %ymm1
; AVX512-NEXT: vpaddw %zmm1, %zmm0, %zmm0
; AVX512-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX512-NEXT: vpaddw %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
; AVX512-NEXT: vpaddw %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
; AVX512-NEXT: vpaddw %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpsrld $16, %xmm0, %xmm1
; AVX512-NEXT: vpaddw %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovd %xmm0, %eax
; AVX512-NEXT: # kill: def $ax killed $ax killed $eax
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
%1 = call i16 @llvm.experimental.vector.reduce.add.v64i16(<64 x i16> %a0)
ret i16 %1
}
;
; vXi8
;
define i8 @test_v2i8(<2 x i8> %a0) {
; SSE-LABEL: test_v2i8:
; SSE: # %bb.0:
; SSE-NEXT: movdqa %xmm0, %xmm1
; SSE-NEXT: psrlw $8, %xmm1
; SSE-NEXT: paddb %xmm0, %xmm1
; SSE-NEXT: movd %xmm1, %eax
; SSE-NEXT: # kill: def $al killed $al killed $eax
; SSE-NEXT: retq
;
; AVX-LABEL: test_v2i8:
; AVX: # %bb.0:
; AVX-NEXT: vpsrlw $8, %xmm0, %xmm1
; AVX-NEXT: vpaddb %xmm1, %xmm0, %xmm0
; AVX-NEXT: vmovd %xmm0, %eax
; AVX-NEXT: # kill: def $al killed $al killed $eax
; AVX-NEXT: retq
;
; AVX512-LABEL: test_v2i8:
; AVX512: # %bb.0:
; AVX512-NEXT: vpsrlw $8, %xmm0, %xmm1
; AVX512-NEXT: vpaddb %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovd %xmm0, %eax
; AVX512-NEXT: # kill: def $al killed $al killed $eax
; AVX512-NEXT: retq
%1 = call i8 @llvm.experimental.vector.reduce.add.v2i8(<2 x i8> %a0)
ret i8 %1
}
define i8 @test_v2i8_load(<2 x i8>* %p) {
; SSE-LABEL: test_v2i8_load:
; SSE: # %bb.0:
; SSE-NEXT: movzwl (%rdi), %eax
; SSE-NEXT: movd %eax, %xmm0
; SSE-NEXT: movdqa %xmm0, %xmm1
; SSE-NEXT: psrlw $8, %xmm1
; SSE-NEXT: paddb %xmm0, %xmm1
; SSE-NEXT: movd %xmm1, %eax
; SSE-NEXT: # kill: def $al killed $al killed $eax
; SSE-NEXT: retq
;
; AVX-LABEL: test_v2i8_load:
; AVX: # %bb.0:
; AVX-NEXT: movzwl (%rdi), %eax
; AVX-NEXT: vmovd %eax, %xmm0
; AVX-NEXT: vpsrlw $8, %xmm0, %xmm1
; AVX-NEXT: vpaddb %xmm1, %xmm0, %xmm0
; AVX-NEXT: vmovd %xmm0, %eax
; AVX-NEXT: # kill: def $al killed $al killed $eax
; AVX-NEXT: retq
;
; AVX512-LABEL: test_v2i8_load:
; AVX512: # %bb.0:
; AVX512-NEXT: movzwl (%rdi), %eax
; AVX512-NEXT: vmovd %eax, %xmm0
; AVX512-NEXT: vpsrlw $8, %xmm0, %xmm1
; AVX512-NEXT: vpaddb %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovd %xmm0, %eax
; AVX512-NEXT: # kill: def $al killed $al killed $eax
; AVX512-NEXT: retq
%a0 = load <2 x i8>, <2 x i8>* %p
%1 = call i8 @llvm.experimental.vector.reduce.add.v2i8(<2 x i8> %a0)
ret i8 %1
}
define i8 @test_v4i8(<4 x i8> %a0) {
; SSE2-LABEL: test_v4i8:
; SSE2: # %bb.0:
; SSE2-NEXT: pxor %xmm1, %xmm1
; SSE2-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE2-NEXT: psadbw %xmm1, %xmm0
; SSE2-NEXT: movd %xmm0, %eax
; SSE2-NEXT: # kill: def $al killed $al killed $eax
; SSE2-NEXT: retq
;
; SSE41-LABEL: test_v4i8:
; SSE41: # %bb.0:
; SSE41-NEXT: pxor %xmm1, %xmm1
; SSE41-NEXT: pblendw {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3,4,5,6,7]
; SSE41-NEXT: psadbw %xmm1, %xmm0
; SSE41-NEXT: movd %xmm0, %eax
; SSE41-NEXT: # kill: def $al killed $al killed $eax
; SSE41-NEXT: retq
;
; AVX-LABEL: test_v4i8:
; AVX: # %bb.0:
; AVX-NEXT: vpxor %xmm1, %xmm1, %xmm1
; AVX-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3,4,5,6,7]
; AVX-NEXT: vpsadbw %xmm1, %xmm0, %xmm0
; AVX-NEXT: vmovd %xmm0, %eax
; AVX-NEXT: # kill: def $al killed $al killed $eax
; AVX-NEXT: retq
;
; AVX512-LABEL: test_v4i8:
; AVX512: # %bb.0:
; AVX512-NEXT: vpxor %xmm1, %xmm1, %xmm1
; AVX512-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3,4,5,6,7]
; AVX512-NEXT: vpxor %xmm1, %xmm1, %xmm1
; AVX512-NEXT: vpsadbw %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovd %xmm0, %eax
; AVX512-NEXT: # kill: def $al killed $al killed $eax
; AVX512-NEXT: retq
%1 = call i8 @llvm.experimental.vector.reduce.add.v4i8(<4 x i8> %a0)
ret i8 %1
}
define i8 @test_v4i8_load(<4 x i8>* %p) {
; SSE-LABEL: test_v4i8_load:
; SSE: # %bb.0:
; SSE-NEXT: movd {{.*#+}} xmm0 = mem[0],zero,zero,zero
; SSE-NEXT: pxor %xmm1, %xmm1
; SSE-NEXT: psadbw %xmm0, %xmm1
; SSE-NEXT: movd %xmm1, %eax
; SSE-NEXT: # kill: def $al killed $al killed $eax
; SSE-NEXT: retq
;
; AVX-LABEL: test_v4i8_load:
; AVX: # %bb.0:
; AVX-NEXT: vmovd {{.*#+}} xmm0 = mem[0],zero,zero,zero
; AVX-NEXT: vpxor %xmm1, %xmm1, %xmm1
; AVX-NEXT: vpsadbw %xmm1, %xmm0, %xmm0
; AVX-NEXT: vmovd %xmm0, %eax
; AVX-NEXT: # kill: def $al killed $al killed $eax
; AVX-NEXT: retq
;
; AVX512-LABEL: test_v4i8_load:
; AVX512: # %bb.0:
; AVX512-NEXT: vmovd {{.*#+}} xmm0 = mem[0],zero,zero,zero
; AVX512-NEXT: vpxor %xmm1, %xmm1, %xmm1
; AVX512-NEXT: vpsadbw %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovd %xmm0, %eax
; AVX512-NEXT: # kill: def $al killed $al killed $eax
; AVX512-NEXT: retq
%a0 = load <4 x i8>, <4 x i8>* %p
%1 = call i8 @llvm.experimental.vector.reduce.add.v4i8(<4 x i8> %a0)
ret i8 %1
}
define i8 @test_v8i8(<8 x i8> %a0) {
; SSE-LABEL: test_v8i8:
; SSE: # %bb.0:
; SSE-NEXT: pxor %xmm1, %xmm1
; SSE-NEXT: psadbw %xmm0, %xmm1
; SSE-NEXT: movd %xmm1, %eax
; SSE-NEXT: # kill: def $al killed $al killed $eax
; SSE-NEXT: retq
;
; AVX-LABEL: test_v8i8:
; AVX: # %bb.0:
; AVX-NEXT: vpxor %xmm1, %xmm1, %xmm1
; AVX-NEXT: vpsadbw %xmm1, %xmm0, %xmm0
; AVX-NEXT: vmovd %xmm0, %eax
; AVX-NEXT: # kill: def $al killed $al killed $eax
; AVX-NEXT: retq
;
; AVX512-LABEL: test_v8i8:
; AVX512: # %bb.0:
; AVX512-NEXT: vpxor %xmm1, %xmm1, %xmm1
; AVX512-NEXT: vpsadbw %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovd %xmm0, %eax
; AVX512-NEXT: # kill: def $al killed $al killed $eax
; AVX512-NEXT: retq
%1 = call i8 @llvm.experimental.vector.reduce.add.v8i8(<8 x i8> %a0)
ret i8 %1
}
define i8 @test_v8i8_load(<8 x i8>* %p) {
; SSE-LABEL: test_v8i8_load:
; SSE: # %bb.0:
; SSE-NEXT: movq {{.*#+}} xmm0 = mem[0],zero
; SSE-NEXT: pxor %xmm1, %xmm1
; SSE-NEXT: psadbw %xmm0, %xmm1
; SSE-NEXT: movd %xmm1, %eax
; SSE-NEXT: # kill: def $al killed $al killed $eax
; SSE-NEXT: retq
;
; AVX-LABEL: test_v8i8_load:
; AVX: # %bb.0:
; AVX-NEXT: vmovq {{.*#+}} xmm0 = mem[0],zero
; AVX-NEXT: vpxor %xmm1, %xmm1, %xmm1
; AVX-NEXT: vpsadbw %xmm1, %xmm0, %xmm0
; AVX-NEXT: vmovd %xmm0, %eax
; AVX-NEXT: # kill: def $al killed $al killed $eax
; AVX-NEXT: retq
;
; AVX512-LABEL: test_v8i8_load:
; AVX512: # %bb.0:
; AVX512-NEXT: vmovq {{.*#+}} xmm0 = mem[0],zero
; AVX512-NEXT: vpxor %xmm1, %xmm1, %xmm1
; AVX512-NEXT: vpsadbw %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovd %xmm0, %eax
; AVX512-NEXT: # kill: def $al killed $al killed $eax
; AVX512-NEXT: retq
%a0 = load <8 x i8>, <8 x i8>* %p
%1 = call i8 @llvm.experimental.vector.reduce.add.v8i8(<8 x i8> %a0)
ret i8 %1
}
define i8 @test_v16i8(<16 x i8> %a0) {
; SSE-LABEL: test_v16i8:
; SSE: # %bb.0:
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
; SSE-NEXT: paddb %xmm0, %xmm1
; SSE-NEXT: pxor %xmm0, %xmm0
; SSE-NEXT: psadbw %xmm1, %xmm0
; SSE-NEXT: movd %xmm0, %eax
; SSE-NEXT: # kill: def $al killed $al killed $eax
; SSE-NEXT: retq
;
; AVX-LABEL: test_v16i8:
; AVX: # %bb.0:
; AVX-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
; AVX-NEXT: vpaddb %xmm1, %xmm0, %xmm0
; AVX-NEXT: vpxor %xmm1, %xmm1, %xmm1
; AVX-NEXT: vpsadbw %xmm1, %xmm0, %xmm0
; AVX-NEXT: vmovd %xmm0, %eax
; AVX-NEXT: # kill: def $al killed $al killed $eax
; AVX-NEXT: retq
;
; AVX512-LABEL: test_v16i8:
; AVX512: # %bb.0:
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
; AVX512-NEXT: vpaddb %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpxor %xmm1, %xmm1, %xmm1
; AVX512-NEXT: vpsadbw %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovd %xmm0, %eax
; AVX512-NEXT: # kill: def $al killed $al killed $eax
; AVX512-NEXT: retq
%1 = call i8 @llvm.experimental.vector.reduce.add.v16i8(<16 x i8> %a0)
ret i8 %1
}
define i8 @test_v32i8(<32 x i8> %a0) {
; SSE-LABEL: test_v32i8:
; SSE: # %bb.0:
; SSE-NEXT: paddb %xmm1, %xmm0
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
; SSE-NEXT: paddb %xmm0, %xmm1
; SSE-NEXT: pxor %xmm0, %xmm0
; SSE-NEXT: psadbw %xmm1, %xmm0
; SSE-NEXT: movd %xmm0, %eax
; SSE-NEXT: # kill: def $al killed $al killed $eax
; SSE-NEXT: retq
;
; AVX1-LABEL: test_v32i8:
; AVX1: # %bb.0:
; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX1-NEXT: vpaddb %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
; AVX1-NEXT: vpaddb %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vpxor %xmm1, %xmm1, %xmm1
; AVX1-NEXT: vpsadbw %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vmovd %xmm0, %eax
; AVX1-NEXT: # kill: def $al killed $al killed $eax
; AVX1-NEXT: vzeroupper
; AVX1-NEXT: retq
;
; AVX2-LABEL: test_v32i8:
; AVX2: # %bb.0:
; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX2-NEXT: vpaddb %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
; AVX2-NEXT: vpaddb %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vpxor %xmm1, %xmm1, %xmm1
; AVX2-NEXT: vpsadbw %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vmovd %xmm0, %eax
; AVX2-NEXT: # kill: def $al killed $al killed $eax
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
;
; AVX512-LABEL: test_v32i8:
; AVX512: # %bb.0:
; AVX512-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX512-NEXT: vpaddb %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
; AVX512-NEXT: vpaddb %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpxor %xmm1, %xmm1, %xmm1
; AVX512-NEXT: vpsadbw %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovd %xmm0, %eax
; AVX512-NEXT: # kill: def $al killed $al killed $eax
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
%1 = call i8 @llvm.experimental.vector.reduce.add.v32i8(<32 x i8> %a0)
ret i8 %1
}
define i8 @test_v64i8(<64 x i8> %a0) {
; SSE-LABEL: test_v64i8:
; SSE: # %bb.0:
; SSE-NEXT: paddb %xmm3, %xmm1
; SSE-NEXT: paddb %xmm2, %xmm1
; SSE-NEXT: paddb %xmm0, %xmm1
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm1[2,3,2,3]
; SSE-NEXT: paddb %xmm1, %xmm0
; SSE-NEXT: pxor %xmm1, %xmm1
; SSE-NEXT: psadbw %xmm0, %xmm1
; SSE-NEXT: movd %xmm1, %eax
; SSE-NEXT: # kill: def $al killed $al killed $eax
; SSE-NEXT: retq
;
; AVX1-LABEL: test_v64i8:
; AVX1: # %bb.0:
; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm2
; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm3
; AVX1-NEXT: vpaddb %xmm2, %xmm3, %xmm2
; AVX1-NEXT: vpaddb %xmm2, %xmm1, %xmm1
; AVX1-NEXT: vpaddb %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
; AVX1-NEXT: vpaddb %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vpxor %xmm1, %xmm1, %xmm1
; AVX1-NEXT: vpsadbw %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vmovd %xmm0, %eax
; AVX1-NEXT: # kill: def $al killed $al killed $eax
; AVX1-NEXT: vzeroupper
; AVX1-NEXT: retq
;
; AVX2-LABEL: test_v64i8:
; AVX2: # %bb.0:
; AVX2-NEXT: vpaddb %ymm1, %ymm0, %ymm0
; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX2-NEXT: vpaddb %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
; AVX2-NEXT: vpaddb %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vpxor %xmm1, %xmm1, %xmm1
; AVX2-NEXT: vpsadbw %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vmovd %xmm0, %eax
; AVX2-NEXT: # kill: def $al killed $al killed $eax
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
;
; AVX512-LABEL: test_v64i8:
; AVX512: # %bb.0:
; AVX512-NEXT: vextracti64x4 $1, %zmm0, %ymm1
; AVX512-NEXT: vpaddb %ymm1, %ymm0, %ymm0
; AVX512-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX512-NEXT: vpaddb %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
; AVX512-NEXT: vpaddb %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpxor %xmm1, %xmm1, %xmm1
; AVX512-NEXT: vpsadbw %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovd %xmm0, %eax
; AVX512-NEXT: # kill: def $al killed $al killed $eax
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
%1 = call i8 @llvm.experimental.vector.reduce.add.v64i8(<64 x i8> %a0)
ret i8 %1
}
define i8 @test_v128i8(<128 x i8> %a0) {
; SSE-LABEL: test_v128i8:
; SSE: # %bb.0:
; SSE-NEXT: paddb %xmm7, %xmm3
; SSE-NEXT: paddb %xmm5, %xmm3
; SSE-NEXT: paddb %xmm1, %xmm3
; SSE-NEXT: paddb %xmm6, %xmm2
; SSE-NEXT: paddb %xmm4, %xmm2
; SSE-NEXT: paddb %xmm3, %xmm2
; SSE-NEXT: paddb %xmm0, %xmm2
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm2[2,3,2,3]
; SSE-NEXT: paddb %xmm2, %xmm0
; SSE-NEXT: pxor %xmm1, %xmm1
; SSE-NEXT: psadbw %xmm0, %xmm1
; SSE-NEXT: movd %xmm1, %eax
; SSE-NEXT: # kill: def $al killed $al killed $eax
; SSE-NEXT: retq
;
; AVX1-LABEL: test_v128i8:
; AVX1: # %bb.0:
; AVX1-NEXT: vextractf128 $1, %ymm3, %xmm4
; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm5
; AVX1-NEXT: vpaddb %xmm4, %xmm5, %xmm4
; AVX1-NEXT: vextractf128 $1, %ymm2, %xmm5
; AVX1-NEXT: vpaddb %xmm4, %xmm5, %xmm4
; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm5
; AVX1-NEXT: vpaddb %xmm4, %xmm5, %xmm4
; AVX1-NEXT: vpaddb %xmm3, %xmm1, %xmm1
; AVX1-NEXT: vpaddb %xmm1, %xmm2, %xmm1
; AVX1-NEXT: vpaddb %xmm4, %xmm1, %xmm1
; AVX1-NEXT: vpaddb %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
; AVX1-NEXT: vpaddb %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vpxor %xmm1, %xmm1, %xmm1
; AVX1-NEXT: vpsadbw %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vmovd %xmm0, %eax
; AVX1-NEXT: # kill: def $al killed $al killed $eax
; AVX1-NEXT: vzeroupper
; AVX1-NEXT: retq
;
; AVX2-LABEL: test_v128i8:
; AVX2: # %bb.0:
; AVX2-NEXT: vpaddb %ymm3, %ymm1, %ymm1
; AVX2-NEXT: vpaddb %ymm1, %ymm2, %ymm1
; AVX2-NEXT: vpaddb %ymm1, %ymm0, %ymm0
; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX2-NEXT: vpaddb %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
; AVX2-NEXT: vpaddb %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vpxor %xmm1, %xmm1, %xmm1
; AVX2-NEXT: vpsadbw %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vmovd %xmm0, %eax
; AVX2-NEXT: # kill: def $al killed $al killed $eax
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
;
; AVX512-LABEL: test_v128i8:
; AVX512: # %bb.0:
; AVX512-NEXT: vpaddb %zmm1, %zmm0, %zmm0
; AVX512-NEXT: vextracti64x4 $1, %zmm0, %ymm1
; AVX512-NEXT: vpaddb %ymm1, %ymm0, %ymm0
; AVX512-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX512-NEXT: vpaddb %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
; AVX512-NEXT: vpaddb %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vpxor %xmm1, %xmm1, %xmm1
; AVX512-NEXT: vpsadbw %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovd %xmm0, %eax
; AVX512-NEXT: # kill: def $al killed $al killed $eax
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
%1 = call i8 @llvm.experimental.vector.reduce.add.v128i8(<128 x i8> %a0)
ret i8 %1
}
declare i64 @llvm.experimental.vector.reduce.add.v2i64(<2 x i64>)
declare i64 @llvm.experimental.vector.reduce.add.v4i64(<4 x i64>)
declare i64 @llvm.experimental.vector.reduce.add.v8i64(<8 x i64>)
declare i64 @llvm.experimental.vector.reduce.add.v16i64(<16 x i64>)
declare i32 @llvm.experimental.vector.reduce.add.v2i32(<2 x i32>)
declare i32 @llvm.experimental.vector.reduce.add.v4i32(<4 x i32>)
declare i32 @llvm.experimental.vector.reduce.add.v8i32(<8 x i32>)
declare i32 @llvm.experimental.vector.reduce.add.v16i32(<16 x i32>)
declare i32 @llvm.experimental.vector.reduce.add.v32i32(<32 x i32>)
declare i16 @llvm.experimental.vector.reduce.add.v2i16(<2 x i16>)
declare i16 @llvm.experimental.vector.reduce.add.v4i16(<4 x i16>)
declare i16 @llvm.experimental.vector.reduce.add.v8i16(<8 x i16>)
declare i16 @llvm.experimental.vector.reduce.add.v16i16(<16 x i16>)
declare i16 @llvm.experimental.vector.reduce.add.v32i16(<32 x i16>)
declare i16 @llvm.experimental.vector.reduce.add.v64i16(<64 x i16>)
declare i8 @llvm.experimental.vector.reduce.add.v2i8(<2 x i8>)
declare i8 @llvm.experimental.vector.reduce.add.v4i8(<4 x i8>)
declare i8 @llvm.experimental.vector.reduce.add.v8i8(<8 x i8>)
declare i8 @llvm.experimental.vector.reduce.add.v16i8(<16 x i8>)
declare i8 @llvm.experimental.vector.reduce.add.v32i8(<32 x i8>)
declare i8 @llvm.experimental.vector.reduce.add.v64i8(<64 x i8>)
declare i8 @llvm.experimental.vector.reduce.add.v128i8(<128 x i8>)
Reference in New Issue View Git Blame Copy Permalink