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
c9a75c816c
llvm-mirror/test/CodeGen/X86/vector-shuffle-variable-256.ll

722 lines
31 KiB
LLVM
Raw Normal View History

Revert "In visitSTORE, always use FindBetterChain, rather than only when UseAA is enabled." This reverts commit r282600 due to test failues with MCJIT llvm-svn: 282604
2016-09-28 18:37:50 +02:00
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
[X86][SSE] Some basic tests for variable shuffles We don't really support non-constant shuffle masks, but these tests are for cases where BUILD_VECTOR is made up from vector extracts (as well as undef/zero scalars). llvm-svn: 265045
2016-03-31 22:26:30 +02:00
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mcpu=x86-64 -mattr=+avx | FileCheck %s --check-prefix=ALL --check-prefix=AVX --check-prefix=AVX1
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mcpu=x86-64 -mattr=+avx2 | FileCheck %s --check-prefix=ALL --check-prefix=AVX --check-prefix=AVX2
;
; Unary shuffle indices from registers
;
define <4 x double> @var_shuffle_v4f64_v4f64_xxxx_i64(<4 x double> %x, i64 %i0, i64 %i1, i64 %i2, i64 %i3) nounwind {
; ALL-LABEL: var_shuffle_v4f64_v4f64_xxxx_i64:
; ALL: # BB#0:
; ALL-NEXT: pushq %rbp
; ALL-NEXT: movq %rsp, %rbp
; ALL-NEXT: andq $-32, %rsp
; ALL-NEXT: subq $64, %rsp
; ALL-NEXT: vmovaps %ymm0, (%rsp)
; ALL-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; ALL-NEXT: vmovhpd {{.*#+}} xmm0 = xmm0[0],mem[0]
; ALL-NEXT: vmovsd {{.*#+}} xmm1 = mem[0],zero
; ALL-NEXT: vmovhpd {{.*#+}} xmm1 = xmm1[0],mem[0]
Revert "In visitSTORE, always use FindBetterChain, rather than only when UseAA is enabled." This reverts commit r282600 due to test failues with MCJIT llvm-svn: 282604
2016-09-28 18:37:50 +02:00
; ALL-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
[X86][SSE] Some basic tests for variable shuffles We don't really support non-constant shuffle masks, but these tests are for cases where BUILD_VECTOR is made up from vector extracts (as well as undef/zero scalars). llvm-svn: 265045
2016-03-31 22:26:30 +02:00
; ALL-NEXT: movq %rbp, %rsp
; ALL-NEXT: popq %rbp
; ALL-NEXT: retq
%x0 = extractelement <4 x double> %x, i64 %i0
%x1 = extractelement <4 x double> %x, i64 %i1
%x2 = extractelement <4 x double> %x, i64 %i2
%x3 = extractelement <4 x double> %x, i64 %i3
%r0 = insertelement <4 x double> undef, double %x0, i32 0
%r1 = insertelement <4 x double> %r0, double %x1, i32 1
%r2 = insertelement <4 x double> %r1, double %x2, i32 2
%r3 = insertelement <4 x double> %r2, double %x3, i32 3
ret <4 x double> %r3
}
define <4 x double> @var_shuffle_v4f64_v4f64_uxx0_i64(<4 x double> %x, i64 %i0, i64 %i1, i64 %i2, i64 %i3) nounwind {
; ALL-LABEL: var_shuffle_v4f64_v4f64_uxx0_i64:
; ALL: # BB#0:
; ALL-NEXT: pushq %rbp
; ALL-NEXT: movq %rsp, %rbp
; ALL-NEXT: andq $-32, %rsp
; ALL-NEXT: subq $64, %rsp
; ALL-NEXT: vmovaps %ymm0, (%rsp)
[X86][SSE] Add support for combining VZEXT_MOVL target shuffles Includes adding more general support for the pattern: VZEXT_MOVL(VZEXT_LOAD(ptr)) -> VZEXT_LOAD(ptr) This has unearthed a couple of latent poor codegen issues (MINSS/MAXSS scalar load folding and MOVDDUP/BROADCAST load folding patterns), which will be fixed shortly. Its also reduced a couple of tests so that they no longer reach the instruction threshold necessary to be combined to PSHUFB (see PR26183). llvm-svn: 279646
2016-08-24 20:07:53 +02:00
; ALL-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; ALL-NEXT: vmovddup {{.*#+}} xmm0 = xmm0[0,0]
[X86][SSE] Some basic tests for variable shuffles We don't really support non-constant shuffle masks, but these tests are for cases where BUILD_VECTOR is made up from vector extracts (as well as undef/zero scalars). llvm-svn: 265045
2016-03-31 22:26:30 +02:00
; ALL-NEXT: vmovsd {{.*#+}} xmm1 = mem[0],zero
; ALL-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
; ALL-NEXT: movq %rbp, %rsp
; ALL-NEXT: popq %rbp
; ALL-NEXT: retq
%x0 = extractelement <4 x double> %x, i64 %i0
%x1 = extractelement <4 x double> %x, i64 %i1
%x2 = extractelement <4 x double> %x, i64 %i2
%x3 = extractelement <4 x double> %x, i64 %i3
%r0 = insertelement <4 x double> undef, double undef, i32 0
%r1 = insertelement <4 x double> %r0, double %x1, i32 1
%r2 = insertelement <4 x double> %r1, double %x2, i32 2
%r3 = insertelement <4 x double> %r2, double 0.0, i32 3
ret <4 x double> %r3
}
define <4 x double> @var_shuffle_v4f64_v2f64_xxxx_i64(<2 x double> %x, i64 %i0, i64 %i1, i64 %i2, i64 %i3) nounwind {
; ALL-LABEL: var_shuffle_v4f64_v2f64_xxxx_i64:
; ALL: # BB#0:
; ALL-NEXT: vmovaps %xmm0, -{{[0-9]+}}(%rsp)
; ALL-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; ALL-NEXT: vmovhpd {{.*#+}} xmm0 = xmm0[0],mem[0]
; ALL-NEXT: vmovsd {{.*#+}} xmm1 = mem[0],zero
; ALL-NEXT: vmovhpd {{.*#+}} xmm1 = xmm1[0],mem[0]
Revert "In visitSTORE, always use FindBetterChain, rather than only when UseAA is enabled." This reverts commit r282600 due to test failues with MCJIT llvm-svn: 282604
2016-09-28 18:37:50 +02:00
; ALL-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
[X86][SSE] Some basic tests for variable shuffles We don't really support non-constant shuffle masks, but these tests are for cases where BUILD_VECTOR is made up from vector extracts (as well as undef/zero scalars). llvm-svn: 265045
2016-03-31 22:26:30 +02:00
; ALL-NEXT: retq
%x0 = extractelement <2 x double> %x, i64 %i0
%x1 = extractelement <2 x double> %x, i64 %i1
%x2 = extractelement <2 x double> %x, i64 %i2
%x3 = extractelement <2 x double> %x, i64 %i3
%r0 = insertelement <4 x double> undef, double %x0, i32 0
%r1 = insertelement <4 x double> %r0, double %x1, i32 1
%r2 = insertelement <4 x double> %r1, double %x2, i32 2
%r3 = insertelement <4 x double> %r2, double %x3, i32 3
ret <4 x double> %r3
}
define <4 x i64> @var_shuffle_v4i64_v4i64_xxxx_i64(<4 x i64> %x, i64 %i0, i64 %i1, i64 %i2, i64 %i3) nounwind {
; AVX1-LABEL: var_shuffle_v4i64_v4i64_xxxx_i64:
; AVX1: # BB#0:
; AVX1-NEXT: pushq %rbp
; AVX1-NEXT: movq %rsp, %rbp
; AVX1-NEXT: andq $-32, %rsp
; AVX1-NEXT: subq $64, %rsp
; AVX1-NEXT: vmovaps %ymm0, (%rsp)
; AVX1-NEXT: vmovq {{.*#+}} xmm0 = mem[0],zero
; AVX1-NEXT: vmovq {{.*#+}} xmm1 = mem[0],zero
Revert "In visitSTORE, always use FindBetterChain, rather than only when UseAA is enabled." This reverts commit r282600 due to test failues with MCJIT llvm-svn: 282604
2016-09-28 18:37:50 +02:00
; AVX1-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
[X86][SSE] Some basic tests for variable shuffles We don't really support non-constant shuffle masks, but these tests are for cases where BUILD_VECTOR is made up from vector extracts (as well as undef/zero scalars). llvm-svn: 265045
2016-03-31 22:26:30 +02:00
; AVX1-NEXT: vmovq {{.*#+}} xmm1 = mem[0],zero
; AVX1-NEXT: vmovq {{.*#+}} xmm2 = mem[0],zero
Revert "In visitSTORE, always use FindBetterChain, rather than only when UseAA is enabled." This reverts commit r282600 due to test failues with MCJIT llvm-svn: 282604
2016-09-28 18:37:50 +02:00
; AVX1-NEXT: vpunpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm2[0]
; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
[X86][SSE] Some basic tests for variable shuffles We don't really support non-constant shuffle masks, but these tests are for cases where BUILD_VECTOR is made up from vector extracts (as well as undef/zero scalars). llvm-svn: 265045
2016-03-31 22:26:30 +02:00
; AVX1-NEXT: movq %rbp, %rsp
; AVX1-NEXT: popq %rbp
; AVX1-NEXT: retq
;
; AVX2-LABEL: var_shuffle_v4i64_v4i64_xxxx_i64:
; AVX2: # BB#0:
; AVX2-NEXT: pushq %rbp
; AVX2-NEXT: movq %rsp, %rbp
; AVX2-NEXT: andq $-32, %rsp
; AVX2-NEXT: subq $64, %rsp
; AVX2-NEXT: vmovaps %ymm0, (%rsp)
; AVX2-NEXT: vmovq {{.*#+}} xmm0 = mem[0],zero
; AVX2-NEXT: vmovq {{.*#+}} xmm1 = mem[0],zero
Revert "In visitSTORE, always use FindBetterChain, rather than only when UseAA is enabled." This reverts commit r282600 due to test failues with MCJIT llvm-svn: 282604
2016-09-28 18:37:50 +02:00
; AVX2-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
[X86][SSE] Some basic tests for variable shuffles We don't really support non-constant shuffle masks, but these tests are for cases where BUILD_VECTOR is made up from vector extracts (as well as undef/zero scalars). llvm-svn: 265045
2016-03-31 22:26:30 +02:00
; AVX2-NEXT: vmovq {{.*#+}} xmm1 = mem[0],zero
; AVX2-NEXT: vmovq {{.*#+}} xmm2 = mem[0],zero
Revert "In visitSTORE, always use FindBetterChain, rather than only when UseAA is enabled." This reverts commit r282600 due to test failues with MCJIT llvm-svn: 282604
2016-09-28 18:37:50 +02:00
; AVX2-NEXT: vpunpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm2[0]
; AVX2-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm0
[X86][SSE] Some basic tests for variable shuffles We don't really support non-constant shuffle masks, but these tests are for cases where BUILD_VECTOR is made up from vector extracts (as well as undef/zero scalars). llvm-svn: 265045
2016-03-31 22:26:30 +02:00
; AVX2-NEXT: movq %rbp, %rsp
; AVX2-NEXT: popq %rbp
; AVX2-NEXT: retq
%x0 = extractelement <4 x i64> %x, i64 %i0
%x1 = extractelement <4 x i64> %x, i64 %i1
%x2 = extractelement <4 x i64> %x, i64 %i2
%x3 = extractelement <4 x i64> %x, i64 %i3
%r0 = insertelement <4 x i64> undef, i64 %x0, i32 0
%r1 = insertelement <4 x i64> %r0, i64 %x1, i32 1
%r2 = insertelement <4 x i64> %r1, i64 %x2, i32 2
%r3 = insertelement <4 x i64> %r2, i64 %x3, i32 3
ret <4 x i64> %r3
}
define <4 x i64> @var_shuffle_v4i64_v4i64_xx00_i64(<4 x i64> %x, i64 %i0, i64 %i1, i64 %i2, i64 %i3) nounwind {
; AVX1-LABEL: var_shuffle_v4i64_v4i64_xx00_i64:
; AVX1: # BB#0:
; AVX1-NEXT: pushq %rbp
; AVX1-NEXT: movq %rsp, %rbp
; AVX1-NEXT: andq $-32, %rsp
; AVX1-NEXT: subq $64, %rsp
; AVX1-NEXT: vmovaps %ymm0, (%rsp)
; AVX1-NEXT: vmovq {{.*#+}} xmm0 = mem[0],zero
; AVX1-NEXT: vmovq {{.*#+}} xmm1 = mem[0],zero
Revert "In visitSTORE, always use FindBetterChain, rather than only when UseAA is enabled." This reverts commit r282600 due to test failues with MCJIT llvm-svn: 282604
2016-09-28 18:37:50 +02:00
; AVX1-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
[X86][SSE] Some basic tests for variable shuffles We don't really support non-constant shuffle masks, but these tests are for cases where BUILD_VECTOR is made up from vector extracts (as well as undef/zero scalars). llvm-svn: 265045
2016-03-31 22:26:30 +02:00
; AVX1-NEXT: vpxor %xmm1, %xmm1, %xmm1
; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
; AVX1-NEXT: movq %rbp, %rsp
; AVX1-NEXT: popq %rbp
; AVX1-NEXT: retq
;
; AVX2-LABEL: var_shuffle_v4i64_v4i64_xx00_i64:
; AVX2: # BB#0:
; AVX2-NEXT: pushq %rbp
; AVX2-NEXT: movq %rsp, %rbp
; AVX2-NEXT: andq $-32, %rsp
; AVX2-NEXT: subq $64, %rsp
; AVX2-NEXT: vmovaps %ymm0, (%rsp)
; AVX2-NEXT: vmovq {{.*#+}} xmm0 = mem[0],zero
; AVX2-NEXT: vmovq {{.*#+}} xmm1 = mem[0],zero
Revert "In visitSTORE, always use FindBetterChain, rather than only when UseAA is enabled." This reverts commit r282600 due to test failues with MCJIT llvm-svn: 282604
2016-09-28 18:37:50 +02:00
; AVX2-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
[X86][SSE] Some basic tests for variable shuffles We don't really support non-constant shuffle masks, but these tests are for cases where BUILD_VECTOR is made up from vector extracts (as well as undef/zero scalars). llvm-svn: 265045
2016-03-31 22:26:30 +02:00
; AVX2-NEXT: vpxor %xmm1, %xmm1, %xmm1
; AVX2-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm0
; AVX2-NEXT: movq %rbp, %rsp
; AVX2-NEXT: popq %rbp
; AVX2-NEXT: retq
%x0 = extractelement <4 x i64> %x, i64 %i0
%x1 = extractelement <4 x i64> %x, i64 %i1
%x2 = extractelement <4 x i64> %x, i64 %i2
%x3 = extractelement <4 x i64> %x, i64 %i3
%r0 = insertelement <4 x i64> undef, i64 %x0, i32 0
%r1 = insertelement <4 x i64> %r0, i64 %x1, i32 1
%r2 = insertelement <4 x i64> %r1, i64 0, i32 2
%r3 = insertelement <4 x i64> %r2, i64 0, i32 3
ret <4 x i64> %r3
}
define <4 x i64> @var_shuffle_v4i64_v2i64_xxxx_i64(<2 x i64> %x, i64 %i0, i64 %i1, i64 %i2, i64 %i3) nounwind {
; AVX1-LABEL: var_shuffle_v4i64_v2i64_xxxx_i64:
; AVX1: # BB#0:
; AVX1-NEXT: vmovaps %xmm0, -{{[0-9]+}}(%rsp)
; AVX1-NEXT: vmovq {{.*#+}} xmm0 = mem[0],zero
; AVX1-NEXT: vmovq {{.*#+}} xmm1 = mem[0],zero
Revert "In visitSTORE, always use FindBetterChain, rather than only when UseAA is enabled." This reverts commit r282600 due to test failues with MCJIT llvm-svn: 282604
2016-09-28 18:37:50 +02:00
; AVX1-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
[X86][SSE] Some basic tests for variable shuffles We don't really support non-constant shuffle masks, but these tests are for cases where BUILD_VECTOR is made up from vector extracts (as well as undef/zero scalars). llvm-svn: 265045
2016-03-31 22:26:30 +02:00
; AVX1-NEXT: vmovq {{.*#+}} xmm1 = mem[0],zero
; AVX1-NEXT: vmovq {{.*#+}} xmm2 = mem[0],zero
Revert "In visitSTORE, always use FindBetterChain, rather than only when UseAA is enabled." This reverts commit r282600 due to test failues with MCJIT llvm-svn: 282604
2016-09-28 18:37:50 +02:00
; AVX1-NEXT: vpunpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm2[0]
; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
[X86][SSE] Some basic tests for variable shuffles We don't really support non-constant shuffle masks, but these tests are for cases where BUILD_VECTOR is made up from vector extracts (as well as undef/zero scalars). llvm-svn: 265045
2016-03-31 22:26:30 +02:00
; AVX1-NEXT: retq
;
; AVX2-LABEL: var_shuffle_v4i64_v2i64_xxxx_i64:
; AVX2: # BB#0:
; AVX2-NEXT: vmovaps %xmm0, -{{[0-9]+}}(%rsp)
; AVX2-NEXT: vmovq {{.*#+}} xmm0 = mem[0],zero
; AVX2-NEXT: vmovq {{.*#+}} xmm1 = mem[0],zero
Revert "In visitSTORE, always use FindBetterChain, rather than only when UseAA is enabled." This reverts commit r282600 due to test failues with MCJIT llvm-svn: 282604
2016-09-28 18:37:50 +02:00
; AVX2-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
[X86][SSE] Some basic tests for variable shuffles We don't really support non-constant shuffle masks, but these tests are for cases where BUILD_VECTOR is made up from vector extracts (as well as undef/zero scalars). llvm-svn: 265045
2016-03-31 22:26:30 +02:00
; AVX2-NEXT: vmovq {{.*#+}} xmm1 = mem[0],zero
; AVX2-NEXT: vmovq {{.*#+}} xmm2 = mem[0],zero
Revert "In visitSTORE, always use FindBetterChain, rather than only when UseAA is enabled." This reverts commit r282600 due to test failues with MCJIT llvm-svn: 282604
2016-09-28 18:37:50 +02:00
; AVX2-NEXT: vpunpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm2[0]
; AVX2-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm0
[X86][SSE] Some basic tests for variable shuffles We don't really support non-constant shuffle masks, but these tests are for cases where BUILD_VECTOR is made up from vector extracts (as well as undef/zero scalars). llvm-svn: 265045
2016-03-31 22:26:30 +02:00
; AVX2-NEXT: retq
%x0 = extractelement <2 x i64> %x, i64 %i0
%x1 = extractelement <2 x i64> %x, i64 %i1
%x2 = extractelement <2 x i64> %x, i64 %i2
%x3 = extractelement <2 x i64> %x, i64 %i3
%r0 = insertelement <4 x i64> undef, i64 %x0, i32 0
%r1 = insertelement <4 x i64> %r0, i64 %x1, i32 1
%r2 = insertelement <4 x i64> %r1, i64 %x2, i32 2
%r3 = insertelement <4 x i64> %r2, i64 %x3, i32 3
ret <4 x i64> %r3
}
define <8 x float> @var_shuffle_v8f32_v8f32_xxxxxxxx_i32(<8 x float> %x, i32 %i0, i32 %i1, i32 %i2, i32 %i3, i32 %i4, i32 %i5, i32 %i6, i32 %i7) nounwind {
; AVX1-LABEL: var_shuffle_v8f32_v8f32_xxxxxxxx_i32:
; AVX1: # BB#0:
; AVX1-NEXT: pushq %rbp
; AVX1-NEXT: movq %rsp, %rbp
; AVX1-NEXT: andq $-32, %rsp
; AVX1-NEXT: subq $64, %rsp
; AVX1-NEXT: movslq %edi, %rax
Revert "In visitSTORE, always use FindBetterChain, rather than only when UseAA is enabled." This reverts commit r282600 due to test failues with MCJIT llvm-svn: 282604
2016-09-28 18:37:50 +02:00
; AVX1-NEXT: movslq %esi, %rsi
; AVX1-NEXT: movslq %edx, %rdx
; AVX1-NEXT: movslq %ecx, %r11
; AVX1-NEXT: movslq %r8d, %r10
[X86][SSE] Some basic tests for variable shuffles We don't really support non-constant shuffle masks, but these tests are for cases where BUILD_VECTOR is made up from vector extracts (as well as undef/zero scalars). llvm-svn: 265045
2016-03-31 22:26:30 +02:00
; AVX1-NEXT: vmovaps %ymm0, (%rsp)
Revert "In visitSTORE, always use FindBetterChain, rather than only when UseAA is enabled." This reverts commit r282600 due to test failues with MCJIT llvm-svn: 282604
2016-09-28 18:37:50 +02:00
; AVX1-NEXT: movslq %r9d, %r8
; AVX1-NEXT: movslq 16(%rbp), %rdi
; AVX1-NEXT: movslq 24(%rbp), %rcx
[X86][SSE] Some basic tests for variable shuffles We don't really support non-constant shuffle masks, but these tests are for cases where BUILD_VECTOR is made up from vector extracts (as well as undef/zero scalars). llvm-svn: 265045
2016-03-31 22:26:30 +02:00
; AVX1-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; AVX1-NEXT: vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
Revert "In visitSTORE, always use FindBetterChain, rather than only when UseAA is enabled." This reverts commit r282600 due to test failues with MCJIT llvm-svn: 282604
2016-09-28 18:37:50 +02:00
; AVX1-NEXT: vmovss {{.*#+}} xmm2 = mem[0],zero,zero,zero
; AVX1-NEXT: vinsertps {{.*#+}} xmm2 = xmm2[0],mem[0],xmm2[2,3]
; AVX1-NEXT: vinsertps {{.*#+}} xmm2 = xmm2[0,1],mem[0],xmm2[3]
; AVX1-NEXT: vinsertps {{.*#+}} xmm2 = xmm2[0,1,2],mem[0]
; AVX1-NEXT: vmovss {{.*#+}} xmm3 = mem[0],zero,zero,zero
; AVX1-NEXT: vinsertps {{.*#+}} xmm3 = xmm3[0],mem[0],xmm3[2,3]
; AVX1-NEXT: vinsertps {{.*#+}} xmm0 = xmm3[0,1],xmm0[0],xmm3[3]
; AVX1-NEXT: vinsertps {{.*#+}} xmm0 = xmm0[0,1,2],xmm1[0]
; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm2, %ymm0
; AVX1-NEXT: movq %rbp, %rsp
[X86][SSE] Some basic tests for variable shuffles We don't really support non-constant shuffle masks, but these tests are for cases where BUILD_VECTOR is made up from vector extracts (as well as undef/zero scalars). llvm-svn: 265045
2016-03-31 22:26:30 +02:00
; AVX1-NEXT: popq %rbp
; AVX1-NEXT: retq
;
; AVX2-LABEL: var_shuffle_v8f32_v8f32_xxxxxxxx_i32:
; AVX2: # BB#0:
; AVX2-NEXT: vmovd %edi, %xmm1
; AVX2-NEXT: vpermps %ymm0, %ymm1, %ymm1
; AVX2-NEXT: vmovd %esi, %xmm2
; AVX2-NEXT: vpermps %ymm0, %ymm2, %ymm2
; AVX2-NEXT: vmovd %edx, %xmm3
; AVX2-NEXT: vpermps %ymm0, %ymm3, %ymm3
; AVX2-NEXT: vmovd %ecx, %xmm4
; AVX2-NEXT: vpermps %ymm0, %ymm4, %ymm4
; AVX2-NEXT: vmovd %r8d, %xmm5
; AVX2-NEXT: vpermps %ymm0, %ymm5, %ymm5
; AVX2-NEXT: vmovd %r9d, %xmm6
; AVX2-NEXT: vpermps %ymm0, %ymm6, %ymm6
; AVX2-NEXT: vmovd {{.*#+}} xmm7 = mem[0],zero,zero,zero
; AVX2-NEXT: vpermps %ymm0, %ymm7, %ymm7
; AVX2-NEXT: vmovd {{.*#+}} xmm8 = mem[0],zero,zero,zero
; AVX2-NEXT: vpermps %ymm0, %ymm8, %ymm0
; AVX2-NEXT: vinsertps {{.*#+}} xmm5 = xmm5[0],xmm6[0],xmm5[2,3]
; AVX2-NEXT: vinsertps {{.*#+}} xmm5 = xmm5[0,1],xmm7[0],xmm5[3]
; AVX2-NEXT: vinsertps {{.*#+}} xmm0 = xmm5[0,1,2],xmm0[0]
; AVX2-NEXT: vinsertps {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[2,3]
; AVX2-NEXT: vinsertps {{.*#+}} xmm1 = xmm1[0,1],xmm3[0],xmm1[3]
; AVX2-NEXT: vinsertps {{.*#+}} xmm1 = xmm1[0,1,2],xmm4[0]
; AVX2-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
; AVX2-NEXT: retq
%x0 = extractelement <8 x float> %x, i32 %i0
%x1 = extractelement <8 x float> %x, i32 %i1
%x2 = extractelement <8 x float> %x, i32 %i2
%x3 = extractelement <8 x float> %x, i32 %i3
%x4 = extractelement <8 x float> %x, i32 %i4
%x5 = extractelement <8 x float> %x, i32 %i5
%x6 = extractelement <8 x float> %x, i32 %i6
%x7 = extractelement <8 x float> %x, i32 %i7
%r0 = insertelement <8 x float> undef, float %x0, i32 0
%r1 = insertelement <8 x float> %r0, float %x1, i32 1
%r2 = insertelement <8 x float> %r1, float %x2, i32 2
%r3 = insertelement <8 x float> %r2, float %x3, i32 3
%r4 = insertelement <8 x float> %r3, float %x4, i32 4
%r5 = insertelement <8 x float> %r4, float %x5, i32 5
%r6 = insertelement <8 x float> %r5, float %x6, i32 6
%r7 = insertelement <8 x float> %r6, float %x7, i32 7
ret <8 x float> %r7
}
define <8 x float> @var_shuffle_v8f32_v4f32_xxxxxxxx_i32(<4 x float> %x, i32 %i0, i32 %i1, i32 %i2, i32 %i3, i32 %i4, i32 %i5, i32 %i6, i32 %i7) nounwind {
; ALL-LABEL: var_shuffle_v8f32_v4f32_xxxxxxxx_i32:
; ALL: # BB#0:
; ALL-NEXT: movslq %edi, %rax
Revert "In visitSTORE, always use FindBetterChain, rather than only when UseAA is enabled." This reverts commit r282600 due to test failues with MCJIT llvm-svn: 282604
2016-09-28 18:37:50 +02:00
; ALL-NEXT: movslq %esi, %rsi
; ALL-NEXT: movslq %edx, %rdx
; ALL-NEXT: movslq %ecx, %r11
; ALL-NEXT: movslq %r8d, %r10
[X86][SSE] Some basic tests for variable shuffles We don't really support non-constant shuffle masks, but these tests are for cases where BUILD_VECTOR is made up from vector extracts (as well as undef/zero scalars). llvm-svn: 265045
2016-03-31 22:26:30 +02:00
; ALL-NEXT: vmovaps %xmm0, -{{[0-9]+}}(%rsp)
Revert "In visitSTORE, always use FindBetterChain, rather than only when UseAA is enabled." This reverts commit r282600 due to test failues with MCJIT llvm-svn: 282604
2016-09-28 18:37:50 +02:00
; ALL-NEXT: movslq %r9d, %r8
; ALL-NEXT: movslq {{[0-9]+}}(%rsp), %rdi
; ALL-NEXT: movslq {{[0-9]+}}(%rsp), %rcx
[X86][SSE] Some basic tests for variable shuffles We don't really support non-constant shuffle masks, but these tests are for cases where BUILD_VECTOR is made up from vector extracts (as well as undef/zero scalars). llvm-svn: 265045
2016-03-31 22:26:30 +02:00
; ALL-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; ALL-NEXT: vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
Revert "In visitSTORE, always use FindBetterChain, rather than only when UseAA is enabled." This reverts commit r282600 due to test failues with MCJIT llvm-svn: 282604
2016-09-28 18:37:50 +02:00
; ALL-NEXT: vmovss {{.*#+}} xmm2 = mem[0],zero,zero,zero
; ALL-NEXT: vinsertps {{.*#+}} xmm2 = xmm2[0],mem[0],xmm2[2,3]
; ALL-NEXT: vinsertps {{.*#+}} xmm2 = xmm2[0,1],mem[0],xmm2[3]
; ALL-NEXT: vinsertps {{.*#+}} xmm2 = xmm2[0,1,2],mem[0]
; ALL-NEXT: vmovss {{.*#+}} xmm3 = mem[0],zero,zero,zero
; ALL-NEXT: vinsertps {{.*#+}} xmm3 = xmm3[0],mem[0],xmm3[2,3]
; ALL-NEXT: vinsertps {{.*#+}} xmm0 = xmm3[0,1],xmm0[0],xmm3[3]
; ALL-NEXT: vinsertps {{.*#+}} xmm0 = xmm0[0,1,2],xmm1[0]
; ALL-NEXT: vinsertf128 $1, %xmm0, %ymm2, %ymm0
[X86][SSE] Some basic tests for variable shuffles We don't really support non-constant shuffle masks, but these tests are for cases where BUILD_VECTOR is made up from vector extracts (as well as undef/zero scalars). llvm-svn: 265045
2016-03-31 22:26:30 +02:00
; ALL-NEXT: retq
%x0 = extractelement <4 x float> %x, i32 %i0
%x1 = extractelement <4 x float> %x, i32 %i1
%x2 = extractelement <4 x float> %x, i32 %i2
%x3 = extractelement <4 x float> %x, i32 %i3
%x4 = extractelement <4 x float> %x, i32 %i4
%x5 = extractelement <4 x float> %x, i32 %i5
%x6 = extractelement <4 x float> %x, i32 %i6
%x7 = extractelement <4 x float> %x, i32 %i7
%r0 = insertelement <8 x float> undef, float %x0, i32 0
%r1 = insertelement <8 x float> %r0, float %x1, i32 1
%r2 = insertelement <8 x float> %r1, float %x2, i32 2
%r3 = insertelement <8 x float> %r2, float %x3, i32 3
%r4 = insertelement <8 x float> %r3, float %x4, i32 4
%r5 = insertelement <8 x float> %r4, float %x5, i32 5
%r6 = insertelement <8 x float> %r5, float %x6, i32 6
%r7 = insertelement <8 x float> %r6, float %x7, i32 7
ret <8 x float> %r7
}
define <16 x i16> @var_shuffle_v16i16_v16i16_xxxxxxxxxxxxxxxx_i16(<16 x i16> %x, i32 %i0, i32 %i1, i32 %i2, i32 %i3, i32 %i4, i32 %i5, i32 %i6, i32 %i7, i32 %i8, i32 %i9, i32 %i10, i32 %i11, i32 %i12, i32 %i13, i32 %i14, i32 %i15) nounwind {
; AVX1-LABEL: var_shuffle_v16i16_v16i16_xxxxxxxxxxxxxxxx_i16:
; AVX1: # BB#0:
; AVX1-NEXT: pushq %rbp
; AVX1-NEXT: movq %rsp, %rbp
; AVX1-NEXT: andq $-32, %rsp
; AVX1-NEXT: subq $64, %rsp
; AVX1-NEXT: vmovaps %ymm0, (%rsp)
; AVX1-NEXT: movslq 32(%rbp), %rax
; AVX1-NEXT: movzwl (%rsp,%rax,2), %eax
; AVX1-NEXT: vmovd %eax, %xmm0
; AVX1-NEXT: movslq 40(%rbp), %rax
Revert "In visitSTORE, always use FindBetterChain, rather than only when UseAA is enabled." This reverts commit r282600 due to test failues with MCJIT llvm-svn: 282604
2016-09-28 18:37:50 +02:00
; AVX1-NEXT: movzwl (%rsp,%rax,2), %eax
; AVX1-NEXT: vpinsrw $1, %eax, %xmm0, %xmm0
[X86][SSE] Some basic tests for variable shuffles We don't really support non-constant shuffle masks, but these tests are for cases where BUILD_VECTOR is made up from vector extracts (as well as undef/zero scalars). llvm-svn: 265045
2016-03-31 22:26:30 +02:00
; AVX1-NEXT: movslq 48(%rbp), %rax
Revert "In visitSTORE, always use FindBetterChain, rather than only when UseAA is enabled." This reverts commit r282600 due to test failues with MCJIT llvm-svn: 282604
2016-09-28 18:37:50 +02:00
; AVX1-NEXT: movzwl (%rsp,%rax,2), %eax
; AVX1-NEXT: vpinsrw $2, %eax, %xmm0, %xmm0
[X86][SSE] Some basic tests for variable shuffles We don't really support non-constant shuffle masks, but these tests are for cases where BUILD_VECTOR is made up from vector extracts (as well as undef/zero scalars). llvm-svn: 265045
2016-03-31 22:26:30 +02:00
; AVX1-NEXT: movslq 56(%rbp), %rax
Revert "In visitSTORE, always use FindBetterChain, rather than only when UseAA is enabled." This reverts commit r282600 due to test failues with MCJIT llvm-svn: 282604
2016-09-28 18:37:50 +02:00
; AVX1-NEXT: movzwl (%rsp,%rax,2), %eax
; AVX1-NEXT: vpinsrw $3, %eax, %xmm0, %xmm0
[X86][SSE] Some basic tests for variable shuffles We don't really support non-constant shuffle masks, but these tests are for cases where BUILD_VECTOR is made up from vector extracts (as well as undef/zero scalars). llvm-svn: 265045
2016-03-31 22:26:30 +02:00
; AVX1-NEXT: movslq 64(%rbp), %rax
Revert "In visitSTORE, always use FindBetterChain, rather than only when UseAA is enabled." This reverts commit r282600 due to test failues with MCJIT llvm-svn: 282604
2016-09-28 18:37:50 +02:00
; AVX1-NEXT: movzwl (%rsp,%rax,2), %eax
; AVX1-NEXT: vpinsrw $4, %eax, %xmm0, %xmm0
[X86][SSE] Some basic tests for variable shuffles We don't really support non-constant shuffle masks, but these tests are for cases where BUILD_VECTOR is made up from vector extracts (as well as undef/zero scalars). llvm-svn: 265045
2016-03-31 22:26:30 +02:00
; AVX1-NEXT: movslq 72(%rbp), %rax
Revert "In visitSTORE, always use FindBetterChain, rather than only when UseAA is enabled." This reverts commit r282600 due to test failues with MCJIT llvm-svn: 282604
2016-09-28 18:37:50 +02:00
; AVX1-NEXT: movzwl (%rsp,%rax,2), %eax
; AVX1-NEXT: vpinsrw $5, %eax, %xmm0, %xmm0
[X86][SSE] Some basic tests for variable shuffles We don't really support non-constant shuffle masks, but these tests are for cases where BUILD_VECTOR is made up from vector extracts (as well as undef/zero scalars). llvm-svn: 265045
2016-03-31 22:26:30 +02:00
; AVX1-NEXT: movslq 80(%rbp), %rax
Revert "In visitSTORE, always use FindBetterChain, rather than only when UseAA is enabled." This reverts commit r282600 due to test failues with MCJIT llvm-svn: 282604
2016-09-28 18:37:50 +02:00
; AVX1-NEXT: movzwl (%rsp,%rax,2), %eax
; AVX1-NEXT: vpinsrw $6, %eax, %xmm0, %xmm0
[X86][SSE] Some basic tests for variable shuffles We don't really support non-constant shuffle masks, but these tests are for cases where BUILD_VECTOR is made up from vector extracts (as well as undef/zero scalars). llvm-svn: 265045
2016-03-31 22:26:30 +02:00
; AVX1-NEXT: movslq 88(%rbp), %rax
Revert "In visitSTORE, always use FindBetterChain, rather than only when UseAA is enabled." This reverts commit r282600 due to test failues with MCJIT llvm-svn: 282604
2016-09-28 18:37:50 +02:00
; AVX1-NEXT: movzwl (%rsp,%rax,2), %eax
; AVX1-NEXT: vpinsrw $7, %eax, %xmm0, %xmm0
[X86][SSE] Some basic tests for variable shuffles We don't really support non-constant shuffle masks, but these tests are for cases where BUILD_VECTOR is made up from vector extracts (as well as undef/zero scalars). llvm-svn: 265045
2016-03-31 22:26:30 +02:00
; AVX1-NEXT: movslq %edi, %rax
; AVX1-NEXT: movzwl (%rsp,%rax,2), %eax
; AVX1-NEXT: vmovd %eax, %xmm1
; AVX1-NEXT: movslq %esi, %rax
; AVX1-NEXT: vpinsrw $1, (%rsp,%rax,2), %xmm1, %xmm1
; AVX1-NEXT: movslq %edx, %rax
; AVX1-NEXT: vpinsrw $2, (%rsp,%rax,2), %xmm1, %xmm1
; AVX1-NEXT: movslq %ecx, %rax
; AVX1-NEXT: vpinsrw $3, (%rsp,%rax,2), %xmm1, %xmm1
; AVX1-NEXT: movslq %r8d, %rax
; AVX1-NEXT: vpinsrw $4, (%rsp,%rax,2), %xmm1, %xmm1
; AVX1-NEXT: movslq %r9d, %rax
; AVX1-NEXT: vpinsrw $5, (%rsp,%rax,2), %xmm1, %xmm1
; AVX1-NEXT: movslq 16(%rbp), %rax
Revert "In visitSTORE, always use FindBetterChain, rather than only when UseAA is enabled." This reverts commit r282600 due to test failues with MCJIT llvm-svn: 282604
2016-09-28 18:37:50 +02:00
; AVX1-NEXT: movzwl (%rsp,%rax,2), %eax
; AVX1-NEXT: vpinsrw $6, %eax, %xmm1, %xmm1
[X86][SSE] Some basic tests for variable shuffles We don't really support non-constant shuffle masks, but these tests are for cases where BUILD_VECTOR is made up from vector extracts (as well as undef/zero scalars). llvm-svn: 265045
2016-03-31 22:26:30 +02:00
; AVX1-NEXT: movslq 24(%rbp), %rax
Revert "In visitSTORE, always use FindBetterChain, rather than only when UseAA is enabled." This reverts commit r282600 due to test failues with MCJIT llvm-svn: 282604
2016-09-28 18:37:50 +02:00
; AVX1-NEXT: movzwl (%rsp,%rax,2), %eax
; AVX1-NEXT: vpinsrw $7, %eax, %xmm1, %xmm1
[X86][SSE] Some basic tests for variable shuffles We don't really support non-constant shuffle masks, but these tests are for cases where BUILD_VECTOR is made up from vector extracts (as well as undef/zero scalars). llvm-svn: 265045
2016-03-31 22:26:30 +02:00
; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
; AVX1-NEXT: movq %rbp, %rsp
; AVX1-NEXT: popq %rbp
; AVX1-NEXT: retq
;
; AVX2-LABEL: var_shuffle_v16i16_v16i16_xxxxxxxxxxxxxxxx_i16:
; AVX2: # BB#0:
; AVX2-NEXT: pushq %rbp
; AVX2-NEXT: movq %rsp, %rbp
; AVX2-NEXT: andq $-32, %rsp
; AVX2-NEXT: subq $64, %rsp
; AVX2-NEXT: vmovaps %ymm0, (%rsp)
; AVX2-NEXT: movslq 32(%rbp), %rax
; AVX2-NEXT: movzwl (%rsp,%rax,2), %eax
; AVX2-NEXT: vmovd %eax, %xmm0
; AVX2-NEXT: movslq 40(%rbp), %rax
Revert "In visitSTORE, always use FindBetterChain, rather than only when UseAA is enabled." This reverts commit r282600 due to test failues with MCJIT llvm-svn: 282604
2016-09-28 18:37:50 +02:00
; AVX2-NEXT: movzwl (%rsp,%rax,2), %eax
; AVX2-NEXT: vpinsrw $1, %eax, %xmm0, %xmm0
[X86][SSE] Some basic tests for variable shuffles We don't really support non-constant shuffle masks, but these tests are for cases where BUILD_VECTOR is made up from vector extracts (as well as undef/zero scalars). llvm-svn: 265045
2016-03-31 22:26:30 +02:00
; AVX2-NEXT: movslq 48(%rbp), %rax
Revert "In visitSTORE, always use FindBetterChain, rather than only when UseAA is enabled." This reverts commit r282600 due to test failues with MCJIT llvm-svn: 282604
2016-09-28 18:37:50 +02:00
; AVX2-NEXT: movzwl (%rsp,%rax,2), %eax
; AVX2-NEXT: vpinsrw $2, %eax, %xmm0, %xmm0
[X86][SSE] Some basic tests for variable shuffles We don't really support non-constant shuffle masks, but these tests are for cases where BUILD_VECTOR is made up from vector extracts (as well as undef/zero scalars). llvm-svn: 265045
2016-03-31 22:26:30 +02:00
; AVX2-NEXT: movslq 56(%rbp), %rax
Revert "In visitSTORE, always use FindBetterChain, rather than only when UseAA is enabled." This reverts commit r282600 due to test failues with MCJIT llvm-svn: 282604
2016-09-28 18:37:50 +02:00
; AVX2-NEXT: movzwl (%rsp,%rax,2), %eax
; AVX2-NEXT: vpinsrw $3, %eax, %xmm0, %xmm0
[X86][SSE] Some basic tests for variable shuffles We don't really support non-constant shuffle masks, but these tests are for cases where BUILD_VECTOR is made up from vector extracts (as well as undef/zero scalars). llvm-svn: 265045
2016-03-31 22:26:30 +02:00
; AVX2-NEXT: movslq 64(%rbp), %rax
Revert "In visitSTORE, always use FindBetterChain, rather than only when UseAA is enabled." This reverts commit r282600 due to test failues with MCJIT llvm-svn: 282604
2016-09-28 18:37:50 +02:00
; AVX2-NEXT: movzwl (%rsp,%rax,2), %eax
; AVX2-NEXT: vpinsrw $4, %eax, %xmm0, %xmm0
[X86][SSE] Some basic tests for variable shuffles We don't really support non-constant shuffle masks, but these tests are for cases where BUILD_VECTOR is made up from vector extracts (as well as undef/zero scalars). llvm-svn: 265045
2016-03-31 22:26:30 +02:00
; AVX2-NEXT: movslq 72(%rbp), %rax
Revert "In visitSTORE, always use FindBetterChain, rather than only when UseAA is enabled." This reverts commit r282600 due to test failues with MCJIT llvm-svn: 282604
2016-09-28 18:37:50 +02:00
; AVX2-NEXT: movzwl (%rsp,%rax,2), %eax
; AVX2-NEXT: vpinsrw $5, %eax, %xmm0, %xmm0
[X86][SSE] Some basic tests for variable shuffles We don't really support non-constant shuffle masks, but these tests are for cases where BUILD_VECTOR is made up from vector extracts (as well as undef/zero scalars). llvm-svn: 265045
2016-03-31 22:26:30 +02:00
; AVX2-NEXT: movslq 80(%rbp), %rax
Revert "In visitSTORE, always use FindBetterChain, rather than only when UseAA is enabled." This reverts commit r282600 due to test failues with MCJIT llvm-svn: 282604
2016-09-28 18:37:50 +02:00
; AVX2-NEXT: movzwl (%rsp,%rax,2), %eax
; AVX2-NEXT: vpinsrw $6, %eax, %xmm0, %xmm0
[X86][SSE] Some basic tests for variable shuffles We don't really support non-constant shuffle masks, but these tests are for cases where BUILD_VECTOR is made up from vector extracts (as well as undef/zero scalars). llvm-svn: 265045
2016-03-31 22:26:30 +02:00
; AVX2-NEXT: movslq 88(%rbp), %rax
Revert "In visitSTORE, always use FindBetterChain, rather than only when UseAA is enabled." This reverts commit r282600 due to test failues with MCJIT llvm-svn: 282604
2016-09-28 18:37:50 +02:00
; AVX2-NEXT: movzwl (%rsp,%rax,2), %eax
; AVX2-NEXT: vpinsrw $7, %eax, %xmm0, %xmm0
[X86][SSE] Some basic tests for variable shuffles We don't really support non-constant shuffle masks, but these tests are for cases where BUILD_VECTOR is made up from vector extracts (as well as undef/zero scalars). llvm-svn: 265045
2016-03-31 22:26:30 +02:00
; AVX2-NEXT: movslq %edi, %rax
; AVX2-NEXT: movzwl (%rsp,%rax,2), %eax
; AVX2-NEXT: vmovd %eax, %xmm1
; AVX2-NEXT: movslq %esi, %rax
; AVX2-NEXT: vpinsrw $1, (%rsp,%rax,2), %xmm1, %xmm1
; AVX2-NEXT: movslq %edx, %rax
; AVX2-NEXT: vpinsrw $2, (%rsp,%rax,2), %xmm1, %xmm1
; AVX2-NEXT: movslq %ecx, %rax
; AVX2-NEXT: vpinsrw $3, (%rsp,%rax,2), %xmm1, %xmm1
; AVX2-NEXT: movslq %r8d, %rax
; AVX2-NEXT: vpinsrw $4, (%rsp,%rax,2), %xmm1, %xmm1
; AVX2-NEXT: movslq %r9d, %rax
; AVX2-NEXT: vpinsrw $5, (%rsp,%rax,2), %xmm1, %xmm1
; AVX2-NEXT: movslq 16(%rbp), %rax
Revert "In visitSTORE, always use FindBetterChain, rather than only when UseAA is enabled." This reverts commit r282600 due to test failues with MCJIT llvm-svn: 282604
2016-09-28 18:37:50 +02:00
; AVX2-NEXT: movzwl (%rsp,%rax,2), %eax
; AVX2-NEXT: vpinsrw $6, %eax, %xmm1, %xmm1
[X86][SSE] Some basic tests for variable shuffles We don't really support non-constant shuffle masks, but these tests are for cases where BUILD_VECTOR is made up from vector extracts (as well as undef/zero scalars). llvm-svn: 265045
2016-03-31 22:26:30 +02:00
; AVX2-NEXT: movslq 24(%rbp), %rax
Revert "In visitSTORE, always use FindBetterChain, rather than only when UseAA is enabled." This reverts commit r282600 due to test failues with MCJIT llvm-svn: 282604
2016-09-28 18:37:50 +02:00
; AVX2-NEXT: movzwl (%rsp,%rax,2), %eax
; AVX2-NEXT: vpinsrw $7, %eax, %xmm1, %xmm1
[X86][SSE] Some basic tests for variable shuffles We don't really support non-constant shuffle masks, but these tests are for cases where BUILD_VECTOR is made up from vector extracts (as well as undef/zero scalars). llvm-svn: 265045
2016-03-31 22:26:30 +02:00
; AVX2-NEXT: vinserti128 $1, %xmm0, %ymm1, %ymm0
; AVX2-NEXT: movq %rbp, %rsp
; AVX2-NEXT: popq %rbp
; AVX2-NEXT: retq
%x0 = extractelement <16 x i16> %x, i32 %i0
%x1 = extractelement <16 x i16> %x, i32 %i1
%x2 = extractelement <16 x i16> %x, i32 %i2
%x3 = extractelement <16 x i16> %x, i32 %i3
%x4 = extractelement <16 x i16> %x, i32 %i4
%x5 = extractelement <16 x i16> %x, i32 %i5
%x6 = extractelement <16 x i16> %x, i32 %i6
%x7 = extractelement <16 x i16> %x, i32 %i7
%x8 = extractelement <16 x i16> %x, i32 %i8
%x9 = extractelement <16 x i16> %x, i32 %i9
%x10 = extractelement <16 x i16> %x, i32 %i10
%x11 = extractelement <16 x i16> %x, i32 %i11
%x12 = extractelement <16 x i16> %x, i32 %i12
%x13 = extractelement <16 x i16> %x, i32 %i13
%x14 = extractelement <16 x i16> %x, i32 %i14
%x15 = extractelement <16 x i16> %x, i32 %i15
%r0 = insertelement <16 x i16> undef, i16 %x0 , i32 0
%r1 = insertelement <16 x i16> %r0 , i16 %x1 , i32 1
%r2 = insertelement <16 x i16> %r1 , i16 %x2 , i32 2
%r3 = insertelement <16 x i16> %r2 , i16 %x3 , i32 3
%r4 = insertelement <16 x i16> %r3 , i16 %x4 , i32 4
%r5 = insertelement <16 x i16> %r4 , i16 %x5 , i32 5
%r6 = insertelement <16 x i16> %r5 , i16 %x6 , i32 6
%r7 = insertelement <16 x i16> %r6 , i16 %x7 , i32 7
%r8 = insertelement <16 x i16> %r7 , i16 %x8 , i32 8
%r9 = insertelement <16 x i16> %r8 , i16 %x9 , i32 9
%r10 = insertelement <16 x i16> %r9 , i16 %x10, i32 10
%r11 = insertelement <16 x i16> %r10, i16 %x11, i32 11
%r12 = insertelement <16 x i16> %r11, i16 %x12, i32 12
%r13 = insertelement <16 x i16> %r12, i16 %x13, i32 13
%r14 = insertelement <16 x i16> %r13, i16 %x14, i32 14
%r15 = insertelement <16 x i16> %r14, i16 %x15, i32 15
ret <16 x i16> %r15
}
define <16 x i16> @var_shuffle_v16i16_v8i16_xxxxxxxxxxxxxxxx_i16(<8 x i16> %x, i32 %i0, i32 %i1, i32 %i2, i32 %i3, i32 %i4, i32 %i5, i32 %i6, i32 %i7, i32 %i8, i32 %i9, i32 %i10, i32 %i11, i32 %i12, i32 %i13, i32 %i14, i32 %i15) nounwind {
; AVX1-LABEL: var_shuffle_v16i16_v8i16_xxxxxxxxxxxxxxxx_i16:
; AVX1: # BB#0:
; AVX1-NEXT: vmovaps %xmm0, -{{[0-9]+}}(%rsp)
; AVX1-NEXT: movslq {{[0-9]+}}(%rsp), %rax
; AVX1-NEXT: movzwl -24(%rsp,%rax,2), %eax
; AVX1-NEXT: vmovd %eax, %xmm0
; AVX1-NEXT: movslq {{[0-9]+}}(%rsp), %rax
Revert "In visitSTORE, always use FindBetterChain, rather than only when UseAA is enabled." This reverts commit r282600 due to test failues with MCJIT llvm-svn: 282604
2016-09-28 18:37:50 +02:00
; AVX1-NEXT: movzwl -24(%rsp,%rax,2), %eax
; AVX1-NEXT: vpinsrw $1, %eax, %xmm0, %xmm0
[X86][SSE] Some basic tests for variable shuffles We don't really support non-constant shuffle masks, but these tests are for cases where BUILD_VECTOR is made up from vector extracts (as well as undef/zero scalars). llvm-svn: 265045
2016-03-31 22:26:30 +02:00
; AVX1-NEXT: movslq {{[0-9]+}}(%rsp), %rax
Revert "In visitSTORE, always use FindBetterChain, rather than only when UseAA is enabled." This reverts commit r282600 due to test failues with MCJIT llvm-svn: 282604
2016-09-28 18:37:50 +02:00
; AVX1-NEXT: movzwl -24(%rsp,%rax,2), %eax
; AVX1-NEXT: vpinsrw $2, %eax, %xmm0, %xmm0
[X86][SSE] Some basic tests for variable shuffles We don't really support non-constant shuffle masks, but these tests are for cases where BUILD_VECTOR is made up from vector extracts (as well as undef/zero scalars). llvm-svn: 265045
2016-03-31 22:26:30 +02:00
; AVX1-NEXT: movslq {{[0-9]+}}(%rsp), %rax
Revert "In visitSTORE, always use FindBetterChain, rather than only when UseAA is enabled." This reverts commit r282600 due to test failues with MCJIT llvm-svn: 282604
2016-09-28 18:37:50 +02:00
; AVX1-NEXT: movzwl -24(%rsp,%rax,2), %eax
; AVX1-NEXT: vpinsrw $3, %eax, %xmm0, %xmm0
[X86][SSE] Some basic tests for variable shuffles We don't really support non-constant shuffle masks, but these tests are for cases where BUILD_VECTOR is made up from vector extracts (as well as undef/zero scalars). llvm-svn: 265045
2016-03-31 22:26:30 +02:00
; AVX1-NEXT: movslq {{[0-9]+}}(%rsp), %rax
Revert "In visitSTORE, always use FindBetterChain, rather than only when UseAA is enabled." This reverts commit r282600 due to test failues with MCJIT llvm-svn: 282604
2016-09-28 18:37:50 +02:00
; AVX1-NEXT: movzwl -24(%rsp,%rax,2), %eax
; AVX1-NEXT: vpinsrw $4, %eax, %xmm0, %xmm0
[X86][SSE] Some basic tests for variable shuffles We don't really support non-constant shuffle masks, but these tests are for cases where BUILD_VECTOR is made up from vector extracts (as well as undef/zero scalars). llvm-svn: 265045
2016-03-31 22:26:30 +02:00
; AVX1-NEXT: movslq {{[0-9]+}}(%rsp), %rax
Revert "In visitSTORE, always use FindBetterChain, rather than only when UseAA is enabled." This reverts commit r282600 due to test failues with MCJIT llvm-svn: 282604
2016-09-28 18:37:50 +02:00
; AVX1-NEXT: movzwl -24(%rsp,%rax,2), %eax
; AVX1-NEXT: vpinsrw $5, %eax, %xmm0, %xmm0
[X86][SSE] Some basic tests for variable shuffles We don't really support non-constant shuffle masks, but these tests are for cases where BUILD_VECTOR is made up from vector extracts (as well as undef/zero scalars). llvm-svn: 265045
2016-03-31 22:26:30 +02:00
; AVX1-NEXT: movslq {{[0-9]+}}(%rsp), %rax
Revert "In visitSTORE, always use FindBetterChain, rather than only when UseAA is enabled." This reverts commit r282600 due to test failues with MCJIT llvm-svn: 282604
2016-09-28 18:37:50 +02:00
; AVX1-NEXT: movzwl -24(%rsp,%rax,2), %eax
; AVX1-NEXT: vpinsrw $6, %eax, %xmm0, %xmm0
[X86][SSE] Some basic tests for variable shuffles We don't really support non-constant shuffle masks, but these tests are for cases where BUILD_VECTOR is made up from vector extracts (as well as undef/zero scalars). llvm-svn: 265045
2016-03-31 22:26:30 +02:00
; AVX1-NEXT: movslq {{[0-9]+}}(%rsp), %rax
Revert "In visitSTORE, always use FindBetterChain, rather than only when UseAA is enabled." This reverts commit r282600 due to test failues with MCJIT llvm-svn: 282604
2016-09-28 18:37:50 +02:00
; AVX1-NEXT: movzwl -24(%rsp,%rax,2), %eax
; AVX1-NEXT: vpinsrw $7, %eax, %xmm0, %xmm0
[X86][SSE] Some basic tests for variable shuffles We don't really support non-constant shuffle masks, but these tests are for cases where BUILD_VECTOR is made up from vector extracts (as well as undef/zero scalars). llvm-svn: 265045
2016-03-31 22:26:30 +02:00
; AVX1-NEXT: movslq %edi, %rax
; AVX1-NEXT: movzwl -24(%rsp,%rax,2), %eax
; AVX1-NEXT: vmovd %eax, %xmm1
; AVX1-NEXT: movslq %esi, %rax
; AVX1-NEXT: vpinsrw $1, -24(%rsp,%rax,2), %xmm1, %xmm1
; AVX1-NEXT: movslq %edx, %rax
; AVX1-NEXT: vpinsrw $2, -24(%rsp,%rax,2), %xmm1, %xmm1
; AVX1-NEXT: movslq %ecx, %rax
; AVX1-NEXT: vpinsrw $3, -24(%rsp,%rax,2), %xmm1, %xmm1
; AVX1-NEXT: movslq %r8d, %rax
; AVX1-NEXT: vpinsrw $4, -24(%rsp,%rax,2), %xmm1, %xmm1
; AVX1-NEXT: movslq %r9d, %rax
; AVX1-NEXT: vpinsrw $5, -24(%rsp,%rax,2), %xmm1, %xmm1
; AVX1-NEXT: movslq {{[0-9]+}}(%rsp), %rax
Revert "In visitSTORE, always use FindBetterChain, rather than only when UseAA is enabled." This reverts commit r282600 due to test failues with MCJIT llvm-svn: 282604
2016-09-28 18:37:50 +02:00
; AVX1-NEXT: movzwl -24(%rsp,%rax,2), %eax
; AVX1-NEXT: vpinsrw $6, %eax, %xmm1, %xmm1
[X86][SSE] Some basic tests for variable shuffles We don't really support non-constant shuffle masks, but these tests are for cases where BUILD_VECTOR is made up from vector extracts (as well as undef/zero scalars). llvm-svn: 265045
2016-03-31 22:26:30 +02:00
; AVX1-NEXT: movslq {{[0-9]+}}(%rsp), %rax
Revert "In visitSTORE, always use FindBetterChain, rather than only when UseAA is enabled." This reverts commit r282600 due to test failues with MCJIT llvm-svn: 282604
2016-09-28 18:37:50 +02:00
; AVX1-NEXT: movzwl -24(%rsp,%rax,2), %eax
; AVX1-NEXT: vpinsrw $7, %eax, %xmm1, %xmm1
[X86][SSE] Some basic tests for variable shuffles We don't really support non-constant shuffle masks, but these tests are for cases where BUILD_VECTOR is made up from vector extracts (as well as undef/zero scalars). llvm-svn: 265045
2016-03-31 22:26:30 +02:00
; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
; AVX1-NEXT: retq
;
; AVX2-LABEL: var_shuffle_v16i16_v8i16_xxxxxxxxxxxxxxxx_i16:
; AVX2: # BB#0:
; AVX2-NEXT: vmovaps %xmm0, -{{[0-9]+}}(%rsp)
; AVX2-NEXT: movslq {{[0-9]+}}(%rsp), %rax
; AVX2-NEXT: movzwl -24(%rsp,%rax,2), %eax
; AVX2-NEXT: vmovd %eax, %xmm0
; AVX2-NEXT: movslq {{[0-9]+}}(%rsp), %rax
Revert "In visitSTORE, always use FindBetterChain, rather than only when UseAA is enabled." This reverts commit r282600 due to test failues with MCJIT llvm-svn: 282604
2016-09-28 18:37:50 +02:00
; AVX2-NEXT: movzwl -24(%rsp,%rax,2), %eax
; AVX2-NEXT: vpinsrw $1, %eax, %xmm0, %xmm0
[X86][SSE] Some basic tests for variable shuffles We don't really support non-constant shuffle masks, but these tests are for cases where BUILD_VECTOR is made up from vector extracts (as well as undef/zero scalars). llvm-svn: 265045
2016-03-31 22:26:30 +02:00
; AVX2-NEXT: movslq {{[0-9]+}}(%rsp), %rax
Revert "In visitSTORE, always use FindBetterChain, rather than only when UseAA is enabled." This reverts commit r282600 due to test failues with MCJIT llvm-svn: 282604
2016-09-28 18:37:50 +02:00
; AVX2-NEXT: movzwl -24(%rsp,%rax,2), %eax
; AVX2-NEXT: vpinsrw $2, %eax, %xmm0, %xmm0
[X86][SSE] Some basic tests for variable shuffles We don't really support non-constant shuffle masks, but these tests are for cases where BUILD_VECTOR is made up from vector extracts (as well as undef/zero scalars). llvm-svn: 265045
2016-03-31 22:26:30 +02:00
; AVX2-NEXT: movslq {{[0-9]+}}(%rsp), %rax
Revert "In visitSTORE, always use FindBetterChain, rather than only when UseAA is enabled." This reverts commit r282600 due to test failues with MCJIT llvm-svn: 282604
2016-09-28 18:37:50 +02:00
; AVX2-NEXT: movzwl -24(%rsp,%rax,2), %eax
; AVX2-NEXT: vpinsrw $3, %eax, %xmm0, %xmm0
[X86][SSE] Some basic tests for variable shuffles We don't really support non-constant shuffle masks, but these tests are for cases where BUILD_VECTOR is made up from vector extracts (as well as undef/zero scalars). llvm-svn: 265045
2016-03-31 22:26:30 +02:00
; AVX2-NEXT: movslq {{[0-9]+}}(%rsp), %rax
Revert "In visitSTORE, always use FindBetterChain, rather than only when UseAA is enabled." This reverts commit r282600 due to test failues with MCJIT llvm-svn: 282604
2016-09-28 18:37:50 +02:00
; AVX2-NEXT: movzwl -24(%rsp,%rax,2), %eax
; AVX2-NEXT: vpinsrw $4, %eax, %xmm0, %xmm0
[X86][SSE] Some basic tests for variable shuffles We don't really support non-constant shuffle masks, but these tests are for cases where BUILD_VECTOR is made up from vector extracts (as well as undef/zero scalars). llvm-svn: 265045
2016-03-31 22:26:30 +02:00
; AVX2-NEXT: movslq {{[0-9]+}}(%rsp), %rax
Revert "In visitSTORE, always use FindBetterChain, rather than only when UseAA is enabled." This reverts commit r282600 due to test failues with MCJIT llvm-svn: 282604
2016-09-28 18:37:50 +02:00
; AVX2-NEXT: movzwl -24(%rsp,%rax,2), %eax
; AVX2-NEXT: vpinsrw $5, %eax, %xmm0, %xmm0
[X86][SSE] Some basic tests for variable shuffles We don't really support non-constant shuffle masks, but these tests are for cases where BUILD_VECTOR is made up from vector extracts (as well as undef/zero scalars). llvm-svn: 265045
2016-03-31 22:26:30 +02:00
; AVX2-NEXT: movslq {{[0-9]+}}(%rsp), %rax
Revert "In visitSTORE, always use FindBetterChain, rather than only when UseAA is enabled." This reverts commit r282600 due to test failues with MCJIT llvm-svn: 282604
2016-09-28 18:37:50 +02:00
; AVX2-NEXT: movzwl -24(%rsp,%rax,2), %eax
; AVX2-NEXT: vpinsrw $6, %eax, %xmm0, %xmm0
[X86][SSE] Some basic tests for variable shuffles We don't really support non-constant shuffle masks, but these tests are for cases where BUILD_VECTOR is made up from vector extracts (as well as undef/zero scalars). llvm-svn: 265045
2016-03-31 22:26:30 +02:00
; AVX2-NEXT: movslq {{[0-9]+}}(%rsp), %rax
Revert "In visitSTORE, always use FindBetterChain, rather than only when UseAA is enabled." This reverts commit r282600 due to test failues with MCJIT llvm-svn: 282604
2016-09-28 18:37:50 +02:00
; AVX2-NEXT: movzwl -24(%rsp,%rax,2), %eax
; AVX2-NEXT: vpinsrw $7, %eax, %xmm0, %xmm0
[X86][SSE] Some basic tests for variable shuffles We don't really support non-constant shuffle masks, but these tests are for cases where BUILD_VECTOR is made up from vector extracts (as well as undef/zero scalars). llvm-svn: 265045
2016-03-31 22:26:30 +02:00
; AVX2-NEXT: movslq %edi, %rax
; AVX2-NEXT: movzwl -24(%rsp,%rax,2), %eax
; AVX2-NEXT: vmovd %eax, %xmm1
; AVX2-NEXT: movslq %esi, %rax
; AVX2-NEXT: vpinsrw $1, -24(%rsp,%rax,2), %xmm1, %xmm1
; AVX2-NEXT: movslq %edx, %rax
; AVX2-NEXT: vpinsrw $2, -24(%rsp,%rax,2), %xmm1, %xmm1
; AVX2-NEXT: movslq %ecx, %rax
; AVX2-NEXT: vpinsrw $3, -24(%rsp,%rax,2), %xmm1, %xmm1
; AVX2-NEXT: movslq %r8d, %rax
; AVX2-NEXT: vpinsrw $4, -24(%rsp,%rax,2), %xmm1, %xmm1
; AVX2-NEXT: movslq %r9d, %rax
; AVX2-NEXT: vpinsrw $5, -24(%rsp,%rax,2), %xmm1, %xmm1
; AVX2-NEXT: movslq {{[0-9]+}}(%rsp), %rax
Revert "In visitSTORE, always use FindBetterChain, rather than only when UseAA is enabled." This reverts commit r282600 due to test failues with MCJIT llvm-svn: 282604
2016-09-28 18:37:50 +02:00
; AVX2-NEXT: movzwl -24(%rsp,%rax,2), %eax
; AVX2-NEXT: vpinsrw $6, %eax, %xmm1, %xmm1
[X86][SSE] Some basic tests for variable shuffles We don't really support non-constant shuffle masks, but these tests are for cases where BUILD_VECTOR is made up from vector extracts (as well as undef/zero scalars). llvm-svn: 265045
2016-03-31 22:26:30 +02:00
; AVX2-NEXT: movslq {{[0-9]+}}(%rsp), %rax
Revert "In visitSTORE, always use FindBetterChain, rather than only when UseAA is enabled." This reverts commit r282600 due to test failues with MCJIT llvm-svn: 282604
2016-09-28 18:37:50 +02:00
; AVX2-NEXT: movzwl -24(%rsp,%rax,2), %eax
; AVX2-NEXT: vpinsrw $7, %eax, %xmm1, %xmm1
[X86][SSE] Some basic tests for variable shuffles We don't really support non-constant shuffle masks, but these tests are for cases where BUILD_VECTOR is made up from vector extracts (as well as undef/zero scalars). llvm-svn: 265045
2016-03-31 22:26:30 +02:00
; AVX2-NEXT: vinserti128 $1, %xmm0, %ymm1, %ymm0
; AVX2-NEXT: retq
%x0 = extractelement <8 x i16> %x, i32 %i0
%x1 = extractelement <8 x i16> %x, i32 %i1
%x2 = extractelement <8 x i16> %x, i32 %i2
%x3 = extractelement <8 x i16> %x, i32 %i3
%x4 = extractelement <8 x i16> %x, i32 %i4
%x5 = extractelement <8 x i16> %x, i32 %i5
%x6 = extractelement <8 x i16> %x, i32 %i6
%x7 = extractelement <8 x i16> %x, i32 %i7
%x8 = extractelement <8 x i16> %x, i32 %i8
%x9 = extractelement <8 x i16> %x, i32 %i9
%x10 = extractelement <8 x i16> %x, i32 %i10
%x11 = extractelement <8 x i16> %x, i32 %i11
%x12 = extractelement <8 x i16> %x, i32 %i12
%x13 = extractelement <8 x i16> %x, i32 %i13
%x14 = extractelement <8 x i16> %x, i32 %i14
%x15 = extractelement <8 x i16> %x, i32 %i15
%r0 = insertelement <16 x i16> undef, i16 %x0 , i32 0
%r1 = insertelement <16 x i16> %r0 , i16 %x1 , i32 1
%r2 = insertelement <16 x i16> %r1 , i16 %x2 , i32 2
%r3 = insertelement <16 x i16> %r2 , i16 %x3 , i32 3
%r4 = insertelement <16 x i16> %r3 , i16 %x4 , i32 4
%r5 = insertelement <16 x i16> %r4 , i16 %x5 , i32 5
%r6 = insertelement <16 x i16> %r5 , i16 %x6 , i32 6
%r7 = insertelement <16 x i16> %r6 , i16 %x7 , i32 7
%r8 = insertelement <16 x i16> %r7 , i16 %x8 , i32 8
%r9 = insertelement <16 x i16> %r8 , i16 %x9 , i32 9
%r10 = insertelement <16 x i16> %r9 , i16 %x10, i32 10
%r11 = insertelement <16 x i16> %r10, i16 %x11, i32 11
%r12 = insertelement <16 x i16> %r11, i16 %x12, i32 12
%r13 = insertelement <16 x i16> %r12, i16 %x13, i32 13
%r14 = insertelement <16 x i16> %r13, i16 %x14, i32 14
%r15 = insertelement <16 x i16> %r14, i16 %x15, i32 15
ret <16 x i16> %r15
}
;
; Unary shuffle indices from memory
;
define <4 x i64> @mem_shuffle_v4i64_v4i64_xxxx_i64(<4 x i64> %x, i64* %i) nounwind {
; AVX1-LABEL: mem_shuffle_v4i64_v4i64_xxxx_i64:
; AVX1: # BB#0:
; AVX1-NEXT: pushq %rbp
; AVX1-NEXT: movq %rsp, %rbp
; AVX1-NEXT: andq $-32, %rsp
; AVX1-NEXT: subq $64, %rsp
; AVX1-NEXT: movq (%rdi), %rax
; AVX1-NEXT: movq 8(%rdi), %rcx
; AVX1-NEXT: movq 16(%rdi), %rdx
; AVX1-NEXT: movq 24(%rdi), %rsi
; AVX1-NEXT: vmovaps %ymm0, (%rsp)
; AVX1-NEXT: vmovq {{.*#+}} xmm0 = mem[0],zero
; AVX1-NEXT: vmovq {{.*#+}} xmm1 = mem[0],zero
; AVX1-NEXT: vmovq {{.*#+}} xmm2 = mem[0],zero
; AVX1-NEXT: vpunpcklqdq {{.*#+}} xmm1 = xmm2[0],xmm1[0]
Revert "In visitSTORE, always use FindBetterChain, rather than only when UseAA is enabled." This reverts commit r282600 due to test failues with MCJIT llvm-svn: 282604
2016-09-28 18:37:50 +02:00
; AVX1-NEXT: vmovq {{.*#+}} xmm2 = mem[0],zero
; AVX1-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm2[0]
; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
[X86][SSE] Some basic tests for variable shuffles We don't really support non-constant shuffle masks, but these tests are for cases where BUILD_VECTOR is made up from vector extracts (as well as undef/zero scalars). llvm-svn: 265045
2016-03-31 22:26:30 +02:00
; AVX1-NEXT: movq %rbp, %rsp
; AVX1-NEXT: popq %rbp
; AVX1-NEXT: retq
;
; AVX2-LABEL: mem_shuffle_v4i64_v4i64_xxxx_i64:
; AVX2: # BB#0:
; AVX2-NEXT: pushq %rbp
; AVX2-NEXT: movq %rsp, %rbp
; AVX2-NEXT: andq $-32, %rsp
; AVX2-NEXT: subq $64, %rsp
; AVX2-NEXT: movq (%rdi), %rax
; AVX2-NEXT: movq 8(%rdi), %rcx
; AVX2-NEXT: movq 16(%rdi), %rdx
; AVX2-NEXT: movq 24(%rdi), %rsi
; AVX2-NEXT: vmovaps %ymm0, (%rsp)
; AVX2-NEXT: vmovq {{.*#+}} xmm0 = mem[0],zero
; AVX2-NEXT: vmovq {{.*#+}} xmm1 = mem[0],zero
; AVX2-NEXT: vmovq {{.*#+}} xmm2 = mem[0],zero
; AVX2-NEXT: vpunpcklqdq {{.*#+}} xmm1 = xmm2[0],xmm1[0]
Revert "In visitSTORE, always use FindBetterChain, rather than only when UseAA is enabled." This reverts commit r282600 due to test failues with MCJIT llvm-svn: 282604
2016-09-28 18:37:50 +02:00
; AVX2-NEXT: vmovq {{.*#+}} xmm2 = mem[0],zero
; AVX2-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm2[0]
; AVX2-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm0
[X86][SSE] Some basic tests for variable shuffles We don't really support non-constant shuffle masks, but these tests are for cases where BUILD_VECTOR is made up from vector extracts (as well as undef/zero scalars). llvm-svn: 265045
2016-03-31 22:26:30 +02:00
; AVX2-NEXT: movq %rbp, %rsp
; AVX2-NEXT: popq %rbp
; AVX2-NEXT: retq
%p0 = getelementptr inbounds i64, i64* %i, i32 0
%p1 = getelementptr inbounds i64, i64* %i, i32 1
%p2 = getelementptr inbounds i64, i64* %i, i32 2
%p3 = getelementptr inbounds i64, i64* %i, i32 3
%i0 = load i64, i64* %p0, align 4
%i1 = load i64, i64* %p1, align 4
%i2 = load i64, i64* %p2, align 4
%i3 = load i64, i64* %p3, align 4
%x0 = extractelement <4 x i64> %x, i64 %i0
%x1 = extractelement <4 x i64> %x, i64 %i1
%x2 = extractelement <4 x i64> %x, i64 %i2
%x3 = extractelement <4 x i64> %x, i64 %i3
%r0 = insertelement <4 x i64> undef, i64 %x0, i32 0
%r1 = insertelement <4 x i64> %r0, i64 %x1, i32 1
%r2 = insertelement <4 x i64> %r1, i64 %x2, i32 2
%r3 = insertelement <4 x i64> %r2, i64 %x3, i32 3
ret <4 x i64> %r3
}
define <4 x i64> @mem_shuffle_v4i64_v2i64_xxxx_i64(<2 x i64> %x, i64* %i) nounwind {
; AVX1-LABEL: mem_shuffle_v4i64_v2i64_xxxx_i64:
; AVX1: # BB#0:
; AVX1-NEXT: movq (%rdi), %rax
; AVX1-NEXT: movq 8(%rdi), %rcx
; AVX1-NEXT: movq 16(%rdi), %rdx
; AVX1-NEXT: movq 24(%rdi), %rsi
; AVX1-NEXT: vmovaps %xmm0, -{{[0-9]+}}(%rsp)
; AVX1-NEXT: vmovq {{.*#+}} xmm0 = mem[0],zero
; AVX1-NEXT: vmovq {{.*#+}} xmm1 = mem[0],zero
; AVX1-NEXT: vmovq {{.*#+}} xmm2 = mem[0],zero
; AVX1-NEXT: vpunpcklqdq {{.*#+}} xmm1 = xmm2[0],xmm1[0]
Revert "In visitSTORE, always use FindBetterChain, rather than only when UseAA is enabled." This reverts commit r282600 due to test failues with MCJIT llvm-svn: 282604
2016-09-28 18:37:50 +02:00
; AVX1-NEXT: vmovq {{.*#+}} xmm2 = mem[0],zero
; AVX1-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm2[0]
; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
[X86][SSE] Some basic tests for variable shuffles We don't really support non-constant shuffle masks, but these tests are for cases where BUILD_VECTOR is made up from vector extracts (as well as undef/zero scalars). llvm-svn: 265045
2016-03-31 22:26:30 +02:00
; AVX1-NEXT: retq
;
; AVX2-LABEL: mem_shuffle_v4i64_v2i64_xxxx_i64:
; AVX2: # BB#0:
; AVX2-NEXT: movq (%rdi), %rax
; AVX2-NEXT: movq 8(%rdi), %rcx
; AVX2-NEXT: movq 16(%rdi), %rdx
; AVX2-NEXT: movq 24(%rdi), %rsi
; AVX2-NEXT: vmovaps %xmm0, -{{[0-9]+}}(%rsp)
; AVX2-NEXT: vmovq {{.*#+}} xmm0 = mem[0],zero
; AVX2-NEXT: vmovq {{.*#+}} xmm1 = mem[0],zero
; AVX2-NEXT: vmovq {{.*#+}} xmm2 = mem[0],zero
; AVX2-NEXT: vpunpcklqdq {{.*#+}} xmm1 = xmm2[0],xmm1[0]
Revert "In visitSTORE, always use FindBetterChain, rather than only when UseAA is enabled." This reverts commit r282600 due to test failues with MCJIT llvm-svn: 282604
2016-09-28 18:37:50 +02:00
; AVX2-NEXT: vmovq {{.*#+}} xmm2 = mem[0],zero
; AVX2-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm2[0]
; AVX2-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm0
[X86][SSE] Some basic tests for variable shuffles We don't really support non-constant shuffle masks, but these tests are for cases where BUILD_VECTOR is made up from vector extracts (as well as undef/zero scalars). llvm-svn: 265045
2016-03-31 22:26:30 +02:00
; AVX2-NEXT: retq
%p0 = getelementptr inbounds i64, i64* %i, i32 0
%p1 = getelementptr inbounds i64, i64* %i, i32 1
%p2 = getelementptr inbounds i64, i64* %i, i32 2
%p3 = getelementptr inbounds i64, i64* %i, i32 3
%i0 = load i64, i64* %p0, align 4
%i1 = load i64, i64* %p1, align 4
%i2 = load i64, i64* %p2, align 4
%i3 = load i64, i64* %p3, align 4
%x0 = extractelement <2 x i64> %x, i64 %i0
%x1 = extractelement <2 x i64> %x, i64 %i1
%x2 = extractelement <2 x i64> %x, i64 %i2
%x3 = extractelement <2 x i64> %x, i64 %i3
%r0 = insertelement <4 x i64> undef, i64 %x0, i32 0
%r1 = insertelement <4 x i64> %r0, i64 %x1, i32 1
%r2 = insertelement <4 x i64> %r1, i64 %x2, i32 2
%r3 = insertelement <4 x i64> %r2, i64 %x3, i32 3
ret <4 x i64> %r3
}
Reference in New Issue Copy Permalink