maxjhandsome
/
llvm-project
forked from OSchip/llvm-project
1
0
Fork 0
Code Issues Pull Requests Packages Releases Wiki Activity
llvm-project/llvm/test/CodeGen/X86/bitcast-int-to-vector-bool-...

706 lines
27 KiB
LLVM
Raw Normal View History

[X86][SSE] Tests for bitcasting iX integers to vXi1 boolean vectors Including sign/zero extension to legal types llvm-svn: 307301
2017-07-07 03:33:10 +08:00
; 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=SSE2-SSSE3,SSE2
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+ssse3 | FileCheck %s --check-prefixes=SSE2-SSSE3,SSSE3
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx | FileCheck %s --check-prefixes=AVX12,AVX1
X86 Tests: Update more isel tests with FastVariableShuffle feature Summary: Added the FastVariableShuffle feature to cases that resembled processors for which this fearure is on. For AVX2 there are processors with and w/o this fearue enable. For AVX512 only KNL does enable this feature so cases which only have +avx512f were left without the FastVariableShuffle enabled. Reviewers: RKSimon, craig.topper Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D41851 llvm-svn: 322090
2018-01-10 00:26:06 +08:00
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2 | FileCheck %s --check-prefixes=AVX12,AVX2,AVX2-SLOW
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2,+fast-variable-shuffle | FileCheck %s --check-prefixes=AVX12,AVX2,AVX2-FAST
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512f,+avx512vl,+avx512bw,+fast-variable-shuffle | FileCheck %s --check-prefix=AVX512
[X86][SSE] Tests for bitcasting iX integers to vXi1 boolean vectors Including sign/zero extension to legal types llvm-svn: 307301
2017-07-07 03:33:10 +08:00
;
; 128-bit vectors
;
define <2 x i64> @ext_i2_2i64(i2 %a0) {
; SSE2-SSSE3-LABEL: ext_i2_2i64:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; SSE2-SSSE3: # %bb.0:
Followup on Proposal to move MIR physical register namespace to '$' sigil. Discussed here: http://lists.llvm.org/pipermail/llvm-dev/2018-January/120320.html In preparation for adding support for named vregs we are changing the sigil for physical registers in MIR to '$' from '%'. This will prevent name clashes of named physical register with named vregs. llvm-svn: 323922
2018-02-01 06:04:26 +08:00
; SSE2-SSSE3-NEXT: # kill: def $edi killed $edi def $rdi
[X86][SSE] Add support for extending bool vectors bitcasted from scalars This patch acts as a reverse to combineBitcastvxi1 - bitcasting a scalar integer to a boolean vector and extending it 'in place' to the requested legal type. Currently this doesn't handle AVX512 at all - but the current mask register approach is lacking for some cases. Differential Revision: https://reviews.llvm.org/D35320 llvm-svn: 314076
2017-09-24 21:42:31 +08:00
; SSE2-SSSE3-NEXT: movq %rdi, %xmm0
; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm1 = xmm0[0,1,0,1]
; SSE2-SSSE3-NEXT: movdqa {{.*#+}} xmm0 = [1,2]
; SSE2-SSSE3-NEXT: pand %xmm0, %xmm1
; SSE2-SSSE3-NEXT: pcmpeqd %xmm0, %xmm1
; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm1[1,0,3,2]
; SSE2-SSSE3-NEXT: pand %xmm1, %xmm0
[X86][SSE] Tests for bitcasting iX integers to vXi1 boolean vectors Including sign/zero extension to legal types llvm-svn: 307301
2017-07-07 03:33:10 +08:00
; SSE2-SSSE3-NEXT: retq
;
[X86][SSE] Add support for extending bool vectors bitcasted from scalars This patch acts as a reverse to combineBitcastvxi1 - bitcasting a scalar integer to a boolean vector and extending it 'in place' to the requested legal type. Currently this doesn't handle AVX512 at all - but the current mask register approach is lacking for some cases. Differential Revision: https://reviews.llvm.org/D35320 llvm-svn: 314076
2017-09-24 21:42:31 +08:00
; AVX1-LABEL: ext_i2_2i64:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; AVX1: # %bb.0:
Followup on Proposal to move MIR physical register namespace to '$' sigil. Discussed here: http://lists.llvm.org/pipermail/llvm-dev/2018-January/120320.html In preparation for adding support for named vregs we are changing the sigil for physical registers in MIR to '$' from '%'. This will prevent name clashes of named physical register with named vregs. llvm-svn: 323922
2018-02-01 06:04:26 +08:00
; AVX1-NEXT: # kill: def $edi killed $edi def $rdi
[X86][SSE] Add support for extending bool vectors bitcasted from scalars This patch acts as a reverse to combineBitcastvxi1 - bitcasting a scalar integer to a boolean vector and extending it 'in place' to the requested legal type. Currently this doesn't handle AVX512 at all - but the current mask register approach is lacking for some cases. Differential Revision: https://reviews.llvm.org/D35320 llvm-svn: 314076
2017-09-24 21:42:31 +08:00
; AVX1-NEXT: vmovq %rdi, %xmm0
; AVX1-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[0,1,0,1]
; AVX1-NEXT: vmovdqa {{.*#+}} xmm1 = [1,2]
; AVX1-NEXT: vpand %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vpcmpeqq %xmm1, %xmm0, %xmm0
; AVX1-NEXT: retq
;
; AVX2-LABEL: ext_i2_2i64:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; AVX2: # %bb.0:
Followup on Proposal to move MIR physical register namespace to '$' sigil. Discussed here: http://lists.llvm.org/pipermail/llvm-dev/2018-January/120320.html In preparation for adding support for named vregs we are changing the sigil for physical registers in MIR to '$' from '%'. This will prevent name clashes of named physical register with named vregs. llvm-svn: 323922
2018-02-01 06:04:26 +08:00
; AVX2-NEXT: # kill: def $edi killed $edi def $rdi
[X86][SSE] Add support for extending bool vectors bitcasted from scalars This patch acts as a reverse to combineBitcastvxi1 - bitcasting a scalar integer to a boolean vector and extending it 'in place' to the requested legal type. Currently this doesn't handle AVX512 at all - but the current mask register approach is lacking for some cases. Differential Revision: https://reviews.llvm.org/D35320 llvm-svn: 314076
2017-09-24 21:42:31 +08:00
; AVX2-NEXT: vmovq %rdi, %xmm0
; AVX2-NEXT: vpbroadcastq %xmm0, %xmm0
; AVX2-NEXT: vmovdqa {{.*#+}} xmm1 = [1,2]
; AVX2-NEXT: vpand %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vpcmpeqq %xmm1, %xmm0, %xmm0
; AVX2-NEXT: retq
[X86][SSE] Tests for bitcasting iX integers to vXi1 boolean vectors Including sign/zero extension to legal types llvm-svn: 307301
2017-07-07 03:33:10 +08:00
;
; AVX512-LABEL: ext_i2_2i64:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; AVX512: # %bb.0:
[X86] Prevent combining (v8i1 (bitconvert (i8 load)))->(v8i1 load) if we don't have DQI. We end up using an i8 load via an isel pattern from v8i1 anyway. This just makes it more explicit. This seems to improve codgen in some cases and I'd like to kill off some of the load patterns. llvm-svn: 321598
2017-12-31 15:38:41 +08:00
; AVX512-NEXT: kmovd %edi, %k1
[X86] When lowering extending loads from v2i1/v4i1, if we have VLX, use a narrower extend. Previously we used an extend from v8i1 to v8i32/v8i64. Then extracted to the final width. But if we have VLX we should extract first. This way we don't end up with an overly large extend. This allows us to use vcmpeq to make all ones for the sign extend when DQI isn't available. Otherwise we get a VPTERNLOG. If we make v2i1/v4i1 legal like proposed in D41560, we could always do this and rely on the lowering of the extend to widen when necessary. llvm-svn: 321538
2017-12-29 03:46:11 +08:00
; AVX512-NEXT: vpcmpeqd %xmm0, %xmm0, %xmm0
; AVX512-NEXT: vmovdqa64 %xmm0, %xmm0 {%k1} {z}
[X86][SSE] Tests for bitcasting iX integers to vXi1 boolean vectors Including sign/zero extension to legal types llvm-svn: 307301
2017-07-07 03:33:10 +08:00
; AVX512-NEXT: retq
%1 = bitcast i2 %a0 to <2 x i1>
%2 = sext <2 x i1> %1 to <2 x i64>
ret <2 x i64> %2
}
define <4 x i32> @ext_i4_4i32(i4 %a0) {
; SSE2-SSSE3-LABEL: ext_i4_4i32:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; SSE2-SSSE3: # %bb.0:
[X86][SSE] Add support for extending bool vectors bitcasted from scalars This patch acts as a reverse to combineBitcastvxi1 - bitcasting a scalar integer to a boolean vector and extending it 'in place' to the requested legal type. Currently this doesn't handle AVX512 at all - but the current mask register approach is lacking for some cases. Differential Revision: https://reviews.llvm.org/D35320 llvm-svn: 314076
2017-09-24 21:42:31 +08:00
; SSE2-SSSE3-NEXT: movd %edi, %xmm0
; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
; SSE2-SSSE3-NEXT: movdqa {{.*#+}} xmm1 = [1,2,4,8]
; SSE2-SSSE3-NEXT: pand %xmm1, %xmm0
; SSE2-SSSE3-NEXT: pcmpeqd %xmm1, %xmm0
[X86][SSE] Tests for bitcasting iX integers to vXi1 boolean vectors Including sign/zero extension to legal types llvm-svn: 307301
2017-07-07 03:33:10 +08:00
; SSE2-SSSE3-NEXT: retq
;
[X86][SSE] Add support for extending bool vectors bitcasted from scalars This patch acts as a reverse to combineBitcastvxi1 - bitcasting a scalar integer to a boolean vector and extending it 'in place' to the requested legal type. Currently this doesn't handle AVX512 at all - but the current mask register approach is lacking for some cases. Differential Revision: https://reviews.llvm.org/D35320 llvm-svn: 314076
2017-09-24 21:42:31 +08:00
; AVX1-LABEL: ext_i4_4i32:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; AVX1: # %bb.0:
[X86][SSE] Add support for extending bool vectors bitcasted from scalars This patch acts as a reverse to combineBitcastvxi1 - bitcasting a scalar integer to a boolean vector and extending it 'in place' to the requested legal type. Currently this doesn't handle AVX512 at all - but the current mask register approach is lacking for some cases. Differential Revision: https://reviews.llvm.org/D35320 llvm-svn: 314076
2017-09-24 21:42:31 +08:00
; AVX1-NEXT: vmovd %edi, %xmm0
; AVX1-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
; AVX1-NEXT: vmovdqa {{.*#+}} xmm1 = [1,2,4,8]
; AVX1-NEXT: vpand %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vpcmpeqd %xmm1, %xmm0, %xmm0
; AVX1-NEXT: retq
;
; AVX2-LABEL: ext_i4_4i32:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; AVX2: # %bb.0:
[X86][SSE] Add support for extending bool vectors bitcasted from scalars This patch acts as a reverse to combineBitcastvxi1 - bitcasting a scalar integer to a boolean vector and extending it 'in place' to the requested legal type. Currently this doesn't handle AVX512 at all - but the current mask register approach is lacking for some cases. Differential Revision: https://reviews.llvm.org/D35320 llvm-svn: 314076
2017-09-24 21:42:31 +08:00
; AVX2-NEXT: vmovd %edi, %xmm0
; AVX2-NEXT: vpbroadcastd %xmm0, %xmm0
; AVX2-NEXT: vmovdqa {{.*#+}} xmm1 = [1,2,4,8]
; AVX2-NEXT: vpand %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vpcmpeqd %xmm1, %xmm0, %xmm0
; AVX2-NEXT: retq
[X86][SSE] Tests for bitcasting iX integers to vXi1 boolean vectors Including sign/zero extension to legal types llvm-svn: 307301
2017-07-07 03:33:10 +08:00
;
; AVX512-LABEL: ext_i4_4i32:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; AVX512: # %bb.0:
[X86] Prevent combining (v8i1 (bitconvert (i8 load)))->(v8i1 load) if we don't have DQI. We end up using an i8 load via an isel pattern from v8i1 anyway. This just makes it more explicit. This seems to improve codgen in some cases and I'd like to kill off some of the load patterns. llvm-svn: 321598
2017-12-31 15:38:41 +08:00
; AVX512-NEXT: kmovd %edi, %k1
[X86] When lowering extending loads from v2i1/v4i1, if we have VLX, use a narrower extend. Previously we used an extend from v8i1 to v8i32/v8i64. Then extracted to the final width. But if we have VLX we should extract first. This way we don't end up with an overly large extend. This allows us to use vcmpeq to make all ones for the sign extend when DQI isn't available. Otherwise we get a VPTERNLOG. If we make v2i1/v4i1 legal like proposed in D41560, we could always do this and rely on the lowering of the extend to widen when necessary. llvm-svn: 321538
2017-12-29 03:46:11 +08:00
; AVX512-NEXT: vpcmpeqd %xmm0, %xmm0, %xmm0
; AVX512-NEXT: vmovdqa32 %xmm0, %xmm0 {%k1} {z}
[X86][SSE] Tests for bitcasting iX integers to vXi1 boolean vectors Including sign/zero extension to legal types llvm-svn: 307301
2017-07-07 03:33:10 +08:00
; AVX512-NEXT: retq
%1 = bitcast i4 %a0 to <4 x i1>
%2 = sext <4 x i1> %1 to <4 x i32>
ret <4 x i32> %2
}
define <8 x i16> @ext_i8_8i16(i8 %a0) {
; SSE2-SSSE3-LABEL: ext_i8_8i16:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; SSE2-SSSE3: # %bb.0:
[X86][SSE] Add support for extending bool vectors bitcasted from scalars This patch acts as a reverse to combineBitcastvxi1 - bitcasting a scalar integer to a boolean vector and extending it 'in place' to the requested legal type. Currently this doesn't handle AVX512 at all - but the current mask register approach is lacking for some cases. Differential Revision: https://reviews.llvm.org/D35320 llvm-svn: 314076
2017-09-24 21:42:31 +08:00
; SSE2-SSSE3-NEXT: movd %edi, %xmm0
[X86][SSE] Match PSHUFLW/PSHUFHW + PSHUFD vXi16 shuffle patterns (PR34686) As noted in PR34686, we are relying on a PSHUFD+PSHUFLW+PSHUFHW shuffle chain for most general vXi16 unary shuffles. This patch checks for simpler PSHUFLW+PSHUFD and PSHUFHW+PSHUFD cases beforehand, building on some existing code that just handled splat shuffles. By doing so we also prevent premature use of PSHUFB shuffles which can be slower and require the creation/loading of constant shuffle masks. We now have the 'fast-variable-shuffle' option for hardware that prefers combining 2 or more shuffles to VPSHUFB etc. Differential Revision: https://reviews.llvm.org/D38318 llvm-svn: 321553
2017-12-29 22:41:50 +08:00
; SSE2-SSSE3-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,0,2,3,4,5,6,7]
; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
[X86][SSE] Add support for extending bool vectors bitcasted from scalars This patch acts as a reverse to combineBitcastvxi1 - bitcasting a scalar integer to a boolean vector and extending it 'in place' to the requested legal type. Currently this doesn't handle AVX512 at all - but the current mask register approach is lacking for some cases. Differential Revision: https://reviews.llvm.org/D35320 llvm-svn: 314076
2017-09-24 21:42:31 +08:00
; SSE2-SSSE3-NEXT: movdqa {{.*#+}} xmm1 = [1,2,4,8,16,32,64,128]
; SSE2-SSSE3-NEXT: pand %xmm1, %xmm0
; SSE2-SSSE3-NEXT: pcmpeqw %xmm1, %xmm0
[X86][SSE] Tests for bitcasting iX integers to vXi1 boolean vectors Including sign/zero extension to legal types llvm-svn: 307301
2017-07-07 03:33:10 +08:00
; SSE2-SSSE3-NEXT: retq
;
[X86][SSE] Add support for extending bool vectors bitcasted from scalars This patch acts as a reverse to combineBitcastvxi1 - bitcasting a scalar integer to a boolean vector and extending it 'in place' to the requested legal type. Currently this doesn't handle AVX512 at all - but the current mask register approach is lacking for some cases. Differential Revision: https://reviews.llvm.org/D35320 llvm-svn: 314076
2017-09-24 21:42:31 +08:00
; AVX1-LABEL: ext_i8_8i16:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; AVX1: # %bb.0:
[X86][SSE] Add support for extending bool vectors bitcasted from scalars This patch acts as a reverse to combineBitcastvxi1 - bitcasting a scalar integer to a boolean vector and extending it 'in place' to the requested legal type. Currently this doesn't handle AVX512 at all - but the current mask register approach is lacking for some cases. Differential Revision: https://reviews.llvm.org/D35320 llvm-svn: 314076
2017-09-24 21:42:31 +08:00
; AVX1-NEXT: vmovd %edi, %xmm0
[X86][SSE] Match PSHUFLW/PSHUFHW + PSHUFD vXi16 shuffle patterns (PR34686) As noted in PR34686, we are relying on a PSHUFD+PSHUFLW+PSHUFHW shuffle chain for most general vXi16 unary shuffles. This patch checks for simpler PSHUFLW+PSHUFD and PSHUFHW+PSHUFD cases beforehand, building on some existing code that just handled splat shuffles. By doing so we also prevent premature use of PSHUFB shuffles which can be slower and require the creation/loading of constant shuffle masks. We now have the 'fast-variable-shuffle' option for hardware that prefers combining 2 or more shuffles to VPSHUFB etc. Differential Revision: https://reviews.llvm.org/D38318 llvm-svn: 321553
2017-12-29 22:41:50 +08:00
; AVX1-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[0,0,2,3,4,5,6,7]
; AVX1-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
[X86][SSE] Add support for extending bool vectors bitcasted from scalars This patch acts as a reverse to combineBitcastvxi1 - bitcasting a scalar integer to a boolean vector and extending it 'in place' to the requested legal type. Currently this doesn't handle AVX512 at all - but the current mask register approach is lacking for some cases. Differential Revision: https://reviews.llvm.org/D35320 llvm-svn: 314076
2017-09-24 21:42:31 +08:00
; AVX1-NEXT: vmovdqa {{.*#+}} xmm1 = [1,2,4,8,16,32,64,128]
; AVX1-NEXT: vpand %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vpcmpeqw %xmm1, %xmm0, %xmm0
; AVX1-NEXT: retq
;
; AVX2-LABEL: ext_i8_8i16:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; AVX2: # %bb.0:
[X86][SSE] Add support for extending bool vectors bitcasted from scalars This patch acts as a reverse to combineBitcastvxi1 - bitcasting a scalar integer to a boolean vector and extending it 'in place' to the requested legal type. Currently this doesn't handle AVX512 at all - but the current mask register approach is lacking for some cases. Differential Revision: https://reviews.llvm.org/D35320 llvm-svn: 314076
2017-09-24 21:42:31 +08:00
; AVX2-NEXT: vmovd %edi, %xmm0
; AVX2-NEXT: vpbroadcastw %xmm0, %xmm0
; AVX2-NEXT: vmovdqa {{.*#+}} xmm1 = [1,2,4,8,16,32,64,128]
; AVX2-NEXT: vpand %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vpcmpeqw %xmm1, %xmm0, %xmm0
; AVX2-NEXT: retq
[X86][SSE] Tests for bitcasting iX integers to vXi1 boolean vectors Including sign/zero extension to legal types llvm-svn: 307301
2017-07-07 03:33:10 +08:00
;
; AVX512-LABEL: ext_i8_8i16:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; AVX512: # %bb.0:
[X86][SSE] Tests for bitcasting iX integers to vXi1 boolean vectors Including sign/zero extension to legal types llvm-svn: 307301
2017-07-07 03:33:10 +08:00
; AVX512-NEXT: kmovd %edi, %k0
; AVX512-NEXT: vpmovm2w %k0, %xmm0
; AVX512-NEXT: retq
%1 = bitcast i8 %a0 to <8 x i1>
%2 = sext <8 x i1> %1 to <8 x i16>
ret <8 x i16> %2
}
define <16 x i8> @ext_i16_16i8(i16 %a0) {
[X86][SSE] Add support for extending bool vectors bitcasted from scalars This patch acts as a reverse to combineBitcastvxi1 - bitcasting a scalar integer to a boolean vector and extending it 'in place' to the requested legal type. Currently this doesn't handle AVX512 at all - but the current mask register approach is lacking for some cases. Differential Revision: https://reviews.llvm.org/D35320 llvm-svn: 314076
2017-09-24 21:42:31 +08:00
; SSE2-LABEL: ext_i16_16i8:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; SSE2: # %bb.0:
[X86][SSE] Add support for extending bool vectors bitcasted from scalars This patch acts as a reverse to combineBitcastvxi1 - bitcasting a scalar integer to a boolean vector and extending it 'in place' to the requested legal type. Currently this doesn't handle AVX512 at all - but the current mask register approach is lacking for some cases. Differential Revision: https://reviews.llvm.org/D35320 llvm-svn: 314076
2017-09-24 21:42:31 +08:00
; SSE2-NEXT: movd %edi, %xmm0
; SSE2-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
; SSE2-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,0,1,1,4,5,6,7]
; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
; SSE2-NEXT: movdqa {{.*#+}} xmm1 = [1,2,4,8,16,32,64,128,1,2,4,8,16,32,64,128]
; SSE2-NEXT: pand %xmm1, %xmm0
; SSE2-NEXT: pcmpeqb %xmm1, %xmm0
; SSE2-NEXT: retq
;
; SSSE3-LABEL: ext_i16_16i8:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; SSSE3: # %bb.0:
[X86][SSE] Add support for extending bool vectors bitcasted from scalars This patch acts as a reverse to combineBitcastvxi1 - bitcasting a scalar integer to a boolean vector and extending it 'in place' to the requested legal type. Currently this doesn't handle AVX512 at all - but the current mask register approach is lacking for some cases. Differential Revision: https://reviews.llvm.org/D35320 llvm-svn: 314076
2017-09-24 21:42:31 +08:00
; SSSE3-NEXT: movd %edi, %xmm0
; SSSE3-NEXT: pshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,1,1,1,1,1,1,1,1]
; SSSE3-NEXT: movdqa {{.*#+}} xmm1 = [1,2,4,8,16,32,64,128,1,2,4,8,16,32,64,128]
; SSSE3-NEXT: pand %xmm1, %xmm0
; SSSE3-NEXT: pcmpeqb %xmm1, %xmm0
; SSSE3-NEXT: retq
;
; AVX1-LABEL: ext_i16_16i8:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; AVX1: # %bb.0:
[X86][SSE] Add support for extending bool vectors bitcasted from scalars This patch acts as a reverse to combineBitcastvxi1 - bitcasting a scalar integer to a boolean vector and extending it 'in place' to the requested legal type. Currently this doesn't handle AVX512 at all - but the current mask register approach is lacking for some cases. Differential Revision: https://reviews.llvm.org/D35320 llvm-svn: 314076
2017-09-24 21:42:31 +08:00
; AVX1-NEXT: vmovd %edi, %xmm0
; AVX1-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,1,1,1,1,1,1,1,1]
[X86] Add 128 MOVDDUP to the constant pool printing in X86AsmPrinter::EmitInstruction. We use this instruction to broadcast a single 64-bit value to a v2i64/v2f64 vector. llvm-svn: 344486
2018-10-15 09:51:53 +08:00
; AVX1-NEXT: vmovddup {{.*#+}} xmm1 = [-1.7939930131212661E-307,-1.7939930131212661E-307]
; AVX1-NEXT: # xmm1 = mem[0,0]
[X86][SSE] Add support for extending bool vectors bitcasted from scalars This patch acts as a reverse to combineBitcastvxi1 - bitcasting a scalar integer to a boolean vector and extending it 'in place' to the requested legal type. Currently this doesn't handle AVX512 at all - but the current mask register approach is lacking for some cases. Differential Revision: https://reviews.llvm.org/D35320 llvm-svn: 314076
2017-09-24 21:42:31 +08:00
; AVX1-NEXT: vpand %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vpcmpeqb %xmm1, %xmm0, %xmm0
; AVX1-NEXT: retq
[X86][SSE] Tests for bitcasting iX integers to vXi1 boolean vectors Including sign/zero extension to legal types llvm-svn: 307301
2017-07-07 03:33:10 +08:00
;
[X86][SSE] Add support for extending bool vectors bitcasted from scalars This patch acts as a reverse to combineBitcastvxi1 - bitcasting a scalar integer to a boolean vector and extending it 'in place' to the requested legal type. Currently this doesn't handle AVX512 at all - but the current mask register approach is lacking for some cases. Differential Revision: https://reviews.llvm.org/D35320 llvm-svn: 314076
2017-09-24 21:42:31 +08:00
; AVX2-LABEL: ext_i16_16i8:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; AVX2: # %bb.0:
[X86][SSE] Add support for extending bool vectors bitcasted from scalars This patch acts as a reverse to combineBitcastvxi1 - bitcasting a scalar integer to a boolean vector and extending it 'in place' to the requested legal type. Currently this doesn't handle AVX512 at all - but the current mask register approach is lacking for some cases. Differential Revision: https://reviews.llvm.org/D35320 llvm-svn: 314076
2017-09-24 21:42:31 +08:00
; AVX2-NEXT: vmovd %edi, %xmm0
; AVX2-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,0,0,0,0,0,0,0,1,1,1,1,1,1,1,1]
; AVX2-NEXT: vpbroadcastq {{.*#+}} xmm1 = [9241421688590303745,9241421688590303745]
; AVX2-NEXT: vpand %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vpcmpeqb %xmm1, %xmm0, %xmm0
; AVX2-NEXT: retq
[X86][SSE] Tests for bitcasting iX integers to vXi1 boolean vectors Including sign/zero extension to legal types llvm-svn: 307301
2017-07-07 03:33:10 +08:00
;
; AVX512-LABEL: ext_i16_16i8:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; AVX512: # %bb.0:
[X86][SSE] Tests for bitcasting iX integers to vXi1 boolean vectors Including sign/zero extension to legal types llvm-svn: 307301
2017-07-07 03:33:10 +08:00
; AVX512-NEXT: kmovd %edi, %k0
; AVX512-NEXT: vpmovm2b %k0, %xmm0
; AVX512-NEXT: retq
%1 = bitcast i16 %a0 to <16 x i1>
%2 = sext <16 x i1> %1 to <16 x i8>
ret <16 x i8> %2
}
;
; 256-bit vectors
;
define <4 x i64> @ext_i4_4i64(i4 %a0) {
; SSE2-SSSE3-LABEL: ext_i4_4i64:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; SSE2-SSSE3: # %bb.0:
Followup on Proposal to move MIR physical register namespace to '$' sigil. Discussed here: http://lists.llvm.org/pipermail/llvm-dev/2018-January/120320.html In preparation for adding support for named vregs we are changing the sigil for physical registers in MIR to '$' from '%'. This will prevent name clashes of named physical register with named vregs. llvm-svn: 323922
2018-02-01 06:04:26 +08:00
; SSE2-SSSE3-NEXT: # kill: def $edi killed $edi def $rdi
[X86][SSE] Add support for extending bool vectors bitcasted from scalars This patch acts as a reverse to combineBitcastvxi1 - bitcasting a scalar integer to a boolean vector and extending it 'in place' to the requested legal type. Currently this doesn't handle AVX512 at all - but the current mask register approach is lacking for some cases. Differential Revision: https://reviews.llvm.org/D35320 llvm-svn: 314076
2017-09-24 21:42:31 +08:00
; SSE2-SSSE3-NEXT: movq %rdi, %xmm0
; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm2 = xmm0[0,1,0,1]
; SSE2-SSSE3-NEXT: movdqa {{.*#+}} xmm0 = [1,2]
; SSE2-SSSE3-NEXT: movdqa %xmm2, %xmm1
; SSE2-SSSE3-NEXT: pand %xmm0, %xmm1
; SSE2-SSSE3-NEXT: pcmpeqd %xmm0, %xmm1
; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm1[1,0,3,2]
; SSE2-SSSE3-NEXT: pand %xmm1, %xmm0
; SSE2-SSSE3-NEXT: movdqa {{.*#+}} xmm1 = [4,8]
; SSE2-SSSE3-NEXT: pand %xmm1, %xmm2
; SSE2-SSSE3-NEXT: pcmpeqd %xmm1, %xmm2
; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm1 = xmm2[1,0,3,2]
; SSE2-SSSE3-NEXT: pand %xmm2, %xmm1
[X86][SSE] Tests for bitcasting iX integers to vXi1 boolean vectors Including sign/zero extension to legal types llvm-svn: 307301
2017-07-07 03:33:10 +08:00
; SSE2-SSSE3-NEXT: retq
;
; AVX1-LABEL: ext_i4_4i64:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; AVX1: # %bb.0:
Followup on Proposal to move MIR physical register namespace to '$' sigil. Discussed here: http://lists.llvm.org/pipermail/llvm-dev/2018-January/120320.html In preparation for adding support for named vregs we are changing the sigil for physical registers in MIR to '$' from '%'. This will prevent name clashes of named physical register with named vregs. llvm-svn: 323922
2018-02-01 06:04:26 +08:00
; AVX1-NEXT: # kill: def $edi killed $edi def $rdi
[X86][SSE] Add support for extending bool vectors bitcasted from scalars This patch acts as a reverse to combineBitcastvxi1 - bitcasting a scalar integer to a boolean vector and extending it 'in place' to the requested legal type. Currently this doesn't handle AVX512 at all - but the current mask register approach is lacking for some cases. Differential Revision: https://reviews.llvm.org/D35320 llvm-svn: 314076
2017-09-24 21:42:31 +08:00
; AVX1-NEXT: vmovq %rdi, %xmm0
; AVX1-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[0,1,0,1]
; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
; AVX1-NEXT: vandps {{.*}}(%rip), %ymm0, %ymm0
; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX1-NEXT: vpxor %xmm2, %xmm2, %xmm2
; AVX1-NEXT: vpcmpeqq %xmm2, %xmm1, %xmm1
; AVX1-NEXT: vpcmpeqd %xmm3, %xmm3, %xmm3
; AVX1-NEXT: vpxor %xmm3, %xmm1, %xmm1
; AVX1-NEXT: vpcmpeqq %xmm2, %xmm0, %xmm0
; AVX1-NEXT: vpxor %xmm3, %xmm0, %xmm0
; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
[X86][SSE] Tests for bitcasting iX integers to vXi1 boolean vectors Including sign/zero extension to legal types llvm-svn: 307301
2017-07-07 03:33:10 +08:00
; AVX1-NEXT: retq
;
; AVX2-LABEL: ext_i4_4i64:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; AVX2: # %bb.0:
Followup on Proposal to move MIR physical register namespace to '$' sigil. Discussed here: http://lists.llvm.org/pipermail/llvm-dev/2018-January/120320.html In preparation for adding support for named vregs we are changing the sigil for physical registers in MIR to '$' from '%'. This will prevent name clashes of named physical register with named vregs. llvm-svn: 323922
2018-02-01 06:04:26 +08:00
; AVX2-NEXT: # kill: def $edi killed $edi def $rdi
[X86][SSE] Add support for extending bool vectors bitcasted from scalars This patch acts as a reverse to combineBitcastvxi1 - bitcasting a scalar integer to a boolean vector and extending it 'in place' to the requested legal type. Currently this doesn't handle AVX512 at all - but the current mask register approach is lacking for some cases. Differential Revision: https://reviews.llvm.org/D35320 llvm-svn: 314076
2017-09-24 21:42:31 +08:00
; AVX2-NEXT: vmovq %rdi, %xmm0
; AVX2-NEXT: vpbroadcastq %xmm0, %ymm0
; AVX2-NEXT: vmovdqa {{.*#+}} ymm1 = [1,2,4,8]
; AVX2-NEXT: vpand %ymm1, %ymm0, %ymm0
; AVX2-NEXT: vpcmpeqq %ymm1, %ymm0, %ymm0
[X86][SSE] Tests for bitcasting iX integers to vXi1 boolean vectors Including sign/zero extension to legal types llvm-svn: 307301
2017-07-07 03:33:10 +08:00
; AVX2-NEXT: retq
;
; AVX512-LABEL: ext_i4_4i64:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; AVX512: # %bb.0:
[X86] Prevent combining (v8i1 (bitconvert (i8 load)))->(v8i1 load) if we don't have DQI. We end up using an i8 load via an isel pattern from v8i1 anyway. This just makes it more explicit. This seems to improve codgen in some cases and I'd like to kill off some of the load patterns. llvm-svn: 321598
2017-12-31 15:38:41 +08:00
; AVX512-NEXT: kmovd %edi, %k1
[X86] When lowering extending loads from v2i1/v4i1, if we have VLX, use a narrower extend. Previously we used an extend from v8i1 to v8i32/v8i64. Then extracted to the final width. But if we have VLX we should extract first. This way we don't end up with an overly large extend. This allows us to use vcmpeq to make all ones for the sign extend when DQI isn't available. Otherwise we get a VPTERNLOG. If we make v2i1/v4i1 legal like proposed in D41560, we could always do this and rely on the lowering of the extend to widen when necessary. llvm-svn: 321538
2017-12-29 03:46:11 +08:00
; AVX512-NEXT: vpcmpeqd %ymm0, %ymm0, %ymm0
; AVX512-NEXT: vmovdqa64 %ymm0, %ymm0 {%k1} {z}
[X86][SSE] Tests for bitcasting iX integers to vXi1 boolean vectors Including sign/zero extension to legal types llvm-svn: 307301
2017-07-07 03:33:10 +08:00
; AVX512-NEXT: retq
%1 = bitcast i4 %a0 to <4 x i1>
%2 = sext <4 x i1> %1 to <4 x i64>
ret <4 x i64> %2
}
define <8 x i32> @ext_i8_8i32(i8 %a0) {
; SSE2-SSSE3-LABEL: ext_i8_8i32:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; SSE2-SSSE3: # %bb.0:
[X86][SSE] Add support for extending bool vectors bitcasted from scalars This patch acts as a reverse to combineBitcastvxi1 - bitcasting a scalar integer to a boolean vector and extending it 'in place' to the requested legal type. Currently this doesn't handle AVX512 at all - but the current mask register approach is lacking for some cases. Differential Revision: https://reviews.llvm.org/D35320 llvm-svn: 314076
2017-09-24 21:42:31 +08:00
; SSE2-SSSE3-NEXT: movd %edi, %xmm0
; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm1 = xmm0[0,0,0,0]
; SSE2-SSSE3-NEXT: movdqa {{.*#+}} xmm2 = [1,2,4,8]
[X86][SSE] Tests for bitcasting iX integers to vXi1 boolean vectors Including sign/zero extension to legal types llvm-svn: 307301
2017-07-07 03:33:10 +08:00
; SSE2-SSSE3-NEXT: movdqa %xmm1, %xmm0
[X86][SSE] Add support for extending bool vectors bitcasted from scalars This patch acts as a reverse to combineBitcastvxi1 - bitcasting a scalar integer to a boolean vector and extending it 'in place' to the requested legal type. Currently this doesn't handle AVX512 at all - but the current mask register approach is lacking for some cases. Differential Revision: https://reviews.llvm.org/D35320 llvm-svn: 314076
2017-09-24 21:42:31 +08:00
; SSE2-SSSE3-NEXT: pand %xmm2, %xmm0
; SSE2-SSSE3-NEXT: pcmpeqd %xmm2, %xmm0
; SSE2-SSSE3-NEXT: movdqa {{.*#+}} xmm2 = [16,32,64,128]
; SSE2-SSSE3-NEXT: pand %xmm2, %xmm1
; SSE2-SSSE3-NEXT: pcmpeqd %xmm2, %xmm1
[X86][SSE] Tests for bitcasting iX integers to vXi1 boolean vectors Including sign/zero extension to legal types llvm-svn: 307301
2017-07-07 03:33:10 +08:00
; SSE2-SSSE3-NEXT: retq
;
; AVX1-LABEL: ext_i8_8i32:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; AVX1: # %bb.0:
[X86][SSE] Add support for extending bool vectors bitcasted from scalars This patch acts as a reverse to combineBitcastvxi1 - bitcasting a scalar integer to a boolean vector and extending it 'in place' to the requested legal type. Currently this doesn't handle AVX512 at all - but the current mask register approach is lacking for some cases. Differential Revision: https://reviews.llvm.org/D35320 llvm-svn: 314076
2017-09-24 21:42:31 +08:00
; AVX1-NEXT: vmovd %edi, %xmm0
; AVX1-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
; AVX1-NEXT: vandps {{.*}}(%rip), %ymm0, %ymm0
; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX1-NEXT: vpxor %xmm2, %xmm2, %xmm2
; AVX1-NEXT: vpcmpeqd %xmm2, %xmm1, %xmm1
; AVX1-NEXT: vpcmpeqd %xmm3, %xmm3, %xmm3
; AVX1-NEXT: vpxor %xmm3, %xmm1, %xmm1
; AVX1-NEXT: vpcmpeqd %xmm2, %xmm0, %xmm0
; AVX1-NEXT: vpxor %xmm3, %xmm0, %xmm0
; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
[X86][SSE] Tests for bitcasting iX integers to vXi1 boolean vectors Including sign/zero extension to legal types llvm-svn: 307301
2017-07-07 03:33:10 +08:00
; AVX1-NEXT: retq
;
; AVX2-LABEL: ext_i8_8i32:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; AVX2: # %bb.0:
[X86][SSE] Add support for extending bool vectors bitcasted from scalars This patch acts as a reverse to combineBitcastvxi1 - bitcasting a scalar integer to a boolean vector and extending it 'in place' to the requested legal type. Currently this doesn't handle AVX512 at all - but the current mask register approach is lacking for some cases. Differential Revision: https://reviews.llvm.org/D35320 llvm-svn: 314076
2017-09-24 21:42:31 +08:00
; AVX2-NEXT: vmovd %edi, %xmm0
; AVX2-NEXT: vpbroadcastd %xmm0, %ymm0
; AVX2-NEXT: vmovdqa {{.*#+}} ymm1 = [1,2,4,8,16,32,64,128]
; AVX2-NEXT: vpand %ymm1, %ymm0, %ymm0
; AVX2-NEXT: vpcmpeqd %ymm1, %ymm0, %ymm0
[X86][SSE] Tests for bitcasting iX integers to vXi1 boolean vectors Including sign/zero extension to legal types llvm-svn: 307301
2017-07-07 03:33:10 +08:00
; AVX2-NEXT: retq
;
; AVX512-LABEL: ext_i8_8i32:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; AVX512: # %bb.0:
[X86][SSE] Tests for bitcasting iX integers to vXi1 boolean vectors Including sign/zero extension to legal types llvm-svn: 307301
2017-07-07 03:33:10 +08:00
; AVX512-NEXT: kmovd %edi, %k1
; AVX512-NEXT: vpcmpeqd %ymm0, %ymm0, %ymm0
; AVX512-NEXT: vmovdqa32 %ymm0, %ymm0 {%k1} {z}
; AVX512-NEXT: retq
%1 = bitcast i8 %a0 to <8 x i1>
%2 = sext <8 x i1> %1 to <8 x i32>
ret <8 x i32> %2
}
define <16 x i16> @ext_i16_16i16(i16 %a0) {
; SSE2-SSSE3-LABEL: ext_i16_16i16:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; SSE2-SSSE3: # %bb.0:
[X86][SSE] Add support for extending bool vectors bitcasted from scalars This patch acts as a reverse to combineBitcastvxi1 - bitcasting a scalar integer to a boolean vector and extending it 'in place' to the requested legal type. Currently this doesn't handle AVX512 at all - but the current mask register approach is lacking for some cases. Differential Revision: https://reviews.llvm.org/D35320 llvm-svn: 314076
2017-09-24 21:42:31 +08:00
; SSE2-SSSE3-NEXT: movd %edi, %xmm0
[X86][SSE] Match PSHUFLW/PSHUFHW + PSHUFD vXi16 shuffle patterns (PR34686) As noted in PR34686, we are relying on a PSHUFD+PSHUFLW+PSHUFHW shuffle chain for most general vXi16 unary shuffles. This patch checks for simpler PSHUFLW+PSHUFD and PSHUFHW+PSHUFD cases beforehand, building on some existing code that just handled splat shuffles. By doing so we also prevent premature use of PSHUFB shuffles which can be slower and require the creation/loading of constant shuffle masks. We now have the 'fast-variable-shuffle' option for hardware that prefers combining 2 or more shuffles to VPSHUFB etc. Differential Revision: https://reviews.llvm.org/D38318 llvm-svn: 321553
2017-12-29 22:41:50 +08:00
; SSE2-SSSE3-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,0,2,3,4,5,6,7]
; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm1 = xmm0[0,0,0,0]
[X86][SSE] Add support for extending bool vectors bitcasted from scalars This patch acts as a reverse to combineBitcastvxi1 - bitcasting a scalar integer to a boolean vector and extending it 'in place' to the requested legal type. Currently this doesn't handle AVX512 at all - but the current mask register approach is lacking for some cases. Differential Revision: https://reviews.llvm.org/D35320 llvm-svn: 314076
2017-09-24 21:42:31 +08:00
; SSE2-SSSE3-NEXT: movdqa {{.*#+}} xmm2 = [1,2,4,8,16,32,64,128]
[X86][SSE] Tests for bitcasting iX integers to vXi1 boolean vectors Including sign/zero extension to legal types llvm-svn: 307301
2017-07-07 03:33:10 +08:00
; SSE2-SSSE3-NEXT: movdqa %xmm1, %xmm0
[X86][SSE] Add support for extending bool vectors bitcasted from scalars This patch acts as a reverse to combineBitcastvxi1 - bitcasting a scalar integer to a boolean vector and extending it 'in place' to the requested legal type. Currently this doesn't handle AVX512 at all - but the current mask register approach is lacking for some cases. Differential Revision: https://reviews.llvm.org/D35320 llvm-svn: 314076
2017-09-24 21:42:31 +08:00
; SSE2-SSSE3-NEXT: pand %xmm2, %xmm0
; SSE2-SSSE3-NEXT: pcmpeqw %xmm2, %xmm0
; SSE2-SSSE3-NEXT: movdqa {{.*#+}} xmm2 = [256,512,1024,2048,4096,8192,16384,32768]
; SSE2-SSSE3-NEXT: pand %xmm2, %xmm1
; SSE2-SSSE3-NEXT: pcmpeqw %xmm2, %xmm1
[X86][SSE] Tests for bitcasting iX integers to vXi1 boolean vectors Including sign/zero extension to legal types llvm-svn: 307301
2017-07-07 03:33:10 +08:00
; SSE2-SSSE3-NEXT: retq
;
; AVX1-LABEL: ext_i16_16i16:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; AVX1: # %bb.0:
[X86][SSE] Add support for extending bool vectors bitcasted from scalars This patch acts as a reverse to combineBitcastvxi1 - bitcasting a scalar integer to a boolean vector and extending it 'in place' to the requested legal type. Currently this doesn't handle AVX512 at all - but the current mask register approach is lacking for some cases. Differential Revision: https://reviews.llvm.org/D35320 llvm-svn: 314076
2017-09-24 21:42:31 +08:00
; AVX1-NEXT: vmovd %edi, %xmm0
[X86][SSE] Match PSHUFLW/PSHUFHW + PSHUFD vXi16 shuffle patterns (PR34686) As noted in PR34686, we are relying on a PSHUFD+PSHUFLW+PSHUFHW shuffle chain for most general vXi16 unary shuffles. This patch checks for simpler PSHUFLW+PSHUFD and PSHUFHW+PSHUFD cases beforehand, building on some existing code that just handled splat shuffles. By doing so we also prevent premature use of PSHUFB shuffles which can be slower and require the creation/loading of constant shuffle masks. We now have the 'fast-variable-shuffle' option for hardware that prefers combining 2 or more shuffles to VPSHUFB etc. Differential Revision: https://reviews.llvm.org/D38318 llvm-svn: 321553
2017-12-29 22:41:50 +08:00
; AVX1-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[0,0,2,3,4,5,6,7]
; AVX1-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
[X86][SSE] Add support for extending bool vectors bitcasted from scalars This patch acts as a reverse to combineBitcastvxi1 - bitcasting a scalar integer to a boolean vector and extending it 'in place' to the requested legal type. Currently this doesn't handle AVX512 at all - but the current mask register approach is lacking for some cases. Differential Revision: https://reviews.llvm.org/D35320 llvm-svn: 314076
2017-09-24 21:42:31 +08:00
; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
; AVX1-NEXT: vandps {{.*}}(%rip), %ymm0, %ymm0
; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX1-NEXT: vpxor %xmm2, %xmm2, %xmm2
; AVX1-NEXT: vpcmpeqw %xmm2, %xmm1, %xmm1
; AVX1-NEXT: vpcmpeqd %xmm3, %xmm3, %xmm3
; AVX1-NEXT: vpxor %xmm3, %xmm1, %xmm1
; AVX1-NEXT: vpcmpeqw %xmm2, %xmm0, %xmm0
; AVX1-NEXT: vpxor %xmm3, %xmm0, %xmm0
; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
[X86][SSE] Tests for bitcasting iX integers to vXi1 boolean vectors Including sign/zero extension to legal types llvm-svn: 307301
2017-07-07 03:33:10 +08:00
; AVX1-NEXT: retq
;
; AVX2-LABEL: ext_i16_16i16:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; AVX2: # %bb.0:
[X86][SSE] Add support for extending bool vectors bitcasted from scalars This patch acts as a reverse to combineBitcastvxi1 - bitcasting a scalar integer to a boolean vector and extending it 'in place' to the requested legal type. Currently this doesn't handle AVX512 at all - but the current mask register approach is lacking for some cases. Differential Revision: https://reviews.llvm.org/D35320 llvm-svn: 314076
2017-09-24 21:42:31 +08:00
; AVX2-NEXT: vmovd %edi, %xmm0
; AVX2-NEXT: vpbroadcastw %xmm0, %ymm0
; AVX2-NEXT: vmovdqa {{.*#+}} ymm1 = [1,2,4,8,16,32,64,128,256,512,1024,2048,4096,8192,16384,32768]
; AVX2-NEXT: vpand %ymm1, %ymm0, %ymm0
; AVX2-NEXT: vpcmpeqw %ymm1, %ymm0, %ymm0
[X86][SSE] Tests for bitcasting iX integers to vXi1 boolean vectors Including sign/zero extension to legal types llvm-svn: 307301
2017-07-07 03:33:10 +08:00
; AVX2-NEXT: retq
;
; AVX512-LABEL: ext_i16_16i16:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; AVX512: # %bb.0:
[X86][SSE] Tests for bitcasting iX integers to vXi1 boolean vectors Including sign/zero extension to legal types llvm-svn: 307301
2017-07-07 03:33:10 +08:00
; AVX512-NEXT: kmovd %edi, %k0
; AVX512-NEXT: vpmovm2w %k0, %ymm0
; AVX512-NEXT: retq
%1 = bitcast i16 %a0 to <16 x i1>
%2 = sext <16 x i1> %1 to <16 x i16>
ret <16 x i16> %2
}
define <32 x i8> @ext_i32_32i8(i32 %a0) {
; SSE2-SSSE3-LABEL: ext_i32_32i8:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; SSE2-SSSE3: # %bb.0:
[X86][SSE] Add support for extending bool vectors bitcasted from scalars This patch acts as a reverse to combineBitcastvxi1 - bitcasting a scalar integer to a boolean vector and extending it 'in place' to the requested legal type. Currently this doesn't handle AVX512 at all - but the current mask register approach is lacking for some cases. Differential Revision: https://reviews.llvm.org/D35320 llvm-svn: 314076
2017-09-24 21:42:31 +08:00
; SSE2-SSSE3-NEXT: movd %edi, %xmm1
; SSE2-SSSE3-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
; SSE2-SSSE3-NEXT: pshuflw {{.*#+}} xmm0 = xmm1[0,0,1,1,4,5,6,7]
; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
; SSE2-SSSE3-NEXT: movdqa {{.*#+}} xmm2 = [1,2,4,8,16,32,64,128,1,2,4,8,16,32,64,128]
; SSE2-SSSE3-NEXT: pand %xmm2, %xmm0
; SSE2-SSSE3-NEXT: pcmpeqb %xmm2, %xmm0
; SSE2-SSSE3-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[2,2,3,3,4,5,6,7]
; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,0,1,1]
; SSE2-SSSE3-NEXT: pand %xmm2, %xmm1
; SSE2-SSSE3-NEXT: pcmpeqb %xmm2, %xmm1
[X86][SSE] Tests for bitcasting iX integers to vXi1 boolean vectors Including sign/zero extension to legal types llvm-svn: 307301
2017-07-07 03:33:10 +08:00
; SSE2-SSSE3-NEXT: retq
;
; AVX1-LABEL: ext_i32_32i8:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; AVX1: # %bb.0:
[X86][SSE] Add support for extending bool vectors bitcasted from scalars This patch acts as a reverse to combineBitcastvxi1 - bitcasting a scalar integer to a boolean vector and extending it 'in place' to the requested legal type. Currently this doesn't handle AVX512 at all - but the current mask register approach is lacking for some cases. Differential Revision: https://reviews.llvm.org/D35320 llvm-svn: 314076
2017-09-24 21:42:31 +08:00
; AVX1-NEXT: vmovd %edi, %xmm0
; AVX1-NEXT: vpunpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
; AVX1-NEXT: vpshuflw {{.*#+}} xmm1 = xmm0[0,0,1,1,4,5,6,7]
; AVX1-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[0,0,1,1]
; AVX1-NEXT: vpshuflw {{.*#+}} xmm0 = xmm0[2,2,3,3,4,5,6,7]
; AVX1-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
[X86][SSE] Tests for bitcasting iX integers to vXi1 boolean vectors Including sign/zero extension to legal types llvm-svn: 307301
2017-07-07 03:33:10 +08:00
; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
[X86][SSE] Add support for extending bool vectors bitcasted from scalars This patch acts as a reverse to combineBitcastvxi1 - bitcasting a scalar integer to a boolean vector and extending it 'in place' to the requested legal type. Currently this doesn't handle AVX512 at all - but the current mask register approach is lacking for some cases. Differential Revision: https://reviews.llvm.org/D35320 llvm-svn: 314076
2017-09-24 21:42:31 +08:00
; AVX1-NEXT: vandps {{.*}}(%rip), %ymm0, %ymm0
; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX1-NEXT: vpxor %xmm2, %xmm2, %xmm2
; AVX1-NEXT: vpcmpeqb %xmm2, %xmm1, %xmm1
; AVX1-NEXT: vpcmpeqd %xmm3, %xmm3, %xmm3
; AVX1-NEXT: vpxor %xmm3, %xmm1, %xmm1
; AVX1-NEXT: vpcmpeqb %xmm2, %xmm0, %xmm0
; AVX1-NEXT: vpxor %xmm3, %xmm0, %xmm0
; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
[X86][SSE] Tests for bitcasting iX integers to vXi1 boolean vectors Including sign/zero extension to legal types llvm-svn: 307301
2017-07-07 03:33:10 +08:00
; AVX1-NEXT: retq
;
[X86][AVX] Enable lowerVectorShuffleAsLanePermuteAndPermute v16i16/v32i8 unary shuffle lowering llvm-svn: 344868
2018-10-22 01:07:50 +08:00
; AVX2-LABEL: ext_i32_32i8:
; AVX2: # %bb.0:
; AVX2-NEXT: vmovd %edi, %xmm0
; AVX2-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,1,0,1]
; AVX2-NEXT: vpshufb {{.*#+}} ymm0 = ymm0[0,0,0,0,0,0,0,0,1,1,1,1,1,1,1,1,18,18,18,18,18,18,18,18,19,19,19,19,19,19,19,19]
; AVX2-NEXT: vpbroadcastq {{.*#+}} ymm1 = [9241421688590303745,9241421688590303745,9241421688590303745,9241421688590303745]
; AVX2-NEXT: vpand %ymm1, %ymm0, %ymm0
; AVX2-NEXT: vpcmpeqb %ymm1, %ymm0, %ymm0
; AVX2-NEXT: retq
[X86][SSE] Tests for bitcasting iX integers to vXi1 boolean vectors Including sign/zero extension to legal types llvm-svn: 307301
2017-07-07 03:33:10 +08:00
;
; AVX512-LABEL: ext_i32_32i8:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; AVX512: # %bb.0:
[X86][SSE] Tests for bitcasting iX integers to vXi1 boolean vectors Including sign/zero extension to legal types llvm-svn: 307301
2017-07-07 03:33:10 +08:00
; AVX512-NEXT: kmovd %edi, %k0
; AVX512-NEXT: vpmovm2b %k0, %ymm0
; AVX512-NEXT: retq
%1 = bitcast i32 %a0 to <32 x i1>
%2 = sext <32 x i1> %1 to <32 x i8>
ret <32 x i8> %2
}
;
; 512-bit vectors
;
define <8 x i64> @ext_i8_8i64(i8 %a0) {
; SSE2-SSSE3-LABEL: ext_i8_8i64:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; SSE2-SSSE3: # %bb.0:
Followup on Proposal to move MIR physical register namespace to '$' sigil. Discussed here: http://lists.llvm.org/pipermail/llvm-dev/2018-January/120320.html In preparation for adding support for named vregs we are changing the sigil for physical registers in MIR to '$' from '%'. This will prevent name clashes of named physical register with named vregs. llvm-svn: 323922
2018-02-01 06:04:26 +08:00
; SSE2-SSSE3-NEXT: # kill: def $edi killed $edi def $rdi
[X86][SSE] Add support for extending bool vectors bitcasted from scalars This patch acts as a reverse to combineBitcastvxi1 - bitcasting a scalar integer to a boolean vector and extending it 'in place' to the requested legal type. Currently this doesn't handle AVX512 at all - but the current mask register approach is lacking for some cases. Differential Revision: https://reviews.llvm.org/D35320 llvm-svn: 314076
2017-09-24 21:42:31 +08:00
; SSE2-SSSE3-NEXT: movq %rdi, %xmm0
; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm4 = xmm0[0,1,0,1]
; SSE2-SSSE3-NEXT: movdqa {{.*#+}} xmm0 = [1,2]
; SSE2-SSSE3-NEXT: movdqa %xmm4, %xmm1
; SSE2-SSSE3-NEXT: pand %xmm0, %xmm1
; SSE2-SSSE3-NEXT: pcmpeqd %xmm0, %xmm1
; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm1[1,0,3,2]
; SSE2-SSSE3-NEXT: pand %xmm1, %xmm0
; SSE2-SSSE3-NEXT: movdqa {{.*#+}} xmm1 = [4,8]
; SSE2-SSSE3-NEXT: movdqa %xmm4, %xmm2
; SSE2-SSSE3-NEXT: pand %xmm1, %xmm2
; SSE2-SSSE3-NEXT: pcmpeqd %xmm1, %xmm2
; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm1 = xmm2[1,0,3,2]
; SSE2-SSSE3-NEXT: pand %xmm2, %xmm1
; SSE2-SSSE3-NEXT: movdqa {{.*#+}} xmm2 = [16,32]
; SSE2-SSSE3-NEXT: movdqa %xmm4, %xmm3
; SSE2-SSSE3-NEXT: pand %xmm2, %xmm3
; SSE2-SSSE3-NEXT: pcmpeqd %xmm2, %xmm3
; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm2 = xmm3[1,0,3,2]
; SSE2-SSSE3-NEXT: pand %xmm3, %xmm2
; SSE2-SSSE3-NEXT: movdqa {{.*#+}} xmm3 = [64,128]
; SSE2-SSSE3-NEXT: pand %xmm3, %xmm4
; SSE2-SSSE3-NEXT: pcmpeqd %xmm3, %xmm4
; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm3 = xmm4[1,0,3,2]
; SSE2-SSSE3-NEXT: pand %xmm4, %xmm3
[X86][SSE] Tests for bitcasting iX integers to vXi1 boolean vectors Including sign/zero extension to legal types llvm-svn: 307301
2017-07-07 03:33:10 +08:00
; SSE2-SSSE3-NEXT: retq
;
; AVX1-LABEL: ext_i8_8i64:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; AVX1: # %bb.0:
Followup on Proposal to move MIR physical register namespace to '$' sigil. Discussed here: http://lists.llvm.org/pipermail/llvm-dev/2018-January/120320.html In preparation for adding support for named vregs we are changing the sigil for physical registers in MIR to '$' from '%'. This will prevent name clashes of named physical register with named vregs. llvm-svn: 323922
2018-02-01 06:04:26 +08:00
; AVX1-NEXT: # kill: def $edi killed $edi def $rdi
[X86][SSE] Add support for extending bool vectors bitcasted from scalars This patch acts as a reverse to combineBitcastvxi1 - bitcasting a scalar integer to a boolean vector and extending it 'in place' to the requested legal type. Currently this doesn't handle AVX512 at all - but the current mask register approach is lacking for some cases. Differential Revision: https://reviews.llvm.org/D35320 llvm-svn: 314076
2017-09-24 21:42:31 +08:00
; AVX1-NEXT: vmovq %rdi, %xmm0
; AVX1-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[0,1,0,1]
; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm1
; AVX1-NEXT: vandps {{.*}}(%rip), %ymm1, %ymm0
; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm2
; AVX1-NEXT: vpxor %xmm3, %xmm3, %xmm3
; AVX1-NEXT: vpcmpeqq %xmm3, %xmm2, %xmm2
; AVX1-NEXT: vpcmpeqd %xmm4, %xmm4, %xmm4
; AVX1-NEXT: vpxor %xmm4, %xmm2, %xmm2
; AVX1-NEXT: vpcmpeqq %xmm3, %xmm0, %xmm0
; AVX1-NEXT: vpxor %xmm4, %xmm0, %xmm0
; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
; AVX1-NEXT: vandps {{.*}}(%rip), %ymm1, %ymm1
; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm2
; AVX1-NEXT: vpcmpeqq %xmm3, %xmm2, %xmm2
; AVX1-NEXT: vpxor %xmm4, %xmm2, %xmm2
; AVX1-NEXT: vpcmpeqq %xmm3, %xmm1, %xmm1
; AVX1-NEXT: vpxor %xmm4, %xmm1, %xmm1
; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm1, %ymm1
[X86][SSE] Tests for bitcasting iX integers to vXi1 boolean vectors Including sign/zero extension to legal types llvm-svn: 307301
2017-07-07 03:33:10 +08:00
; AVX1-NEXT: retq
;
; AVX2-LABEL: ext_i8_8i64:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; AVX2: # %bb.0:
Followup on Proposal to move MIR physical register namespace to '$' sigil. Discussed here: http://lists.llvm.org/pipermail/llvm-dev/2018-January/120320.html In preparation for adding support for named vregs we are changing the sigil for physical registers in MIR to '$' from '%'. This will prevent name clashes of named physical register with named vregs. llvm-svn: 323922
2018-02-01 06:04:26 +08:00
; AVX2-NEXT: # kill: def $edi killed $edi def $rdi
[X86][SSE] Add support for extending bool vectors bitcasted from scalars This patch acts as a reverse to combineBitcastvxi1 - bitcasting a scalar integer to a boolean vector and extending it 'in place' to the requested legal type. Currently this doesn't handle AVX512 at all - but the current mask register approach is lacking for some cases. Differential Revision: https://reviews.llvm.org/D35320 llvm-svn: 314076
2017-09-24 21:42:31 +08:00
; AVX2-NEXT: vmovq %rdi, %xmm0
; AVX2-NEXT: vpbroadcastq %xmm0, %ymm1
; AVX2-NEXT: vmovdqa {{.*#+}} ymm0 = [1,2,4,8]
; AVX2-NEXT: vpand %ymm0, %ymm1, %ymm2
; AVX2-NEXT: vpcmpeqq %ymm0, %ymm2, %ymm0
; AVX2-NEXT: vmovdqa {{.*#+}} ymm2 = [16,32,64,128]
; AVX2-NEXT: vpand %ymm2, %ymm1, %ymm1
; AVX2-NEXT: vpcmpeqq %ymm2, %ymm1, %ymm1
[X86][SSE] Tests for bitcasting iX integers to vXi1 boolean vectors Including sign/zero extension to legal types llvm-svn: 307301
2017-07-07 03:33:10 +08:00
; AVX2-NEXT: retq
;
; AVX512-LABEL: ext_i8_8i64:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; AVX512: # %bb.0:
[X86][SSE] Tests for bitcasting iX integers to vXi1 boolean vectors Including sign/zero extension to legal types llvm-svn: 307301
2017-07-07 03:33:10 +08:00
; AVX512-NEXT: kmovd %edi, %k1
; AVX512-NEXT: vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
; AVX512-NEXT: retq
%1 = bitcast i8 %a0 to <8 x i1>
%2 = sext <8 x i1> %1 to <8 x i64>
ret <8 x i64> %2
}
define <16 x i32> @ext_i16_16i32(i16 %a0) {
; SSE2-SSSE3-LABEL: ext_i16_16i32:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; SSE2-SSSE3: # %bb.0:
[X86][SSE] Add support for extending bool vectors bitcasted from scalars This patch acts as a reverse to combineBitcastvxi1 - bitcasting a scalar integer to a boolean vector and extending it 'in place' to the requested legal type. Currently this doesn't handle AVX512 at all - but the current mask register approach is lacking for some cases. Differential Revision: https://reviews.llvm.org/D35320 llvm-svn: 314076
2017-09-24 21:42:31 +08:00
; SSE2-SSSE3-NEXT: movd %edi, %xmm0
; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm3 = xmm0[0,0,0,0]
; SSE2-SSSE3-NEXT: movdqa {{.*#+}} xmm1 = [1,2,4,8]
; SSE2-SSSE3-NEXT: movdqa %xmm3, %xmm0
; SSE2-SSSE3-NEXT: pand %xmm1, %xmm0
; SSE2-SSSE3-NEXT: pcmpeqd %xmm1, %xmm0
; SSE2-SSSE3-NEXT: movdqa {{.*#+}} xmm2 = [16,32,64,128]
[X86][SSE] Tests for bitcasting iX integers to vXi1 boolean vectors Including sign/zero extension to legal types llvm-svn: 307301
2017-07-07 03:33:10 +08:00
; SSE2-SSSE3-NEXT: movdqa %xmm3, %xmm1
[X86][SSE] Add support for extending bool vectors bitcasted from scalars This patch acts as a reverse to combineBitcastvxi1 - bitcasting a scalar integer to a boolean vector and extending it 'in place' to the requested legal type. Currently this doesn't handle AVX512 at all - but the current mask register approach is lacking for some cases. Differential Revision: https://reviews.llvm.org/D35320 llvm-svn: 314076
2017-09-24 21:42:31 +08:00
; SSE2-SSSE3-NEXT: pand %xmm2, %xmm1
; SSE2-SSSE3-NEXT: pcmpeqd %xmm2, %xmm1
; SSE2-SSSE3-NEXT: movdqa {{.*#+}} xmm4 = [256,512,1024,2048]
[X86][SSE] Tests for bitcasting iX integers to vXi1 boolean vectors Including sign/zero extension to legal types llvm-svn: 307301
2017-07-07 03:33:10 +08:00
; SSE2-SSSE3-NEXT: movdqa %xmm3, %xmm2
[X86][SSE] Add support for extending bool vectors bitcasted from scalars This patch acts as a reverse to combineBitcastvxi1 - bitcasting a scalar integer to a boolean vector and extending it 'in place' to the requested legal type. Currently this doesn't handle AVX512 at all - but the current mask register approach is lacking for some cases. Differential Revision: https://reviews.llvm.org/D35320 llvm-svn: 314076
2017-09-24 21:42:31 +08:00
; SSE2-SSSE3-NEXT: pand %xmm4, %xmm2
; SSE2-SSSE3-NEXT: pcmpeqd %xmm4, %xmm2
; SSE2-SSSE3-NEXT: movdqa {{.*#+}} xmm4 = [4096,8192,16384,32768]
; SSE2-SSSE3-NEXT: pand %xmm4, %xmm3
; SSE2-SSSE3-NEXT: pcmpeqd %xmm4, %xmm3
[X86][SSE] Tests for bitcasting iX integers to vXi1 boolean vectors Including sign/zero extension to legal types llvm-svn: 307301
2017-07-07 03:33:10 +08:00
; SSE2-SSSE3-NEXT: retq
;
; AVX1-LABEL: ext_i16_16i32:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; AVX1: # %bb.0:
[X86][SSE] Add support for extending bool vectors bitcasted from scalars This patch acts as a reverse to combineBitcastvxi1 - bitcasting a scalar integer to a boolean vector and extending it 'in place' to the requested legal type. Currently this doesn't handle AVX512 at all - but the current mask register approach is lacking for some cases. Differential Revision: https://reviews.llvm.org/D35320 llvm-svn: 314076
2017-09-24 21:42:31 +08:00
; AVX1-NEXT: vmovd %edi, %xmm0
; AVX1-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm1
; AVX1-NEXT: vandps {{.*}}(%rip), %ymm1, %ymm0
; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm2
; AVX1-NEXT: vpxor %xmm3, %xmm3, %xmm3
; AVX1-NEXT: vpcmpeqd %xmm3, %xmm2, %xmm2
; AVX1-NEXT: vpcmpeqd %xmm4, %xmm4, %xmm4
; AVX1-NEXT: vpxor %xmm4, %xmm2, %xmm2
; AVX1-NEXT: vpcmpeqd %xmm3, %xmm0, %xmm0
; AVX1-NEXT: vpxor %xmm4, %xmm0, %xmm0
; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
; AVX1-NEXT: vandps {{.*}}(%rip), %ymm1, %ymm1
; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm2
; AVX1-NEXT: vpcmpeqd %xmm3, %xmm2, %xmm2
; AVX1-NEXT: vpxor %xmm4, %xmm2, %xmm2
; AVX1-NEXT: vpcmpeqd %xmm3, %xmm1, %xmm1
; AVX1-NEXT: vpxor %xmm4, %xmm1, %xmm1
; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm1, %ymm1
[X86][SSE] Tests for bitcasting iX integers to vXi1 boolean vectors Including sign/zero extension to legal types llvm-svn: 307301
2017-07-07 03:33:10 +08:00
; AVX1-NEXT: retq
;
; AVX2-LABEL: ext_i16_16i32:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; AVX2: # %bb.0:
[X86][SSE] Add support for extending bool vectors bitcasted from scalars This patch acts as a reverse to combineBitcastvxi1 - bitcasting a scalar integer to a boolean vector and extending it 'in place' to the requested legal type. Currently this doesn't handle AVX512 at all - but the current mask register approach is lacking for some cases. Differential Revision: https://reviews.llvm.org/D35320 llvm-svn: 314076
2017-09-24 21:42:31 +08:00
; AVX2-NEXT: vmovd %edi, %xmm0
; AVX2-NEXT: vpbroadcastd %xmm0, %ymm1
; AVX2-NEXT: vmovdqa {{.*#+}} ymm0 = [1,2,4,8,16,32,64,128]
; AVX2-NEXT: vpand %ymm0, %ymm1, %ymm2
; AVX2-NEXT: vpcmpeqd %ymm0, %ymm2, %ymm0
; AVX2-NEXT: vmovdqa {{.*#+}} ymm2 = [256,512,1024,2048,4096,8192,16384,32768]
; AVX2-NEXT: vpand %ymm2, %ymm1, %ymm1
; AVX2-NEXT: vpcmpeqd %ymm2, %ymm1, %ymm1
[X86][SSE] Tests for bitcasting iX integers to vXi1 boolean vectors Including sign/zero extension to legal types llvm-svn: 307301
2017-07-07 03:33:10 +08:00
; AVX2-NEXT: retq
;
; AVX512-LABEL: ext_i16_16i32:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; AVX512: # %bb.0:
[X86][SSE] Tests for bitcasting iX integers to vXi1 boolean vectors Including sign/zero extension to legal types llvm-svn: 307301
2017-07-07 03:33:10 +08:00
; AVX512-NEXT: kmovd %edi, %k1
; AVX512-NEXT: vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
; AVX512-NEXT: retq
%1 = bitcast i16 %a0 to <16 x i1>
%2 = sext <16 x i1> %1 to <16 x i32>
ret <16 x i32> %2
}
define <32 x i16> @ext_i32_32i16(i32 %a0) {
; SSE2-SSSE3-LABEL: ext_i32_32i16:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; SSE2-SSSE3: # %bb.0:
[X86][SSE] Add support for extending bool vectors bitcasted from scalars This patch acts as a reverse to combineBitcastvxi1 - bitcasting a scalar integer to a boolean vector and extending it 'in place' to the requested legal type. Currently this doesn't handle AVX512 at all - but the current mask register approach is lacking for some cases. Differential Revision: https://reviews.llvm.org/D35320 llvm-svn: 314076
2017-09-24 21:42:31 +08:00
; SSE2-SSSE3-NEXT: movd %edi, %xmm2
[X86][SSE] Match PSHUFLW/PSHUFHW + PSHUFD vXi16 shuffle patterns (PR34686) As noted in PR34686, we are relying on a PSHUFD+PSHUFLW+PSHUFHW shuffle chain for most general vXi16 unary shuffles. This patch checks for simpler PSHUFLW+PSHUFD and PSHUFHW+PSHUFD cases beforehand, building on some existing code that just handled splat shuffles. By doing so we also prevent premature use of PSHUFB shuffles which can be slower and require the creation/loading of constant shuffle masks. We now have the 'fast-variable-shuffle' option for hardware that prefers combining 2 or more shuffles to VPSHUFB etc. Differential Revision: https://reviews.llvm.org/D38318 llvm-svn: 321553
2017-12-29 22:41:50 +08:00
; SSE2-SSSE3-NEXT: pshuflw {{.*#+}} xmm0 = xmm2[0,0,2,3,4,5,6,7]
; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm1 = xmm0[0,0,0,0]
[X86][SSE] Add support for extending bool vectors bitcasted from scalars This patch acts as a reverse to combineBitcastvxi1 - bitcasting a scalar integer to a boolean vector and extending it 'in place' to the requested legal type. Currently this doesn't handle AVX512 at all - but the current mask register approach is lacking for some cases. Differential Revision: https://reviews.llvm.org/D35320 llvm-svn: 314076
2017-09-24 21:42:31 +08:00
; SSE2-SSSE3-NEXT: movdqa {{.*#+}} xmm4 = [1,2,4,8,16,32,64,128]
[X86][SSE] Tests for bitcasting iX integers to vXi1 boolean vectors Including sign/zero extension to legal types llvm-svn: 307301
2017-07-07 03:33:10 +08:00
; SSE2-SSSE3-NEXT: movdqa %xmm1, %xmm0
[X86][SSE] Add support for extending bool vectors bitcasted from scalars This patch acts as a reverse to combineBitcastvxi1 - bitcasting a scalar integer to a boolean vector and extending it 'in place' to the requested legal type. Currently this doesn't handle AVX512 at all - but the current mask register approach is lacking for some cases. Differential Revision: https://reviews.llvm.org/D35320 llvm-svn: 314076
2017-09-24 21:42:31 +08:00
; SSE2-SSSE3-NEXT: pand %xmm4, %xmm0
; SSE2-SSSE3-NEXT: pcmpeqw %xmm4, %xmm0
; SSE2-SSSE3-NEXT: movdqa {{.*#+}} xmm5 = [256,512,1024,2048,4096,8192,16384,32768]
; SSE2-SSSE3-NEXT: pand %xmm5, %xmm1
; SSE2-SSSE3-NEXT: pcmpeqw %xmm5, %xmm1
[X86][SSE] Match PSHUFLW/PSHUFHW + PSHUFD vXi16 shuffle patterns (PR34686) As noted in PR34686, we are relying on a PSHUFD+PSHUFLW+PSHUFHW shuffle chain for most general vXi16 unary shuffles. This patch checks for simpler PSHUFLW+PSHUFD and PSHUFHW+PSHUFD cases beforehand, building on some existing code that just handled splat shuffles. By doing so we also prevent premature use of PSHUFB shuffles which can be slower and require the creation/loading of constant shuffle masks. We now have the 'fast-variable-shuffle' option for hardware that prefers combining 2 or more shuffles to VPSHUFB etc. Differential Revision: https://reviews.llvm.org/D38318 llvm-svn: 321553
2017-12-29 22:41:50 +08:00
; SSE2-SSSE3-NEXT: pshuflw {{.*#+}} xmm2 = xmm2[1,1,2,3,4,5,6,7]
; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm3 = xmm2[0,0,0,0]
[X86][SSE] Tests for bitcasting iX integers to vXi1 boolean vectors Including sign/zero extension to legal types llvm-svn: 307301
2017-07-07 03:33:10 +08:00
; SSE2-SSSE3-NEXT: movdqa %xmm3, %xmm2
[X86][SSE] Add support for extending bool vectors bitcasted from scalars This patch acts as a reverse to combineBitcastvxi1 - bitcasting a scalar integer to a boolean vector and extending it 'in place' to the requested legal type. Currently this doesn't handle AVX512 at all - but the current mask register approach is lacking for some cases. Differential Revision: https://reviews.llvm.org/D35320 llvm-svn: 314076
2017-09-24 21:42:31 +08:00
; SSE2-SSSE3-NEXT: pand %xmm4, %xmm2
; SSE2-SSSE3-NEXT: pcmpeqw %xmm4, %xmm2
; SSE2-SSSE3-NEXT: pand %xmm5, %xmm3
; SSE2-SSSE3-NEXT: pcmpeqw %xmm5, %xmm3
[X86][SSE] Tests for bitcasting iX integers to vXi1 boolean vectors Including sign/zero extension to legal types llvm-svn: 307301
2017-07-07 03:33:10 +08:00
; SSE2-SSSE3-NEXT: retq
;
; AVX1-LABEL: ext_i32_32i16:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; AVX1: # %bb.0:
[X86][SSE] Add support for extending bool vectors bitcasted from scalars This patch acts as a reverse to combineBitcastvxi1 - bitcasting a scalar integer to a boolean vector and extending it 'in place' to the requested legal type. Currently this doesn't handle AVX512 at all - but the current mask register approach is lacking for some cases. Differential Revision: https://reviews.llvm.org/D35320 llvm-svn: 314076
2017-09-24 21:42:31 +08:00
; AVX1-NEXT: vmovd %edi, %xmm1
[X86][SSE] Match PSHUFLW/PSHUFHW + PSHUFD vXi16 shuffle patterns (PR34686) As noted in PR34686, we are relying on a PSHUFD+PSHUFLW+PSHUFHW shuffle chain for most general vXi16 unary shuffles. This patch checks for simpler PSHUFLW+PSHUFD and PSHUFHW+PSHUFD cases beforehand, building on some existing code that just handled splat shuffles. By doing so we also prevent premature use of PSHUFB shuffles which can be slower and require the creation/loading of constant shuffle masks. We now have the 'fast-variable-shuffle' option for hardware that prefers combining 2 or more shuffles to VPSHUFB etc. Differential Revision: https://reviews.llvm.org/D38318 llvm-svn: 321553
2017-12-29 22:41:50 +08:00
; AVX1-NEXT: vpshuflw {{.*#+}} xmm0 = xmm1[0,0,2,3,4,5,6,7]
; AVX1-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
[X86][SSE] Add support for extending bool vectors bitcasted from scalars This patch acts as a reverse to combineBitcastvxi1 - bitcasting a scalar integer to a boolean vector and extending it 'in place' to the requested legal type. Currently this doesn't handle AVX512 at all - but the current mask register approach is lacking for some cases. Differential Revision: https://reviews.llvm.org/D35320 llvm-svn: 314076
2017-09-24 21:42:31 +08:00
; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
; AVX1-NEXT: vmovaps {{.*#+}} ymm2 = [1,2,4,8,16,32,64,128,256,512,1024,2048,4096,8192,16384,32768]
; AVX1-NEXT: vandps %ymm2, %ymm0, %ymm0
; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm3
; AVX1-NEXT: vpxor %xmm4, %xmm4, %xmm4
; AVX1-NEXT: vpcmpeqw %xmm4, %xmm3, %xmm3
; AVX1-NEXT: vpcmpeqd %xmm5, %xmm5, %xmm5
; AVX1-NEXT: vpxor %xmm5, %xmm3, %xmm3
; AVX1-NEXT: vpcmpeqw %xmm4, %xmm0, %xmm0
; AVX1-NEXT: vpxor %xmm5, %xmm0, %xmm0
; AVX1-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm0
[X86][SSE] Match PSHUFLW/PSHUFHW + PSHUFD vXi16 shuffle patterns (PR34686) As noted in PR34686, we are relying on a PSHUFD+PSHUFLW+PSHUFHW shuffle chain for most general vXi16 unary shuffles. This patch checks for simpler PSHUFLW+PSHUFD and PSHUFHW+PSHUFD cases beforehand, building on some existing code that just handled splat shuffles. By doing so we also prevent premature use of PSHUFB shuffles which can be slower and require the creation/loading of constant shuffle masks. We now have the 'fast-variable-shuffle' option for hardware that prefers combining 2 or more shuffles to VPSHUFB etc. Differential Revision: https://reviews.llvm.org/D38318 llvm-svn: 321553
2017-12-29 22:41:50 +08:00
; AVX1-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[1,1,2,3,4,5,6,7]
; AVX1-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[0,0,0,0]
[X86][SSE] Add support for extending bool vectors bitcasted from scalars This patch acts as a reverse to combineBitcastvxi1 - bitcasting a scalar integer to a boolean vector and extending it 'in place' to the requested legal type. Currently this doesn't handle AVX512 at all - but the current mask register approach is lacking for some cases. Differential Revision: https://reviews.llvm.org/D35320 llvm-svn: 314076
2017-09-24 21:42:31 +08:00
; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm1, %ymm1
; AVX1-NEXT: vandps %ymm2, %ymm1, %ymm1
; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm2
; AVX1-NEXT: vpcmpeqw %xmm4, %xmm2, %xmm2
; AVX1-NEXT: vpxor %xmm5, %xmm2, %xmm2
; AVX1-NEXT: vpcmpeqw %xmm4, %xmm1, %xmm1
; AVX1-NEXT: vpxor %xmm5, %xmm1, %xmm1
; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm1, %ymm1
[X86][SSE] Tests for bitcasting iX integers to vXi1 boolean vectors Including sign/zero extension to legal types llvm-svn: 307301
2017-07-07 03:33:10 +08:00
; AVX1-NEXT: retq
;
; AVX2-LABEL: ext_i32_32i16:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; AVX2: # %bb.0:
[X86][SSE] Tests for bitcasting iX integers to vXi1 boolean vectors Including sign/zero extension to legal types llvm-svn: 307301
2017-07-07 03:33:10 +08:00
; AVX2-NEXT: vmovd %edi, %xmm0
[X86][SSE] Add support for extending bool vectors bitcasted from scalars This patch acts as a reverse to combineBitcastvxi1 - bitcasting a scalar integer to a boolean vector and extending it 'in place' to the requested legal type. Currently this doesn't handle AVX512 at all - but the current mask register approach is lacking for some cases. Differential Revision: https://reviews.llvm.org/D35320 llvm-svn: 314076
2017-09-24 21:42:31 +08:00
; AVX2-NEXT: vpbroadcastw %xmm0, %ymm0
; AVX2-NEXT: vmovdqa {{.*#+}} ymm1 = [1,2,4,8,16,32,64,128,256,512,1024,2048,4096,8192,16384,32768]
; AVX2-NEXT: vpand %ymm1, %ymm0, %ymm0
; AVX2-NEXT: vpcmpeqw %ymm1, %ymm0, %ymm0
; AVX2-NEXT: shrl $16, %edi
; AVX2-NEXT: vmovd %edi, %xmm2
; AVX2-NEXT: vpbroadcastw %xmm2, %ymm2
; AVX2-NEXT: vpand %ymm1, %ymm2, %ymm2
; AVX2-NEXT: vpcmpeqw %ymm1, %ymm2, %ymm1
[X86][SSE] Tests for bitcasting iX integers to vXi1 boolean vectors Including sign/zero extension to legal types llvm-svn: 307301
2017-07-07 03:33:10 +08:00
; AVX2-NEXT: retq
;
; AVX512-LABEL: ext_i32_32i16:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; AVX512: # %bb.0:
[X86][SSE] Tests for bitcasting iX integers to vXi1 boolean vectors Including sign/zero extension to legal types llvm-svn: 307301
2017-07-07 03:33:10 +08:00
; AVX512-NEXT: kmovd %edi, %k0
; AVX512-NEXT: vpmovm2w %k0, %zmm0
; AVX512-NEXT: retq
%1 = bitcast i32 %a0 to <32 x i1>
%2 = sext <32 x i1> %1 to <32 x i16>
ret <32 x i16> %2
}
define <64 x i8> @ext_i64_64i8(i64 %a0) {
; SSE2-SSSE3-LABEL: ext_i64_64i8:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; SSE2-SSSE3: # %bb.0:
[X86][SSE] Add support for extending bool vectors bitcasted from scalars This patch acts as a reverse to combineBitcastvxi1 - bitcasting a scalar integer to a boolean vector and extending it 'in place' to the requested legal type. Currently this doesn't handle AVX512 at all - but the current mask register approach is lacking for some cases. Differential Revision: https://reviews.llvm.org/D35320 llvm-svn: 314076
2017-09-24 21:42:31 +08:00
; SSE2-SSSE3-NEXT: movq %rdi, %xmm3
; SSE2-SSSE3-NEXT: punpcklbw {{.*#+}} xmm3 = xmm3[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
; SSE2-SSSE3-NEXT: pshuflw {{.*#+}} xmm0 = xmm3[0,0,1,1,4,5,6,7]
; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
; SSE2-SSSE3-NEXT: movdqa {{.*#+}} xmm4 = [1,2,4,8,16,32,64,128,1,2,4,8,16,32,64,128]
; SSE2-SSSE3-NEXT: pand %xmm4, %xmm0
; SSE2-SSSE3-NEXT: pcmpeqb %xmm4, %xmm0
; SSE2-SSSE3-NEXT: pshuflw {{.*#+}} xmm1 = xmm3[2,2,3,3,4,5,6,7]
; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,0,1,1]
; SSE2-SSSE3-NEXT: pand %xmm4, %xmm1
; SSE2-SSSE3-NEXT: pcmpeqb %xmm4, %xmm1
; SSE2-SSSE3-NEXT: pshufhw {{.*#+}} xmm2 = xmm3[0,1,2,3,4,4,5,5]
; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm2 = xmm2[2,2,3,3]
; SSE2-SSSE3-NEXT: pand %xmm4, %xmm2
; SSE2-SSSE3-NEXT: pcmpeqb %xmm4, %xmm2
; SSE2-SSSE3-NEXT: pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,6,6,7,7]
; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm3 = xmm3[2,2,3,3]
; SSE2-SSSE3-NEXT: pand %xmm4, %xmm3
; SSE2-SSSE3-NEXT: pcmpeqb %xmm4, %xmm3
[X86][SSE] Tests for bitcasting iX integers to vXi1 boolean vectors Including sign/zero extension to legal types llvm-svn: 307301
2017-07-07 03:33:10 +08:00
; SSE2-SSSE3-NEXT: retq
;
; AVX1-LABEL: ext_i64_64i8:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; AVX1: # %bb.0:
[X86][SSE] Add support for extending bool vectors bitcasted from scalars This patch acts as a reverse to combineBitcastvxi1 - bitcasting a scalar integer to a boolean vector and extending it 'in place' to the requested legal type. Currently this doesn't handle AVX512 at all - but the current mask register approach is lacking for some cases. Differential Revision: https://reviews.llvm.org/D35320 llvm-svn: 314076
2017-09-24 21:42:31 +08:00
; AVX1-NEXT: vmovq %rdi, %xmm0
; AVX1-NEXT: vpunpcklbw {{.*#+}} xmm1 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
; AVX1-NEXT: vpshuflw {{.*#+}} xmm0 = xmm1[0,0,1,1,4,5,6,7]
; AVX1-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
; AVX1-NEXT: vpshuflw {{.*#+}} xmm2 = xmm1[2,2,3,3,4,5,6,7]
; AVX1-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[0,0,1,1]
; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
[X86] Stop promoting vector and/or/xor/andn to vXi64. These promotions add additional bitcasts to the SelectionDAG that can pessimize computeKnownBits/computeNumSignBits. It also seems to interfere with broadcast formation. This patch removes the promotion and adds isel patterns instead. The increased table size is more than I would like, but hopefully we can find some canonicalizations or other tricks to start pruning out patterns going forward. Differential Revision: https://reviews.llvm.org/D53268 llvm-svn: 345408
2018-10-27 01:21:26 +08:00
; AVX1-NEXT: vbroadcastsd {{.*#+}} ymm2 = [-1.7939930131212661E-307,-1.7939930131212661E-307,-1.7939930131212661E-307,-1.7939930131212661E-307]
[X86][SSE] Add support for extending bool vectors bitcasted from scalars This patch acts as a reverse to combineBitcastvxi1 - bitcasting a scalar integer to a boolean vector and extending it 'in place' to the requested legal type. Currently this doesn't handle AVX512 at all - but the current mask register approach is lacking for some cases. Differential Revision: https://reviews.llvm.org/D35320 llvm-svn: 314076
2017-09-24 21:42:31 +08:00
; AVX1-NEXT: vandps %ymm2, %ymm0, %ymm0
; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm3
; AVX1-NEXT: vpxor %xmm4, %xmm4, %xmm4
; AVX1-NEXT: vpcmpeqb %xmm4, %xmm3, %xmm3
; AVX1-NEXT: vpcmpeqd %xmm5, %xmm5, %xmm5
; AVX1-NEXT: vpxor %xmm5, %xmm3, %xmm3
; AVX1-NEXT: vpcmpeqb %xmm4, %xmm0, %xmm0
; AVX1-NEXT: vpxor %xmm5, %xmm0, %xmm0
; AVX1-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm0
; AVX1-NEXT: vpshufhw {{.*#+}} xmm3 = xmm1[0,1,2,3,4,4,5,5]
; AVX1-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[2,2,3,3]
; AVX1-NEXT: vpshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,6,7,7]
; AVX1-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[2,2,3,3]
; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm3, %ymm1
; AVX1-NEXT: vandps %ymm2, %ymm1, %ymm1
; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm2
; AVX1-NEXT: vpcmpeqb %xmm4, %xmm2, %xmm2
; AVX1-NEXT: vpxor %xmm5, %xmm2, %xmm2
; AVX1-NEXT: vpcmpeqb %xmm4, %xmm1, %xmm1
; AVX1-NEXT: vpxor %xmm5, %xmm1, %xmm1
[X86][SSE] Tests for bitcasting iX integers to vXi1 boolean vectors Including sign/zero extension to legal types llvm-svn: 307301
2017-07-07 03:33:10 +08:00
; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm1, %ymm1
; AVX1-NEXT: retq
;
[X86][AVX] Enable lowerVectorShuffleAsLanePermuteAndPermute v16i16/v32i8 unary shuffle lowering llvm-svn: 344868
2018-10-22 01:07:50 +08:00
; AVX2-LABEL: ext_i64_64i8:
; AVX2: # %bb.0:
; AVX2-NEXT: vmovq %rdi, %xmm0
; AVX2-NEXT: vpermq {{.*#+}} ymm1 = ymm0[0,1,0,1]
; AVX2-NEXT: vpshufb {{.*#+}} ymm0 = ymm1[0,0,0,0,0,0,0,0,1,1,1,1,1,1,1,1,18,18,18,18,18,18,18,18,19,19,19,19,19,19,19,19]
; AVX2-NEXT: vpbroadcastq {{.*#+}} ymm2 = [9241421688590303745,9241421688590303745,9241421688590303745,9241421688590303745]
; AVX2-NEXT: vpand %ymm2, %ymm0, %ymm0
; AVX2-NEXT: vpcmpeqb %ymm2, %ymm0, %ymm0
; AVX2-NEXT: vpshufb {{.*#+}} ymm1 = ymm1[4,4,4,4,4,4,4,4,5,5,5,5,5,5,5,5,22,22,22,22,22,22,22,22,23,23,23,23,23,23,23,23]
; AVX2-NEXT: vpand %ymm2, %ymm1, %ymm1
; AVX2-NEXT: vpcmpeqb %ymm2, %ymm1, %ymm1
; AVX2-NEXT: retq
[X86][SSE] Tests for bitcasting iX integers to vXi1 boolean vectors Including sign/zero extension to legal types llvm-svn: 307301
2017-07-07 03:33:10 +08:00
;
; AVX512-LABEL: ext_i64_64i8:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; AVX512: # %bb.0:
[X86][SSE] Tests for bitcasting iX integers to vXi1 boolean vectors Including sign/zero extension to legal types llvm-svn: 307301
2017-07-07 03:33:10 +08:00
; AVX512-NEXT: kmovq %rdi, %k0
; AVX512-NEXT: vpmovm2b %k0, %zmm0
; AVX512-NEXT: retq
%1 = bitcast i64 %a0 to <64 x i1>
%2 = sext <64 x i1> %1 to <64 x i8>
ret <64 x i8> %2
}