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-and-setcc-256.ll

446 lines
17 KiB
LLVM
Raw Normal View History

[X86] Add tests for (ix bitcast (vxi1 and ...)). NFC. To be improved by D33311. llvm-svn: 304171
2017-05-30 03:00:57 +08:00
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
[X86][SSE] Dropped -mcpu from bitcast+setcc mask tests Use triple and attribute only for consistency llvm-svn: 307176
2017-07-06 01:30:30 +08:00
; 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
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2 | FileCheck %s --check-prefixes=AVX12,AVX2
[X86][AVX512] Split AVX512F and AVX512BW bool-vector bitcast tests llvm-svn: 317020
2017-11-01 02:41:48 +08:00
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512f,+avx512vl | FileCheck %s --check-prefix=AVX512 --check-prefix=AVX512F
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512f,+avx512vl,+avx512bw | FileCheck %s --check-prefix=AVX512 --check-prefix=AVX512BW
[X86] Add tests for (ix bitcast (vxi1 and ...)). NFC. To be improved by D33311. llvm-svn: 304171
2017-05-30 03:00:57 +08:00
define i4 @v4i64(<4 x i64> %a, <4 x i64> %b, <4 x i64> %c, <4 x i64> %d) {
[X86] Match bitcast of vxi1 to pmovmsk Summary: Add an early combine to match patterns such as: (i16 bitcast (v16i1 x)) -> (i16 movmsk (v16i8 sext (v16i1 x))) This combine needs to happen early enough before type-legalization scalarizes the result of the setcc. Reviewers: igorb, craig.topper, RKSimon Subscribers: delena, llvm-commits Differential Revision: https://reviews.llvm.org/D33311 llvm-svn: 304406
2017-06-01 19:27:57 +08:00
; SSE2-SSSE3-LABEL: v4i64:
[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] When lowering unsigned v2i64 setcc without SSE42, flip the sign bits in the v2i64 type then bitcast to v4i32. This may give slightly better opportunities for DAG combine to simplify with the operations before the setcc. It also matches the type the xors will eventually be promoted to anyway so it saves a legalization step. Almost all of the test changes are because our constant pool entry is now v2i64 instead of v4i32 on 64-bit targets. On 32-bit targets getConstant should be emitting a v4i32 build_vector and a v4i32->v2i64 bitcast. There are a couple test cases where it appears we now combine a bitwise not with one of these xors which caused a new constant vector to be generated. This prevented a constant pool entry from being shared. But if that's an issue we're concerned about, it seems we need to address it another way that just relying a bitcast to hide it. This came about from experiments I've been trying with pushing the promotion of and/or/xor to vXi64 later than LegalizeVectorOps where it is today. We run LegalizeVectorOps in a bottom up order. So the and/or/xor are promoted before their users are legalized. The bitcasts added for the promotion act as a barrier to computeKnownBits if we try to use it during vector legalization of a later operation. So by moving the promotion out we can hopefully get better results from computeKnownBits/computeNumSignBits like in LowerTruncate on AVX512. I've also looked at running LegalizeVectorOps in a top down order like LegalizeDAG, but thats showing some other issues. llvm-svn: 344071
2018-10-10 03:05:50 +08:00
; SSE2-SSSE3-NEXT: movdqa {{.*#+}} xmm8 = [2147483648,2147483648]
[X86] Match bitcast of vxi1 to pmovmsk Summary: Add an early combine to match patterns such as: (i16 bitcast (v16i1 x)) -> (i16 movmsk (v16i8 sext (v16i1 x))) This combine needs to happen early enough before type-legalization scalarizes the result of the setcc. Reviewers: igorb, craig.topper, RKSimon Subscribers: delena, llvm-commits Differential Revision: https://reviews.llvm.org/D33311 llvm-svn: 304406
2017-06-01 19:27:57 +08:00
; SSE2-SSSE3-NEXT: pxor %xmm8, %xmm3
; SSE2-SSSE3-NEXT: pxor %xmm8, %xmm1
; SSE2-SSSE3-NEXT: movdqa %xmm1, %xmm9
; SSE2-SSSE3-NEXT: pcmpgtd %xmm3, %xmm9
; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm10 = xmm9[0,0,2,2]
; SSE2-SSSE3-NEXT: pcmpeqd %xmm3, %xmm1
; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
; SSE2-SSSE3-NEXT: pand %xmm10, %xmm1
; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm3 = xmm9[1,1,3,3]
; SSE2-SSSE3-NEXT: por %xmm1, %xmm3
; SSE2-SSSE3-NEXT: pxor %xmm8, %xmm2
; SSE2-SSSE3-NEXT: pxor %xmm8, %xmm0
; SSE2-SSSE3-NEXT: movdqa %xmm0, %xmm1
; SSE2-SSSE3-NEXT: pcmpgtd %xmm2, %xmm1
; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm9 = xmm1[0,0,2,2]
; SSE2-SSSE3-NEXT: pcmpeqd %xmm2, %xmm0
; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm2 = xmm0[1,1,3,3]
; SSE2-SSSE3-NEXT: pand %xmm9, %xmm2
; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm0 = xmm1[1,1,3,3]
; SSE2-SSSE3-NEXT: por %xmm2, %xmm0
; SSE2-SSSE3-NEXT: shufps {{.*#+}} xmm0 = xmm0[0,2],xmm3[0,2]
; SSE2-SSSE3-NEXT: pxor %xmm8, %xmm7
; SSE2-SSSE3-NEXT: pxor %xmm8, %xmm5
; SSE2-SSSE3-NEXT: movdqa %xmm5, %xmm1
; SSE2-SSSE3-NEXT: pcmpgtd %xmm7, %xmm1
; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm2 = xmm1[0,0,2,2]
; SSE2-SSSE3-NEXT: pcmpeqd %xmm7, %xmm5
; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm3 = xmm5[1,1,3,3]
; SSE2-SSSE3-NEXT: pand %xmm2, %xmm3
; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
; SSE2-SSSE3-NEXT: por %xmm3, %xmm1
; SSE2-SSSE3-NEXT: pxor %xmm8, %xmm6
; SSE2-SSSE3-NEXT: pxor %xmm8, %xmm4
; SSE2-SSSE3-NEXT: movdqa %xmm4, %xmm2
; SSE2-SSSE3-NEXT: pcmpgtd %xmm6, %xmm2
; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
; SSE2-SSSE3-NEXT: pcmpeqd %xmm6, %xmm4
; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm4 = xmm4[1,1,3,3]
; SSE2-SSSE3-NEXT: pand %xmm3, %xmm4
; SSE2-SSSE3-NEXT: pshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
; SSE2-SSSE3-NEXT: por %xmm4, %xmm2
; SSE2-SSSE3-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,2],xmm1[0,2]
[SelectionDAG] Add BITCAST handling to ComputeNumSignBits for splatted sign bits. For cases where we are BITCASTing to vectors of smaller elements, then if the entire source was a splatted sign (src's NumSignBits == SrcBitWidth) we can say that the dst's NumSignBit == DstBitWidth, as we're just splitting those sign bits across multiple elements. We could generalize this but at the moment the only use case I have is to peek through bitcasts to vector comparison results. Differential Revision: https://reviews.llvm.org/D37849 llvm-svn: 313543
2017-09-19 00:45:05 +08:00
; SSE2-SSSE3-NEXT: andps %xmm0, %xmm2
[X86] Match bitcast of vxi1 to pmovmsk Summary: Add an early combine to match patterns such as: (i16 bitcast (v16i1 x)) -> (i16 movmsk (v16i8 sext (v16i1 x))) This combine needs to happen early enough before type-legalization scalarizes the result of the setcc. Reviewers: igorb, craig.topper, RKSimon Subscribers: delena, llvm-commits Differential Revision: https://reviews.llvm.org/D33311 llvm-svn: 304406
2017-06-01 19:27:57 +08:00
; SSE2-SSSE3-NEXT: movmskps %xmm2, %eax
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 $al killed $al killed $eax
[X86][AVX512] Regenerate tests to remove retl/retq regex These are only testing 64-bit targets so we don't need the regex llvm-svn: 317021
2017-11-01 02:43:24 +08:00
; SSE2-SSSE3-NEXT: retq
[X86] Match bitcast of vxi1 to pmovmsk Summary: Add an early combine to match patterns such as: (i16 bitcast (v16i1 x)) -> (i16 movmsk (v16i8 sext (v16i1 x))) This combine needs to happen early enough before type-legalization scalarizes the result of the setcc. Reviewers: igorb, craig.topper, RKSimon Subscribers: delena, llvm-commits Differential Revision: https://reviews.llvm.org/D33311 llvm-svn: 304406
2017-06-01 19:27:57 +08:00
;
; AVX1-LABEL: v4i64:
[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] Match bitcast of vxi1 to pmovmsk Summary: Add an early combine to match patterns such as: (i16 bitcast (v16i1 x)) -> (i16 movmsk (v16i8 sext (v16i1 x))) This combine needs to happen early enough before type-legalization scalarizes the result of the setcc. Reviewers: igorb, craig.topper, RKSimon Subscribers: delena, llvm-commits Differential Revision: https://reviews.llvm.org/D33311 llvm-svn: 304406
2017-06-01 19:27:57 +08:00
; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm4
; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm5
; AVX1-NEXT: vpcmpgtq %xmm4, %xmm5, %xmm4
; AVX1-NEXT: vpcmpgtq %xmm1, %xmm0, %xmm0
[X86] truncateVectorCompareWithPACKSS - use PACKSSDW/PACKSSWB instead of just PACKSSWB. By using the widest type possible for PACKSS truncation we have a better chance of being able to peek through bitcasts and improves other combines driven by ComputeNumSignBits. llvm-svn: 316448
2017-10-24 23:38:16 +08:00
; AVX1-NEXT: vpackssdw %xmm4, %xmm0, %xmm0
[X86] Match bitcast of vxi1 to pmovmsk Summary: Add an early combine to match patterns such as: (i16 bitcast (v16i1 x)) -> (i16 movmsk (v16i8 sext (v16i1 x))) This combine needs to happen early enough before type-legalization scalarizes the result of the setcc. Reviewers: igorb, craig.topper, RKSimon Subscribers: delena, llvm-commits Differential Revision: https://reviews.llvm.org/D33311 llvm-svn: 304406
2017-06-01 19:27:57 +08:00
; AVX1-NEXT: vextractf128 $1, %ymm3, %xmm1
; AVX1-NEXT: vextractf128 $1, %ymm2, %xmm4
; AVX1-NEXT: vpcmpgtq %xmm1, %xmm4, %xmm1
; AVX1-NEXT: vpcmpgtq %xmm3, %xmm2, %xmm2
[X86] truncateVectorCompareWithPACKSS - use PACKSSDW/PACKSSWB instead of just PACKSSWB. By using the widest type possible for PACKSS truncation we have a better chance of being able to peek through bitcasts and improves other combines driven by ComputeNumSignBits. llvm-svn: 316448
2017-10-24 23:38:16 +08:00
; AVX1-NEXT: vpackssdw %xmm1, %xmm2, %xmm1
[X86] Match bitcast of vxi1 to pmovmsk Summary: Add an early combine to match patterns such as: (i16 bitcast (v16i1 x)) -> (i16 movmsk (v16i8 sext (v16i1 x))) This combine needs to happen early enough before type-legalization scalarizes the result of the setcc. Reviewers: igorb, craig.topper, RKSimon Subscribers: delena, llvm-commits Differential Revision: https://reviews.llvm.org/D33311 llvm-svn: 304406
2017-06-01 19:27:57 +08:00
; AVX1-NEXT: vpand %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vmovmskps %xmm0, %eax
Followup on Proposal to move MIR physical register namespace to '$' 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 $al killed $al killed $eax
[X86] Match bitcast of vxi1 to pmovmsk Summary: Add an early combine to match patterns such as: (i16 bitcast (v16i1 x)) -> (i16 movmsk (v16i8 sext (v16i1 x))) This combine needs to happen early enough before type-legalization scalarizes the result of the setcc. Reviewers: igorb, craig.topper, RKSimon Subscribers: delena, llvm-commits Differential Revision: https://reviews.llvm.org/D33311 llvm-svn: 304406
2017-06-01 19:27:57 +08:00
; AVX1-NEXT: vzeroupper
[X86][AVX512] Regenerate tests to remove retl/retq regex These are only testing 64-bit targets so we don't need the regex llvm-svn: 317021
2017-11-01 02:43:24 +08:00
; AVX1-NEXT: retq
[X86] Match bitcast of vxi1 to pmovmsk Summary: Add an early combine to match patterns such as: (i16 bitcast (v16i1 x)) -> (i16 movmsk (v16i8 sext (v16i1 x))) This combine needs to happen early enough before type-legalization scalarizes the result of the setcc. Reviewers: igorb, craig.topper, RKSimon Subscribers: delena, llvm-commits Differential Revision: https://reviews.llvm.org/D33311 llvm-svn: 304406
2017-06-01 19:27:57 +08:00
;
[X86] Add tests for (ix bitcast (vxi1 and ...)). NFC. To be improved by D33311. llvm-svn: 304171
2017-05-30 03:00:57 +08:00
; AVX2-LABEL: v4i64:
[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] Add tests for (ix bitcast (vxi1 and ...)). NFC. To be improved by D33311. llvm-svn: 304171
2017-05-30 03:00:57 +08:00
; AVX2-NEXT: vpcmpgtq %ymm1, %ymm0, %ymm0
; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1
[X86] truncateVectorCompareWithPACKSS - use PACKSSDW/PACKSSWB instead of just PACKSSWB. By using the widest type possible for PACKSS truncation we have a better chance of being able to peek through bitcasts and improves other combines driven by ComputeNumSignBits. llvm-svn: 316448
2017-10-24 23:38:16 +08:00
; AVX2-NEXT: vpackssdw %xmm1, %xmm0, %xmm0
[X86] Add tests for (ix bitcast (vxi1 and ...)). NFC. To be improved by D33311. llvm-svn: 304171
2017-05-30 03:00:57 +08:00
; AVX2-NEXT: vpcmpgtq %ymm3, %ymm2, %ymm1
; AVX2-NEXT: vextracti128 $1, %ymm1, %xmm2
[X86] truncateVectorCompareWithPACKSS - use PACKSSDW/PACKSSWB instead of just PACKSSWB. By using the widest type possible for PACKSS truncation we have a better chance of being able to peek through bitcasts and improves other combines driven by ComputeNumSignBits. llvm-svn: 316448
2017-10-24 23:38:16 +08:00
; AVX2-NEXT: vpackssdw %xmm2, %xmm1, %xmm1
[X86] Add tests for (ix bitcast (vxi1 and ...)). NFC. To be improved by D33311. llvm-svn: 304171
2017-05-30 03:00:57 +08:00
; AVX2-NEXT: vpand %xmm1, %xmm0, %xmm0
[X86] Match bitcast of vxi1 to pmovmsk Summary: Add an early combine to match patterns such as: (i16 bitcast (v16i1 x)) -> (i16 movmsk (v16i8 sext (v16i1 x))) This combine needs to happen early enough before type-legalization scalarizes the result of the setcc. Reviewers: igorb, craig.topper, RKSimon Subscribers: delena, llvm-commits Differential Revision: https://reviews.llvm.org/D33311 llvm-svn: 304406
2017-06-01 19:27:57 +08:00
; AVX2-NEXT: vmovmskps %xmm0, %eax
Followup on Proposal to move MIR physical register namespace to '$' 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 $al killed $al killed $eax
[X86] Add tests for (ix bitcast (vxi1 and ...)). NFC. To be improved by D33311. llvm-svn: 304171
2017-05-30 03:00:57 +08:00
; AVX2-NEXT: vzeroupper
[X86][AVX512] Regenerate tests to remove retl/retq regex These are only testing 64-bit targets so we don't need the regex llvm-svn: 317021
2017-11-01 02:43:24 +08:00
; AVX2-NEXT: retq
[X86] Add tests for (ix bitcast (vxi1 and ...)). NFC. To be improved by D33311. llvm-svn: 304171
2017-05-30 03:00:57 +08:00
;
[X86][AVX512] Split AVX512F and AVX512BW bool-vector bitcast tests llvm-svn: 317020
2017-11-01 02:41:48 +08:00
; AVX512F-LABEL: v4i64:
[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
; AVX512F: # %bb.0:
[X86][AVX512] Split AVX512F and AVX512BW bool-vector bitcast tests llvm-svn: 317020
2017-11-01 02:41:48 +08:00
; AVX512F-NEXT: vpcmpgtq %ymm1, %ymm0, %k1
; AVX512F-NEXT: vpcmpgtq %ymm3, %ymm2, %k0 {%k1}
; AVX512F-NEXT: kmovw %k0, %eax
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
; AVX512F-NEXT: # kill: def $al killed $al killed $eax
[X86][AVX512] Split AVX512F and AVX512BW bool-vector bitcast tests llvm-svn: 317020
2017-11-01 02:41:48 +08:00
; AVX512F-NEXT: vzeroupper
[X86][AVX512] Regenerate tests to remove retl/retq regex These are only testing 64-bit targets so we don't need the regex llvm-svn: 317021
2017-11-01 02:43:24 +08:00
; AVX512F-NEXT: retq
[X86][AVX512] Split AVX512F and AVX512BW bool-vector bitcast tests llvm-svn: 317020
2017-11-01 02:41:48 +08:00
;
; AVX512BW-LABEL: v4i64:
[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
; AVX512BW: # %bb.0:
[X86][AVX512] Split AVX512F and AVX512BW bool-vector bitcast tests llvm-svn: 317020
2017-11-01 02:41:48 +08:00
; AVX512BW-NEXT: vpcmpgtq %ymm1, %ymm0, %k1
; AVX512BW-NEXT: vpcmpgtq %ymm3, %ymm2, %k0 {%k1}
; AVX512BW-NEXT: kmovd %k0, %eax
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
; AVX512BW-NEXT: # kill: def $al killed $al killed $eax
[X86][AVX512] Split AVX512F and AVX512BW bool-vector bitcast tests llvm-svn: 317020
2017-11-01 02:41:48 +08:00
; AVX512BW-NEXT: vzeroupper
[X86][AVX512] Regenerate tests to remove retl/retq regex These are only testing 64-bit targets so we don't need the regex llvm-svn: 317021
2017-11-01 02:43:24 +08:00
; AVX512BW-NEXT: retq
[X86] Add tests for (ix bitcast (vxi1 and ...)). NFC. To be improved by D33311. llvm-svn: 304171
2017-05-30 03:00:57 +08:00
%x0 = icmp sgt <4 x i64> %a, %b
%x1 = icmp sgt <4 x i64> %c, %d
%y = and <4 x i1> %x0, %x1
%res = bitcast <4 x i1> %y to i4
ret i4 %res
}
define i4 @v4f64(<4 x double> %a, <4 x double> %b, <4 x double> %c, <4 x double> %d) {
[X86] Match bitcast of vxi1 to pmovmsk Summary: Add an early combine to match patterns such as: (i16 bitcast (v16i1 x)) -> (i16 movmsk (v16i8 sext (v16i1 x))) This combine needs to happen early enough before type-legalization scalarizes the result of the setcc. Reviewers: igorb, craig.topper, RKSimon Subscribers: delena, llvm-commits Differential Revision: https://reviews.llvm.org/D33311 llvm-svn: 304406
2017-06-01 19:27:57 +08:00
; SSE2-SSSE3-LABEL: v4f64:
[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] Match bitcast of vxi1 to pmovmsk Summary: Add an early combine to match patterns such as: (i16 bitcast (v16i1 x)) -> (i16 movmsk (v16i8 sext (v16i1 x))) This combine needs to happen early enough before type-legalization scalarizes the result of the setcc. Reviewers: igorb, craig.topper, RKSimon Subscribers: delena, llvm-commits Differential Revision: https://reviews.llvm.org/D33311 llvm-svn: 304406
2017-06-01 19:27:57 +08:00
; SSE2-SSSE3-NEXT: cmpltpd %xmm1, %xmm3
; SSE2-SSSE3-NEXT: cmpltpd %xmm0, %xmm2
; SSE2-SSSE3-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,2],xmm3[0,2]
; SSE2-SSSE3-NEXT: cmpltpd %xmm5, %xmm7
; SSE2-SSSE3-NEXT: cmpltpd %xmm4, %xmm6
; SSE2-SSSE3-NEXT: shufps {{.*#+}} xmm6 = xmm6[0,2],xmm7[0,2]
[SelectionDAG] Add BITCAST handling to ComputeNumSignBits for splatted sign bits. For cases where we are BITCASTing to vectors of smaller elements, then if the entire source was a splatted sign (src's NumSignBits == SrcBitWidth) we can say that the dst's NumSignBit == DstBitWidth, as we're just splitting those sign bits across multiple elements. We could generalize this but at the moment the only use case I have is to peek through bitcasts to vector comparison results. Differential Revision: https://reviews.llvm.org/D37849 llvm-svn: 313543
2017-09-19 00:45:05 +08:00
; SSE2-SSSE3-NEXT: andps %xmm2, %xmm6
[X86] Match bitcast of vxi1 to pmovmsk Summary: Add an early combine to match patterns such as: (i16 bitcast (v16i1 x)) -> (i16 movmsk (v16i8 sext (v16i1 x))) This combine needs to happen early enough before type-legalization scalarizes the result of the setcc. Reviewers: igorb, craig.topper, RKSimon Subscribers: delena, llvm-commits Differential Revision: https://reviews.llvm.org/D33311 llvm-svn: 304406
2017-06-01 19:27:57 +08:00
; SSE2-SSSE3-NEXT: movmskps %xmm6, %eax
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 $al killed $al killed $eax
[X86][AVX512] Regenerate tests to remove retl/retq regex These are only testing 64-bit targets so we don't need the regex llvm-svn: 317021
2017-11-01 02:43:24 +08:00
; SSE2-SSSE3-NEXT: retq
[X86] Match bitcast of vxi1 to pmovmsk Summary: Add an early combine to match patterns such as: (i16 bitcast (v16i1 x)) -> (i16 movmsk (v16i8 sext (v16i1 x))) This combine needs to happen early enough before type-legalization scalarizes the result of the setcc. Reviewers: igorb, craig.topper, RKSimon Subscribers: delena, llvm-commits Differential Revision: https://reviews.llvm.org/D33311 llvm-svn: 304406
2017-06-01 19:27:57 +08:00
;
; AVX12-LABEL: v4f64:
[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
; AVX12: # %bb.0:
[X86] Match bitcast of vxi1 to pmovmsk Summary: Add an early combine to match patterns such as: (i16 bitcast (v16i1 x)) -> (i16 movmsk (v16i8 sext (v16i1 x))) This combine needs to happen early enough before type-legalization scalarizes the result of the setcc. Reviewers: igorb, craig.topper, RKSimon Subscribers: delena, llvm-commits Differential Revision: https://reviews.llvm.org/D33311 llvm-svn: 304406
2017-06-01 19:27:57 +08:00
; AVX12-NEXT: vcmpltpd %ymm0, %ymm1, %ymm0
; AVX12-NEXT: vextractf128 $1, %ymm0, %xmm1
[X86] truncateVectorCompareWithPACKSS - use PACKSSDW/PACKSSWB instead of just PACKSSWB. By using the widest type possible for PACKSS truncation we have a better chance of being able to peek through bitcasts and improves other combines driven by ComputeNumSignBits. llvm-svn: 316448
2017-10-24 23:38:16 +08:00
; AVX12-NEXT: vpackssdw %xmm1, %xmm0, %xmm0
[X86] Match bitcast of vxi1 to pmovmsk Summary: Add an early combine to match patterns such as: (i16 bitcast (v16i1 x)) -> (i16 movmsk (v16i8 sext (v16i1 x))) This combine needs to happen early enough before type-legalization scalarizes the result of the setcc. Reviewers: igorb, craig.topper, RKSimon Subscribers: delena, llvm-commits Differential Revision: https://reviews.llvm.org/D33311 llvm-svn: 304406
2017-06-01 19:27:57 +08:00
; AVX12-NEXT: vcmpltpd %ymm2, %ymm3, %ymm1
; AVX12-NEXT: vextractf128 $1, %ymm1, %xmm2
[X86] truncateVectorCompareWithPACKSS - use PACKSSDW/PACKSSWB instead of just PACKSSWB. By using the widest type possible for PACKSS truncation we have a better chance of being able to peek through bitcasts and improves other combines driven by ComputeNumSignBits. llvm-svn: 316448
2017-10-24 23:38:16 +08:00
; AVX12-NEXT: vpackssdw %xmm2, %xmm1, %xmm1
[X86] Match bitcast of vxi1 to pmovmsk Summary: Add an early combine to match patterns such as: (i16 bitcast (v16i1 x)) -> (i16 movmsk (v16i8 sext (v16i1 x))) This combine needs to happen early enough before type-legalization scalarizes the result of the setcc. Reviewers: igorb, craig.topper, RKSimon Subscribers: delena, llvm-commits Differential Revision: https://reviews.llvm.org/D33311 llvm-svn: 304406
2017-06-01 19:27:57 +08:00
; AVX12-NEXT: vpand %xmm1, %xmm0, %xmm0
; AVX12-NEXT: vmovmskps %xmm0, %eax
Followup on Proposal to move MIR physical register namespace to '$' 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
; AVX12-NEXT: # kill: def $al killed $al killed $eax
[X86] Match bitcast of vxi1 to pmovmsk Summary: Add an early combine to match patterns such as: (i16 bitcast (v16i1 x)) -> (i16 movmsk (v16i8 sext (v16i1 x))) This combine needs to happen early enough before type-legalization scalarizes the result of the setcc. Reviewers: igorb, craig.topper, RKSimon Subscribers: delena, llvm-commits Differential Revision: https://reviews.llvm.org/D33311 llvm-svn: 304406
2017-06-01 19:27:57 +08:00
; AVX12-NEXT: vzeroupper
[X86][AVX512] Regenerate tests to remove retl/retq regex These are only testing 64-bit targets so we don't need the regex llvm-svn: 317021
2017-11-01 02:43:24 +08:00
; AVX12-NEXT: retq
[X86] Add tests for (ix bitcast (vxi1 and ...)). NFC. To be improved by D33311. llvm-svn: 304171
2017-05-30 03:00:57 +08:00
;
[X86][AVX512] Split AVX512F and AVX512BW bool-vector bitcast tests llvm-svn: 317020
2017-11-01 02:41:48 +08:00
; AVX512F-LABEL: v4f64:
[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
; AVX512F: # %bb.0:
[X86][AVX512] Split AVX512F and AVX512BW bool-vector bitcast tests llvm-svn: 317020
2017-11-01 02:41:48 +08:00
; AVX512F-NEXT: vcmpltpd %ymm0, %ymm1, %k1
; AVX512F-NEXT: vcmpltpd %ymm2, %ymm3, %k0 {%k1}
; AVX512F-NEXT: kmovw %k0, %eax
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
; AVX512F-NEXT: # kill: def $al killed $al killed $eax
[X86][AVX512] Split AVX512F and AVX512BW bool-vector bitcast tests llvm-svn: 317020
2017-11-01 02:41:48 +08:00
; AVX512F-NEXT: vzeroupper
[X86][AVX512] Regenerate tests to remove retl/retq regex These are only testing 64-bit targets so we don't need the regex llvm-svn: 317021
2017-11-01 02:43:24 +08:00
; AVX512F-NEXT: retq
[X86][AVX512] Split AVX512F and AVX512BW bool-vector bitcast tests llvm-svn: 317020
2017-11-01 02:41:48 +08:00
;
; AVX512BW-LABEL: v4f64:
[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
; AVX512BW: # %bb.0:
[X86][AVX512] Split AVX512F and AVX512BW bool-vector bitcast tests llvm-svn: 317020
2017-11-01 02:41:48 +08:00
; AVX512BW-NEXT: vcmpltpd %ymm0, %ymm1, %k1
; AVX512BW-NEXT: vcmpltpd %ymm2, %ymm3, %k0 {%k1}
; AVX512BW-NEXT: kmovd %k0, %eax
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
; AVX512BW-NEXT: # kill: def $al killed $al killed $eax
[X86][AVX512] Split AVX512F and AVX512BW bool-vector bitcast tests llvm-svn: 317020
2017-11-01 02:41:48 +08:00
; AVX512BW-NEXT: vzeroupper
[X86][AVX512] Regenerate tests to remove retl/retq regex These are only testing 64-bit targets so we don't need the regex llvm-svn: 317021
2017-11-01 02:43:24 +08:00
; AVX512BW-NEXT: retq
[X86] Add tests for (ix bitcast (vxi1 and ...)). NFC. To be improved by D33311. llvm-svn: 304171
2017-05-30 03:00:57 +08:00
%x0 = fcmp ogt <4 x double> %a, %b
%x1 = fcmp ogt <4 x double> %c, %d
%y = and <4 x i1> %x0, %x1
%res = bitcast <4 x i1> %y to i4
ret i4 %res
}
define i16 @v16i16(<16 x i16> %a, <16 x i16> %b, <16 x i16> %c, <16 x i16> %d) {
[X86][SSE] Add support for lowering shuffles to PACKSS/PACKUS If the upper bits of a truncation shuffle patterns have at least the minimum number of sign/zero bits on their inputs then we can safely use PACKSS/PACKUS as shuffles. Partial fix for https://bugs.llvm.org/show_bug.cgi?id=34773 Differential Revision: https://reviews.llvm.org/D38472 llvm-svn: 314788
2017-10-03 20:01:31 +08:00
; SSE2-SSSE3-LABEL: v16i16:
[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 lowering shuffles to PACKSS/PACKUS If the upper bits of a truncation shuffle patterns have at least the minimum number of sign/zero bits on their inputs then we can safely use PACKSS/PACKUS as shuffles. Partial fix for https://bugs.llvm.org/show_bug.cgi?id=34773 Differential Revision: https://reviews.llvm.org/D38472 llvm-svn: 314788
2017-10-03 20:01:31 +08:00
; SSE2-SSSE3-NEXT: pcmpgtw %xmm3, %xmm1
; SSE2-SSSE3-NEXT: pcmpgtw %xmm2, %xmm0
; SSE2-SSSE3-NEXT: packsswb %xmm1, %xmm0
; SSE2-SSSE3-NEXT: pcmpgtw %xmm7, %xmm5
; SSE2-SSSE3-NEXT: pcmpgtw %xmm6, %xmm4
; SSE2-SSSE3-NEXT: packsswb %xmm5, %xmm4
; SSE2-SSSE3-NEXT: pand %xmm0, %xmm4
; SSE2-SSSE3-NEXT: pmovmskb %xmm4, %eax
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 $ax killed $ax killed $eax
[X86][AVX512] Regenerate tests to remove retl/retq regex These are only testing 64-bit targets so we don't need the regex llvm-svn: 317021
2017-11-01 02:43:24 +08:00
; SSE2-SSSE3-NEXT: retq
[X86] Match bitcast of vxi1 to pmovmsk Summary: Add an early combine to match patterns such as: (i16 bitcast (v16i1 x)) -> (i16 movmsk (v16i8 sext (v16i1 x))) This combine needs to happen early enough before type-legalization scalarizes the result of the setcc. Reviewers: igorb, craig.topper, RKSimon Subscribers: delena, llvm-commits Differential Revision: https://reviews.llvm.org/D33311 llvm-svn: 304406
2017-06-01 19:27:57 +08:00
;
; AVX1-LABEL: v16i16:
[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] Match bitcast of vxi1 to pmovmsk Summary: Add an early combine to match patterns such as: (i16 bitcast (v16i1 x)) -> (i16 movmsk (v16i8 sext (v16i1 x))) This combine needs to happen early enough before type-legalization scalarizes the result of the setcc. Reviewers: igorb, craig.topper, RKSimon Subscribers: delena, llvm-commits Differential Revision: https://reviews.llvm.org/D33311 llvm-svn: 304406
2017-06-01 19:27:57 +08:00
; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm4
; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm5
; AVX1-NEXT: vpcmpgtw %xmm4, %xmm5, %xmm4
; AVX1-NEXT: vpcmpgtw %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vpacksswb %xmm4, %xmm0, %xmm0
; AVX1-NEXT: vextractf128 $1, %ymm3, %xmm1
; AVX1-NEXT: vextractf128 $1, %ymm2, %xmm4
; AVX1-NEXT: vpcmpgtw %xmm1, %xmm4, %xmm1
; AVX1-NEXT: vpcmpgtw %xmm3, %xmm2, %xmm2
; AVX1-NEXT: vpacksswb %xmm1, %xmm2, %xmm1
; AVX1-NEXT: vpand %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vpmovmskb %xmm0, %eax
Followup on Proposal to move MIR physical register namespace to '$' 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 $ax killed $ax killed $eax
[X86] Match bitcast of vxi1 to pmovmsk Summary: Add an early combine to match patterns such as: (i16 bitcast (v16i1 x)) -> (i16 movmsk (v16i8 sext (v16i1 x))) This combine needs to happen early enough before type-legalization scalarizes the result of the setcc. Reviewers: igorb, craig.topper, RKSimon Subscribers: delena, llvm-commits Differential Revision: https://reviews.llvm.org/D33311 llvm-svn: 304406
2017-06-01 19:27:57 +08:00
; AVX1-NEXT: vzeroupper
[X86][AVX512] Regenerate tests to remove retl/retq regex These are only testing 64-bit targets so we don't need the regex llvm-svn: 317021
2017-11-01 02:43:24 +08:00
; AVX1-NEXT: retq
[X86] Match bitcast of vxi1 to pmovmsk Summary: Add an early combine to match patterns such as: (i16 bitcast (v16i1 x)) -> (i16 movmsk (v16i8 sext (v16i1 x))) This combine needs to happen early enough before type-legalization scalarizes the result of the setcc. Reviewers: igorb, craig.topper, RKSimon Subscribers: delena, llvm-commits Differential Revision: https://reviews.llvm.org/D33311 llvm-svn: 304406
2017-06-01 19:27:57 +08:00
;
[X86] Add tests for (ix bitcast (vxi1 and ...)). NFC. To be improved by D33311. llvm-svn: 304171
2017-05-30 03:00:57 +08:00
; AVX2-LABEL: v16i16:
[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] Add tests for (ix bitcast (vxi1 and ...)). NFC. To be improved by D33311. llvm-svn: 304171
2017-05-30 03:00:57 +08:00
; AVX2-NEXT: vpcmpgtw %ymm1, %ymm0, %ymm0
; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX2-NEXT: vpacksswb %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vpcmpgtw %ymm3, %ymm2, %ymm1
; AVX2-NEXT: vextracti128 $1, %ymm1, %xmm2
; AVX2-NEXT: vpacksswb %xmm2, %xmm1, %xmm1
; AVX2-NEXT: vpand %xmm1, %xmm0, %xmm0
[X86] Match bitcast of vxi1 to pmovmsk Summary: Add an early combine to match patterns such as: (i16 bitcast (v16i1 x)) -> (i16 movmsk (v16i8 sext (v16i1 x))) This combine needs to happen early enough before type-legalization scalarizes the result of the setcc. Reviewers: igorb, craig.topper, RKSimon Subscribers: delena, llvm-commits Differential Revision: https://reviews.llvm.org/D33311 llvm-svn: 304406
2017-06-01 19:27:57 +08:00
; AVX2-NEXT: vpmovmskb %xmm0, %eax
Followup on Proposal to move MIR physical register namespace to '$' 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 $ax killed $ax killed $eax
[X86] Add tests for (ix bitcast (vxi1 and ...)). NFC. To be improved by D33311. llvm-svn: 304171
2017-05-30 03:00:57 +08:00
; AVX2-NEXT: vzeroupper
[X86][AVX512] Regenerate tests to remove retl/retq regex These are only testing 64-bit targets so we don't need the regex llvm-svn: 317021
2017-11-01 02:43:24 +08:00
; AVX2-NEXT: retq
[X86] Add tests for (ix bitcast (vxi1 and ...)). NFC. To be improved by D33311. llvm-svn: 304171
2017-05-30 03:00:57 +08:00
;
[X86][AVX512] Split AVX512F and AVX512BW bool-vector bitcast tests llvm-svn: 317020
2017-11-01 02:41:48 +08:00
; AVX512F-LABEL: v16i16:
[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
; AVX512F: # %bb.0:
[X86][AVX512] Split AVX512F and AVX512BW bool-vector bitcast tests llvm-svn: 317020
2017-11-01 02:41:48 +08:00
; AVX512F-NEXT: vpcmpgtw %ymm1, %ymm0, %ymm0
; AVX512F-NEXT: vpmovsxwd %ymm0, %zmm0
; AVX512F-NEXT: vptestmd %zmm0, %zmm0, %k1
; AVX512F-NEXT: vpcmpgtw %ymm3, %ymm2, %ymm0
; AVX512F-NEXT: vpmovsxwd %ymm0, %zmm0
; AVX512F-NEXT: vptestmd %zmm0, %zmm0, %k0 {%k1}
; AVX512F-NEXT: kmovw %k0, %eax
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
; AVX512F-NEXT: # kill: def $ax killed $ax killed $eax
[X86][AVX512] Split AVX512F and AVX512BW bool-vector bitcast tests llvm-svn: 317020
2017-11-01 02:41:48 +08:00
; AVX512F-NEXT: vzeroupper
[X86][AVX512] Regenerate tests to remove retl/retq regex These are only testing 64-bit targets so we don't need the regex llvm-svn: 317021
2017-11-01 02:43:24 +08:00
; AVX512F-NEXT: retq
[X86][AVX512] Split AVX512F and AVX512BW bool-vector bitcast tests llvm-svn: 317020
2017-11-01 02:41:48 +08:00
;
; AVX512BW-LABEL: v16i16:
[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
; AVX512BW: # %bb.0:
[X86][AVX512] Split AVX512F and AVX512BW bool-vector bitcast tests llvm-svn: 317020
2017-11-01 02:41:48 +08:00
; AVX512BW-NEXT: vpcmpgtw %ymm1, %ymm0, %k1
; AVX512BW-NEXT: vpcmpgtw %ymm3, %ymm2, %k0 {%k1}
; AVX512BW-NEXT: kmovd %k0, %eax
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
; AVX512BW-NEXT: # kill: def $ax killed $ax killed $eax
[X86][AVX512] Split AVX512F and AVX512BW bool-vector bitcast tests llvm-svn: 317020
2017-11-01 02:41:48 +08:00
; AVX512BW-NEXT: vzeroupper
[X86][AVX512] Regenerate tests to remove retl/retq regex These are only testing 64-bit targets so we don't need the regex llvm-svn: 317021
2017-11-01 02:43:24 +08:00
; AVX512BW-NEXT: retq
[X86] Add tests for (ix bitcast (vxi1 and ...)). NFC. To be improved by D33311. llvm-svn: 304171
2017-05-30 03:00:57 +08:00
%x0 = icmp sgt <16 x i16> %a, %b
%x1 = icmp sgt <16 x i16> %c, %d
%y = and <16 x i1> %x0, %x1
%res = bitcast <16 x i1> %y to i16
ret i16 %res
}
define i8 @v8i32(<8 x i32> %a, <8 x i32> %b, <8 x i32> %c, <8 x i32> %d) {
[X86][SSE] combineBitcastvxi1 - use PACKSSWB directly to pack v8i16 to v16i8 Avoid difficulties determining the number of sign bits later on in shuffle lowering to lower to PACKSS llvm-svn: 316383
2017-10-24 06:05:02 +08:00
; SSE2-SSSE3-LABEL: v8i32:
[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] combineBitcastvxi1 - use PACKSSWB directly to pack v8i16 to v16i8 Avoid difficulties determining the number of sign bits later on in shuffle lowering to lower to PACKSS llvm-svn: 316383
2017-10-24 06:05:02 +08:00
; SSE2-SSSE3-NEXT: pcmpgtd %xmm3, %xmm1
; SSE2-SSSE3-NEXT: pcmpgtd %xmm2, %xmm0
; SSE2-SSSE3-NEXT: packssdw %xmm1, %xmm0
; SSE2-SSSE3-NEXT: pcmpgtd %xmm7, %xmm5
; SSE2-SSSE3-NEXT: pcmpgtd %xmm6, %xmm4
; SSE2-SSSE3-NEXT: packssdw %xmm5, %xmm4
; SSE2-SSSE3-NEXT: pand %xmm0, %xmm4
; SSE2-SSSE3-NEXT: packsswb %xmm0, %xmm4
; SSE2-SSSE3-NEXT: pmovmskb %xmm4, %eax
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 $al killed $al killed $eax
[X86][AVX512] Regenerate tests to remove retl/retq regex These are only testing 64-bit targets so we don't need the regex llvm-svn: 317021
2017-11-01 02:43:24 +08:00
; SSE2-SSSE3-NEXT: retq
[X86] Match bitcast of vxi1 to pmovmsk Summary: Add an early combine to match patterns such as: (i16 bitcast (v16i1 x)) -> (i16 movmsk (v16i8 sext (v16i1 x))) This combine needs to happen early enough before type-legalization scalarizes the result of the setcc. Reviewers: igorb, craig.topper, RKSimon Subscribers: delena, llvm-commits Differential Revision: https://reviews.llvm.org/D33311 llvm-svn: 304406
2017-06-01 19:27:57 +08:00
;
; AVX1-LABEL: v8i32:
[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] Match bitcast of vxi1 to pmovmsk Summary: Add an early combine to match patterns such as: (i16 bitcast (v16i1 x)) -> (i16 movmsk (v16i8 sext (v16i1 x))) This combine needs to happen early enough before type-legalization scalarizes the result of the setcc. Reviewers: igorb, craig.topper, RKSimon Subscribers: delena, llvm-commits Differential Revision: https://reviews.llvm.org/D33311 llvm-svn: 304406
2017-06-01 19:27:57 +08:00
; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm4
; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm5
; AVX1-NEXT: vpcmpgtd %xmm4, %xmm5, %xmm4
; AVX1-NEXT: vpcmpgtd %xmm1, %xmm0, %xmm0
[X86] truncateVectorCompareWithPACKSS - use PACKSSDW/PACKSSWB instead of just PACKSSWB. By using the widest type possible for PACKSS truncation we have a better chance of being able to peek through bitcasts and improves other combines driven by ComputeNumSignBits. llvm-svn: 316448
2017-10-24 23:38:16 +08:00
; AVX1-NEXT: vpackssdw %xmm4, %xmm0, %xmm0
[X86] Match bitcast of vxi1 to pmovmsk Summary: Add an early combine to match patterns such as: (i16 bitcast (v16i1 x)) -> (i16 movmsk (v16i8 sext (v16i1 x))) This combine needs to happen early enough before type-legalization scalarizes the result of the setcc. Reviewers: igorb, craig.topper, RKSimon Subscribers: delena, llvm-commits Differential Revision: https://reviews.llvm.org/D33311 llvm-svn: 304406
2017-06-01 19:27:57 +08:00
; AVX1-NEXT: vextractf128 $1, %ymm3, %xmm1
; AVX1-NEXT: vextractf128 $1, %ymm2, %xmm4
; AVX1-NEXT: vpcmpgtd %xmm1, %xmm4, %xmm1
; AVX1-NEXT: vpcmpgtd %xmm3, %xmm2, %xmm2
[X86] truncateVectorCompareWithPACKSS - use PACKSSDW/PACKSSWB instead of just PACKSSWB. By using the widest type possible for PACKSS truncation we have a better chance of being able to peek through bitcasts and improves other combines driven by ComputeNumSignBits. llvm-svn: 316448
2017-10-24 23:38:16 +08:00
; AVX1-NEXT: vpackssdw %xmm1, %xmm2, %xmm1
[X86] Match bitcast of vxi1 to pmovmsk Summary: Add an early combine to match patterns such as: (i16 bitcast (v16i1 x)) -> (i16 movmsk (v16i8 sext (v16i1 x))) This combine needs to happen early enough before type-legalization scalarizes the result of the setcc. Reviewers: igorb, craig.topper, RKSimon Subscribers: delena, llvm-commits Differential Revision: https://reviews.llvm.org/D33311 llvm-svn: 304406
2017-06-01 19:27:57 +08:00
; AVX1-NEXT: vpand %xmm1, %xmm0, %xmm0
[X86][SSE] combineBitcastvxi1 - use PACKSSWB directly to pack v8i16 to v16i8 Avoid difficulties determining the number of sign bits later on in shuffle lowering to lower to PACKSS llvm-svn: 316383
2017-10-24 06:05:02 +08:00
; AVX1-NEXT: vpacksswb %xmm0, %xmm0, %xmm0
[X86] Match bitcast of vxi1 to pmovmsk Summary: Add an early combine to match patterns such as: (i16 bitcast (v16i1 x)) -> (i16 movmsk (v16i8 sext (v16i1 x))) This combine needs to happen early enough before type-legalization scalarizes the result of the setcc. Reviewers: igorb, craig.topper, RKSimon Subscribers: delena, llvm-commits Differential Revision: https://reviews.llvm.org/D33311 llvm-svn: 304406
2017-06-01 19:27:57 +08:00
; AVX1-NEXT: vpmovmskb %xmm0, %eax
Followup on Proposal to move MIR physical register namespace to '$' 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 $al killed $al killed $eax
[X86] Match bitcast of vxi1 to pmovmsk Summary: Add an early combine to match patterns such as: (i16 bitcast (v16i1 x)) -> (i16 movmsk (v16i8 sext (v16i1 x))) This combine needs to happen early enough before type-legalization scalarizes the result of the setcc. Reviewers: igorb, craig.topper, RKSimon Subscribers: delena, llvm-commits Differential Revision: https://reviews.llvm.org/D33311 llvm-svn: 304406
2017-06-01 19:27:57 +08:00
; AVX1-NEXT: vzeroupper
[X86][AVX512] Regenerate tests to remove retl/retq regex These are only testing 64-bit targets so we don't need the regex llvm-svn: 317021
2017-11-01 02:43:24 +08:00
; AVX1-NEXT: retq
[X86] Match bitcast of vxi1 to pmovmsk Summary: Add an early combine to match patterns such as: (i16 bitcast (v16i1 x)) -> (i16 movmsk (v16i8 sext (v16i1 x))) This combine needs to happen early enough before type-legalization scalarizes the result of the setcc. Reviewers: igorb, craig.topper, RKSimon Subscribers: delena, llvm-commits Differential Revision: https://reviews.llvm.org/D33311 llvm-svn: 304406
2017-06-01 19:27:57 +08:00
;
[X86] Add tests for (ix bitcast (vxi1 and ...)). NFC. To be improved by D33311. llvm-svn: 304171
2017-05-30 03:00:57 +08:00
; AVX2-LABEL: v8i32:
[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] Add tests for (ix bitcast (vxi1 and ...)). NFC. To be improved by D33311. llvm-svn: 304171
2017-05-30 03:00:57 +08:00
; AVX2-NEXT: vpcmpgtd %ymm1, %ymm0, %ymm0
; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1
[X86] truncateVectorCompareWithPACKSS - use PACKSSDW/PACKSSWB instead of just PACKSSWB. By using the widest type possible for PACKSS truncation we have a better chance of being able to peek through bitcasts and improves other combines driven by ComputeNumSignBits. llvm-svn: 316448
2017-10-24 23:38:16 +08:00
; AVX2-NEXT: vpackssdw %xmm1, %xmm0, %xmm0
[X86] Add tests for (ix bitcast (vxi1 and ...)). NFC. To be improved by D33311. llvm-svn: 304171
2017-05-30 03:00:57 +08:00
; AVX2-NEXT: vpcmpgtd %ymm3, %ymm2, %ymm1
; AVX2-NEXT: vextracti128 $1, %ymm1, %xmm2
[X86] truncateVectorCompareWithPACKSS - use PACKSSDW/PACKSSWB instead of just PACKSSWB. By using the widest type possible for PACKSS truncation we have a better chance of being able to peek through bitcasts and improves other combines driven by ComputeNumSignBits. llvm-svn: 316448
2017-10-24 23:38:16 +08:00
; AVX2-NEXT: vpackssdw %xmm2, %xmm1, %xmm1
[X86] Add tests for (ix bitcast (vxi1 and ...)). NFC. To be improved by D33311. llvm-svn: 304171
2017-05-30 03:00:57 +08:00
; AVX2-NEXT: vpand %xmm1, %xmm0, %xmm0
[X86][SSE] combineBitcastvxi1 - use PACKSSWB directly to pack v8i16 to v16i8 Avoid difficulties determining the number of sign bits later on in shuffle lowering to lower to PACKSS llvm-svn: 316383
2017-10-24 06:05:02 +08:00
; AVX2-NEXT: vpacksswb %xmm0, %xmm0, %xmm0
[X86] Match bitcast of vxi1 to pmovmsk Summary: Add an early combine to match patterns such as: (i16 bitcast (v16i1 x)) -> (i16 movmsk (v16i8 sext (v16i1 x))) This combine needs to happen early enough before type-legalization scalarizes the result of the setcc. Reviewers: igorb, craig.topper, RKSimon Subscribers: delena, llvm-commits Differential Revision: https://reviews.llvm.org/D33311 llvm-svn: 304406
2017-06-01 19:27:57 +08:00
; AVX2-NEXT: vpmovmskb %xmm0, %eax
Followup on Proposal to move MIR physical register namespace to '$' 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 $al killed $al killed $eax
[X86] Add tests for (ix bitcast (vxi1 and ...)). NFC. To be improved by D33311. llvm-svn: 304171
2017-05-30 03:00:57 +08:00
; AVX2-NEXT: vzeroupper
[X86][AVX512] Regenerate tests to remove retl/retq regex These are only testing 64-bit targets so we don't need the regex llvm-svn: 317021
2017-11-01 02:43:24 +08:00
; AVX2-NEXT: retq
[X86] Add tests for (ix bitcast (vxi1 and ...)). NFC. To be improved by D33311. llvm-svn: 304171
2017-05-30 03:00:57 +08:00
;
[X86][AVX512] Split AVX512F and AVX512BW bool-vector bitcast tests llvm-svn: 317020
2017-11-01 02:41:48 +08:00
; AVX512F-LABEL: v8i32:
[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
; AVX512F: # %bb.0:
[X86][AVX512] Split AVX512F and AVX512BW bool-vector bitcast tests llvm-svn: 317020
2017-11-01 02:41:48 +08:00
; AVX512F-NEXT: vpcmpgtd %ymm1, %ymm0, %k1
; AVX512F-NEXT: vpcmpgtd %ymm3, %ymm2, %k0 {%k1}
; AVX512F-NEXT: kmovw %k0, %eax
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
; AVX512F-NEXT: # kill: def $al killed $al killed $eax
[X86][AVX512] Split AVX512F and AVX512BW bool-vector bitcast tests llvm-svn: 317020
2017-11-01 02:41:48 +08:00
; AVX512F-NEXT: vzeroupper
[X86][AVX512] Regenerate tests to remove retl/retq regex These are only testing 64-bit targets so we don't need the regex llvm-svn: 317021
2017-11-01 02:43:24 +08:00
; AVX512F-NEXT: retq
[X86][AVX512] Split AVX512F and AVX512BW bool-vector bitcast tests llvm-svn: 317020
2017-11-01 02:41:48 +08:00
;
; AVX512BW-LABEL: v8i32:
[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
; AVX512BW: # %bb.0:
[X86][AVX512] Split AVX512F and AVX512BW bool-vector bitcast tests llvm-svn: 317020
2017-11-01 02:41:48 +08:00
; AVX512BW-NEXT: vpcmpgtd %ymm1, %ymm0, %k1
; AVX512BW-NEXT: vpcmpgtd %ymm3, %ymm2, %k0 {%k1}
; AVX512BW-NEXT: kmovd %k0, %eax
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
; AVX512BW-NEXT: # kill: def $al killed $al killed $eax
[X86][AVX512] Split AVX512F and AVX512BW bool-vector bitcast tests llvm-svn: 317020
2017-11-01 02:41:48 +08:00
; AVX512BW-NEXT: vzeroupper
[X86][AVX512] Regenerate tests to remove retl/retq regex These are only testing 64-bit targets so we don't need the regex llvm-svn: 317021
2017-11-01 02:43:24 +08:00
; AVX512BW-NEXT: retq
[X86] Add tests for (ix bitcast (vxi1 and ...)). NFC. To be improved by D33311. llvm-svn: 304171
2017-05-30 03:00:57 +08:00
%x0 = icmp sgt <8 x i32> %a, %b
%x1 = icmp sgt <8 x i32> %c, %d
%y = and <8 x i1> %x0, %x1
%res = bitcast <8 x i1> %y to i8
ret i8 %res
}
define i8 @v8f32(<8 x float> %a, <8 x float> %b, <8 x float> %c, <8 x float> %d) {
[SelectionDAG] Support 'bit preserving' floating points bitcasts on computeKnownBits/ComputeNumSignBits For cases where we know the floating point representations match the bitcasted integer equivalent, allow bitcasting to these types. This is especially useful for the X86 floating point compare results which return all/zero bits but as a floating point type. Differential Revision: https://reviews.llvm.org/D39289 llvm-svn: 316831
2017-10-28 22:27:53 +08:00
; SSE2-SSSE3-LABEL: v8f32:
[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:
[SelectionDAG] Support 'bit preserving' floating points bitcasts on computeKnownBits/ComputeNumSignBits For cases where we know the floating point representations match the bitcasted integer equivalent, allow bitcasting to these types. This is especially useful for the X86 floating point compare results which return all/zero bits but as a floating point type. Differential Revision: https://reviews.llvm.org/D39289 llvm-svn: 316831
2017-10-28 22:27:53 +08:00
; SSE2-SSSE3-NEXT: cmpltps %xmm1, %xmm3
; SSE2-SSSE3-NEXT: cmpltps %xmm0, %xmm2
; SSE2-SSSE3-NEXT: packssdw %xmm3, %xmm2
; SSE2-SSSE3-NEXT: cmpltps %xmm5, %xmm7
; SSE2-SSSE3-NEXT: cmpltps %xmm4, %xmm6
; SSE2-SSSE3-NEXT: packssdw %xmm7, %xmm6
; SSE2-SSSE3-NEXT: pand %xmm2, %xmm6
; SSE2-SSSE3-NEXT: packsswb %xmm0, %xmm6
; SSE2-SSSE3-NEXT: pmovmskb %xmm6, %eax
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 $al killed $al killed $eax
[X86][AVX512] Regenerate tests to remove retl/retq regex These are only testing 64-bit targets so we don't need the regex llvm-svn: 317021
2017-11-01 02:43:24 +08:00
; SSE2-SSSE3-NEXT: retq
[X86] Match bitcast of vxi1 to pmovmsk Summary: Add an early combine to match patterns such as: (i16 bitcast (v16i1 x)) -> (i16 movmsk (v16i8 sext (v16i1 x))) This combine needs to happen early enough before type-legalization scalarizes the result of the setcc. Reviewers: igorb, craig.topper, RKSimon Subscribers: delena, llvm-commits Differential Revision: https://reviews.llvm.org/D33311 llvm-svn: 304406
2017-06-01 19:27:57 +08:00
;
; AVX12-LABEL: v8f32:
[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
; AVX12: # %bb.0:
[X86] Match bitcast of vxi1 to pmovmsk Summary: Add an early combine to match patterns such as: (i16 bitcast (v16i1 x)) -> (i16 movmsk (v16i8 sext (v16i1 x))) This combine needs to happen early enough before type-legalization scalarizes the result of the setcc. Reviewers: igorb, craig.topper, RKSimon Subscribers: delena, llvm-commits Differential Revision: https://reviews.llvm.org/D33311 llvm-svn: 304406
2017-06-01 19:27:57 +08:00
; AVX12-NEXT: vcmpltps %ymm0, %ymm1, %ymm0
; AVX12-NEXT: vextractf128 $1, %ymm0, %xmm1
[X86] truncateVectorCompareWithPACKSS - use PACKSSDW/PACKSSWB instead of just PACKSSWB. By using the widest type possible for PACKSS truncation we have a better chance of being able to peek through bitcasts and improves other combines driven by ComputeNumSignBits. llvm-svn: 316448
2017-10-24 23:38:16 +08:00
; AVX12-NEXT: vpackssdw %xmm1, %xmm0, %xmm0
[X86] Match bitcast of vxi1 to pmovmsk Summary: Add an early combine to match patterns such as: (i16 bitcast (v16i1 x)) -> (i16 movmsk (v16i8 sext (v16i1 x))) This combine needs to happen early enough before type-legalization scalarizes the result of the setcc. Reviewers: igorb, craig.topper, RKSimon Subscribers: delena, llvm-commits Differential Revision: https://reviews.llvm.org/D33311 llvm-svn: 304406
2017-06-01 19:27:57 +08:00
; AVX12-NEXT: vcmpltps %ymm2, %ymm3, %ymm1
; AVX12-NEXT: vextractf128 $1, %ymm1, %xmm2
[X86] truncateVectorCompareWithPACKSS - use PACKSSDW/PACKSSWB instead of just PACKSSWB. By using the widest type possible for PACKSS truncation we have a better chance of being able to peek through bitcasts and improves other combines driven by ComputeNumSignBits. llvm-svn: 316448
2017-10-24 23:38:16 +08:00
; AVX12-NEXT: vpackssdw %xmm2, %xmm1, %xmm1
[X86] Match bitcast of vxi1 to pmovmsk Summary: Add an early combine to match patterns such as: (i16 bitcast (v16i1 x)) -> (i16 movmsk (v16i8 sext (v16i1 x))) This combine needs to happen early enough before type-legalization scalarizes the result of the setcc. Reviewers: igorb, craig.topper, RKSimon Subscribers: delena, llvm-commits Differential Revision: https://reviews.llvm.org/D33311 llvm-svn: 304406
2017-06-01 19:27:57 +08:00
; AVX12-NEXT: vpand %xmm1, %xmm0, %xmm0
[X86][SSE] combineBitcastvxi1 - use PACKSSWB directly to pack v8i16 to v16i8 Avoid difficulties determining the number of sign bits later on in shuffle lowering to lower to PACKSS llvm-svn: 316383
2017-10-24 06:05:02 +08:00
; AVX12-NEXT: vpacksswb %xmm0, %xmm0, %xmm0
[X86] Match bitcast of vxi1 to pmovmsk Summary: Add an early combine to match patterns such as: (i16 bitcast (v16i1 x)) -> (i16 movmsk (v16i8 sext (v16i1 x))) This combine needs to happen early enough before type-legalization scalarizes the result of the setcc. Reviewers: igorb, craig.topper, RKSimon Subscribers: delena, llvm-commits Differential Revision: https://reviews.llvm.org/D33311 llvm-svn: 304406
2017-06-01 19:27:57 +08:00
; AVX12-NEXT: vpmovmskb %xmm0, %eax
Followup on Proposal to move MIR physical register namespace to '$' 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
; AVX12-NEXT: # kill: def $al killed $al killed $eax
[X86] Match bitcast of vxi1 to pmovmsk Summary: Add an early combine to match patterns such as: (i16 bitcast (v16i1 x)) -> (i16 movmsk (v16i8 sext (v16i1 x))) This combine needs to happen early enough before type-legalization scalarizes the result of the setcc. Reviewers: igorb, craig.topper, RKSimon Subscribers: delena, llvm-commits Differential Revision: https://reviews.llvm.org/D33311 llvm-svn: 304406
2017-06-01 19:27:57 +08:00
; AVX12-NEXT: vzeroupper
[X86][AVX512] Regenerate tests to remove retl/retq regex These are only testing 64-bit targets so we don't need the regex llvm-svn: 317021
2017-11-01 02:43:24 +08:00
; AVX12-NEXT: retq
[X86] Add tests for (ix bitcast (vxi1 and ...)). NFC. To be improved by D33311. llvm-svn: 304171
2017-05-30 03:00:57 +08:00
;
[X86][AVX512] Split AVX512F and AVX512BW bool-vector bitcast tests llvm-svn: 317020
2017-11-01 02:41:48 +08:00
; AVX512F-LABEL: v8f32:
[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
; AVX512F: # %bb.0:
[X86][AVX512] Split AVX512F and AVX512BW bool-vector bitcast tests llvm-svn: 317020
2017-11-01 02:41:48 +08:00
; AVX512F-NEXT: vcmpltps %ymm0, %ymm1, %k1
; AVX512F-NEXT: vcmpltps %ymm2, %ymm3, %k0 {%k1}
; AVX512F-NEXT: kmovw %k0, %eax
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
; AVX512F-NEXT: # kill: def $al killed $al killed $eax
[X86][AVX512] Split AVX512F and AVX512BW bool-vector bitcast tests llvm-svn: 317020
2017-11-01 02:41:48 +08:00
; AVX512F-NEXT: vzeroupper
[X86][AVX512] Regenerate tests to remove retl/retq regex These are only testing 64-bit targets so we don't need the regex llvm-svn: 317021
2017-11-01 02:43:24 +08:00
; AVX512F-NEXT: retq
[X86][AVX512] Split AVX512F and AVX512BW bool-vector bitcast tests llvm-svn: 317020
2017-11-01 02:41:48 +08:00
;
; AVX512BW-LABEL: v8f32:
[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
; AVX512BW: # %bb.0:
[X86][AVX512] Split AVX512F and AVX512BW bool-vector bitcast tests llvm-svn: 317020
2017-11-01 02:41:48 +08:00
; AVX512BW-NEXT: vcmpltps %ymm0, %ymm1, %k1
; AVX512BW-NEXT: vcmpltps %ymm2, %ymm3, %k0 {%k1}
; AVX512BW-NEXT: kmovd %k0, %eax
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
; AVX512BW-NEXT: # kill: def $al killed $al killed $eax
[X86][AVX512] Split AVX512F and AVX512BW bool-vector bitcast tests llvm-svn: 317020
2017-11-01 02:41:48 +08:00
; AVX512BW-NEXT: vzeroupper
[X86][AVX512] Regenerate tests to remove retl/retq regex These are only testing 64-bit targets so we don't need the regex llvm-svn: 317021
2017-11-01 02:43:24 +08:00
; AVX512BW-NEXT: retq
[X86] Add tests for (ix bitcast (vxi1 and ...)). NFC. To be improved by D33311. llvm-svn: 304171
2017-05-30 03:00:57 +08:00
%x0 = fcmp ogt <8 x float> %a, %b
%x1 = fcmp ogt <8 x float> %c, %d
%y = and <8 x i1> %x0, %x1
%res = bitcast <8 x i1> %y to i8
ret i8 %res
}
define i32 @v32i8(<32 x i8> %a, <32 x i8> %b, <32 x i8> %c, <32 x i8> %d) {
[X86] Match bitcast of vxi1 to pmovmsk Summary: Add an early combine to match patterns such as: (i16 bitcast (v16i1 x)) -> (i16 movmsk (v16i8 sext (v16i1 x))) This combine needs to happen early enough before type-legalization scalarizes the result of the setcc. Reviewers: igorb, craig.topper, RKSimon Subscribers: delena, llvm-commits Differential Revision: https://reviews.llvm.org/D33311 llvm-svn: 304406
2017-06-01 19:27:57 +08:00
; SSE2-SSSE3-LABEL: v32i8:
[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] Match bitcast of vxi1 to pmovmsk Summary: Add an early combine to match patterns such as: (i16 bitcast (v16i1 x)) -> (i16 movmsk (v16i8 sext (v16i1 x))) This combine needs to happen early enough before type-legalization scalarizes the result of the setcc. Reviewers: igorb, craig.topper, RKSimon Subscribers: delena, llvm-commits Differential Revision: https://reviews.llvm.org/D33311 llvm-svn: 304406
2017-06-01 19:27:57 +08:00
; SSE2-SSSE3-NEXT: pcmpgtb %xmm2, %xmm0
; SSE2-SSSE3-NEXT: pcmpgtb %xmm3, %xmm1
; SSE2-SSSE3-NEXT: pcmpgtb %xmm6, %xmm4
Reverting commit 306414 on behalf of @gadi.haber llvm-svn: 306532
2017-06-28 19:23:31 +08:00
; SSE2-SSSE3-NEXT: pand %xmm0, %xmm4
[X86] Match bitcast of vxi1 to pmovmsk Summary: Add an early combine to match patterns such as: (i16 bitcast (v16i1 x)) -> (i16 movmsk (v16i8 sext (v16i1 x))) This combine needs to happen early enough before type-legalization scalarizes the result of the setcc. Reviewers: igorb, craig.topper, RKSimon Subscribers: delena, llvm-commits Differential Revision: https://reviews.llvm.org/D33311 llvm-svn: 304406
2017-06-01 19:27:57 +08:00
; SSE2-SSSE3-NEXT: pcmpgtb %xmm7, %xmm5
; SSE2-SSSE3-NEXT: pand %xmm1, %xmm5
[X86][SSE] Add pre-AVX2 support for (i32 bitcast(v32i1)) -> 2xMOVMSK Currently we only support (i32 bitcast(v32i1)) using the AVX2 VPMOVMSKB ymm instruction. This patch adds support for splitting pre-AVX2 targets into 2 x (V)PMOVMSKB xmm instructions and merging the integer results. In future we could probably generalize this to handle more cases. Differential Revision: https://reviews.llvm.org/D35303 llvm-svn: 308723
2017-07-21 17:58:50 +08:00
; SSE2-SSSE3-NEXT: pmovmskb %xmm4, %ecx
; SSE2-SSSE3-NEXT: pmovmskb %xmm5, %eax
; SSE2-SSSE3-NEXT: shll $16, %eax
[X86] Match bitcast of vxi1 to pmovmsk Summary: Add an early combine to match patterns such as: (i16 bitcast (v16i1 x)) -> (i16 movmsk (v16i8 sext (v16i1 x))) This combine needs to happen early enough before type-legalization scalarizes the result of the setcc. Reviewers: igorb, craig.topper, RKSimon Subscribers: delena, llvm-commits Differential Revision: https://reviews.llvm.org/D33311 llvm-svn: 304406
2017-06-01 19:27:57 +08:00
; SSE2-SSSE3-NEXT: orl %ecx, %eax
[X86][AVX512] Regenerate tests to remove retl/retq regex These are only testing 64-bit targets so we don't need the regex llvm-svn: 317021
2017-11-01 02:43:24 +08:00
; SSE2-SSSE3-NEXT: retq
[X86] Match bitcast of vxi1 to pmovmsk Summary: Add an early combine to match patterns such as: (i16 bitcast (v16i1 x)) -> (i16 movmsk (v16i8 sext (v16i1 x))) This combine needs to happen early enough before type-legalization scalarizes the result of the setcc. Reviewers: igorb, craig.topper, RKSimon Subscribers: delena, llvm-commits Differential Revision: https://reviews.llvm.org/D33311 llvm-svn: 304406
2017-06-01 19:27:57 +08:00
;
; AVX1-LABEL: v32i8:
[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] Match bitcast of vxi1 to pmovmsk Summary: Add an early combine to match patterns such as: (i16 bitcast (v16i1 x)) -> (i16 movmsk (v16i8 sext (v16i1 x))) This combine needs to happen early enough before type-legalization scalarizes the result of the setcc. Reviewers: igorb, craig.topper, RKSimon Subscribers: delena, llvm-commits Differential Revision: https://reviews.llvm.org/D33311 llvm-svn: 304406
2017-06-01 19:27:57 +08:00
; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm4
; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm5
; AVX1-NEXT: vpcmpgtb %xmm4, %xmm5, %xmm4
; AVX1-NEXT: vpcmpgtb %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vextractf128 $1, %ymm3, %xmm1
[X86][SSE] Add pre-AVX2 support for (i32 bitcast(v32i1)) -> 2xMOVMSK Currently we only support (i32 bitcast(v32i1)) using the AVX2 VPMOVMSKB ymm instruction. This patch adds support for splitting pre-AVX2 targets into 2 x (V)PMOVMSKB xmm instructions and merging the integer results. In future we could probably generalize this to handle more cases. Differential Revision: https://reviews.llvm.org/D35303 llvm-svn: 308723
2017-07-21 17:58:50 +08:00
; AVX1-NEXT: vextractf128 $1, %ymm2, %xmm5
; AVX1-NEXT: vpcmpgtb %xmm1, %xmm5, %xmm1
; AVX1-NEXT: vpand %xmm1, %xmm4, %xmm1
[X86] Match bitcast of vxi1 to pmovmsk Summary: Add an early combine to match patterns such as: (i16 bitcast (v16i1 x)) -> (i16 movmsk (v16i8 sext (v16i1 x))) This combine needs to happen early enough before type-legalization scalarizes the result of the setcc. Reviewers: igorb, craig.topper, RKSimon Subscribers: delena, llvm-commits Differential Revision: https://reviews.llvm.org/D33311 llvm-svn: 304406
2017-06-01 19:27:57 +08:00
; AVX1-NEXT: vpcmpgtb %xmm3, %xmm2, %xmm2
[X86][SSE] Add pre-AVX2 support for (i32 bitcast(v32i1)) -> 2xMOVMSK Currently we only support (i32 bitcast(v32i1)) using the AVX2 VPMOVMSKB ymm instruction. This patch adds support for splitting pre-AVX2 targets into 2 x (V)PMOVMSKB xmm instructions and merging the integer results. In future we could probably generalize this to handle more cases. Differential Revision: https://reviews.llvm.org/D35303 llvm-svn: 308723
2017-07-21 17:58:50 +08:00
; AVX1-NEXT: vpand %xmm2, %xmm0, %xmm0
; AVX1-NEXT: vpmovmskb %xmm0, %ecx
; AVX1-NEXT: vpmovmskb %xmm1, %eax
; AVX1-NEXT: shll $16, %eax
; AVX1-NEXT: orl %ecx, %eax
[X86] Match bitcast of vxi1 to pmovmsk Summary: Add an early combine to match patterns such as: (i16 bitcast (v16i1 x)) -> (i16 movmsk (v16i8 sext (v16i1 x))) This combine needs to happen early enough before type-legalization scalarizes the result of the setcc. Reviewers: igorb, craig.topper, RKSimon Subscribers: delena, llvm-commits Differential Revision: https://reviews.llvm.org/D33311 llvm-svn: 304406
2017-06-01 19:27:57 +08:00
; AVX1-NEXT: vzeroupper
[X86][AVX512] Regenerate tests to remove retl/retq regex These are only testing 64-bit targets so we don't need the regex llvm-svn: 317021
2017-11-01 02:43:24 +08:00
; AVX1-NEXT: retq
[X86] Match bitcast of vxi1 to pmovmsk Summary: Add an early combine to match patterns such as: (i16 bitcast (v16i1 x)) -> (i16 movmsk (v16i8 sext (v16i1 x))) This combine needs to happen early enough before type-legalization scalarizes the result of the setcc. Reviewers: igorb, craig.topper, RKSimon Subscribers: delena, llvm-commits Differential Revision: https://reviews.llvm.org/D33311 llvm-svn: 304406
2017-06-01 19:27:57 +08:00
;
[X86] Add tests for (ix bitcast (vxi1 and ...)). NFC. To be improved by D33311. llvm-svn: 304171
2017-05-30 03:00:57 +08:00
; AVX2-LABEL: v32i8:
[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] Add tests for (ix bitcast (vxi1 and ...)). NFC. To be improved by D33311. llvm-svn: 304171
2017-05-30 03:00:57 +08:00
; AVX2-NEXT: vpcmpgtb %ymm1, %ymm0, %ymm0
; AVX2-NEXT: vpcmpgtb %ymm3, %ymm2, %ymm1
; AVX2-NEXT: vpand %ymm1, %ymm0, %ymm0
[X86] Match bitcast of vxi1 to pmovmsk Summary: Add an early combine to match patterns such as: (i16 bitcast (v16i1 x)) -> (i16 movmsk (v16i8 sext (v16i1 x))) This combine needs to happen early enough before type-legalization scalarizes the result of the setcc. Reviewers: igorb, craig.topper, RKSimon Subscribers: delena, llvm-commits Differential Revision: https://reviews.llvm.org/D33311 llvm-svn: 304406
2017-06-01 19:27:57 +08:00
; AVX2-NEXT: vpmovmskb %ymm0, %eax
[X86] Add tests for (ix bitcast (vxi1 and ...)). NFC. To be improved by D33311. llvm-svn: 304171
2017-05-30 03:00:57 +08:00
; AVX2-NEXT: vzeroupper
[X86][AVX512] Regenerate tests to remove retl/retq regex These are only testing 64-bit targets so we don't need the regex llvm-svn: 317021
2017-11-01 02:43:24 +08:00
; AVX2-NEXT: retq
[X86] Add tests for (ix bitcast (vxi1 and ...)). NFC. To be improved by D33311. llvm-svn: 304171
2017-05-30 03:00:57 +08:00
;
[X86][AVX512] Split AVX512F and AVX512BW bool-vector bitcast tests llvm-svn: 317020
2017-11-01 02:41:48 +08:00
; AVX512F-LABEL: v32i8:
[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
; AVX512F: # %bb.0:
[X86][AVX512] Split AVX512F and AVX512BW bool-vector bitcast tests llvm-svn: 317020
2017-11-01 02:41:48 +08:00
; AVX512F-NEXT: vpcmpgtb %ymm1, %ymm0, %ymm0
; AVX512F-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX512F-NEXT: vpmovsxbd %xmm1, %zmm1
[X86] Legalize v32i1 without BWI via splitting to v16i1 rather than the default of promoting to v32i8. Summary: For the most part its better to keep v32i1 as a mask type of a narrower width than trying to promote it to a ymm register. I had to add some overrides to the methods that get the types for the calling convention so that we still use v32i8 for argument/return purposes. There are still some regressions in here. I definitely saw some around shuffles. I think we probably should move vXi1 shuffle from lowering to a DAG combine where I think the extend and truncate we have to emit would be better combined. I think we also need a DAG combine to remove trunc from (extract_vector_elt (trunc)) Overall this removes something like 13000 CHECK lines from lit tests. Reviewers: zvi, RKSimon, delena, spatel Reviewed By: RKSimon Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D42031 llvm-svn: 323201
2018-01-23 22:25:39 +08:00
; AVX512F-NEXT: vptestmd %zmm1, %zmm1, %k1
[X86][AVX512] Split AVX512F and AVX512BW bool-vector bitcast tests llvm-svn: 317020
2017-11-01 02:41:48 +08:00
; AVX512F-NEXT: vpmovsxbd %xmm0, %zmm0
[X86] Legalize v32i1 without BWI via splitting to v16i1 rather than the default of promoting to v32i8. Summary: For the most part its better to keep v32i1 as a mask type of a narrower width than trying to promote it to a ymm register. I had to add some overrides to the methods that get the types for the calling convention so that we still use v32i8 for argument/return purposes. There are still some regressions in here. I definitely saw some around shuffles. I think we probably should move vXi1 shuffle from lowering to a DAG combine where I think the extend and truncate we have to emit would be better combined. I think we also need a DAG combine to remove trunc from (extract_vector_elt (trunc)) Overall this removes something like 13000 CHECK lines from lit tests. Reviewers: zvi, RKSimon, delena, spatel Reviewed By: RKSimon Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D42031 llvm-svn: 323201
2018-01-23 22:25:39 +08:00
; AVX512F-NEXT: vptestmd %zmm0, %zmm0, %k2
; AVX512F-NEXT: vpcmpgtb %ymm3, %ymm2, %ymm0
; AVX512F-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX512F-NEXT: vpmovsxbd %xmm1, %zmm1
; AVX512F-NEXT: vpmovsxbd %xmm0, %zmm0
; AVX512F-NEXT: vptestmd %zmm0, %zmm0, %k0 {%k2}
; AVX512F-NEXT: kmovw %k0, %ecx
; AVX512F-NEXT: vptestmd %zmm1, %zmm1, %k0 {%k1}
; AVX512F-NEXT: kmovw %k0, %eax
; AVX512F-NEXT: shll $16, %eax
; AVX512F-NEXT: orl %ecx, %eax
[X86][AVX512] Split AVX512F and AVX512BW bool-vector bitcast tests llvm-svn: 317020
2017-11-01 02:41:48 +08:00
; AVX512F-NEXT: vzeroupper
[X86][AVX512] Regenerate tests to remove retl/retq regex These are only testing 64-bit targets so we don't need the regex llvm-svn: 317021
2017-11-01 02:43:24 +08:00
; AVX512F-NEXT: retq
[X86][AVX512] Split AVX512F and AVX512BW bool-vector bitcast tests llvm-svn: 317020
2017-11-01 02:41:48 +08:00
;
; AVX512BW-LABEL: v32i8:
[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
; AVX512BW: # %bb.0:
[X86][AVX512] Split AVX512F and AVX512BW bool-vector bitcast tests llvm-svn: 317020
2017-11-01 02:41:48 +08:00
; AVX512BW-NEXT: vpcmpgtb %ymm1, %ymm0, %k1
; AVX512BW-NEXT: vpcmpgtb %ymm3, %ymm2, %k0 {%k1}
; AVX512BW-NEXT: kmovd %k0, %eax
; AVX512BW-NEXT: vzeroupper
[X86][AVX512] Regenerate tests to remove retl/retq regex These are only testing 64-bit targets so we don't need the regex llvm-svn: 317021
2017-11-01 02:43:24 +08:00
; AVX512BW-NEXT: retq
[X86] Add tests for (ix bitcast (vxi1 and ...)). NFC. To be improved by D33311. llvm-svn: 304171
2017-05-30 03:00:57 +08:00
%x0 = icmp sgt <32 x i8> %a, %b
%x1 = icmp sgt <32 x i8> %c, %d
%y = and <32 x i1> %x0, %x1
%res = bitcast <32 x i1> %y to i32
ret i32 %res
}