llvm-project.git/llvm/test/Transforms/SLPVectorizer/alt-cmp-vectorize.ll, branch users/nico/python-2 Unnamed repository; edit this file 'description' to name the repository. [SLP]Improve minbitwidth analysis. 2024-03-19T15:19:45+00:00 Alexey Bataev 5361294+alexey-bataev@users.noreply.github.com 2024-03-08T18:57:02+00:00 31eaf86a1e8f1870e6ee4c42088a5213bde294b8 This improves overall analysis for minbitwidth in SLP. It allows to analyze the trees with store/insertelement root nodes. Also, instead of using single minbitwidth, detected from the very first analysis stage, it tries to detect the best one for each trunc/ext subtree in the graph and use it for the subtree. Results in better code and less vector register pressure. Metric: size..text Program size..text results results0 diff test-suite :: SingleSource/Benchmarks/Adobe-C++/simple_types_loop_invariant.test 92549.00 92609.00 0.1% test-suite :: External/SPEC/CINT2017speed/625.x264_s/625.x264_s.test 663381.00 663493.00 0.0% test-suite :: External/SPEC/CINT2017rate/525.x264_r/525.x264_r.test 663381.00 663493.00 0.0% test-suite :: MultiSource/Benchmarks/Bullet/bullet.test 307182.00 307214.00 0.0% test-suite :: External/SPEC/CFP2017speed/638.imagick_s/638.imagick_s.test 1394420.00 1394484.00 0.0% test-suite :: External/SPEC/CFP2017rate/538.imagick_r/538.imagick_r.test 1394420.00 1394484.00 0.0% test-suite :: External/SPEC/CFP2017rate/510.parest_r/510.parest_r.test 2040257.00 2040273.00 0.0% test-suite :: External/SPEC/CFP2017rate/526.blender_r/526.blender_r.test 12396098.00 12395858.00 -0.0% test-suite :: External/SPEC/CINT2006/445.gobmk/445.gobmk.test 909944.00 909768.00 -0.0% SingleSource/Benchmarks/Adobe-C++/simple_types_loop_invariant - 4 scalar instructions remain scalar (good). Spec2017/x264 - the whole function idct4x4dc is vectorized using <16 x i16> instead of <16 x i32>, also zext/trunc are removed. In other places last vector zext/sext removed and replaced by extractelement + scalar zext/sext pair. MultiSource/Benchmarks/Bullet/bullet - reduce or <4 x i32> replaced by reduce or <4 x i8> Spec2017/imagick - Removed extra zext from 2 packs of the operations. Spec2017/parest - Removed extra zext, replaced by extractelement+scalar zext Spec2017/blender - the whole bunch of vector zext/sext replaced by extractelement+scalar zext/sext, some extra code vectorized in smaller types. Spec2006/gobmk - fixed cost estimation, some small code remains scalar. Original Pull Request: https://github.com/llvm/llvm-project/pull/84334 The patch has the same functionality (no test changes, no changes in benchmarks) as the original patch, just has some compile time improvements + fixes for xxhash unittest, discovered earlier in the previous version of the patch. Reviewers: Pull Request: https://github.com/llvm/llvm-project/pull/84536 
This improves overall analysis for minbitwidth in SLP. It allows to
analyze the trees with store/insertelement root nodes. Also, instead of
using single minbitwidth, detected from the very first analysis stage,
it tries to detect the best one for each trunc/ext subtree in the graph
and use it for the subtree.
Results in better code and less vector register pressure.

Metric: size..text

Program                                                                                                                                                size..text
                                                                                                                                                       results     results0    diff
                                                                      test-suite :: SingleSource/Benchmarks/Adobe-C++/simple_types_loop_invariant.test    92549.00    92609.00  0.1%
                                                                                  test-suite :: External/SPEC/CINT2017speed/625.x264_s/625.x264_s.test   663381.00   663493.00  0.0%
                                                                                   test-suite :: External/SPEC/CINT2017rate/525.x264_r/525.x264_r.test   663381.00   663493.00  0.0%
                                                                                               test-suite :: MultiSource/Benchmarks/Bullet/bullet.test   307182.00   307214.00  0.0%
                                                                             test-suite :: External/SPEC/CFP2017speed/638.imagick_s/638.imagick_s.test  1394420.00  1394484.00  0.0%
                                                                              test-suite :: External/SPEC/CFP2017rate/538.imagick_r/538.imagick_r.test  1394420.00  1394484.00  0.0%
                                                                                test-suite :: External/SPEC/CFP2017rate/510.parest_r/510.parest_r.test  2040257.00  2040273.00  0.0%

                                                                              test-suite :: External/SPEC/CFP2017rate/526.blender_r/526.blender_r.test 12396098.00 12395858.00 -0.0%
                                                                                         test-suite :: External/SPEC/CINT2006/445.gobmk/445.gobmk.test   909944.00   909768.00 -0.0%

SingleSource/Benchmarks/Adobe-C++/simple_types_loop_invariant - 4 scalar
instructions remain scalar (good).
Spec2017/x264 - the whole function idct4x4dc is vectorized using <16
x i16> instead of <16 x i32>, also zext/trunc are removed. In other
places last vector zext/sext removed and replaced by
extractelement + scalar zext/sext pair.
MultiSource/Benchmarks/Bullet/bullet - reduce or <4 x i32> replaced by
reduce or <4 x i8>
Spec2017/imagick - Removed extra zext from 2 packs of the operations.
Spec2017/parest - Removed extra zext, replaced by extractelement+scalar
zext
Spec2017/blender - the whole bunch of vector zext/sext replaced by
extractelement+scalar zext/sext, some extra code vectorized in smaller
types.
Spec2006/gobmk - fixed cost estimation, some small code remains scalar.

Original Pull Request: https://github.com/llvm/llvm-project/pull/84334

The patch has the same functionality (no test changes, no changes in
benchmarks) as the original patch, just has some compile time
improvements + fixes for xxhash unittest, discovered earlier in the
previous version of the patch.

Reviewers:

Pull Request: https://github.com/llvm/llvm-project/pull/84536
Revert "[SLP]Improve minbitwidth analysis." 2024-03-15T10:59:48+00:00 Alexey Bataev a.bataev@outlook.com 2024-03-15T10:59:48+00:00 37898707585af6df9545620fa8053e7acd23be9f This reverts commit 7f2167868d8c1cedd3915883412b9c787a2f01db to fix issues reported in https://github.com/llvm/llvm-project/pull/84536. 
This reverts commit 7f2167868d8c1cedd3915883412b9c787a2f01db to fix
issues reported in https://github.com/llvm/llvm-project/pull/84536.
[SLP]Improve minbitwidth analysis. 2024-03-14T13:23:14+00:00 Alexey Bataev 5361294+alexey-bataev@users.noreply.github.com 2024-03-08T18:57:02+00:00 7f2167868d8c1cedd3915883412b9c787a2f01db This improves overall analysis for minbitwidth in SLP. It allows to analyze the trees with store/insertelement root nodes. Also, instead of using single minbitwidth, detected from the very first analysis stage, it tries to detect the best one for each trunc/ext subtree in the graph and use it for the subtree. Results in better code and less vector register pressure. Metric: size..text Program size..text results results0 diff test-suite :: SingleSource/Benchmarks/Adobe-C++/simple_types_loop_invariant.test 92549.00 92609.00 0.1% test-suite :: External/SPEC/CINT2017speed/625.x264_s/625.x264_s.test 663381.00 663493.00 0.0% test-suite :: External/SPEC/CINT2017rate/525.x264_r/525.x264_r.test 663381.00 663493.00 0.0% test-suite :: MultiSource/Benchmarks/Bullet/bullet.test 307182.00 307214.00 0.0% test-suite :: External/SPEC/CFP2017speed/638.imagick_s/638.imagick_s.test 1394420.00 1394484.00 0.0% test-suite :: External/SPEC/CFP2017rate/538.imagick_r/538.imagick_r.test 1394420.00 1394484.00 0.0% test-suite :: External/SPEC/CFP2017rate/510.parest_r/510.parest_r.test 2040257.00 2040273.00 0.0% test-suite :: External/SPEC/CFP2017rate/526.blender_r/526.blender_r.test 12396098.00 12395858.00 -0.0% test-suite :: External/SPEC/CINT2006/445.gobmk/445.gobmk.test 909944.00 909768.00 -0.0% SingleSource/Benchmarks/Adobe-C++/simple_types_loop_invariant - 4 scalar instructions remain scalar (good). Spec2017/x264 - the whole function idct4x4dc is vectorized using <16 x i16> instead of <16 x i32>, also zext/trunc are removed. In other places last vector zext/sext removed and replaced by extractelement + scalar zext/sext pair. MultiSource/Benchmarks/Bullet/bullet - reduce or <4 x i32> replaced by reduce or <4 x i8> Spec2017/imagick - Removed extra zext from 2 packs of the operations. Spec2017/parest - Removed extra zext, replaced by extractelement+scalar zext Spec2017/blender - the whole bunch of vector zext/sext replaced by extractelement+scalar zext/sext, some extra code vectorized in smaller types. Spec2006/gobmk - fixed cost estimation, some small code remains scalar. Original Pull Request: https://github.com/llvm/llvm-project/pull/84334 The patch has the same functionality (no test changes, no changes in benchmarks) as the original patch, just has some compile time improvements + fixes for xxhash unittest, discovered earlier in the previous version of the patch. Reviewers: Pull Request: https://github.com/llvm/llvm-project/pull/84536 
This improves overall analysis for minbitwidth in SLP. It allows to
analyze the trees with store/insertelement root nodes. Also, instead of
using single minbitwidth, detected from the very first analysis stage,
it tries to detect the best one for each trunc/ext subtree in the graph
and use it for the subtree.
Results in better code and less vector register pressure.

Metric: size..text

Program                                                                                                                                                size..text
                                                                                                                                                       results     results0    diff
                                                                      test-suite :: SingleSource/Benchmarks/Adobe-C++/simple_types_loop_invariant.test    92549.00    92609.00  0.1%
                                                                                  test-suite :: External/SPEC/CINT2017speed/625.x264_s/625.x264_s.test   663381.00   663493.00  0.0%
                                                                                   test-suite :: External/SPEC/CINT2017rate/525.x264_r/525.x264_r.test   663381.00   663493.00  0.0%
                                                                                               test-suite :: MultiSource/Benchmarks/Bullet/bullet.test   307182.00   307214.00  0.0%
                                                                             test-suite :: External/SPEC/CFP2017speed/638.imagick_s/638.imagick_s.test  1394420.00  1394484.00  0.0%
                                                                              test-suite :: External/SPEC/CFP2017rate/538.imagick_r/538.imagick_r.test  1394420.00  1394484.00  0.0%
                                                                                test-suite :: External/SPEC/CFP2017rate/510.parest_r/510.parest_r.test  2040257.00  2040273.00  0.0%

                                                                              test-suite :: External/SPEC/CFP2017rate/526.blender_r/526.blender_r.test 12396098.00 12395858.00 -0.0%
                                                                                         test-suite :: External/SPEC/CINT2006/445.gobmk/445.gobmk.test   909944.00   909768.00 -0.0%

SingleSource/Benchmarks/Adobe-C++/simple_types_loop_invariant - 4 scalar
instructions remain scalar (good).
Spec2017/x264 - the whole function idct4x4dc is vectorized using <16
x i16> instead of <16 x i32>, also zext/trunc are removed. In other
places last vector zext/sext removed and replaced by
extractelement + scalar zext/sext pair.
MultiSource/Benchmarks/Bullet/bullet - reduce or <4 x i32> replaced by
reduce or <4 x i8>
Spec2017/imagick - Removed extra zext from 2 packs of the operations.
Spec2017/parest - Removed extra zext, replaced by extractelement+scalar
zext
Spec2017/blender - the whole bunch of vector zext/sext replaced by
extractelement+scalar zext/sext, some extra code vectorized in smaller
types.
Spec2006/gobmk - fixed cost estimation, some small code remains scalar.

Original Pull Request: https://github.com/llvm/llvm-project/pull/84334

The patch has the same functionality (no test changes, no changes in
benchmarks) as the original patch, just has some compile time
improvements + fixes for xxhash unittest, discovered earlier in the
previous version of the patch.

Reviewers:

Pull Request: https://github.com/llvm/llvm-project/pull/84536
Revert "[SLP]Improve minbitwidth analysis." 2024-03-09T11:53:13+00:00 Martin Storsjö martin@martin.st 2024-03-09T10:57:20+00:00 5b5c21d772d20320fd876edddfc204cca93fae0f This reverts commit 2bd369b48dbf0bc3128becb7ef8f8a1b82514b87. That commit triggered failed assertions: $ cat repro.c short *a; int b; void h() { short *c = a; b = 0; for (; b < 4; b++) { unsigned d = a[b] + a[b + 4 * 2], e = a[b] - a[b + 4 * 2], f = (a[b + 4] >> 1) - a[b + 4 * 3], g = a[b + 4] + (a[b + 4 * 3] >> 1); c[b] = g; c[b + 4] = e + f; c[b + 4 * 2] = e - f; c[b + 4 * 3] = d - g; } } $ clang -target aarch64-linux-gnu -c -O2 repro.c clang: ../lib/Transforms/Vectorize/SLPVectorizer.cpp:12503: llvm::Value* llvm::slpvectorizer::BoUpSLP::vectorizeTree(llvm::slpvectorizer::BoUpSLP::TreeEntry*, bool): Assertion `(MinBWs.contains(getOperandEntry(E, 0)) || MinBWs.contains(getOperandEntry(E, 1))) && "Expected item in MinBWs."' failed. 
This reverts commit 2bd369b48dbf0bc3128becb7ef8f8a1b82514b87.

That commit triggered failed assertions:
$ cat repro.c
short *a;
int b;
void h() {
  short *c = a;
  b = 0;
  for (; b < 4; b++) {
    unsigned d = a[b] + a[b + 4 * 2], e = a[b] - a[b + 4 * 2],
             f = (a[b + 4] >> 1) - a[b + 4 * 3],
             g = a[b + 4] + (a[b + 4 * 3] >> 1);
    c[b] = g;
    c[b + 4] = e + f;
    c[b + 4 * 2] = e - f;
    c[b + 4 * 3] = d - g;
  }
}
$ clang -target aarch64-linux-gnu -c -O2 repro.c
clang: ../lib/Transforms/Vectorize/SLPVectorizer.cpp:12503: llvm::Value* llvm::slpvectorizer::BoUpSLP::vectorizeTree(llvm::slpvectorizer::BoUpSLP::TreeEntry*, bool): Assertion `(MinBWs.contains(getOperandEntry(E, 0)) || MinBWs.contains(getOperandEntry(E, 1))) && "Expected item in MinBWs."' failed.
[SLP]Improve minbitwidth analysis. 2024-03-08T18:57:02+00:00 Alexey Bataev 5361294+alexey-bataev@users.noreply.github.com 2024-03-08T18:57:02+00:00 2bd369b48dbf0bc3128becb7ef8f8a1b82514b87 This improves overall analysis for minbitwidth in SLP. It allows to analyze the trees with store/insertelement root nodes. Also, instead of using single minbitwidth, detected from the very first analysis stage, it tries to detect the best one for each trunc/ext subtree in the graph and use it for the subtree. Results in better code and less vector register pressure. Metric: size..text Program size..text results results0 diff test-suite :: SingleSource/Benchmarks/Adobe-C++/simple_types_loop_invariant.test 92549.00 92609.00 0.1% test-suite :: External/SPEC/CINT2017speed/625.x264_s/625.x264_s.test 663381.00 663493.00 0.0% test-suite :: External/SPEC/CINT2017rate/525.x264_r/525.x264_r.test 663381.00 663493.00 0.0% test-suite :: MultiSource/Benchmarks/Bullet/bullet.test 307182.00 307214.00 0.0% test-suite :: External/SPEC/CFP2017speed/638.imagick_s/638.imagick_s.test 1394420.00 1394484.00 0.0% test-suite :: External/SPEC/CFP2017rate/538.imagick_r/538.imagick_r.test 1394420.00 1394484.00 0.0% test-suite :: External/SPEC/CFP2017rate/510.parest_r/510.parest_r.test 2040257.00 2040273.00 0.0% test-suite :: External/SPEC/CFP2017rate/526.blender_r/526.blender_r.test 12396098.00 12395858.00 -0.0% test-suite :: External/SPEC/CINT2006/445.gobmk/445.gobmk.test 909944.00 909768.00 -0.0% SingleSource/Benchmarks/Adobe-C++/simple_types_loop_invariant - 4 scalar instructions remain scalar (good). Spec2017/x264 - the whole function idct4x4dc is vectorized using <16 x i16> instead of <16 x i32>, also zext/trunc are removed. In other places last vector zext/sext removed and replaced by extractelement + scalar zext/sext pair. MultiSource/Benchmarks/Bullet/bullet - reduce or <4 x i32> replaced by reduce or <4 x i8> Spec2017/imagick - Removed extra zext from 2 packs of the operations. Spec2017/parest - Removed extra zext, replaced by extractelement+scalar zext Spec2017/blender - the whole bunch of vector zext/sext replaced by extractelement+scalar zext/sext, some extra code vectorized in smaller types. Spec2006/gobmk - fixed cost estimation, some small code remains scalar. Original Pull Request: https://github.com/llvm/llvm-project/pull/84334 The patch has the same functionality (no test changes, no changes in benchmarks) as the original patch, just has some compile time improvements + fixes for xxhash unittest, discovered earlier in the previous version of the patch. Reviewers: Pull Request: https://github.com/llvm/llvm-project/pull/84536 
This improves overall analysis for minbitwidth in SLP. It allows to
analyze the trees with store/insertelement root nodes. Also, instead of
using single minbitwidth, detected from the very first analysis stage,
it tries to detect the best one for each trunc/ext subtree in the graph
and use it for the subtree.
Results in better code and less vector register pressure.

Metric: size..text

Program                                                                                                                                                size..text
                                                                                                                                                       results     results0    diff
                                                                      test-suite :: SingleSource/Benchmarks/Adobe-C++/simple_types_loop_invariant.test    92549.00    92609.00  0.1%
                                                                                  test-suite :: External/SPEC/CINT2017speed/625.x264_s/625.x264_s.test   663381.00   663493.00  0.0%
                                                                                   test-suite :: External/SPEC/CINT2017rate/525.x264_r/525.x264_r.test   663381.00   663493.00  0.0%
                                                                                               test-suite :: MultiSource/Benchmarks/Bullet/bullet.test   307182.00   307214.00  0.0%
                                                                             test-suite :: External/SPEC/CFP2017speed/638.imagick_s/638.imagick_s.test  1394420.00  1394484.00  0.0%
                                                                              test-suite :: External/SPEC/CFP2017rate/538.imagick_r/538.imagick_r.test  1394420.00  1394484.00  0.0%
                                                                                test-suite :: External/SPEC/CFP2017rate/510.parest_r/510.parest_r.test  2040257.00  2040273.00  0.0%

                                                                              test-suite :: External/SPEC/CFP2017rate/526.blender_r/526.blender_r.test 12396098.00 12395858.00 -0.0%
                                                                                         test-suite :: External/SPEC/CINT2006/445.gobmk/445.gobmk.test   909944.00   909768.00 -0.0%

SingleSource/Benchmarks/Adobe-C++/simple_types_loop_invariant - 4 scalar
instructions remain scalar (good).
Spec2017/x264 - the whole function idct4x4dc is vectorized using <16
x i16> instead of <16 x i32>, also zext/trunc are removed. In other
places last vector zext/sext removed and replaced by
extractelement + scalar zext/sext pair.
MultiSource/Benchmarks/Bullet/bullet - reduce or <4 x i32> replaced by
reduce or <4 x i8>
Spec2017/imagick - Removed extra zext from 2 packs of the operations.
Spec2017/parest - Removed extra zext, replaced by extractelement+scalar
zext
Spec2017/blender - the whole bunch of vector zext/sext replaced by
extractelement+scalar zext/sext, some extra code vectorized in smaller
types.
Spec2006/gobmk - fixed cost estimation, some small code remains scalar.

Original Pull Request: https://github.com/llvm/llvm-project/pull/84334

The patch has the same functionality (no test changes, no changes in
benchmarks) as the original patch, just has some compile time
improvements + fixes for xxhash unittest, discovered earlier in the
previous version of the patch.

Reviewers: 

Pull Request: https://github.com/llvm/llvm-project/pull/84536
Revert "[SLP]Improve minbitwidth analysis." 2024-03-07T20:44:53+00:00 Alexey Bataev a.bataev@outlook.com 2024-03-07T20:44:25+00:00 11185715a28c6592ca6fe247fe693b305c85627a This reverts commit 4ce52e2d576937fe930294cae883a0daa17eeced to fix issues detected by https://lab.llvm.org/buildbot/#/builders/74/builds/26470/steps/12/logs/stdio. 
This reverts commit 4ce52e2d576937fe930294cae883a0daa17eeced to fix
issues detected by https://lab.llvm.org/buildbot/#/builders/74/builds/26470/steps/12/logs/stdio.
[SLP]Improve minbitwidth analysis. 2024-03-07T15:36:41+00:00 Alexey Bataev 5361294+alexey-bataev@users.noreply.github.com 2024-03-07T15:36:41+00:00 4ce52e2d576937fe930294cae883a0daa17eeced This improves overall analysis for minbitwidth in SLP. It allows to analyze the trees with store/insertelement root nodes. Also, instead of using single minbitwidth, detected from the very first analysis stage, it tries to detect the best one for each trunc/ext subtree in the graph and use it for the subtree. Results in better code and less vector register pressure. Metric: size..text Program size..text results results0 diff test-suite :: SingleSource/Benchmarks/Adobe-C++/simple_types_loop_invariant.test 92549.00 92609.00 0.1% test-suite :: External/SPEC/CINT2017speed/625.x264_s/625.x264_s.test 663381.00 663493.00 0.0% test-suite :: External/SPEC/CINT2017rate/525.x264_r/525.x264_r.test 663381.00 663493.00 0.0% test-suite :: MultiSource/Benchmarks/Bullet/bullet.test 307182.00 307214.00 0.0% test-suite :: External/SPEC/CFP2017speed/638.imagick_s/638.imagick_s.test 1394420.00 1394484.00 0.0% test-suite :: External/SPEC/CFP2017rate/538.imagick_r/538.imagick_r.test 1394420.00 1394484.00 0.0% test-suite :: External/SPEC/CFP2017rate/510.parest_r/510.parest_r.test 2040257.00 2040273.00 0.0% test-suite :: External/SPEC/CFP2017rate/526.blender_r/526.blender_r.test 12396098.00 12395858.00 -0.0% test-suite :: External/SPEC/CINT2006/445.gobmk/445.gobmk.test 909944.00 909768.00 -0.0% SingleSource/Benchmarks/Adobe-C++/simple_types_loop_invariant - 4 scalar instructions remain scalar (good). Spec2017/x264 - the whole function idct4x4dc is vectorized using <16 x i16> instead of <16 x i32>, also zext/trunc are removed. In other places last vector zext/sext removed and replaced by extractelement + scalar zext/sext pair. MultiSource/Benchmarks/Bullet/bullet - reduce or <4 x i32> replaced by reduce or <4 x i8> Spec2017/imagick - Removed extra zext from 2 packs of the operations. Spec2017/parest - Removed extra zext, replaced by extractelement+scalar zext Spec2017/blender - the whole bunch of vector zext/sext replaced by extractelement+scalar zext/sext, some extra code vectorized in smaller types. Spec2006/gobmk - fixed cost estimation, some small code remains scalar. Reviewers: RKSimon Pull Request: https://github.com/llvm/llvm-project/pull/84334 
This improves overall analysis for minbitwidth in SLP. It allows to
analyze the trees with store/insertelement root nodes. Also, instead of
using single minbitwidth, detected from the very first analysis stage,
it tries to detect the best one for each trunc/ext subtree in the graph
and use it for the subtree.
Results in better code and less vector register pressure.

Metric: size..text

Program                                                                                                                                                size..text
                                                                                                                                                       results     results0    diff
                                                                      test-suite :: SingleSource/Benchmarks/Adobe-C++/simple_types_loop_invariant.test    92549.00    92609.00  0.1%
                                                                                  test-suite :: External/SPEC/CINT2017speed/625.x264_s/625.x264_s.test   663381.00   663493.00  0.0%
                                                                                   test-suite :: External/SPEC/CINT2017rate/525.x264_r/525.x264_r.test   663381.00   663493.00  0.0%
                                                                                               test-suite :: MultiSource/Benchmarks/Bullet/bullet.test   307182.00   307214.00  0.0%
                                                                             test-suite :: External/SPEC/CFP2017speed/638.imagick_s/638.imagick_s.test  1394420.00  1394484.00  0.0%
                                                                              test-suite :: External/SPEC/CFP2017rate/538.imagick_r/538.imagick_r.test  1394420.00  1394484.00  0.0%
                                                                                test-suite :: External/SPEC/CFP2017rate/510.parest_r/510.parest_r.test  2040257.00  2040273.00  0.0%

                                                                              test-suite :: External/SPEC/CFP2017rate/526.blender_r/526.blender_r.test 12396098.00 12395858.00 -0.0%
                                                                                         test-suite :: External/SPEC/CINT2006/445.gobmk/445.gobmk.test   909944.00   909768.00 -0.0%

SingleSource/Benchmarks/Adobe-C++/simple_types_loop_invariant - 4 scalar
instructions remain scalar (good).
Spec2017/x264 - the whole function idct4x4dc is vectorized using <16
x i16> instead of <16 x i32>, also zext/trunc are removed. In other
places last vector zext/sext removed and replaced by
extractelement + scalar zext/sext pair.
MultiSource/Benchmarks/Bullet/bullet - reduce or <4 x i32> replaced by
reduce or <4 x i8>
Spec2017/imagick - Removed extra zext from 2 packs of the operations.
Spec2017/parest - Removed extra zext, replaced by extractelement+scalar
zext
Spec2017/blender - the whole bunch of vector zext/sext replaced by
extractelement+scalar zext/sext, some extra code vectorized in smaller
types.
Spec2006/gobmk - fixed cost estimation, some small code remains scalar.

Reviewers: RKSimon

Pull Request: https://github.com/llvm/llvm-project/pull/84334
Revert "[SLP]Improve minbitwidth analysis." 2024-03-05T20:13:45+00:00 Alexey Bataev a.bataev@outlook.com 2024-03-05T20:08:02+00:00 aae152f1be1db1b4760c7464a3accd3d72b5f4bd This reverts commit a730ed7c1a4a35f5219df720ffb0ba6122d64fe4 to fix compile time issue. 
This reverts commit a730ed7c1a4a35f5219df720ffb0ba6122d64fe4 to fix
compile time issue.
[SLP]Improve minbitwidth analysis. 2024-03-05T17:20:28+00:00 Alexey Bataev 5361294+alexey-bataev@users.noreply.github.com 2024-03-05T17:20:28+00:00 a730ed7c1a4a35f5219df720ffb0ba6122d64fe4 This improves overall analysis for minbitwidth in SLP. It allows to analyze the trees with store/insertelement root nodes. Also, instead of using single minbitwidth, detected from the very first analysis stage, it tries to detect the best one for each trunc/ext subtree in the graph and use it for the subtree. Results in better code and less vector register pressure. Metric: size..text Program size..text results results0 diff test-suite :: SingleSource/Benchmarks/Adobe-C++/simple_types_loop_invariant.test 92549.00 92609.00 0.1% test-suite :: External/SPEC/CINT2017speed/625.x264_s/625.x264_s.test 663381.00 663493.00 0.0% test-suite :: External/SPEC/CINT2017rate/525.x264_r/525.x264_r.test 663381.00 663493.00 0.0% test-suite :: MultiSource/Benchmarks/Bullet/bullet.test 307182.00 307214.00 0.0% test-suite :: External/SPEC/CFP2017speed/638.imagick_s/638.imagick_s.test 1394420.00 1394484.00 0.0% test-suite :: External/SPEC/CFP2017rate/538.imagick_r/538.imagick_r.test 1394420.00 1394484.00 0.0% test-suite :: External/SPEC/CFP2017rate/510.parest_r/510.parest_r.test 2040257.00 2040273.00 0.0% test-suite :: External/SPEC/CFP2017rate/526.blender_r/526.blender_r.test 12396098.00 12395858.00 -0.0% test-suite :: External/SPEC/CINT2006/445.gobmk/445.gobmk.test 909944.00 909768.00 -0.0% SingleSource/Benchmarks/Adobe-C++/simple_types_loop_invariant - 4 scalar instructions remain scalar (good). Spec2017/x264 - the whole function idct4x4dc is vectorized using <16 x i16> instead of <16 x i32>, also zext/trunc are removed. In other places last vector zext/sext removed and replaced by extractelement + scalar zext/sext pair. MultiSource/Benchmarks/Bullet/bullet - reduce or <4 x i32> replaced by reduce or <4 x i8> Spec2017/imagick - Removed extra zext from 2 packs of the operations. Spec2017/parest - Removed extra zext, replaced by extractelement+scalar zext Spec2017/blender - the whole bunch of vector zext/sext replaced by extractelement+scalar zext/sext, some extra code vectorized in smaller types. Spec2006/gobmk - fixed cost estimation, some small code remains scalar. Reviewers: RKSimon Reviewed By: RKSimon Pull Request: https://github.com/llvm/llvm-project/pull/78976 
This improves overall analysis for minbitwidth in SLP. It allows to
analyze the trees with store/insertelement root nodes. Also, instead of
using single minbitwidth, detected from the very first analysis stage,
it tries to detect the best one for each trunc/ext subtree in the graph
and use it for the subtree.
Results in better code and less vector register pressure.

Metric: size..text

Program                                                                                                                                                size..text
                                                                                                                                                       results     results0    diff
                                                                      test-suite :: SingleSource/Benchmarks/Adobe-C++/simple_types_loop_invariant.test    92549.00    92609.00  0.1%
                                                                                  test-suite :: External/SPEC/CINT2017speed/625.x264_s/625.x264_s.test   663381.00   663493.00  0.0%
                                                                                   test-suite :: External/SPEC/CINT2017rate/525.x264_r/525.x264_r.test   663381.00   663493.00  0.0%
                                                                                               test-suite :: MultiSource/Benchmarks/Bullet/bullet.test   307182.00   307214.00  0.0%
                                                                             test-suite :: External/SPEC/CFP2017speed/638.imagick_s/638.imagick_s.test  1394420.00  1394484.00  0.0%
                                                                              test-suite :: External/SPEC/CFP2017rate/538.imagick_r/538.imagick_r.test  1394420.00  1394484.00  0.0%
                                                                                test-suite :: External/SPEC/CFP2017rate/510.parest_r/510.parest_r.test  2040257.00  2040273.00  0.0%

                                                                              test-suite :: External/SPEC/CFP2017rate/526.blender_r/526.blender_r.test 12396098.00 12395858.00 -0.0%
                                                                                         test-suite :: External/SPEC/CINT2006/445.gobmk/445.gobmk.test   909944.00   909768.00 -0.0%

SingleSource/Benchmarks/Adobe-C++/simple_types_loop_invariant - 4 scalar
instructions remain scalar (good).
Spec2017/x264 - the whole function idct4x4dc is vectorized using <16
x i16> instead of <16 x i32>, also zext/trunc are removed. In other
places last vector zext/sext removed and replaced by
extractelement + scalar zext/sext pair.
MultiSource/Benchmarks/Bullet/bullet - reduce or <4 x i32> replaced by
reduce or <4 x i8>
Spec2017/imagick - Removed extra zext from 2 packs of the operations.
Spec2017/parest - Removed extra zext, replaced by extractelement+scalar
zext
Spec2017/blender - the whole bunch of vector zext/sext replaced by
extractelement+scalar zext/sext, some extra code vectorized in smaller
types.
Spec2006/gobmk - fixed cost estimation, some small code remains scalar.

Reviewers: RKSimon

Reviewed By: RKSimon

Pull Request: https://github.com/llvm/llvm-project/pull/78976
[SLP]Fix a crash in the analysis of the compatible cmp operands. 2022-10-31T16:47:25+00:00 Alexey Bataev a.bataev@outlook.com 2022-10-31T15:54:49+00:00 99f9bd4807d6132b603a1910902e4586e20638e0 We can skip the analysis of the operands opcodes, can compare directly them in some cases. 
We can skip the analysis of the operands opcodes, can compare directly
them in some cases.