[SLP]Try partial store vectorization if supported by target.

[ABataev]

Summary

We can try to vectorize number of stores less than MinVecRegSize
/ scalar_value_size, if it is allowed by target. Gives an extra
opportunity for the vectorization.

Fixes PR54985.

[xbolva00]

llvm/test/Transforms/SLPVectorizer/X86/pr49933.ll

2–3

Remove please.

4

Great!

This patch also fixes https://github.com/llvm/llvm-project/issues/49277

[ABataev]

Address comment

[ABataev]

I’m going to do some extra testing next week for this patch (if the bugs with non-typed pointers are fixed).

[ABataev]

Metric: SLP.NumVectorInstructions

Program                                                                                  SLP.NumVectorInstructions
                                                                                         results                   results0 diff
                    test-suite :: MultiSource/Benchmarks/Prolangs-C++/shapes/shapes.test     0.00                      6.00     inf%
                      test-suite :: MultiSource/Benchmarks/Prolangs-C/bison/mybison.test     0.00                     20.00     inf%
                        test-suite :: MultiSource/Benchmarks/Prolangs-C++/city/city.test     0.00                      4.00     inf%
              test-suite :: MultiSource/Benchmarks/Prolangs-C/unix-smail/unix-smail.test     0.00                      2.00     inf%
                    test-suite :: MultiSource/Benchmarks/BitBench/uuencode/uuencode.test     0.00                      3.00     inf%
                              test-suite :: SingleSource/Benchmarks/Stanford/Towers.test     0.00                      1.00     inf%
                                        test-suite :: SingleSource/UnitTests/initp1.test     0.00                     20.00     inf%
                       test-suite :: SingleSource/UnitTests/ms_struct-bitfield-init.test     0.00                      1.00     inf%
     test-suite :: MultiSource/Benchmarks/MiBench/network-dijkstra/network-dijkstra.test     0.00                      2.00     inf%
     test-suite :: MultiSource/Benchmarks/MiBench/automotive-susan/automotive-susan.test     0.00                      4.00     inf%
                        test-suite :: MultiSource/Benchmarks/McCat/01-qbsort/qbsort.test     0.00                      2.00     inf%
                        test-suite :: MultiSource/Benchmarks/McCat/12-IOtest/iotest.test     0.00                      1.00     inf%
                                test-suite :: SingleSource/Benchmarks/Dhrystone/dry.test     0.00                      3.00     inf%
                                 test-suite :: MultiSource/Applications/sgefa/sgefa.test     0.00                      1.00     inf%
                       test-suite :: MultiSource/Benchmarks/Rodinia/hotspot/hotspot.test     0.00                      2.00     inf%
                 test-suite :: MultiSource/Benchmarks/Rodinia/pathfinder/pathfinder.test     0.00                      2.00     inf%
                       test-suite :: MultiSource/Benchmarks/MallocBench/cfrac/cfrac.test     0.00                      4.00     inf%
                      test-suite :: MultiSource/Benchmarks/Trimaran/enc-rc4/enc-rc4.test     0.00                      1.00     inf%
                     test-suite :: MultiSource/Benchmarks/Rodinia/backprop/backprop.test     0.00                      3.00     inf%
                                     test-suite :: MultiSource/Applications/aha/aha.test     0.00                      4.00     inf%
                                 test-suite :: MultiSource/Applications/spiff/spiff.test     0.00                      5.00     inf%
      test-suite :: MultiSource/Benchmarks/TSVC/NodeSplitting-flt/NodeSplitting-flt.test     0.00                      4.00     inf%
                         test-suite :: MultiSource/Applications/lambda-0.1.3/lambda.test     0.00                      2.00     inf%
            test-suite :: MultiSource/Benchmarks/Trimaran/netbench-url/netbench-url.test     0.00                      1.00     inf%
                           test-suite :: MultiSource/Applications/hexxagon/hexxagon.test     0.00                      8.00     inf%
                  test-suite :: MultiSource/Benchmarks/Prolangs-C/unix-tbl/unix-tbl.test     0.00                      2.00     inf%
                        test-suite :: MultiSource/Benchmarks/Prolangs-C++/life/life.test     0.00                     20.00     inf%
                  test-suite :: MultiSource/Benchmarks/Prolangs-C++/objects/objects.test     0.00                      3.00     inf%
           test-suite :: MultiSource/Benchmarks/MiBench/office-ispell/office-ispell.test     0.00                      5.00     inf%
                test-suite :: MultiSource/Benchmarks/Prolangs-C/assembler/assembler.test     0.00                      2.00     inf%
                                   test-suite :: MultiSource/Applications/siod/siod.test     2.00                    209.00 10350.0%
                                     test-suite :: MultiSource/Applications/lua/lua.test     1.00                     46.00  4500.0%
                             test-suite :: MultiSource/Applications/sqlite3/sqlite3.test    21.00                    438.00  1985.7%
                               test-suite :: SingleSource/Benchmarks/Stanford/Oscar.test     2.00                     30.00  1400.0%
                 test-suite :: MultiSource/Benchmarks/MallocBench/espresso/espresso.test     3.00                     41.00  1266.7%
                      test-suite :: MultiSource/Benchmarks/Prolangs-C++/ocean/ocean.test     2.00                     26.00  1200.0%
              test-suite :: MultiSource/Benchmarks/VersaBench/beamformer/beamformer.test    32.00                    378.00  1081.2%
                                   test-suite :: MultiSource/Benchmarks/PAQ8p/paq8p.test    10.00                     77.00   670.0%
                              test-suite :: MultiSource/Applications/d/make_dparser.test     2.00                     15.00   650.0%
                   test-suite :: External/SPEC/CINT2017rate/541.leela_r/541.leela_r.test    20.00                    141.00   605.0%
                  test-suite :: External/SPEC/CINT2017speed/641.leela_s/641.leela_s.test    20.00                    141.00   605.0%
              test-suite :: MultiSource/Applications/ALAC/decode/alacconvert-decode.test     2.00                     13.00   550.0%
              test-suite :: MultiSource/Applications/ALAC/encode/alacconvert-encode.test     2.00                     13.00   550.0%
             test-suite :: MultiSource/Benchmarks/mediabench/g721/g721encode/encode.test     5.00                     32.00   540.0%
           test-suite :: External/SPEC/CINT2017rate/500.perlbench_r/500.perlbench_r.test   100.00                    507.00   407.0%
          test-suite :: External/SPEC/CINT2017speed/600.perlbench_s/600.perlbench_s.test   100.00                    507.00   407.0%
                               test-suite :: SingleSource/Benchmarks/McGill/exptree.test     1.00                      5.00   400.0%
           test-suite :: MultiSource/Benchmarks/DOE-ProxyApps-C/SimpleMOC/SimpleMOC.test     2.00                     10.00   400.0%
                              test-suite :: MultiSource/Benchmarks/McCat/18-imp/imp.test    11.00                     44.00   300.0%
                                 test-suite :: MultiSource/Benchmarks/Ptrdist/bc/bc.test     5.00                     18.00   260.0%
                   test-suite :: MultiSource/Benchmarks/ASCI_Purple/SMG2000/smg2000.test   131.00                    469.00   258.0%
                       test-suite :: MultiSource/Benchmarks/Fhourstones/fhourstones.test     8.00                     28.00   250.0%
                   test-suite :: External/SPEC/CINT2006/400.perlbench/400.perlbench.test    73.00                    222.00   204.1%
         test-suite :: MultiSource/Benchmarks/DOE-ProxyApps-C/Pathfinder/PathFinder.test     2.00                      6.00   200.0%
                                 test-suite :: MultiSource/Applications/lemon/lemon.test     5.00                     15.00   200.0%
                             test-suite :: MultiSource/Applications/SIBsim4/SIBsim4.test    12.00                     35.00   191.7%
                           test-suite :: External/SPEC/CINT2006/456.hmmer/456.hmmer.test    97.00                    282.00   190.7%
                test-suite :: External/SPEC/CFP2017rate/526.blender_r/526.blender_r.test 11194.00                  31786.00   184.0%
               test-suite :: MultiSource/Benchmarks/Prolangs-C/archie-client/archie.test     4.00                     11.00   175.0%
                test-suite :: MultiSource/Benchmarks/Fhourstones-3.1/fhourstones3.1.test     7.00                     18.00   157.1%
     test-suite :: MultiSource/Benchmarks/MiBench/consumer-typeset/consumer-typeset.test    47.00                    120.00   155.3%
                              test-suite :: SingleSource/Benchmarks/Stanford/Puzzle.test    11.00                     28.00   154.5%
                                     test-suite :: MultiSource/Applications/hbd/hbd.test    41.00                    104.00   153.7%
                      test-suite :: External/SPEC/CINT2017speed/602.gcc_s/602.gcc_s.test   698.00                   1681.00   140.8%
                       test-suite :: External/SPEC/CINT2017rate/502.gcc_r/502.gcc_r.test   698.00                   1681.00   140.8%
                             test-suite :: MultiSource/Applications/ClamAV/clamscan.test    85.00                    195.00   129.4%
                           test-suite :: MultiSource/Benchmarks/7zip/7zip-benchmark.test   396.00                    907.00   129.0%
                           test-suite :: External/SPEC/CINT2006/401.bzip2/401.bzip2.test    31.00                     71.00   129.0%
                           test-suite :: External/SPEC/CINT2006/473.astar/473.astar.test    45.00                    101.00   124.4%
                           test-suite :: External/SPEC/CINT2006/445.gobmk/445.gobmk.test   101.00                    214.00   111.9%
                         test-suite :: MultiSource/Benchmarks/tramp3d-v4/tramp3d-v4.test  1762.00                   3598.00   104.2%
                          test-suite :: External/SPEC/CFP2006/450.soplex/450.soplex.test    64.00                    130.00   103.1%
                                test-suite :: SingleSource/Benchmarks/Stanford/Perm.test     6.00                     12.00   100.0%
                              test-suite :: SingleSource/Benchmarks/Dhrystone/fldry.test     1.00                      2.00   100.0%
                             test-suite :: MultiSource/Applications/viterbi/viterbi.test     1.00                      2.00   100.0%
                             test-suite :: MultiSource/Applications/obsequi/Obsequi.test     2.00                      4.00   100.0%
                       test-suite :: External/SPEC/CINT2006/471.omnetpp/471.omnetpp.test   120.00                    233.00    94.2%
                        test-suite :: External/SPEC/CFP2006/482.sphinx3/482.sphinx3.test    56.00                    107.00    91.1%
                               test-suite :: External/SPEC/CINT2006/403.gcc/403.gcc.test   522.00                    976.00    87.0%
                             test-suite :: MultiSource/Benchmarks/MallocBench/gs/gs.test   165.00                    295.00    78.8%
                                 test-suite :: MultiSource/Applications/SPASS/SPASS.test   176.00                    307.00    74.4%
                             test-suite :: MultiSource/Benchmarks/Rodinia/srad/srad.test     3.00                      5.00    66.7%
                                 test-suite :: MultiSource/Benchmarks/Bullet/bullet.test  6965.00                  11467.00    64.6%
                  test-suite :: MultiSource/Benchmarks/Prolangs-C/football/football.test    45.00                     73.00    62.2%
                      test-suite :: SingleSource/Benchmarks/Misc/richards_benchmark.test    10.00                     16.00    60.0%
          test-suite :: MultiSource/Benchmarks/Prolangs-C/TimberWolfMC/timberwolfmc.test   681.00                   1080.00    58.6%
                            test-suite :: MultiSource/Applications/JM/lencod/lencod.test  1175.00                   1814.00    54.4%
                        test-suite :: MultiSource/Benchmarks/Prolangs-C/agrep/agrep.test    24.00                     37.00    54.2%
                       test-suite :: External/SPEC/CINT2006/464.h264ref/464.h264ref.test   980.00                   1508.00    53.9%
                           test-suite :: External/SPEC/CINT2006/458.sjeng/458.sjeng.test    32.00                     49.00    53.1%
               test-suite :: MultiSource/Benchmarks/DOE-ProxyApps-C/miniAMR/miniAMR.test    49.00                     74.00    51.0%
                 test-suite :: External/SPEC/CINT2006/462.libquantum/462.libquantum.test   107.00                    161.00    50.5%
                test-suite :: External/SPEC/CFP2017rate/538.imagick_r/538.imagick_r.test  3638.00                   5474.00    50.5%
               test-suite :: External/SPEC/CFP2017speed/638.imagick_s/638.imagick_s.test  3638.00                   5474.00    50.5%
                       test-suite :: SingleSource/Benchmarks/BenchmarkGame/fannkuch.test     2.00                      3.00    50.0%                                                                                               
                         test-suite :: External/SPEC/CINT2017rate/557.xz_r/557.xz_r.test    88.00                    129.00    46.6%
                        test-suite :: External/SPEC/CINT2017speed/657.xz_s/657.xz_s.test    88.00                    129.00    46.6%
                       test-suite :: MultiSource/Benchmarks/ASC_Sequoia/AMGmk/AMGmk.test    10.00                     14.00    40.0%
                           test-suite :: MultiSource/Benchmarks/Olden/health/health.test     5.00                      7.00    40.0%
                                test-suite :: MultiSource/Applications/kimwitu++/kc.test    58.00                     81.00    39.7%
           test-suite :: MultiSource/Benchmarks/MiBench/consumer-jpeg/consumer-jpeg.test   271.00                    378.00    39.5%
                    test-suite :: External/SPEC/CINT2017speed/625.x264_s/625.x264_s.test   750.00                   1031.00    37.5%
                     test-suite :: External/SPEC/CINT2017rate/525.x264_r/525.x264_r.test   750.00                   1031.00    37.5%
             test-suite :: MultiSource/Benchmarks/MiBench/telecomm-gsm/telecomm-gsm.test    75.00                    103.00    37.3%
                    test-suite :: MultiSource/Benchmarks/mediabench/gsm/toast/toast.test    75.00                    103.00    37.3%
                 test-suite :: MultiSource/Benchmarks/mediabench/jpeg/jpeg-6a/cjpeg.test   273.00                    372.00    36.3%
                   test-suite :: External/SPEC/CINT2006/483.xalancbmk/483.xalancbmk.test  2078.00                   2778.00    33.7%
                               test-suite :: MultiSource/Applications/treecc/treecc.test    12.00                     16.00    33.3%
                                  test-suite :: SingleSource/Benchmarks/McGill/misr.test     6.00                      8.00    33.3%
                             test-suite :: MultiSource/Applications/minisat/minisat.test     3.00                      4.00    33.3%
               test-suite :: External/SPEC/CINT2017rate/520.omnetpp_r/520.omnetpp_r.test   814.00                   1084.00    33.2%
              test-suite :: External/SPEC/CINT2017speed/620.omnetpp_s/620.omnetpp_s.test   814.00                   1084.00    33.2%
                            test-suite :: MultiSource/Applications/JM/ldecod/ldecod.test   584.00                    777.00    33.0%
                               test-suite :: MultiSource/Applications/oggenc/oggenc.test   237.00                    311.00    31.2%
           test-suite :: External/SPEC/CINT2017rate/531.deepsjeng_r/531.deepsjeng_r.test    65.00                     85.00    30.8%
          test-suite :: External/SPEC/CINT2017speed/631.deepsjeng_s/631.deepsjeng_s.test    65.00                     85.00    30.8%
         test-suite :: MultiSource/Benchmarks/mediabench/mpeg2/mpeg2dec/mpeg2decode.test    53.00                     67.00    26.4%
                              test-suite :: SingleSource/Benchmarks/Misc-C++/bigfib.test     4.00                      5.00    25.0%
                                 test-suite :: MultiSource/Benchmarks/nbench/nbench.test   218.00                    271.00    24.3%
           test-suite :: External/SPEC/CINT2017rate/523.xalancbmk_r/523.xalancbmk_r.test  3719.00                   4588.00    23.4%
          test-suite :: External/SPEC/CINT2017speed/623.xalancbmk_s/623.xalancbmk_s.test  3719.00                   4588.00    23.4%
           test-suite :: MultiSource/Benchmarks/MiBench/consumer-lame/consumer-lame.test   670.00                    820.00    22.4%
                       test-suite :: MultiSource/Benchmarks/Ptrdist/anagram/anagram.test    32.00                     39.00    21.9%
                  test-suite :: External/SPEC/CFP2017rate/511.povray_r/511.povray_r.test  4980.00                   6059.00    21.7%
                          test-suite :: External/SPEC/CFP2006/453.povray/453.povray.test  4991.00                   6071.00    21.6%
                           test-suite :: MultiSource/Benchmarks/SciMark2-C/scimark2.test    10.00                     12.00    20.0%
                 test-suite :: MultiSource/Benchmarks/DOE-ProxyApps-C++/CLAMR/CLAMR.test   490.00                    586.00    19.6%
                  test-suite :: External/SPEC/CFP2017rate/510.parest_r/510.parest_r.test 15903.00                  18607.00    17.0%
                          test-suite :: External/SPEC/CFP2006/447.dealII/447.dealII.test  5982.00                   6994.00    16.9%
                       test-suite :: External/SPEC/CFP2017speed/644.nab_s/644.nab_s.test   494.00                    553.00    11.9%
                        test-suite :: External/SPEC/CFP2017rate/544.nab_r/544.nab_r.test   494.00                    553.00    11.9%
               test-suite :: MultiSource/Benchmarks/DOE-ProxyApps-C/XSBench/XSBench.test    27.00                     30.00    11.1%
             test-suite :: MultiSource/Benchmarks/MiBench/security-sha/security-sha.test    18.00                     20.00    11.1%
                             test-suite :: SingleSource/Benchmarks/Misc/ReedSolomon.test    25.00                     27.00     8.0%
               test-suite :: MultiSource/Benchmarks/DOE-ProxyApps-C/miniGMG/miniGMG.test    86.00                     92.00     7.0%
                           test-suite :: SingleSource/Benchmarks/Misc-C++-EH/spirit.test    16.00                     17.00     6.2%
             test-suite :: MultiSource/Benchmarks/DOE-ProxyApps-C++/PENNANT/PENNANT.test   236.00                    247.00     4.7%
                      test-suite :: External/SPEC/CFP2017rate/508.namd_r/508.namd_r.test  6030.00                   6307.00     4.6%
               test-suite :: MultiSource/Benchmarks/DOE-ProxyApps-C++/miniFE/miniFE.test   285.00                    298.00     4.6%
                     test-suite :: MultiSource/Benchmarks/FreeBench/distray/distray.test    89.00                     93.00     4.5%
               test-suite :: MultiSource/Benchmarks/DOE-ProxyApps-C/RSBench/rsbench.test   102.00                    106.00     3.9%
                              test-suite :: External/SPEC/CFP2006/444.namd/444.namd.test  3098.00                   3198.00     3.2%
                             test-suite :: SingleSource/UnitTests/matrix-types-spec.test    31.00                     32.00     3.2%
                     test-suite :: MultiSource/Benchmarks/DOE-ProxyApps-C/CoMD/CoMD.test   143.00                    147.00     2.8%
                 test-suite :: MultiSource/Benchmarks/DOE-ProxyApps-C++/HPCCG/HPCCG.test    49.00                     50.00     2.0%
                            test-suite :: MultiSource/Benchmarks/McCat/08-main/main.test    59.00                     60.00     1.7%
                          test-suite :: MultiSource/Benchmarks/mafft/pairlocalalign.test  1023.00                   1038.00     1.5%
                test-suite :: MultiSource/Benchmarks/Prolangs-C/simulator/simulator.test    84.00                     85.00     1.2%
                              test-suite :: External/SPEC/CFP2006/433.milc/433.milc.test  1020.00                   1029.00     0.9%
                              test-suite :: SingleSource/Benchmarks/SmallPT/smallpt.test   114.00                    115.00     0.9%
                         test-suite :: MultiSource/Benchmarks/FreeBench/pifft/pifft.test  1560.00                   1564.00     0.3%

Statistics. All numbers are improvements.

[xbolva00]

Looks great. Could you also observe some runtime perf improvements?

[ABataev]

In D124284#3472395, @xbolva00 wrote:

Looks great. Could you also observe some runtime perf improvements?

My system is pretty busy so the perf numbers would not be correct most probably.

[ABataev]

In D124284#3472397, @ABataev wrote:

In D124284#3472395, @xbolva00 wrote:

Looks great. Could you also observe some runtime perf improvements?

My system is pretty busy so the perf numbers would not be correct most probably.

Just an example:

test-suite :: MultiSource/Benchmarks/llubenchmark/llu.test  10.33     17.52     69.7%

The test is not affected at all.

As to some long run tests:

test-suite :: External/SPEC/CINT2006/401.bzip2/401.bzip2.test  28.09     29.60      5.4%
test-suite :: External/SPEC/CINT2017rate/502.gcc_r/502.gcc_r.test  32.65     34.00      4.1%
test-suite :: External/SPEC/CFP2017rate/526.blender_r/526.blender_r.test  96.71    100.30      3.7%
test-suite :: External/SPEC/CINT2017speed/605.mcf_s/605.mcf_s.test  49.65     50.82      2.4%
test-suite :: External/SPEC/CINT2017rate/505.mcf_r/505.mcf_r.test  50.21     50.93      1.4%
test-suite :: External/SPEC/CINT2017speed/602.gcc_s/602.gcc_s.test  33.50     31.27     -6.6%
test-suite :: External/SPEC/CINT2017rate/520.omnetpp_r/520.omnetpp_r.test  60.73     55.16     -9.2%
test-suite :: External/SPEC/CINT2017rate/557.xz_r/557.xz_r.test  36.94     32.76    -11.3%

The less %, the better. Geomean is -100% but just like I said I would not trust these numbers.

[xbolva00]

Yeah, understood.

[ABataev]

Increased priority, some numbers for long run tests:
Regressions

Metric: exec_time
test-suite :: External/SPEC/CINT2006/401.bzip2/401.bzip2.test  26.98     27.49      1.9%
test-suite :: External/SPEC/CFP2017rate/526.blender_r/526.blender_r.test  94.46     95.73      1.3%
test-suite :: External/SPEC/CFP2017speed/644.nab_s/644.nab_s.test  63.61     64.22      1.0%

Gains

test-suite :: External/SPEC/CFP2017speed/619.lbm_s/619.lbm_s.test 234.34    231.17     -1.4%
test-suite :: External/SPEC/CFP2017rate/519.lbm_r/519.lbm_r.test  27.07     26.70     -1.4%
test-suite :: External/SPEC/CINT2017speed/631.deepsjeng_s/631.deepsjeng_s.test  59.34     57.95     -2.3%
test-suite :: External/SPEC/CINT2017rate/531.deepsjeng_r/531.deepsjeng_r.test  45.51     44.42     -2.4%
test-suite :: External/SPEC/CINT2006/464.h264ref/464.h264ref.test  36.46     35.40     -2.9%
test-suite :: External/SPEC/CINT2006/445.gobmk/445.gobmk.test  57.70     55.93     -3.1%
test-suite :: External/SPEC/CINT2006/473.astar/473.astar.test  57.02     53.86     -5.5%
test-suite :: External/SPEC/CINT2017rate/520.omnetpp_r/520.omnetpp_r.test  45.46     40.61    -10.7%
test-suite :: External/SPEC/CINT2017speed/620.omnetpp_s/620.omnetpp_s.test  51.19     41.87    -18.2%

Geomean is still -100%, which means that still the performance with the patch is better than before.
Tested for O3 + LTO, generic CPU.

[xbolva00]

Very nice numbers!

Maybe @asbirlea or @fhahn want to run own internal tests?

[asbirlea]

Noted, I will run some testing this week.

[RKSimon]

llvm/test/Transforms/GVN/no_speculative_loads_with_asan.ll

53

its weird that none of tests actually CHECK-NOT for loads any more - given its the name of the test file :-|

[ABataev]

llvm/test/Transforms/GVN/no_speculative_loads_with_asan.ll

53

I can keep CHECK-NOT, if needed

[RKSimon]

llvm/test/Transforms/GVN/no_speculative_loads_with_asan.ll

53

Either that or we move to auto-generating the test file checks - I have no strong preference

[RKSimon]

rebase?

[ABataev]

Rebase

[dmgreen]

llvm/include/llvm/Analysis/TargetTransformInfo.h

1361

I don't think that exposing isLegalOrCustom to the midend is the right way to go - I feel it sets a bad precedent

I don't think that "Custom" means enough to base mid-end optimizations on. It can mean anything from "this can be custom lowered to a single instruction", to "this can _sometimes_ be custom lowered to a single instruction, in specific situations, otherwise it will expand", to "this has to be custom expanded into 150 instructions". The variance between them is just too large.

It also created a dependency between the mid-end and SDAG ISel lowering that isn't good to introduce - considering that there are other ISel's like Global ISel, there might be a point in the future where SDAG is entirely unused in certain backends.

From what I can tell (correct me if I'm wrong), what you want to add for this specific patch is a way to override/ignore getMinVectorRegisterBitWidth for stores that the target can efficiently handle. But you don't just want to change getMinVectorRegisterBitWidth? Can we add a method for doing that? shouldOverrideMinStoreVectorRegisterBitwidth(Type *Ty). The default implementation can still be the same as the current BasicTTI::isLegalOrCustomInstruction method, but it allows the target to override it if desired, and doesn't expose LegalOrCustom to the midend. Which I think is better in the long run.

[dmgreen]

llvm/include/llvm/CodeGen/BasicTTIImpl.h

318

Should this be getTypeLegalizationCost or getValueType? Otherwise we are asking for the isOperationLegalOrCustom on a legal type (LT.second) below, which you would hope was always Legal and won't really tell you much about how legal a store to Ty is.

[ABataev]

llvm/include/llvm/Analysis/TargetTransformInfo.h

1361

I'll try to invent something better.

[ABataev]

Address comments

[ABataev]

Fix test checks

[dmgreen]

Thanks for the update, this looks better to me. The perf results I have were looking OK too, except for one fp16 case where it was choosing to use v2f16 vectorization. It could be OK, it's essentially trading unrolling vs vectorizing and one isn't obviously better or worse than the other. But small vectorization factors can be difficult at times.

I think the problem is that getOperationAction will get the data from these OpActions, which will all be initialize to 0 (=Legal) and targets do not usually overrides that for illegal types:
https://github.com/llvm/llvm-project/blob/a9d68a5524dea113cace5983697786599cbdce9a/llvm/lib/CodeGen/TargetLoweringBase.cpp#L733

So it can pick up "Legal" stores just because the default initialization and the target has never had to set them to anything else in the past. There are TruncStoreActions that should be set though. What do you think of using something like this, based on whether there is a legal trunk store?

unsigned getStoreMinimumVF(unsigned VF, Type *ScalarTy) const {
  auto &&IsSupportedByTarget = [this, ScalarTy](unsigned VF) {
    auto *SrcTy = FixedVectorType::get(ScalarTy, VF / 2);
    EVT VT = getTLI()->getValueType(DL, SrcTy);
    TargetLowering::LegalizeAction LA =
        getTLI()->getOperationAction(ISD::STORE, VT);
    if (getTLI()->isTypeLegal(VT) &&
        (LA == TargetLowering::Legal || LA == TargetLowering::Custom))
      return true;

    auto LT = getTLI()->getTypeLegalizationCost(DL, SrcTy);
    LA = getTLI()->getTruncStoreAction(LT.second, VT);
    return LA == TargetLowering::Legal || LA == TargetLowering::Custom;
  };
  while (VF > 2 && IsSupportedByTarget(VF))
    VF /= 2;
  return VF;
}

Would that work for the cases you are interested in? The target can override it in any case, so at least it is controllable. But if that works for your use-cases it should hopefully match the target lowering a little better.

[ABataev]

In D124284#3484061, @dmgreen wrote:

Thanks for the update, this looks better to me. The perf results I have were looking OK too, except for one fp16 case where it was choosing to use v2f16 vectorization. It could be OK, it's essentially trading unrolling vs vectorizing and one isn't obviously better or worse than the other. But small vectorization factors can be difficult at times.

I think the problem is that getOperationAction will get the data from these OpActions, which will all be initialize to 0 (=Legal) and targets do not usually overrides that for illegal types:
https://github.com/llvm/llvm-project/blob/a9d68a5524dea113cace5983697786599cbdce9a/llvm/lib/CodeGen/TargetLoweringBase.cpp#L733

So it can pick up "Legal" stores just because the default initialization and the target has never had to set them to anything else in the past. There are TruncStoreActions that should be set though. What do you think of using something like this, based on whether there is a legal trunk store?

unsigned getStoreMinimumVF(unsigned VF, Type *ScalarTy) const {
  auto &&IsSupportedByTarget = [this, ScalarTy](unsigned VF) {
    auto *SrcTy = FixedVectorType::get(ScalarTy, VF / 2);
    EVT VT = getTLI()->getValueType(DL, SrcTy);
    TargetLowering::LegalizeAction LA =
        getTLI()->getOperationAction(ISD::STORE, VT);
    if (getTLI()->isTypeLegal(VT) &&
        (LA == TargetLowering::Legal || LA == TargetLowering::Custom))
      return true;

    auto LT = getTLI()->getTypeLegalizationCost(DL, SrcTy);
    LA = getTLI()->getTruncStoreAction(LT.second, VT);
    return LA == TargetLowering::Legal || LA == TargetLowering::Custom;
  };
  while (VF > 2 && IsSupportedByTarget(VF))
    VF /= 2;
  return VF;
}

Would that work for the cases you are interested in? The target can override it in any case, so at least it is controllable. But if that works for your use-cases it should hopefully match the target lowering a little better.

I tried something similar already, it won't work. Plus, trunc store is not the case we're looking at here, it is different. This function just says to vectorizer that it might be worth trying this vector factor. The cost model should later inform that it is not profitable. If something is not correct in the TTI, it should be fixed in TTI.

[xbolva00]

If something is not correct in the TTI, it should be fixed in TTI.

Strong +1

[xbolva00]

But that issue should not be a blocker IMHO.

[dmgreen]

I tried something similar already, it won't work. Plus, trunc store is not the case we're looking at here, it is different. This function just says to vectorizer that it might be worth trying this vector factor. The cost model should later inform that it is not profitable. If something is not correct in the TTI, it should be fixed in TTI.

Oh OK, that's a shame. There may be something a little off with the f16 costmodel, it is not always perfect, but I don't see anything obvious from what it is printing. There are only v2f16 values, which can't go too wrong.
The issue isn't that SLP vectorization is worse than scalar, it's that in that particular case runtime unrolling is better. The SLP that happens can get in the way of something more profitable, and no pass in llvm operates in a vacuum.

That issue isn't too important though. My worry is that this is currently an expensive way of saying "return 2". I've no strong objection if you want to go with the current method, but perhaps the default should be more "correct" and we can override the targets that want something different/more aggressive? They can choose to spend the extra compile time on factors that might not be expected to be very profitable to other archs.

[RKSimon]

rebase? I'm not sure if rGc5e875f599c25c2ea5a5c3dc6396de17c0c80a45 will have changed due to this patch

[asbirlea]

I'm seeing some fairly big regressions with this patch, specifically on Rome (AMD) architecture.
A couple of examples that are public in the test suite: SingleSource/Benchmarks/Shootout: for sieve I'm seeing a 20% performance regression in an opt build and an xfdo one, and for MicroBenchmarks/ImageProcessing/Dither 10% regression (opt, thinlto and xfdo).
I'm seeing also a couple on Skylake, opt build, in the range of 5-13 %, an example being eigen with 13% regression; this may be harder to track down as it's in a specific configuration but let me know if you want to reproduce this one.

As far as performance improvements, I see a few on Skylake in the range of 3-6%. An example here is MicroBenchmarks/ImageProcessing/Blur, which ranges between 4-5% improvement.

Overall, the regressions outnumber the gains in the testing I've done so far and would likely block our compiler release.

[ABataev]

In D124284#3485220, @dmgreen wrote:

I tried something similar already, it won't work. Plus, trunc store is not the case we're looking at here, it is different. This function just says to vectorizer that it might be worth trying this vector factor. The cost model should later inform that it is not profitable. If something is not correct in the TTI, it should be fixed in TTI.

Oh OK, that's a shame. There may be something a little off with the f16 costmodel, it is not always perfect, but I don't see anything obvious from what it is printing. There are only v2f16 values, which can't go too wrong.
The issue isn't that SLP vectorization is worse than scalar, it's that in that particular case runtime unrolling is better. The SLP that happens can get in the way of something more profitable, and no pass in llvm operates in a vacuum.

That issue isn't too important though. My worry is that this is currently an expensive way of saying "return 2". I've no strong objection if you want to go with the current method, but perhaps the default should be more "correct" and we can override the targets that want something different/more aggressive? They can choose to spend the extra compile time on factors that might not be expected to be very profitable to other archs.

In D124284#3489002, @asbirlea wrote:

I'm seeing some fairly big regressions with this patch, specifically on Rome (AMD) architecture.
A couple of examples that are public in the test suite: SingleSource/Benchmarks/Shootout: for sieve I'm seeing a 20% performance regression in an opt build and an xfdo one, and for MicroBenchmarks/ImageProcessing/Dither 10% regression (opt, thinlto and xfdo).
I'm seeing also a couple on Skylake, opt build, in the range of 5-13 %, an example being eigen with 13% regression; this may be harder to track down as it's in a specific configuration but let me know if you want to reproduce this one.

As far as performance improvements, I see a few on Skylake in the range of 3-6%. An example here is MicroBenchmarks/ImageProcessing/Blur, which ranges between 4-5% improvement.

Overall, the regressions outnumber the gains in the testing I've done so far and would likely block our compiler release.

I'm trying to improve it. But if we have perf regressions, there is something wrong with the cost model.

[ABataev]

Reworked initial implementation to be more conservative. Also, now it is able to handle trunc stores.

[RKSimon]

@asbirlea Do you have an update on the regressions you were seeing vs latest patch?

[asbirlea]

In D124284#3493607, @RKSimon wrote:

@asbirlea Do you have an update on the regressions you were seeing vs latest patch?

Performance testing still ongoing, should be completed by tomorrow.

[dmgreen]

I ran the same set of benchmarks again without issue this time (well, there was an issue, but it turned out that someone had changed the benchmark sources :) ). They might not be the most amazing SLP tests, but no remaining objections from me.

(The reason I suggested the truncate code the way I did was because by default the legalizing rules for smaller than legal power2 types is to promote integers to larger sizes. So under MVE where we only have 128bit vectors, a v4i8 vector will be promoted to a v4i32. We would then need a v4i32->v4i8 truncstore for it to be legal. Which it does have! So it would allow some of the smaller than legal types to be vectorized, essentially treating the v4i8 operations as v4i32's. If we are going for more conservative it might make sense to ignore that though, and float types will always widen as opposed to promote by default.)

[RKSimon]

@asbirlea How are you specifying the SSE/AVX level for your benchmark runs - are you running with -march=native the x86-64-v* levels or something else?

[asbirlea]

In D124284#3496381, @RKSimon wrote:

@asbirlea How are you specifying the SSE/AVX level for your benchmark runs - are you running with -march=native the x86-64-v* levels or something else?

I believe the runs are using -target-cpu k8 and -target-cpu haswell.

The latest performance testing still shows one regression in a benchmark from singlesource, but there are many improvements that offset it in non-public benchmarks. So this latest diff is good to go from my side.

[xbolva00]

LG. Thanks for many perf improvements.

@RKSimon ?

[RKSimon]

Please can you rebase? I added a number of PR tests yesterday and I'm curious how many improve.

[ABataev]

In D124284#3499844, @asbirlea wrote:

In D124284#3496381, @RKSimon wrote:

@asbirlea How are you specifying the SSE/AVX level for your benchmark runs - are you running with -march=native the x86-64-v* levels or something else?

I believe the runs are using -target-cpu k8 and -target-cpu haswell.

The latest performance testing still shows one regression in a benchmark from singlesource, but there are many improvements that offset it in non-public benchmarks. So this latest diff is good to go from my side.

Hi, thanks for the testing. What's the name of the regressed single source test?

[ABataev]

In D124284#3500236, @RKSimon wrote:

Please can you rebase? I added a number of PR tests yesterday and I'm curious how many improve.

Sure, will do later today

[ABataev]

Rebase

[RKSimon]

LGTM - naturally the test-suite regression needs further investigation but I think that can be performed post-commit

llvm/test/Transforms/SLPVectorizer/X86/arith-mul-load.ll

14 ↗

(On Diff #428098)

I'll investigate adding 32-bit vector load/store handling as well (it has the same costs as the codegen for 64-bit anyhow).

[ABataev]

llvm/test/Transforms/SLPVectorizer/X86/arith-mul-load.ll

14 ↗

(On Diff #428098)

TTI does not report that it supports 32 bit stores.

[RKSimon]

llvm/test/Transforms/SLPVectorizer/X86/arith-mul-load.ll

14 ↗

(On Diff #428098)

We never bothered to add it - we mainly use the 64-bit vector load/store to handle f64-i64 handling on 32-bit targets

[asbirlea]

In D124284#3500259, @ABataev wrote:

In D124284#3499844, @asbirlea wrote:

In D124284#3496381, @RKSimon wrote:

@asbirlea How are you specifying the SSE/AVX level for your benchmark runs - are you running with -march=native the x86-64-v* levels or something else?

I believe the runs are using -target-cpu k8 and -target-cpu haswell.

The latest performance testing still shows one regression in a benchmark from singlesource, but there are many improvements that offset it in non-public benchmarks. So this latest diff is good to go from my side.

Hi, thanks for the testing. What's the name of the regressed single source test?

Shootout/sieve for xfdo configuration on Rome looks regressed by 20%, and Shootout/fib2 for opt configuration also on Rome by 10%.
Dither has one -10% (floyd 128) and one +10% (floyd 512) on Rome opt and xfdo, but the same floyd ones plus a few others are +15 to +20% on Skylake opt, thinlto and xfdo and +5% to +10% on Haswell opt, thinlto and xfdo. Net win overall.
Eigen, complex benchmark also has many configs with net wins, e.g. +10% to +30% haswell xfdo and +10 to +25% rome opt.

There are others, but I hope this gives a rough idea.

[vdmitrie]

llvm/include/llvm/CodeGen/BasicTTIImpl.h

330

Hm, there is some discrepancy here.
Let's assume we entered the loop with VF = 8 and target supports 8 but not 4.

VF == 8 , VF >2 && IsSupportedByTarget(8) evaluates to true => VF = VF/2
VF == 4 , VF >2 && IsSupportedByTarget(4) evaluates to false

hence we return 4 which is actually not supported by target. Is that what the intent was?

[ABataev]

llvm/include/llvm/CodeGen/BasicTTIImpl.h

330

IsSupportedByTarget performs checks for VF/2, not for VF, and only if it is supported, sets VF to VF/2.

[vdmitrie]

llvm/include/llvm/CodeGen/BasicTTIImpl.h

330

Ah, I overlooked that. Thanks.

[xbolva00]

llvm/test/Transforms/SLPVectorizer/X86/arith-mul-load.ll

14 ↗

(On Diff #428098)

Do you plan to add them?

[RKSimon]

llvm/test/Transforms/SLPVectorizer/X86/arith-mul-load.ll

14 ↗

(On Diff #428098)

yes - got a few other blockers to deal with first though - that yak has to be shaved.......

[xbolva00]

llvm/test/Transforms/SLPVectorizer/X86/arith-mul-load.ll

14 ↗

(On Diff #428098)

ok, thanks!

[xbolva00]

llvm/test/Transforms/SLPVectorizer/X86/arith-mul-load.ll

14 ↗

(On Diff #428098)

any updates?

void pr(char* __restrict a, char* __restrict r){
    for (int i = 0; i < 4; i++){
        r[i] += a[i];
    }
}

gcc emits nicely paddb.

[RKSimon]

llvm/test/Transforms/SLPVectorizer/X86/arith-mul-load.ll

14 ↗

(On Diff #428098)

https://reviews.llvm.org/D127604 - but I need someone to perf test the patch properly.

llvm/include/llvm/Analysis/TargetTransformInfo.h

   /// \returns True if it is legal to vectorize the given store chain.
   bool isLegalToVectorizeStoreChain(unsigned ChainSizeInBytes, Align Alignment,
                                     unsigned AddrSpace) const;
 
   /// \returns True if it is legal to vectorize the given reduction kind.
   bool isLegalToVectorizeReduction(const RecurrenceDescriptor &RdxDesc,
                                    ElementCount VF) const;
 
+  /// \returns True if it is legal or custom instruction kind for the given
+  /// type.
+  bool isLegalOrCustomInstruction(unsigned Op, Type *Ty) const;
+
   /// \returns True if the given type is supported for scalable vectors
   bool isElementTypeLegalForScalableVector(Type *Ty) const;
 
   /// \returns The new vector factor value if the target doesn't support \p
   /// SizeInBytes loads or has a better vector factor.
   unsigned getLoadVectorFactor(unsigned VF, unsigned LoadSize,
                                unsigned ChainSizeInBytes,
                                VectorType *VecTy) const;
public:
   virtual bool isLegalToVectorizeLoadChain(unsigned ChainSizeInBytes,
                                            Align Alignment,
                                            unsigned AddrSpace) const = 0;
   virtual bool isLegalToVectorizeStoreChain(unsigned ChainSizeInBytes,
                                             Align Alignment,
                                             unsigned AddrSpace) const = 0;
   virtual bool isLegalToVectorizeReduction(const RecurrenceDescriptor &RdxDesc,
                                            ElementCount VF) const = 0;
+  virtual bool isLegalOrCustomInstruction(unsigned Op, Type *Ty) const = 0;
   virtual bool isElementTypeLegalForScalableVector(Type *Ty) const = 0;
   virtual unsigned getLoadVectorFactor(unsigned VF, unsigned LoadSize,
                                        unsigned ChainSizeInBytes,
                                        VectorType *VecTy) const = 0;
   virtual unsigned getStoreVectorFactor(unsigned VF, unsigned StoreSize,
                                         unsigned ChainSizeInBytes,
                                         VectorType *VecTy) const = 0;
   virtual bool preferInLoopReduction(unsigned Opcode, Type *Ty,
bool isLegalToVectorizeStoreChain(unsigned ChainSizeInBytes, Align Alignment,
                                     unsigned AddrSpace) const override {
     return Impl.isLegalToVectorizeStoreChain(ChainSizeInBytes, Alignment,
                                              AddrSpace);
   }
   bool isLegalToVectorizeReduction(const RecurrenceDescriptor &RdxDesc,
                                    ElementCount VF) const override {
     return Impl.isLegalToVectorizeReduction(RdxDesc, VF);
   }
+  bool isLegalOrCustomInstruction(unsigned Op, Type *Ty) const override {
+    return Impl.isLegalOrCustomInstruction(Op, Ty);
+  }
   bool isElementTypeLegalForScalableVector(Type *Ty) const override {
     return Impl.isElementTypeLegalForScalableVector(Ty);
   }
   unsigned getLoadVectorFactor(unsigned VF, unsigned LoadSize,
                                unsigned ChainSizeInBytes,
                                VectorType *VecTy) const override {
     return Impl.getLoadVectorFactor(VF, LoadSize, ChainSizeInBytes, VecTy);
   }

llvm/include/llvm/Analysis/TargetTransformInfoImpl.h

bool isLegalToVectorizeStoreChain(unsigned ChainSizeInBytes, Align Alignment,
     return true;
   }
 
   bool isLegalToVectorizeReduction(const RecurrenceDescriptor &RdxDesc,
                                    ElementCount VF) const {
     return true;
   }
 
+  bool isLegalOrCustomInstruction(unsigned Op, Type *Ty) const { return true; }
   bool isElementTypeLegalForScalableVector(Type *Ty) const { return true; }
 
   unsigned getLoadVectorFactor(unsigned VF, unsigned LoadSize,
                                unsigned ChainSizeInBytes,
                                VectorType *VecTy) const {
     return VF;
   }
 

llvm/include/llvm/CodeGen/BasicTTIImpl.h

bool isLegalAddressingMode(Type *Ty, GlobalValue *BaseGV, int64_t BaseOffset,
     TargetLoweringBase::AddrMode AM;
     AM.BaseGV = BaseGV;
     AM.BaseOffs = BaseOffset;
     AM.HasBaseReg = HasBaseReg;
     AM.Scale = Scale;
     return getTLI()->isLegalAddressingMode(DL, AM, Ty, AddrSpace, I);
   }
 
+  bool isLegalOrCustomInstruction(unsigned Op, Type *Ty) const {
+    int ISD = getTLI()->InstructionOpcodeToISD(Op);
+    std::pair<InstructionCost, MVT> LT =
+        getTLI()->getTypeLegalizationCost(DL, Ty);
+    return getTLI()->isOperationLegalOrCustom(ISD, LT.second);
+  }
+
   bool isIndexedLoadLegal(TTI::MemIndexedMode M, Type *Ty,
                           const DataLayout &DL) const {
     EVT VT = getTLI()->getValueType(DL, Ty);
     return getTLI()->isIndexedLoadLegal(getISDIndexedMode(M), VT);
   }
 
   bool isIndexedStoreLegal(TTI::MemIndexedMode M, Type *Ty,
                            const DataLayout &DL) const {
     EVT VT = getTLI()->getValueType(DL, Ty);
     return getTLI()->isIndexedStoreLegal(getISDIndexedMode(M), VT);
   }
 
   bool isLSRCostLess(TTI::LSRCost C1, TTI::LSRCost C2) {
     return TargetTransformInfoImplBase::isLSRCostLess(C1, C2);
   }
 
   bool isNumRegsMajorCostOfLSR() {

llvm/include/llvm/Transforms/Vectorize/SLPVectorizer.h

bool vectorizeSimpleInstructions(SmallVectorImpl<Instruction *> &Instructions,
                                    BasicBlock *BB, slpvectorizer::BoUpSLP &R,
                                    bool AtTerminator);
 
   /// Scan the basic block and look for patterns that are likely to start
   /// a vectorization chain.
   bool vectorizeChainsInBlock(BasicBlock *BB, slpvectorizer::BoUpSLP &R);
 
   bool vectorizeStoreChain(ArrayRef<Value *> Chain, slpvectorizer::BoUpSLP &R,
-                           unsigned Idx);
+                           unsigned Idx, unsigned MinVF);
 
   bool vectorizeStores(ArrayRef<StoreInst *> Stores, slpvectorizer::BoUpSLP &R);
 
   /// The store instructions in a basic block organized by base pointer.
   StoreListMap Stores;
 
   /// The getelementptr instructions in a basic block organized by base pointer.
   GEPListMap GEPs;
 };
 
 } // end namespace llvm
 
 #endif // LLVM_TRANSFORMS_VECTORIZE_SLPVECTORIZER_H

llvm/lib/Analysis/TargetTransformInfo.cpp

return TTIImpl->isLegalToVectorizeStoreChain(ChainSizeInBytes, Alignment,
                                                AddrSpace);
 }
 
 bool TargetTransformInfo::isLegalToVectorizeReduction(
     const RecurrenceDescriptor &RdxDesc, ElementCount VF) const {
   return TTIImpl->isLegalToVectorizeReduction(RdxDesc, VF);
 }
 
+bool TargetTransformInfo::isLegalOrCustomInstruction(unsigned Op,
+                                                     Type *Ty) const {
+  return TTIImpl->isLegalOrCustomInstruction(Op, Ty);
+}
+
 bool TargetTransformInfo::isElementTypeLegalForScalableVector(Type *Ty) const {
   return TTIImpl->isElementTypeLegalForScalableVector(Ty);
 }
 
 unsigned TargetTransformInfo::getLoadVectorFactor(unsigned VF,
                                                   unsigned LoadSize,
                                                   unsigned ChainSizeInBytes,
                                                   VectorType *VecTy) const {

llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp

#endif
   // Avoid duplicate scheduling of the block.
   BS->ScheduleStart = nullptr;
 }
 
 unsigned BoUpSLP::getVectorElementSize(Value *V) {
   // If V is a store, just return the width of the stored value (or value
   // truncated just before storing) without traversing the expression tree.
   // This is the common case.
-  if (auto *Store = dyn_cast<StoreInst>(V)) {
-    if (auto *Trunc = dyn_cast<TruncInst>(Store->getValueOperand()))
-      return DL->getTypeSizeInBits(Trunc->getSrcTy());
+  if (auto *Store = dyn_cast<StoreInst>(V))
     return DL->getTypeSizeInBits(Store->getValueOperand()->getType());
-  }
 
   if (auto *IEI = dyn_cast<InsertElementInst>(V))
     return getVectorElementSize(IEI->getOperand(1));
 
   auto E = InstrElementSize.find(V);
   if (E != InstrElementSize.end())
     return E->second;
 
bool SLPVectorizerPass::runImpl(Function &F, ScalarEvolution *SE_,
   if (Changed) {
     R.optimizeGatherSequence();
     LLVM_DEBUG(dbgs() << "SLP: vectorized \"" << F.getName() << "\"\n");
   }
   return Changed;
 }
 
 bool SLPVectorizerPass::vectorizeStoreChain(ArrayRef<Value *> Chain, BoUpSLP &R,
-                                            unsigned Idx) {
+                                            unsigned Idx, unsigned MinVF) {
   LLVM_DEBUG(dbgs() << "SLP: Analyzing a store chain of length " << Chain.size()
                     << "\n");
   const unsigned Sz = R.getVectorElementSize(Chain[0]);
-  const unsigned MinVF = R.getMinVecRegSize() / Sz;
   unsigned VF = Chain.size();
 
   if (!isPowerOf2_32(Sz) || !isPowerOf2_32(VF) || VF < 2 || VF < MinVF)
     return false;
 
   LLVM_DEBUG(dbgs() << "SLP: Analyzing " << VF << " stores at offset " << Idx
                     << "\n");
 
while (I != E && !VectorizedStores.count(Stores[I])) {
       I = ConsecutiveChain[I].first;
     }
     assert(!Operands.empty() && "Expected non-empty list of stores.");
 
     unsigned MaxVecRegSize = R.getMaxVecRegSize();
     unsigned EltSize = R.getVectorElementSize(Operands[0]);
     unsigned MaxElts = llvm::PowerOf2Floor(MaxVecRegSize / EltSize);
 
-    unsigned MinVF = R.getMinVF(EltSize);
     unsigned MaxVF = std::min(R.getMaximumVF(EltSize, Instruction::Store),
                               MaxElts);
+    unsigned MinVF = R.getMinVF(EltSize);
+    Type *ValueTy = cast<StoreInst>(Operands[0])->getValueOperand()->getType();
+    while (MinVF > 2 &&
+           TTI->isLegalOrCustomInstruction(
+               Instruction::Store, FixedVectorType::get(ValueTy, MinVF / 2)))
+      MinVF /= 2;
 
     // FIXME: Is division-by-2 the correct step? Should we assert that the
     // register size is a power-of-2?
     unsigned StartIdx = 0;
     for (unsigned Size = MaxVF; Size >= MinVF; Size /= 2) {
       for (unsigned Cnt = StartIdx, E = Operands.size(); Cnt + Size <= E;) {
         ArrayRef<Value *> Slice = makeArrayRef(Operands).slice(Cnt, Size);
         if (!VectorizedStores.count(Slice.front()) &&
             !VectorizedStores.count(Slice.back()) &&
-            vectorizeStoreChain(Slice, R, Cnt)) {
+            vectorizeStoreChain(Slice, R, Cnt, MinVF)) {
           // Mark the vectorized stores so that we don't vectorize them again.
           VectorizedStores.insert(Slice.begin(), Slice.end());
           Changed = true;
           // If we vectorized initial block, no need to try to vectorize it
           // again.
           if (Cnt == StartIdx)
             StartIdx += Size;
           Cnt += Size;

llvm/test/Transforms/GVN/no_speculative_loads_with_asan.ll

; <label>:6                                       ; preds = %0
   br label %11
 
 ; <label>:11                                      ; preds = %0, %6
   %12 = phi i32 [ %10, %6 ], [ 0, %0 ]
   ret i32 %12
 }
 
 ; CHECK-LABEL: @TestAsan
-; CHECK-NOT: %[[LOAD:[^ ]+]] = load i32
-; CHECK: {{.*}} = phi
+; CHECK: ret i32 0
 
 
 define i32 @TestHWAsan() sanitize_hwaddress {
   %1 = tail call noalias i8* @_Znam(i64 2)
   %2 = getelementptr inbounds i8, i8* %1, i64 1
   store i8 0, i8* %2, align 1
   store i8 0, i8* %1, align 1
   %3 = bitcast i8* %1 to i16*
; <label>:6                                       ; preds = %0
   br label %11
 
 ; <label>:11                                      ; preds = %0, %6
   %12 = phi i32 [ %10, %6 ], [ 0, %0 ]
   ret i32 %12
 }
 
 ; CHECK-LABEL: @TestHWAsan
-; CHECK-NOT: %[[LOAD:[^ ]+]] = load i32
-; CHECK: {{.*}} = phi
+; CHECK: ret i32 0
 

llvm/test/Transforms/NewGVN/no_speculative_loads_with_asan.ll

; <label>:6                                       ; preds = %0
   br label %11
 
 ; <label>:11                                      ; preds = %0, %6
   %12 = phi i32 [ %10, %6 ], [ 0, %0 ]
   ret i32 %12
 }
 
 ; CHECK-LABEL: @TestAsan
-; CHECK-NOT: %[[LOAD:[^ ]+]] = load i32
-; CHECK: {{.*}} = phi
+; CHECK: ret i32 0
 
 define i32 @TestHWAsan() sanitize_hwaddress {
   %1 = tail call noalias i8* @_Znam(i64 2)
   %2 = getelementptr inbounds i8, i8* %1, i64 1
   store i8 0, i8* %2, align 1
   store i8 0, i8* %1, align 1
   %3 = bitcast i8* %1 to i16*
   %4 = load i16, i16* %3, align 4
   %5 = icmp eq i16 %4, 0
   br i1 %5, label %11, label %6
 
 ; <label>:6                                       ; preds = %0
   %7 = getelementptr inbounds i8, i8* %1, i64 2
   %8 = bitcast i8* %7 to i16*
   %9 = load i16, i16* %8, align 2
   %10 = sext i16 %9 to i32
   br label %11
 
 ; <label>:11                                      ; preds = %0, %6
   %12 = phi i32 [ %10, %6 ], [ 0, %0 ]
   ret i32 %12
 }
 
 ; CHECK-LABEL: @TestHWAsan
-; CHECK-NOT: %[[LOAD:[^ ]+]] = load i32
-; CHECK: {{.*}} = phi
+; CHECK: ret i32 0
 

llvm/test/Transforms/SLPVectorizer/AArch64/vectorizable-selects-uniform-cmps.ll

entry:
   %s.7 = select i1 %cmp.7, i16 %l.7, i16 %x
   store i16 %s.7, i16* %gep.7, align 2
   ret void
 }
 
 define void @select_uniform_ugt_7xi8(i8* %ptr, i8 %x) {
 ; CHECK-LABEL: @select_uniform_ugt_7xi8(
 ; CHECK-NEXT:  entry:
-; CHECK-NEXT:    [[L_0:%.*]] = load i8, i8* [[PTR:%.*]], align 1
-; CHECK-NEXT:    [[CMP_0:%.*]] = icmp ugt i8 [[L_0]], -1
-; CHECK-NEXT:    [[S_0:%.*]] = select i1 [[CMP_0]], i8 [[L_0]], i8 [[X:%.*]]
-; CHECK-NEXT:    store i8 [[S_0]], i8* [[PTR]], align 2
-; CHECK-NEXT:    [[GEP_1:%.*]] = getelementptr inbounds i8, i8* [[PTR]], i8 1
-; CHECK-NEXT:    [[L_1:%.*]] = load i8, i8* [[GEP_1]], align 1
-; CHECK-NEXT:    [[CMP_1:%.*]] = icmp ugt i8 [[L_1]], -1
-; CHECK-NEXT:    [[S_1:%.*]] = select i1 [[CMP_1]], i8 [[L_1]], i8 [[X]]
-; CHECK-NEXT:    store i8 [[S_1]], i8* [[GEP_1]], align 2
-; CHECK-NEXT:    [[GEP_2:%.*]] = getelementptr inbounds i8, i8* [[PTR]], i8 2
-; CHECK-NEXT:    [[L_2:%.*]] = load i8, i8* [[GEP_2]], align 1
-; CHECK-NEXT:    [[CMP_2:%.*]] = icmp ugt i8 [[L_2]], -1
-; CHECK-NEXT:    [[S_2:%.*]] = select i1 [[CMP_2]], i8 [[L_2]], i8 [[X]]
-; CHECK-NEXT:    store i8 [[S_2]], i8* [[GEP_2]], align 2
-; CHECK-NEXT:    [[GEP_3:%.*]] = getelementptr inbounds i8, i8* [[PTR]], i8 3
-; CHECK-NEXT:    [[L_3:%.*]] = load i8, i8* [[GEP_3]], align 1
-; CHECK-NEXT:    [[CMP_3:%.*]] = icmp ugt i8 [[L_3]], -1
-; CHECK-NEXT:    [[S_3:%.*]] = select i1 [[CMP_3]], i8 [[L_3]], i8 [[X]]
-; CHECK-NEXT:    store i8 [[S_3]], i8* [[GEP_3]], align 2
+; CHECK-NEXT:    [[TMP0:%.*]] = bitcast i8* [[PTR:%.*]] to <4 x i8>*
+; CHECK-NEXT:    [[TMP1:%.*]] = load <4 x i8>, <4 x i8>* [[TMP0]], align 1
+; CHECK-NEXT:    [[TMP2:%.*]] = icmp ugt <4 x i8> [[TMP1]], <i8 -1, i8 -1, i8 -1, i8 -1>
+; CHECK-NEXT:    [[TMP3:%.*]] = insertelement <4 x i8> poison, i8 [[X:%.*]], i32 0
+; CHECK-NEXT:    [[SHUFFLE:%.*]] = shufflevector <4 x i8> [[TMP3]], <4 x i8> poison, <4 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP4:%.*]] = select <4 x i1> [[TMP2]], <4 x i8> [[TMP1]], <4 x i8> [[SHUFFLE]]
+; CHECK-NEXT:    [[TMP5:%.*]] = bitcast i8* [[PTR]] to <4 x i8>*
+; CHECK-NEXT:    store <4 x i8> [[TMP4]], <4 x i8>* [[TMP5]], align 2
 ; CHECK-NEXT:    [[GEP_4:%.*]] = getelementptr inbounds i8, i8* [[PTR]], i8 4
 ; CHECK-NEXT:    [[L_4:%.*]] = load i8, i8* [[GEP_4]], align 1
 ; CHECK-NEXT:    [[CMP_4:%.*]] = icmp ugt i8 [[L_4]], -1
 ; CHECK-NEXT:    [[S_4:%.*]] = select i1 [[CMP_4]], i8 [[L_4]], i8 [[X]]
 ; CHECK-NEXT:    store i8 [[S_4]], i8* [[GEP_4]], align 2
 ; CHECK-NEXT:    [[GEP_5:%.*]] = getelementptr inbounds i8, i8* [[PTR]], i8 5
 ; CHECK-NEXT:    [[L_5:%.*]] = load i8, i8* [[GEP_5]], align 1
 ; CHECK-NEXT:    [[CMP_5:%.*]] = icmp ugt i8 [[L_5]], -1

 ; CHECK-NEXT:    [[GEP_8:%.*]] = getelementptr inbounds i8, i8* [[PTR]], i8 8
 ; CHECK-NEXT:    [[L_8:%.*]] = load i8, i8* [[GEP_8]], align 1
 ; CHECK-NEXT:    [[CMP_8:%.*]] = icmp ugt i8 [[L_8]], -1
 ; CHECK-NEXT:    [[TMP6:%.*]] = extractelement <8 x i8> [[TMP1]], i32 0
 ; CHECK-NEXT:    [[S_8:%.*]] = select i1 [[CMP_8]], i8 [[TMP6]], i8 [[X]]
 ; CHECK-NEXT:    [[TMP7:%.*]] = extractelement <8 x i8> [[TMP4]], i32 0
 ; CHECK-NEXT:    store i8 [[TMP7]], i8* [[GEP_8]], align 2
 ; CHECK-NEXT:    [[GEP_9:%.*]] = getelementptr inbounds i8, i8* [[PTR]], i8 9
-; CHECK-NEXT:    [[L_9:%.*]] = load i8, i8* [[GEP_9]], align 1
-; CHECK-NEXT:    [[CMP_9:%.*]] = icmp ugt i8 [[L_9]], -1
-; CHECK-NEXT:    [[S_9:%.*]] = select i1 [[CMP_9]], i8 [[L_9]], i8 [[X]]
-; CHECK-NEXT:    store i8 [[S_9]], i8* [[GEP_9]], align 2
-; CHECK-NEXT:    [[GEP_10:%.*]] = getelementptr inbounds i8, i8* [[PTR]], i8 10
-; CHECK-NEXT:    [[L_10:%.*]] = load i8, i8* [[GEP_10]], align 1
-; CHECK-NEXT:    [[CMP_10:%.*]] = icmp ugt i8 [[L_10]], -1
-; CHECK-NEXT:    [[S_10:%.*]] = select i1 [[CMP_10]], i8 [[L_10]], i8 [[X]]
-; CHECK-NEXT:    store i8 [[S_10]], i8* [[GEP_10]], align 2
-; CHECK-NEXT:    [[GEP_11:%.*]] = getelementptr inbounds i8, i8* [[PTR]], i8 11
-; CHECK-NEXT:    [[L_11:%.*]] = load i8, i8* [[GEP_11]], align 1
-; CHECK-NEXT:    [[CMP_11:%.*]] = icmp ugt i8 [[L_11]], -1
-; CHECK-NEXT:    [[S_11:%.*]] = select i1 [[CMP_11]], i8 [[L_11]], i8 [[X]]
-; CHECK-NEXT:    store i8 [[S_11]], i8* [[GEP_11]], align 2
-; CHECK-NEXT:    [[GEP_12:%.*]] = getelementptr inbounds i8, i8* [[PTR]], i8 12
-; CHECK-NEXT:    [[L_12:%.*]] = load i8, i8* [[GEP_12]], align 1
-; CHECK-NEXT:    [[CMP_12:%.*]] = icmp ugt i8 [[L_12]], -1
-; CHECK-NEXT:    [[S_12:%.*]] = select i1 [[CMP_12]], i8 [[L_12]], i8 [[X]]
-; CHECK-NEXT:    store i8 [[S_12]], i8* [[GEP_12]], align 2
+; CHECK-NEXT:    [[TMP8:%.*]] = bitcast i8* [[GEP_9]] to <4 x i8>*
+; CHECK-NEXT:    [[TMP9:%.*]] = load <4 x i8>, <4 x i8>* [[TMP8]], align 1
+; CHECK-NEXT:    [[TMP10:%.*]] = icmp ugt <4 x i8> [[TMP9]], <i8 -1, i8 -1, i8 -1, i8 -1>
+; CHECK-NEXT:    [[TMP11:%.*]] = insertelement <4 x i8> poison, i8 [[X]], i32 0
+; CHECK-NEXT:    [[SHUFFLE1:%.*]] = shufflevector <4 x i8> [[TMP11]], <4 x i8> poison, <4 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP12:%.*]] = select <4 x i1> [[TMP10]], <4 x i8> [[TMP9]], <4 x i8> [[SHUFFLE1]]
+; CHECK-NEXT:    [[TMP13:%.*]] = bitcast i8* [[GEP_9]] to <4 x i8>*
+; CHECK-NEXT:    store <4 x i8> [[TMP12]], <4 x i8>* [[TMP13]], align 2
 ; CHECK-NEXT:    [[GEP_13:%.*]] = getelementptr inbounds i8, i8* [[PTR]], i8 13
 ; CHECK-NEXT:    [[L_13:%.*]] = load i8, i8* [[GEP_13]], align 1
 ; CHECK-NEXT:    [[CMP_13:%.*]] = icmp ugt i8 [[L_13]], -1
 ; CHECK-NEXT:    [[S_13:%.*]] = select i1 [[CMP_13]], i8 [[L_13]], i8 [[X]]
 ; CHECK-NEXT:    store i8 [[S_13]], i8* [[GEP_13]], align 2
 ; CHECK-NEXT:    [[GEP_14:%.*]] = getelementptr inbounds i8, i8* [[PTR]], i8 14
 ; CHECK-NEXT:    [[L_14:%.*]] = load i8, i8* [[GEP_14]], align 1
 ; CHECK-NEXT:    [[CMP_14:%.*]] = icmp ugt i8 [[L_14]], -1

llvm/test/Transforms/SLPVectorizer/X86/crash_7zip.ll

 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
 ; RUN: opt < %s -basic-aa -slp-vectorizer -dce -S -mtriple=x86_64-apple-macosx10.8.0 -mcpu=corei7 | FileCheck %s
 
 target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
 target triple = "x86_64-apple-macosx10.8.0"
 
 %struct.CLzmaDec.1.28.55.82.103.124.145.166.181.196.229.259.334 = type { %struct._CLzmaProps.0.27.54.81.102.123.144.165.180.195.228.258.333, i16*, i8*, i8*, i32, i32, i64, i64, i32, i32, i32, [4 x i32], i32, i32, i32, i32, i32, [20 x i8] }
 %struct._CLzmaProps.0.27.54.81.102.123.144.165.180.195.228.258.333 = type { i32, i32, i32, i32 }
 
 define fastcc void @LzmaDec_DecodeReal2(%struct.CLzmaDec.1.28.55.82.103.124.145.166.181.196.229.259.334* %p) {
 ; CHECK-LABEL: @LzmaDec_DecodeReal2(
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[RANGE20_I:%.*]] = getelementptr inbounds [[STRUCT_CLZMADEC_1_28_55_82_103_124_145_166_181_196_229_259_334:%.*]], %struct.CLzmaDec.1.28.55.82.103.124.145.166.181.196.229.259.334* [[P:%.*]], i64 0, i32 4
-; CHECK-NEXT:    [[CODE21_I:%.*]] = getelementptr inbounds [[STRUCT_CLZMADEC_1_28_55_82_103_124_145_166_181_196_229_259_334]], %struct.CLzmaDec.1.28.55.82.103.124.145.166.181.196.229.259.334* [[P]], i64 0, i32 5
 ; CHECK-NEXT:    br label [[DO_BODY66_I:%.*]]
 ; CHECK:       do.body66.i:
-; CHECK-NEXT:    [[RANGE_2_I:%.*]] = phi i32 [ [[RANGE_4_I:%.*]], [[DO_COND_I:%.*]] ], [ undef, [[ENTRY:%.*]] ]
-; CHECK-NEXT:    [[CODE_2_I:%.*]] = phi i32 [ [[CODE_4_I:%.*]], [[DO_COND_I]] ], [ undef, [[ENTRY]] ]
-; CHECK-NEXT:    [[DOTRANGE_2_I:%.*]] = select i1 undef, i32 undef, i32 [[RANGE_2_I]]
-; CHECK-NEXT:    [[DOTCODE_2_I:%.*]] = select i1 undef, i32 undef, i32 [[CODE_2_I]]
+; CHECK-NEXT:    [[TMP0:%.*]] = phi <2 x i32> [ [[TMP5:%.*]], [[DO_COND_I:%.*]] ], [ undef, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[TMP1:%.*]] = select <2 x i1> undef, <2 x i32> undef, <2 x i32> [[TMP0]]
+; CHECK-NEXT:    [[TMP2:%.*]] = extractelement <2 x i32> [[TMP1]], i32 1
+; CHECK-NEXT:    [[TMP3:%.*]] = insertelement <2 x i32> <i32 undef, i32 poison>, i32 [[TMP2]], i32 1
 ; CHECK-NEXT:    br i1 undef, label [[DO_COND_I]], label [[IF_ELSE_I:%.*]]
 ; CHECK:       if.else.i:
-; CHECK-NEXT:    [[SUB91_I:%.*]] = sub i32 [[DOTRANGE_2_I]], undef
-; CHECK-NEXT:    [[SUB92_I:%.*]] = sub i32 [[DOTCODE_2_I]], undef
+; CHECK-NEXT:    [[TMP4:%.*]] = sub <2 x i32> [[TMP1]], undef
 ; CHECK-NEXT:    br label [[DO_COND_I]]
 ; CHECK:       do.cond.i:
-; CHECK-NEXT:    [[RANGE_4_I]] = phi i32 [ [[SUB91_I]], [[IF_ELSE_I]] ], [ undef, [[DO_BODY66_I]] ]
-; CHECK-NEXT:    [[CODE_4_I]] = phi i32 [ [[SUB92_I]], [[IF_ELSE_I]] ], [ [[DOTCODE_2_I]], [[DO_BODY66_I]] ]
+; CHECK-NEXT:    [[TMP5]] = phi <2 x i32> [ [[TMP4]], [[IF_ELSE_I]] ], [ [[TMP3]], [[DO_BODY66_I]] ]
 ; CHECK-NEXT:    br i1 undef, label [[DO_BODY66_I]], label [[DO_END1006_I:%.*]]
 ; CHECK:       do.end1006.i:
-; CHECK-NEXT:    [[DOTRANGE_4_I:%.*]] = select i1 undef, i32 undef, i32 [[RANGE_4_I]]
-; CHECK-NEXT:    [[DOTCODE_4_I:%.*]] = select i1 undef, i32 undef, i32 [[CODE_4_I]]
-; CHECK-NEXT:    store i32 [[DOTRANGE_4_I]], i32* [[RANGE20_I]], align 4
-; CHECK-NEXT:    store i32 [[DOTCODE_4_I]], i32* [[CODE21_I]], align 4
+; CHECK-NEXT:    [[TMP6:%.*]] = select <2 x i1> undef, <2 x i32> undef, <2 x i32> [[TMP5]]
+; CHECK-NEXT:    [[TMP7:%.*]] = bitcast i32* [[RANGE20_I]] to <2 x i32>*
+; CHECK-NEXT:    store <2 x i32> [[TMP6]], <2 x i32>* [[TMP7]], align 4
 ; CHECK-NEXT:    ret void
 ;
 entry:
   %range20.i = getelementptr inbounds %struct.CLzmaDec.1.28.55.82.103.124.145.166.181.196.229.259.334, %struct.CLzmaDec.1.28.55.82.103.124.145.166.181.196.229.259.334* %p, i64 0, i32 4
   %code21.i = getelementptr inbounds %struct.CLzmaDec.1.28.55.82.103.124.145.166.181.196.229.259.334, %struct.CLzmaDec.1.28.55.82.103.124.145.166.181.196.229.259.334* %p, i64 0, i32 5
   br label %do.body66.i
 
 do.body66.i:                                      ; preds = %do.cond.i, %entry

llvm/test/Transforms/SLPVectorizer/X86/crash_bullet.ll

 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
 ; RUN: opt < %s -basic-aa -slp-vectorizer -dce -S -mtriple=x86_64-apple-macosx10.8.0 -mcpu=corei7 | FileCheck %s
 
 target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
 target triple = "x86_64-apple-macosx10.8.0"
 
 %"struct.btTypedConstraint::btConstraintInfo1.17.157.357.417.477.960" = type { i32, i32 }
 
 define void @_ZN23btGeneric6DofConstraint8getInfo1EPN17btTypedConstraint17btConstraintInfo1E(%"struct.btTypedConstraint::btConstraintInfo1.17.157.357.417.477.960"* nocapture %info) {
 ; CHECK-LABEL: @_ZN23btGeneric6DofConstraint8getInfo1EPN17btTypedConstraint17btConstraintInfo1E(
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    br i1 undef, label [[IF_ELSE:%.*]], label [[IF_THEN:%.*]]
 ; CHECK:       if.then:
 ; CHECK-NEXT:    ret void
 ; CHECK:       if.else:
 ; CHECK-NEXT:    [[M_NUMCONSTRAINTROWS4:%.*]] = getelementptr inbounds %"struct.btTypedConstraint::btConstraintInfo1.17.157.357.417.477.960", %"struct.btTypedConstraint::btConstraintInfo1.17.157.357.417.477.960"* [[INFO:%.*]], i64 0, i32 0
-; CHECK-NEXT:    [[NUB5:%.*]] = getelementptr inbounds %"struct.btTypedConstraint::btConstraintInfo1.17.157.357.417.477.960", %"struct.btTypedConstraint::btConstraintInfo1.17.157.357.417.477.960"* [[INFO]], i64 0, i32 1
 ; CHECK-NEXT:    br i1 undef, label [[LAND_LHS_TRUE_I_1:%.*]], label [[IF_THEN7_1:%.*]]
 ; CHECK:       land.lhs.true.i.1:
 ; CHECK-NEXT:    br i1 undef, label [[FOR_INC_1:%.*]], label [[IF_THEN7_1]]
 ; CHECK:       if.then7.1:
-; CHECK-NEXT:    [[INC_1:%.*]] = add nsw i32 0, 1
-; CHECK-NEXT:    store i32 [[INC_1]], i32* [[M_NUMCONSTRAINTROWS4]], align 4
-; CHECK-NEXT:    [[DEC_1:%.*]] = add nsw i32 6, -1
-; CHECK-NEXT:    store i32 [[DEC_1]], i32* [[NUB5]], align 4
+; CHECK-NEXT:    [[TMP0:%.*]] = bitcast i32* [[M_NUMCONSTRAINTROWS4]] to <2 x i32>*
+; CHECK-NEXT:    store <2 x i32> <i32 1, i32 5>, <2 x i32>* [[TMP0]], align 4
 ; CHECK-NEXT:    br label [[FOR_INC_1]]
 ; CHECK:       for.inc.1:
-; CHECK-NEXT:    [[TMP0:%.*]] = phi i32 [ [[DEC_1]], [[IF_THEN7_1]] ], [ 6, [[LAND_LHS_TRUE_I_1]] ]
-; CHECK-NEXT:    [[TMP1:%.*]] = phi i32 [ [[INC_1]], [[IF_THEN7_1]] ], [ 0, [[LAND_LHS_TRUE_I_1]] ]
-; CHECK-NEXT:    [[INC_2:%.*]] = add nsw i32 [[TMP1]], 1
-; CHECK-NEXT:    store i32 [[INC_2]], i32* [[M_NUMCONSTRAINTROWS4]], align 4
-; CHECK-NEXT:    [[DEC_2:%.*]] = add nsw i32 [[TMP0]], -1
-; CHECK-NEXT:    store i32 [[DEC_2]], i32* [[NUB5]], align 4
+; CHECK-NEXT:    [[TMP1:%.*]] = phi <2 x i32> [ <i32 1, i32 5>, [[IF_THEN7_1]] ], [ <i32 0, i32 6>, [[LAND_LHS_TRUE_I_1]] ]
+; CHECK-NEXT:    [[TMP2:%.*]] = add nsw <2 x i32> [[TMP1]], <i32 1, i32 -1>
+; CHECK-NEXT:    [[TMP3:%.*]] = bitcast i32* [[M_NUMCONSTRAINTROWS4]] to <2 x i32>*
+; CHECK-NEXT:    store <2 x i32> [[TMP2]], <2 x i32>* [[TMP3]], align 4
 ; CHECK-NEXT:    unreachable
 ;
 entry:
   br i1 undef, label %if.else, label %if.then
 
 if.then:                                          ; preds = %entry
   ret void
 

 %class.btVector4.7.32.67.92.117.142.177.187.262.282.331 = type { %class.btVector3.5.30.65.90.115.140.175.185.260.280.330 }
 
 define void @_ZN30GIM_TRIANGLE_CALCULATION_CACHE18triangle_collisionERK9btVector3S2_S2_fS2_S2_S2_fR25GIM_TRIANGLE_CONTACT_DATA(%class.GIM_TRIANGLE_CALCULATION_CACHE.9.34.69.94.119.144.179.189.264.284.332* %this) {
 ; CHECK-LABEL: @_ZN30GIM_TRIANGLE_CALCULATION_CACHE18triangle_collisionERK9btVector3S2_S2_fS2_S2_S2_fR25GIM_TRIANGLE_CONTACT_DATA(
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[ARRAYIDX26:%.*]] = getelementptr inbounds [[CLASS_GIM_TRIANGLE_CALCULATION_CACHE_9_34_69_94_119_144_179_189_264_284_332:%.*]], %class.GIM_TRIANGLE_CALCULATION_CACHE.9.34.69.94.119.144.179.189.264.284.332* [[THIS:%.*]], i64 0, i32 2, i64 0, i32 0, i64 1
 ; CHECK-NEXT:    [[ARRAYIDX36:%.*]] = getelementptr inbounds [[CLASS_GIM_TRIANGLE_CALCULATION_CACHE_9_34_69_94_119_144_179_189_264_284_332]], %class.GIM_TRIANGLE_CALCULATION_CACHE.9.34.69.94.119.144.179.189.264.284.332* [[THIS]], i64 0, i32 2, i64 0, i32 0, i64 2
 ; CHECK-NEXT:    [[TMP0:%.*]] = load float, float* [[ARRAYIDX36]], align 4
-; CHECK-NEXT:    [[ADD587:%.*]] = fadd float undef, undef
-; CHECK-NEXT:    [[SUB600:%.*]] = fsub float [[ADD587]], undef
-; CHECK-NEXT:    store float [[SUB600]], float* undef, align 4
-; CHECK-NEXT:    [[SUB613:%.*]] = fsub float [[ADD587]], [[SUB600]]
-; CHECK-NEXT:    store float [[SUB613]], float* [[ARRAYIDX26]], align 4
-; CHECK-NEXT:    [[ADD626:%.*]] = fadd float [[TMP0]], undef
-; CHECK-NEXT:    [[SUB639:%.*]] = fsub float [[ADD626]], undef
-; CHECK-NEXT:    [[SUB652:%.*]] = fsub float [[ADD626]], [[SUB639]]
-; CHECK-NEXT:    store float [[SUB652]], float* [[ARRAYIDX36]], align 4
+; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <2 x float> <float undef, float poison>, float [[TMP0]], i32 1
+; CHECK-NEXT:    [[TMP2:%.*]] = fadd <2 x float> [[TMP1]], undef
+; CHECK-NEXT:    [[TMP3:%.*]] = fsub <2 x float> [[TMP2]], undef
+; CHECK-NEXT:    [[TMP4:%.*]] = extractelement <2 x float> [[TMP3]], i32 0
+; CHECK-NEXT:    store float [[TMP4]], float* undef, align 4
+; CHECK-NEXT:    [[TMP5:%.*]] = fsub <2 x float> [[TMP2]], [[TMP3]]
+; CHECK-NEXT:    [[TMP6:%.*]] = bitcast float* [[ARRAYIDX26]] to <2 x float>*
+; CHECK-NEXT:    store <2 x float> [[TMP5]], <2 x float>* [[TMP6]], align 4
 ; CHECK-NEXT:    br i1 undef, label [[IF_ELSE1609:%.*]], label [[IF_THEN1595:%.*]]
 ; CHECK:       if.then1595:
 ; CHECK-NEXT:    br i1 undef, label [[RETURN:%.*]], label [[FOR_BODY_LR_PH_I_I1702:%.*]]
 ; CHECK:       for.body.lr.ph.i.i1702:
 ; CHECK-NEXT:    unreachable
 ; CHECK:       if.else1609:
 ; CHECK-NEXT:    unreachable
 ; CHECK:       return:

llvm/test/Transforms/SLPVectorizer/X86/crash_bullet3.ll

 ; CHECK-NEXT:    br label [[FOR_BODY144:%.*]]
 ; CHECK:       for.body144:
 ; CHECK-NEXT:    br i1 undef, label [[FOR_END227:%.*]], label [[FOR_BODY144]]
 ; CHECK:       for.end227:
 ; CHECK-NEXT:    br i1 undef, label [[FOR_END271:%.*]], label [[FOR_BODY233:%.*]]
 ; CHECK:       for.body233:
 ; CHECK-NEXT:    br i1 undef, label [[FOR_BODY233]], label [[FOR_END271]]
 ; CHECK:       for.end271:
-; CHECK-NEXT:    [[TMP0:%.*]] = phi float [ 0x47EFFFFFE0000000, [[FOR_END227]] ], [ undef, [[FOR_BODY233]] ]
-; CHECK-NEXT:    [[TMP1:%.*]] = phi float [ 0x47EFFFFFE0000000, [[FOR_END227]] ], [ undef, [[FOR_BODY233]] ]
-; CHECK-NEXT:    [[SUB275:%.*]] = fsub float undef, [[TMP1]]
-; CHECK-NEXT:    [[SUB279:%.*]] = fsub float undef, [[TMP0]]
+; CHECK-NEXT:    [[TMP0:%.*]] = phi <2 x float> [ <float 0x47EFFFFFE0000000, float 0x47EFFFFFE0000000>, [[FOR_END227]] ], [ undef, [[FOR_BODY233]] ]
+; CHECK-NEXT:    [[TMP1:%.*]] = fsub <2 x float> undef, [[TMP0]]
 ; CHECK-NEXT:    br i1 undef, label [[IF_THEN291:%.*]], label [[RETURN]]
 ; CHECK:       if.then291:
-; CHECK-NEXT:    [[MUL292:%.*]] = fmul float [[SUB275]], 5.000000e-01
-; CHECK-NEXT:    [[ADD294:%.*]] = fadd float [[TMP1]], [[MUL292]]
-; CHECK-NEXT:    [[MUL295:%.*]] = fmul float [[SUB279]], 5.000000e-01
-; CHECK-NEXT:    [[ADD297:%.*]] = fadd float [[TMP0]], [[MUL295]]
+; CHECK-NEXT:    [[TMP2:%.*]] = fmul <2 x float> [[TMP1]], <float 5.000000e-01, float 5.000000e-01>
+; CHECK-NEXT:    [[TMP3:%.*]] = fadd <2 x float> [[TMP0]], [[TMP2]]
 ; CHECK-NEXT:    br i1 undef, label [[IF_END332:%.*]], label [[IF_ELSE319:%.*]]
 ; CHECK:       if.else319:
 ; CHECK-NEXT:    br i1 undef, label [[IF_THEN325:%.*]], label [[IF_END327:%.*]]
 ; CHECK:       if.then325:
 ; CHECK-NEXT:    br label [[IF_END327]]
 ; CHECK:       if.end327:
+; CHECK-NEXT:    [[TMP4:%.*]] = extractelement <2 x float> [[TMP1]], i32 0
+; CHECK-NEXT:    [[TMP5:%.*]] = insertelement <2 x float> <float poison, float undef>, float [[TMP4]], i32 0
 ; CHECK-NEXT:    br i1 undef, label [[IF_THEN329:%.*]], label [[IF_END332]]
 ; CHECK:       if.then329:
 ; CHECK-NEXT:    br label [[IF_END332]]
 ; CHECK:       if.end332:
-; CHECK-NEXT:    [[DX272_1:%.*]] = phi float [ [[SUB275]], [[IF_THEN329]] ], [ [[SUB275]], [[IF_END327]] ], [ 0x3F847AE140000000, [[IF_THEN291]] ]
-; CHECK-NEXT:    [[DY276_1:%.*]] = phi float [ undef, [[IF_THEN329]] ], [ undef, [[IF_END327]] ], [ 0x3F847AE140000000, [[IF_THEN291]] ]
-; CHECK-NEXT:    [[SUB334:%.*]] = fsub float [[ADD294]], [[DX272_1]]
-; CHECK-NEXT:    [[SUB338:%.*]] = fsub float [[ADD297]], [[DY276_1]]
+; CHECK-NEXT:    [[TMP6:%.*]] = phi <2 x float> [ [[TMP5]], [[IF_THEN329]] ], [ [[TMP5]], [[IF_END327]] ], [ <float 0x3F847AE140000000, float 0x3F847AE140000000>, [[IF_THEN291]] ]
+; CHECK-NEXT:    [[TMP7:%.*]] = fsub <2 x float> [[TMP3]], [[TMP6]]
 ; CHECK-NEXT:    [[ARRAYIDX_I_I606:%.*]] = getelementptr inbounds [[CLASS_BTVECTOR3_23_221_463_485_507_573_595_683_727_749_815_837_991_1585_1607_1629_1651_1849_2047_2069_2091_2113:%.*]], %class.btVector3.23.221.463.485.507.573.595.683.727.749.815.837.991.1585.1607.1629.1651.1849.2047.2069.2091.2113* [[VERTICES:%.*]], i64 0, i32 0, i64 0
-; CHECK-NEXT:    store float [[SUB334]], float* [[ARRAYIDX_I_I606]], align 4
-; CHECK-NEXT:    [[ARRAYIDX3_I607:%.*]] = getelementptr inbounds [[CLASS_BTVECTOR3_23_221_463_485_507_573_595_683_727_749_815_837_991_1585_1607_1629_1651_1849_2047_2069_2091_2113]], %class.btVector3.23.221.463.485.507.573.595.683.727.749.815.837.991.1585.1607.1629.1651.1849.2047.2069.2091.2113* [[VERTICES]], i64 0, i32 0, i64 1
-; CHECK-NEXT:    store float [[SUB338]], float* [[ARRAYIDX3_I607]], align 4
+; CHECK-NEXT:    [[TMP8:%.*]] = bitcast float* [[ARRAYIDX_I_I606]] to <2 x float>*
+; CHECK-NEXT:    store <2 x float> [[TMP7]], <2 x float>* [[TMP8]], align 4
 ; CHECK-NEXT:    br label [[RETURN]]
 ; CHECK:       return:
 ; CHECK-NEXT:    ret void
 ; CHECK:       if.then17.1:
 ; CHECK-NEXT:    br label [[IF_END22_1]]
 ; CHECK:       if.end22.1:
 ; CHECK-NEXT:    br i1 undef, label [[IF_THEN17_2:%.*]], label [[IF_END22_2:%.*]]
 ; CHECK:       if.then17.2:

llvm/test/Transforms/SLPVectorizer/X86/crash_sim4b1.ll

 ; CHECK-NEXT:    br i1 undef, label [[LAND_LHS_TRUE:%.*]], label [[LAND_LHS_TRUE167:%.*]]
 ; CHECK:       land.lhs.true:
 ; CHECK-NEXT:    br i1 undef, label [[IF_THEN17:%.*]], label [[LAND_LHS_TRUE167]]
 ; CHECK:       if.then17:
 ; CHECK-NEXT:    br i1 undef, label [[IF_END98:%.*]], label [[LAND_RHS_LR_PH:%.*]]
 ; CHECK:       land.rhs.lr.ph:
 ; CHECK-NEXT:    unreachable
 ; CHECK:       if.end98:
-; CHECK-NEXT:    [[FROM299:%.*]] = getelementptr inbounds [[STRUCT__EXON_T_12_103_220_363_480_649_740_857_1039_1065_1078_1091_1117_1130_1156_1169_1195_1221_1234_1286_1299_1312_1338_1429_1455_1468_1494_1520_1884_1897_1975_2066_2105_2170_2171:%.*]], %struct._exon_t.12.103.220.363.480.649.740.857.1039.1065.1078.1091.1117.1130.1156.1169.1195.1221.1234.1286.1299.1312.1338.1429.1455.1468.1494.1520.1884.1897.1975.2066.2105.2170.2171* undef, i64 0, i32 1
 ; CHECK-NEXT:    br i1 undef, label [[LAND_LHS_TRUE167]], label [[IF_THEN103:%.*]]
 ; CHECK:       if.then103:
+; CHECK-NEXT:    [[FROM1115:%.*]] = getelementptr inbounds [[STRUCT__EXON_T_12_103_220_363_480_649_740_857_1039_1065_1078_1091_1117_1130_1156_1169_1195_1221_1234_1286_1299_1312_1338_1429_1455_1468_1494_1520_1884_1897_1975_2066_2105_2170_2171:%.*]], %struct._exon_t.12.103.220.363.480.649.740.857.1039.1065.1078.1091.1117.1130.1156.1169.1195.1221.1234.1286.1299.1312.1338.1429.1455.1468.1494.1520.1884.1897.1975.2066.2105.2170.2171* undef, i64 0, i32 0
 ; CHECK-NEXT:    [[DOTSUB100:%.*]] = select i1 undef, i32 250, i32 undef
 ; CHECK-NEXT:    [[MUL114:%.*]] = shl nsw i32 [[DOTSUB100]], 2
-; CHECK-NEXT:    [[FROM1115:%.*]] = getelementptr inbounds [[STRUCT__EXON_T_12_103_220_363_480_649_740_857_1039_1065_1078_1091_1117_1130_1156_1169_1195_1221_1234_1286_1299_1312_1338_1429_1455_1468_1494_1520_1884_1897_1975_2066_2105_2170_2171]], %struct._exon_t.12.103.220.363.480.649.740.857.1039.1065.1078.1091.1117.1130.1156.1169.1195.1221.1234.1286.1299.1312.1338.1429.1455.1468.1494.1520.1884.1897.1975.2066.2105.2170.2171* undef, i64 0, i32 0
 ; CHECK-NEXT:    [[COND125:%.*]] = select i1 undef, i32 undef, i32 [[MUL114]]
+; CHECK-NEXT:    [[TMP0:%.*]] = insertelement <2 x i32> poison, i32 [[COND125]], i32 0
+; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <2 x i32> [[TMP0]], i32 [[DOTSUB100]], i32 1
 ; CHECK-NEXT:    br label [[FOR_COND_I:%.*]]
 ; CHECK:       for.cond.i:
-; CHECK-NEXT:    [[ROW_0_I:%.*]] = phi i32 [ undef, [[LAND_RHS_I874:%.*]] ], [ [[DOTSUB100]], [[IF_THEN103]] ]
-; CHECK-NEXT:    [[COL_0_I:%.*]] = phi i32 [ undef, [[LAND_RHS_I874]] ], [ [[COND125]], [[IF_THEN103]] ]
+; CHECK-NEXT:    [[TMP2:%.*]] = phi <2 x i32> [ undef, [[LAND_RHS_I874:%.*]] ], [ [[TMP1]], [[IF_THEN103]] ]
 ; CHECK-NEXT:    br i1 undef, label [[LAND_RHS_I874]], label [[FOR_END_I:%.*]]
 ; CHECK:       land.rhs.i874:
 ; CHECK-NEXT:    br i1 undef, label [[FOR_COND_I]], label [[FOR_END_I]]
 ; CHECK:       for.end.i:
 ; CHECK-NEXT:    br i1 undef, label [[IF_THEN_I:%.*]], label [[IF_END_I:%.*]]
 ; CHECK:       if.then.i:
-; CHECK-NEXT:    [[ADD14_I:%.*]] = add nsw i32 [[ROW_0_I]], undef
-; CHECK-NEXT:    [[ADD15_I:%.*]] = add nsw i32 [[COL_0_I]], undef
+; CHECK-NEXT:    [[TMP3:%.*]] = add nsw <2 x i32> [[TMP2]], undef
 ; CHECK-NEXT:    br label [[EXTEND_BW_EXIT:%.*]]
 ; CHECK:       if.end.i:
 ; CHECK-NEXT:    [[ADD16_I:%.*]] = add i32 [[COND125]], [[DOTSUB100]]
 ; CHECK-NEXT:    [[CMP26514_I:%.*]] = icmp slt i32 [[ADD16_I]], 0
 ; CHECK-NEXT:    br i1 [[CMP26514_I]], label [[FOR_END33_I:%.*]], label [[FOR_BODY28_LR_PH_I:%.*]]
 ; CHECK:       for.body28.lr.ph.i:
 ; CHECK-NEXT:    br label [[FOR_END33_I]]
 ; CHECK:       for.end33.i:
 ; CHECK-NEXT:    br i1 undef, label [[FOR_END58_I:%.*]], label [[FOR_BODY52_LR_PH_I:%.*]]
 ; CHECK:       for.body52.lr.ph.i:
 ; CHECK-NEXT:    br label [[FOR_END58_I]]
 ; CHECK:       for.end58.i:
 ; CHECK-NEXT:    br label [[WHILE_COND260_I:%.*]]
 ; CHECK:       while.cond260.i:
 ; CHECK-NEXT:    br i1 undef, label [[LAND_RHS263_I:%.*]], label [[WHILE_END275_I:%.*]]
 ; CHECK:       land.rhs263.i:
 ; CHECK-NEXT:    br i1 undef, label [[WHILE_COND260_I]], label [[WHILE_END275_I]]
 ; CHECK:       while.end275.i:
 ; CHECK-NEXT:    br label [[EXTEND_BW_EXIT]]
 ; CHECK:       extend_bw.exit:
-; CHECK-NEXT:    [[ADD14_I1262:%.*]] = phi i32 [ [[ADD14_I]], [[IF_THEN_I]] ], [ undef, [[WHILE_END275_I]] ]
-; CHECK-NEXT:    [[ADD15_I1261:%.*]] = phi i32 [ [[ADD15_I]], [[IF_THEN_I]] ], [ undef, [[WHILE_END275_I]] ]
+; CHECK-NEXT:    [[TMP4:%.*]] = phi <2 x i32> [ [[TMP3]], [[IF_THEN_I]] ], [ undef, [[WHILE_END275_I]] ]
 ; CHECK-NEXT:    br i1 false, label [[IF_THEN157:%.*]], label [[LAND_LHS_TRUE167]]
 ; CHECK:       if.then157:
-; CHECK-NEXT:    [[ADD158:%.*]] = add nsw i32 [[ADD14_I1262]], 1
-; CHECK-NEXT:    store i32 [[ADD158]], i32* [[FROM299]], align 4
-; CHECK-NEXT:    [[ADD160:%.*]] = add nsw i32 [[ADD15_I1261]], 1
-; CHECK-NEXT:    store i32 [[ADD160]], i32* [[FROM1115]], align 4
+; CHECK-NEXT:    [[TMP5:%.*]] = add nsw <2 x i32> [[TMP4]], <i32 1, i32 1>
+; CHECK-NEXT:    [[TMP6:%.*]] = bitcast i32* [[FROM1115]] to <2 x i32>*
+; CHECK-NEXT:    store <2 x i32> [[TMP5]], <2 x i32>* [[TMP6]], align 4
 ; CHECK-NEXT:    br label [[LAND_LHS_TRUE167]]
 ; CHECK:       land.lhs.true167:
 ; CHECK-NEXT:    unreachable
 ; CHECK:       for.inc603:
 ; CHECK-NEXT:    br i1 undef, label [[FOR_BODY]], label [[FOR_END605]]
 ; CHECK:       for.end605:
 ; CHECK-NEXT:    unreachable
 ; CHECK:       return:

llvm/test/Transforms/SLPVectorizer/X86/fptosi-inseltpoison.ll

;
   store i16 %cvt5, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @dst16, i32 0, i64 5), align 2
   store i16 %cvt6, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @dst16, i32 0, i64 6), align 2
   store i16 %cvt7, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @dst16, i32 0, i64 7), align 2
   ret void
 }
 
 define void @fptosi_8f64_8i8() #0 {
 ; CHECK-LABEL: @fptosi_8f64_8i8(
-; CHECK-NEXT:    [[A0:%.*]] = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 0), align 8
-; CHECK-NEXT:    [[A1:%.*]] = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 1), align 8
-; CHECK-NEXT:    [[A2:%.*]] = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 2), align 8
-; CHECK-NEXT:    [[A3:%.*]] = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 3), align 8
-; CHECK-NEXT:    [[A4:%.*]] = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 4), align 8
-; CHECK-NEXT:    [[A5:%.*]] = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 5), align 8
-; CHECK-NEXT:    [[A6:%.*]] = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 6), align 8
-; CHECK-NEXT:    [[A7:%.*]] = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 7), align 8
-; CHECK-NEXT:    [[CVT0:%.*]] = fptosi double [[A0]] to i8
-; CHECK-NEXT:    [[CVT1:%.*]] = fptosi double [[A1]] to i8
-; CHECK-NEXT:    [[CVT2:%.*]] = fptosi double [[A2]] to i8
-; CHECK-NEXT:    [[CVT3:%.*]] = fptosi double [[A3]] to i8
-; CHECK-NEXT:    [[CVT4:%.*]] = fptosi double [[A4]] to i8
-; CHECK-NEXT:    [[CVT5:%.*]] = fptosi double [[A5]] to i8
-; CHECK-NEXT:    [[CVT6:%.*]] = fptosi double [[A6]] to i8
-; CHECK-NEXT:    [[CVT7:%.*]] = fptosi double [[A7]] to i8
-; CHECK-NEXT:    store i8 [[CVT0]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @dst8, i32 0, i64 0), align 1
-; CHECK-NEXT:    store i8 [[CVT1]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @dst8, i32 0, i64 1), align 1
-; CHECK-NEXT:    store i8 [[CVT2]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @dst8, i32 0, i64 2), align 1
-; CHECK-NEXT:    store i8 [[CVT3]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @dst8, i32 0, i64 3), align 1
-; CHECK-NEXT:    store i8 [[CVT4]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @dst8, i32 0, i64 4), align 1
-; CHECK-NEXT:    store i8 [[CVT5]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @dst8, i32 0, i64 5), align 1
-; CHECK-NEXT:    store i8 [[CVT6]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @dst8, i32 0, i64 6), align 1
-; CHECK-NEXT:    store i8 [[CVT7]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @dst8, i32 0, i64 7), align 1
+; CHECK-NEXT:    [[TMP1:%.*]] = load <8 x double>, <8 x double>* bitcast ([8 x double]* @src64 to <8 x double>*), align 8
+; CHECK-NEXT:    [[TMP2:%.*]] = fptosi <8 x double> [[TMP1]] to <8 x i8>
+; CHECK-NEXT:    store <8 x i8> [[TMP2]], <8 x i8>* bitcast ([64 x i8]* @dst8 to <8 x i8>*), align 1
 ; CHECK-NEXT:    ret void
 ;
   %a0 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 0), align 8
   %a1 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 1), align 8
   %a2 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 2), align 8
   %a3 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 3), align 8
   %a4 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 4), align 8
   %a5 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 5), align 8
;
   store i16 %cvt5, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @dst16, i32 0, i64 5), align 2
   store i16 %cvt6, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @dst16, i32 0, i64 6), align 2
   store i16 %cvt7, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @dst16, i32 0, i64 7), align 2
   ret void
 }
 
 define void @fptosi_8f32_8i8() #0 {
 ; CHECK-LABEL: @fptosi_8f32_8i8(
-; CHECK-NEXT:    [[A0:%.*]] = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 0), align 4
-; CHECK-NEXT:    [[A1:%.*]] = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 1), align 4
-; CHECK-NEXT:    [[A2:%.*]] = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 2), align 4
-; CHECK-NEXT:    [[A3:%.*]] = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 3), align 4
-; CHECK-NEXT:    [[A4:%.*]] = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 4), align 4
-; CHECK-NEXT:    [[A5:%.*]] = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 5), align 4
-; CHECK-NEXT:    [[A6:%.*]] = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 6), align 4
-; CHECK-NEXT:    [[A7:%.*]] = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 7), align 4
-; CHECK-NEXT:    [[CVT0:%.*]] = fptosi float [[A0]] to i8
-; CHECK-NEXT:    [[CVT1:%.*]] = fptosi float [[A1]] to i8
-; CHECK-NEXT:    [[CVT2:%.*]] = fptosi float [[A2]] to i8
-; CHECK-NEXT:    [[CVT3:%.*]] = fptosi float [[A3]] to i8
-; CHECK-NEXT:    [[CVT4:%.*]] = fptosi float [[A4]] to i8
-; CHECK-NEXT:    [[CVT5:%.*]] = fptosi float [[A5]] to i8
-; CHECK-NEXT:    [[CVT6:%.*]] = fptosi float [[A6]] to i8
-; CHECK-NEXT:    [[CVT7:%.*]] = fptosi float [[A7]] to i8
-; CHECK-NEXT:    store i8 [[CVT0]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @dst8, i32 0, i64 0), align 1
-; CHECK-NEXT:    store i8 [[CVT1]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @dst8, i32 0, i64 1), align 1
-; CHECK-NEXT:    store i8 [[CVT2]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @dst8, i32 0, i64 2), align 1
-; CHECK-NEXT:    store i8 [[CVT3]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @dst8, i32 0, i64 3), align 1
-; CHECK-NEXT:    store i8 [[CVT4]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @dst8, i32 0, i64 4), align 1
-; CHECK-NEXT:    store i8 [[CVT5]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @dst8, i32 0, i64 5), align 1
-; CHECK-NEXT:    store i8 [[CVT6]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @dst8, i32 0, i64 6), align 1
-; CHECK-NEXT:    store i8 [[CVT7]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @dst8, i32 0, i64 7), align 1
+; CHECK-NEXT:    [[TMP1:%.*]] = load <8 x float>, <8 x float>* bitcast ([16 x float]* @src32 to <8 x float>*), align 4
+; CHECK-NEXT:    [[TMP2:%.*]] = fptosi <8 x float> [[TMP1]] to <8 x i8>
+; CHECK-NEXT:    store <8 x i8> [[TMP2]], <8 x i8>* bitcast ([64 x i8]* @dst8 to <8 x i8>*), align 1
 ; CHECK-NEXT:    ret void
 ;
   %a0 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 0), align 4
   %a1 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 1), align 4
   %a2 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 2), align 4
   %a3 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 3), align 4
   %a4 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 4), align 4
   %a5 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 5), align 4

llvm/test/Transforms/SLPVectorizer/X86/fptosi.ll

;
   store i16 %cvt5, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @dst16, i32 0, i64 5), align 2
   store i16 %cvt6, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @dst16, i32 0, i64 6), align 2
   store i16 %cvt7, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @dst16, i32 0, i64 7), align 2
   ret void
 }
 
 define void @fptosi_8f64_8i8() #0 {
 ; CHECK-LABEL: @fptosi_8f64_8i8(
-; CHECK-NEXT:    [[A0:%.*]] = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 0), align 8
-; CHECK-NEXT:    [[A1:%.*]] = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 1), align 8
-; CHECK-NEXT:    [[A2:%.*]] = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 2), align 8
-; CHECK-NEXT:    [[A3:%.*]] = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 3), align 8
-; CHECK-NEXT:    [[A4:%.*]] = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 4), align 8
-; CHECK-NEXT:    [[A5:%.*]] = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 5), align 8
-; CHECK-NEXT:    [[A6:%.*]] = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 6), align 8
-; CHECK-NEXT:    [[A7:%.*]] = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 7), align 8
-; CHECK-NEXT:    [[CVT0:%.*]] = fptosi double [[A0]] to i8
-; CHECK-NEXT:    [[CVT1:%.*]] = fptosi double [[A1]] to i8
-; CHECK-NEXT:    [[CVT2:%.*]] = fptosi double [[A2]] to i8
-; CHECK-NEXT:    [[CVT3:%.*]] = fptosi double [[A3]] to i8
-; CHECK-NEXT:    [[CVT4:%.*]] = fptosi double [[A4]] to i8
-; CHECK-NEXT:    [[CVT5:%.*]] = fptosi double [[A5]] to i8
-; CHECK-NEXT:    [[CVT6:%.*]] = fptosi double [[A6]] to i8
-; CHECK-NEXT:    [[CVT7:%.*]] = fptosi double [[A7]] to i8
-; CHECK-NEXT:    store i8 [[CVT0]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @dst8, i32 0, i64 0), align 1
-; CHECK-NEXT:    store i8 [[CVT1]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @dst8, i32 0, i64 1), align 1
-; CHECK-NEXT:    store i8 [[CVT2]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @dst8, i32 0, i64 2), align 1
-; CHECK-NEXT:    store i8 [[CVT3]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @dst8, i32 0, i64 3), align 1
-; CHECK-NEXT:    store i8 [[CVT4]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @dst8, i32 0, i64 4), align 1
-; CHECK-NEXT:    store i8 [[CVT5]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @dst8, i32 0, i64 5), align 1
-; CHECK-NEXT:    store i8 [[CVT6]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @dst8, i32 0, i64 6), align 1
-; CHECK-NEXT:    store i8 [[CVT7]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @dst8, i32 0, i64 7), align 1
+; CHECK-NEXT:    [[TMP1:%.*]] = load <8 x double>, <8 x double>* bitcast ([8 x double]* @src64 to <8 x double>*), align 8
+; CHECK-NEXT:    [[TMP2:%.*]] = fptosi <8 x double> [[TMP1]] to <8 x i8>
+; CHECK-NEXT:    store <8 x i8> [[TMP2]], <8 x i8>* bitcast ([64 x i8]* @dst8 to <8 x i8>*), align 1
 ; CHECK-NEXT:    ret void
 ;
   %a0 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 0), align 8
   %a1 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 1), align 8
   %a2 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 2), align 8
   %a3 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 3), align 8
   %a4 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 4), align 8
   %a5 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 5), align 8
;
   store i16 %cvt5, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @dst16, i32 0, i64 5), align 2
   store i16 %cvt6, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @dst16, i32 0, i64 6), align 2
   store i16 %cvt7, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @dst16, i32 0, i64 7), align 2
   ret void
 }
 
 define void @fptosi_8f32_8i8() #0 {
 ; CHECK-LABEL: @fptosi_8f32_8i8(
-; CHECK-NEXT:    [[A0:%.*]] = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 0), align 4
-; CHECK-NEXT:    [[A1:%.*]] = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 1), align 4
-; CHECK-NEXT:    [[A2:%.*]] = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 2), align 4
-; CHECK-NEXT:    [[A3:%.*]] = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 3), align 4
-; CHECK-NEXT:    [[A4:%.*]] = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 4), align 4
-; CHECK-NEXT:    [[A5:%.*]] = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 5), align 4
-; CHECK-NEXT:    [[A6:%.*]] = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 6), align 4
-; CHECK-NEXT:    [[A7:%.*]] = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 7), align 4
-; CHECK-NEXT:    [[CVT0:%.*]] = fptosi float [[A0]] to i8
-; CHECK-NEXT:    [[CVT1:%.*]] = fptosi float [[A1]] to i8
-; CHECK-NEXT:    [[CVT2:%.*]] = fptosi float [[A2]] to i8
-; CHECK-NEXT:    [[CVT3:%.*]] = fptosi float [[A3]] to i8
-; CHECK-NEXT:    [[CVT4:%.*]] = fptosi float [[A4]] to i8
-; CHECK-NEXT:    [[CVT5:%.*]] = fptosi float [[A5]] to i8
-; CHECK-NEXT:    [[CVT6:%.*]] = fptosi float [[A6]] to i8
-; CHECK-NEXT:    [[CVT7:%.*]] = fptosi float [[A7]] to i8
-; CHECK-NEXT:    store i8 [[CVT0]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @dst8, i32 0, i64 0), align 1
-; CHECK-NEXT:    store i8 [[CVT1]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @dst8, i32 0, i64 1), align 1
-; CHECK-NEXT:    store i8 [[CVT2]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @dst8, i32 0, i64 2), align 1
-; CHECK-NEXT:    store i8 [[CVT3]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @dst8, i32 0, i64 3), align 1
-; CHECK-NEXT:    store i8 [[CVT4]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @dst8, i32 0, i64 4), align 1
-; CHECK-NEXT:    store i8 [[CVT5]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @dst8, i32 0, i64 5), align 1
-; CHECK-NEXT:    store i8 [[CVT6]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @dst8, i32 0, i64 6), align 1
-; CHECK-NEXT:    store i8 [[CVT7]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @dst8, i32 0, i64 7), align 1
+; CHECK-NEXT:    [[TMP1:%.*]] = load <8 x float>, <8 x float>* bitcast ([16 x float]* @src32 to <8 x float>*), align 4
+; CHECK-NEXT:    [[TMP2:%.*]] = fptosi <8 x float> [[TMP1]] to <8 x i8>
+; CHECK-NEXT:    store <8 x i8> [[TMP2]], <8 x i8>* bitcast ([64 x i8]* @dst8 to <8 x i8>*), align 1
 ; CHECK-NEXT:    ret void
 ;
   %a0 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 0), align 4
   %a1 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 1), align 4
   %a2 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 2), align 4
   %a3 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 3), align 4
   %a4 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 4), align 4
   %a5 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 5), align 4

llvm/test/Transforms/SLPVectorizer/X86/fptoui.ll

;
   store i16 %cvt5, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @dst16, i32 0, i64 5), align 2
   store i16 %cvt6, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @dst16, i32 0, i64 6), align 2
   store i16 %cvt7, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @dst16, i32 0, i64 7), align 2
   ret void
 }
 
 define void @fptoui_8f64_8i8() #0 {
 ; CHECK-LABEL: @fptoui_8f64_8i8(
-; CHECK-NEXT:    [[A0:%.*]] = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 0), align 8
-; CHECK-NEXT:    [[A1:%.*]] = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 1), align 8
-; CHECK-NEXT:    [[A2:%.*]] = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 2), align 8
-; CHECK-NEXT:    [[A3:%.*]] = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 3), align 8
-; CHECK-NEXT:    [[A4:%.*]] = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 4), align 8
-; CHECK-NEXT:    [[A5:%.*]] = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 5), align 8
-; CHECK-NEXT:    [[A6:%.*]] = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 6), align 8
-; CHECK-NEXT:    [[A7:%.*]] = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 7), align 8
-; CHECK-NEXT:    [[CVT0:%.*]] = fptoui double [[A0]] to i8
-; CHECK-NEXT:    [[CVT1:%.*]] = fptoui double [[A1]] to i8
-; CHECK-NEXT:    [[CVT2:%.*]] = fptoui double [[A2]] to i8
-; CHECK-NEXT:    [[CVT3:%.*]] = fptoui double [[A3]] to i8
-; CHECK-NEXT:    [[CVT4:%.*]] = fptoui double [[A4]] to i8
-; CHECK-NEXT:    [[CVT5:%.*]] = fptoui double [[A5]] to i8
-; CHECK-NEXT:    [[CVT6:%.*]] = fptoui double [[A6]] to i8
-; CHECK-NEXT:    [[CVT7:%.*]] = fptoui double [[A7]] to i8
-; CHECK-NEXT:    store i8 [[CVT0]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @dst8, i32 0, i64 0), align 1
-; CHECK-NEXT:    store i8 [[CVT1]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @dst8, i32 0, i64 1), align 1
-; CHECK-NEXT:    store i8 [[CVT2]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @dst8, i32 0, i64 2), align 1
-; CHECK-NEXT:    store i8 [[CVT3]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @dst8, i32 0, i64 3), align 1
-; CHECK-NEXT:    store i8 [[CVT4]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @dst8, i32 0, i64 4), align 1
-; CHECK-NEXT:    store i8 [[CVT5]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @dst8, i32 0, i64 5), align 1
-; CHECK-NEXT:    store i8 [[CVT6]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @dst8, i32 0, i64 6), align 1
-; CHECK-NEXT:    store i8 [[CVT7]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @dst8, i32 0, i64 7), align 1
+; CHECK-NEXT:    [[TMP1:%.*]] = load <8 x double>, <8 x double>* bitcast ([8 x double]* @src64 to <8 x double>*), align 8
+; CHECK-NEXT:    [[TMP2:%.*]] = fptoui <8 x double> [[TMP1]] to <8 x i8>
+; CHECK-NEXT:    store <8 x i8> [[TMP2]], <8 x i8>* bitcast ([64 x i8]* @dst8 to <8 x i8>*), align 1
 ; CHECK-NEXT:    ret void
 ;
   %a0 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 0), align 8
   %a1 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 1), align 8
   %a2 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 2), align 8
   %a3 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 3), align 8
   %a4 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 4), align 8
   %a5 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 5), align 8
;
   store i16 %cvt5, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @dst16, i32 0, i64 5), align 2
   store i16 %cvt6, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @dst16, i32 0, i64 6), align 2
   store i16 %cvt7, i16* getelementptr inbounds ([32 x i16], [32 x i16]* @dst16, i32 0, i64 7), align 2
   ret void
 }
 
 define void @fptoui_8f32_8i8() #0 {
 ; CHECK-LABEL: @fptoui_8f32_8i8(
-; CHECK-NEXT:    [[A0:%.*]] = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 0), align 4
-; CHECK-NEXT:    [[A1:%.*]] = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 1), align 4
-; CHECK-NEXT:    [[A2:%.*]] = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 2), align 4
-; CHECK-NEXT:    [[A3:%.*]] = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 3), align 4
-; CHECK-NEXT:    [[A4:%.*]] = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 4), align 4
-; CHECK-NEXT:    [[A5:%.*]] = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 5), align 4
-; CHECK-NEXT:    [[A6:%.*]] = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 6), align 4
-; CHECK-NEXT:    [[A7:%.*]] = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 7), align 4
-; CHECK-NEXT:    [[CVT0:%.*]] = fptoui float [[A0]] to i8
-; CHECK-NEXT:    [[CVT1:%.*]] = fptoui float [[A1]] to i8
-; CHECK-NEXT:    [[CVT2:%.*]] = fptoui float [[A2]] to i8
-; CHECK-NEXT:    [[CVT3:%.*]] = fptoui float [[A3]] to i8
-; CHECK-NEXT:    [[CVT4:%.*]] = fptoui float [[A4]] to i8
-; CHECK-NEXT:    [[CVT5:%.*]] = fptoui float [[A5]] to i8
-; CHECK-NEXT:    [[CVT6:%.*]] = fptoui float [[A6]] to i8
-; CHECK-NEXT:    [[CVT7:%.*]] = fptoui float [[A7]] to i8
-; CHECK-NEXT:    store i8 [[CVT0]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @dst8, i32 0, i64 0), align 1
-; CHECK-NEXT:    store i8 [[CVT1]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @dst8, i32 0, i64 1), align 1
-; CHECK-NEXT:    store i8 [[CVT2]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @dst8, i32 0, i64 2), align 1
-; CHECK-NEXT:    store i8 [[CVT3]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @dst8, i32 0, i64 3), align 1
-; CHECK-NEXT:    store i8 [[CVT4]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @dst8, i32 0, i64 4), align 1
-; CHECK-NEXT:    store i8 [[CVT5]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @dst8, i32 0, i64 5), align 1
-; CHECK-NEXT:    store i8 [[CVT6]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @dst8, i32 0, i64 6), align 1
-; CHECK-NEXT:    store i8 [[CVT7]], i8* getelementptr inbounds ([64 x i8], [64 x i8]* @dst8, i32 0, i64 7), align 1
+; CHECK-NEXT:    [[TMP1:%.*]] = load <8 x float>, <8 x float>* bitcast ([16 x float]* @src32 to <8 x float>*), align 4
+; CHECK-NEXT:    [[TMP2:%.*]] = fptoui <8 x float> [[TMP1]] to <8 x i8>
+; CHECK-NEXT:    store <8 x i8> [[TMP2]], <8 x i8>* bitcast ([64 x i8]* @dst8 to <8 x i8>*), align 1
 ; CHECK-NEXT:    ret void
 ;
   %a0 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 0), align 4
   %a1 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 1), align 4
   %a2 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 2), align 4
   %a3 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 3), align 4
   %a4 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 4), align 4
   %a5 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 5), align 4

llvm/test/Transforms/SLPVectorizer/X86/insert-after-bundle.ll

 entry:
   %cmp = icmp slt i32 %x, %y
   %b.a = select i1 %cmp, i32 %b, i32 %a
   %retval.0 = trunc i32 %b.a to i8
   ret i8 %retval.0
 }
 
 define void @bar(i8* noalias nocapture readonly %a, i8* noalias nocapture readonly %b, i8* noalias nocapture readonly %c, i8* noalias nocapture readonly %d, i8* noalias nocapture %e, i32 %w) local_unnamed_addr #1 {
-; SSE-LABEL: @bar(
-; SSE-NEXT:  entry:
-; SSE-NEXT:    [[TMP0:%.*]] = insertelement <4 x i32> poison, i32 [[W:%.*]], i32 0
-; SSE-NEXT:    [[SHUFFLE:%.*]] = shufflevector <4 x i32> [[TMP0]], <4 x i32> poison, <4 x i32> zeroinitializer
-; SSE-NEXT:    [[TMP1:%.*]] = insertelement <4 x i32> poison, i32 [[W]], i32 0
-; SSE-NEXT:    [[SHUFFLE1:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> poison, <4 x i32> zeroinitializer
-; SSE-NEXT:    [[TMP2:%.*]] = insertelement <4 x i32> poison, i32 [[W]], i32 0
-; SSE-NEXT:    [[SHUFFLE2:%.*]] = shufflevector <4 x i32> [[TMP2]], <4 x i32> poison, <4 x i32> zeroinitializer
-; SSE-NEXT:    [[TMP3:%.*]] = insertelement <4 x i32> poison, i32 [[W]], i32 0
-; SSE-NEXT:    [[SHUFFLE3:%.*]] = shufflevector <4 x i32> [[TMP3]], <4 x i32> poison, <4 x i32> zeroinitializer
-; SSE-NEXT:    br label [[FOR_BODY:%.*]]
-; SSE:       for.body:
-; SSE-NEXT:    [[I_0356:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[INC:%.*]], [[FOR_BODY]] ]
-; SSE-NEXT:    [[A_ADDR_0355:%.*]] = phi i8* [ [[A:%.*]], [[ENTRY]] ], [ [[ADD_PTR:%.*]], [[FOR_BODY]] ]
-; SSE-NEXT:    [[E_ADDR_0354:%.*]] = phi i8* [ [[E:%.*]], [[ENTRY]] ], [ [[ADD_PTR192:%.*]], [[FOR_BODY]] ]
-; SSE-NEXT:    [[D_ADDR_0353:%.*]] = phi i8* [ [[D:%.*]], [[ENTRY]] ], [ [[ADD_PTR191:%.*]], [[FOR_BODY]] ]
-; SSE-NEXT:    [[C_ADDR_0352:%.*]] = phi i8* [ [[C:%.*]], [[ENTRY]] ], [ [[ADD_PTR190:%.*]], [[FOR_BODY]] ]
-; SSE-NEXT:    [[B_ADDR_0351:%.*]] = phi i8* [ [[B:%.*]], [[ENTRY]] ], [ [[ADD_PTR189:%.*]], [[FOR_BODY]] ]
-; SSE-NEXT:    [[TMP4:%.*]] = bitcast i8* [[C_ADDR_0352]] to <4 x i8>*
-; SSE-NEXT:    [[TMP5:%.*]] = load <4 x i8>, <4 x i8>* [[TMP4]], align 1
-; SSE-NEXT:    [[TMP6:%.*]] = bitcast i8* [[D_ADDR_0353]] to <4 x i8>*
-; SSE-NEXT:    [[TMP7:%.*]] = load <4 x i8>, <4 x i8>* [[TMP6]], align 1
-; SSE-NEXT:    [[TMP8:%.*]] = bitcast i8* [[A_ADDR_0355]] to <4 x i8>*
-; SSE-NEXT:    [[TMP9:%.*]] = load <4 x i8>, <4 x i8>* [[TMP8]], align 1
-; SSE-NEXT:    [[TMP10:%.*]] = bitcast i8* [[B_ADDR_0351]] to <4 x i8>*
-; SSE-NEXT:    [[TMP11:%.*]] = load <4 x i8>, <4 x i8>* [[TMP10]], align 1
-; SSE-NEXT:    [[TMP12:%.*]] = icmp ult <4 x i8> [[TMP5]], [[TMP7]]
-; SSE-NEXT:    [[TMP13:%.*]] = select <4 x i1> [[TMP12]], <4 x i8> [[TMP11]], <4 x i8> [[TMP9]]
-; SSE-NEXT:    [[TMP14:%.*]] = zext <4 x i8> [[TMP13]] to <4 x i32>
-; SSE-NEXT:    [[TMP15:%.*]] = mul <4 x i32> [[TMP14]], [[SHUFFLE]]
-; SSE-NEXT:    [[TMP16:%.*]] = trunc <4 x i32> [[TMP15]] to <4 x i8>
-; SSE-NEXT:    [[TMP17:%.*]] = bitcast i8* [[E_ADDR_0354]] to <4 x i8>*
-; SSE-NEXT:    store <4 x i8> [[TMP16]], <4 x i8>* [[TMP17]], align 1
-; SSE-NEXT:    [[ARRAYIDX45:%.*]] = getelementptr inbounds i8, i8* [[C_ADDR_0352]], i64 4
-; SSE-NEXT:    [[ARRAYIDX47:%.*]] = getelementptr inbounds i8, i8* [[D_ADDR_0353]], i64 4
-; SSE-NEXT:    [[ARRAYIDX49:%.*]] = getelementptr inbounds i8, i8* [[A_ADDR_0355]], i64 4
-; SSE-NEXT:    [[ARRAYIDX52:%.*]] = getelementptr inbounds i8, i8* [[B_ADDR_0351]], i64 4
-; SSE-NEXT:    [[ARRAYIDX56:%.*]] = getelementptr inbounds i8, i8* [[E_ADDR_0354]], i64 4
-; SSE-NEXT:    [[TMP18:%.*]] = bitcast i8* [[ARRAYIDX45]] to <4 x i8>*
-; SSE-NEXT:    [[TMP19:%.*]] = load <4 x i8>, <4 x i8>* [[TMP18]], align 1
-; SSE-NEXT:    [[TMP20:%.*]] = bitcast i8* [[ARRAYIDX47]] to <4 x i8>*
-; SSE-NEXT:    [[TMP21:%.*]] = load <4 x i8>, <4 x i8>* [[TMP20]], align 1
-; SSE-NEXT:    [[TMP22:%.*]] = bitcast i8* [[ARRAYIDX49]] to <4 x i8>*
-; SSE-NEXT:    [[TMP23:%.*]] = load <4 x i8>, <4 x i8>* [[TMP22]], align 1
-; SSE-NEXT:    [[TMP24:%.*]] = bitcast i8* [[ARRAYIDX52]] to <4 x i8>*
-; SSE-NEXT:    [[TMP25:%.*]] = load <4 x i8>, <4 x i8>* [[TMP24]], align 1
-; SSE-NEXT:    [[TMP26:%.*]] = icmp ult <4 x i8> [[TMP19]], [[TMP21]]
-; SSE-NEXT:    [[TMP27:%.*]] = select <4 x i1> [[TMP26]], <4 x i8> [[TMP25]], <4 x i8> [[TMP23]]
-; SSE-NEXT:    [[TMP28:%.*]] = zext <4 x i8> [[TMP27]] to <4 x i32>
-; SSE-NEXT:    [[TMP29:%.*]] = mul <4 x i32> [[TMP28]], [[SHUFFLE1]]
-; SSE-NEXT:    [[TMP30:%.*]] = trunc <4 x i32> [[TMP29]] to <4 x i8>
-; SSE-NEXT:    [[TMP31:%.*]] = bitcast i8* [[ARRAYIDX56]] to <4 x i8>*
-; SSE-NEXT:    store <4 x i8> [[TMP30]], <4 x i8>* [[TMP31]], align 1
-; SSE-NEXT:    [[ARRAYIDX93:%.*]] = getelementptr inbounds i8, i8* [[C_ADDR_0352]], i64 8
-; SSE-NEXT:    [[ARRAYIDX95:%.*]] = getelementptr inbounds i8, i8* [[D_ADDR_0353]], i64 8
-; SSE-NEXT:    [[ARRAYIDX97:%.*]] = getelementptr inbounds i8, i8* [[A_ADDR_0355]], i64 8
-; SSE-NEXT:    [[ARRAYIDX100:%.*]] = getelementptr inbounds i8, i8* [[B_ADDR_0351]], i64 8
-; SSE-NEXT:    [[ARRAYIDX104:%.*]] = getelementptr inbounds i8, i8* [[E_ADDR_0354]], i64 8
-; SSE-NEXT:    [[TMP32:%.*]] = bitcast i8* [[ARRAYIDX93]] to <4 x i8>*
-; SSE-NEXT:    [[TMP33:%.*]] = load <4 x i8>, <4 x i8>* [[TMP32]], align 1
-; SSE-NEXT:    [[TMP34:%.*]] = bitcast i8* [[ARRAYIDX95]] to <4 x i8>*
-; SSE-NEXT:    [[TMP35:%.*]] = load <4 x i8>, <4 x i8>* [[TMP34]], align 1
-; SSE-NEXT:    [[TMP36:%.*]] = bitcast i8* [[ARRAYIDX97]] to <4 x i8>*
-; SSE-NEXT:    [[TMP37:%.*]] = load <4 x i8>, <4 x i8>* [[TMP36]], align 1
-; SSE-NEXT:    [[TMP38:%.*]] = bitcast i8* [[ARRAYIDX100]] to <4 x i8>*
-; SSE-NEXT:    [[TMP39:%.*]] = load <4 x i8>, <4 x i8>* [[TMP38]], align 1
-; SSE-NEXT:    [[TMP40:%.*]] = icmp ult <4 x i8> [[TMP33]], [[TMP35]]
-; SSE-NEXT:    [[TMP41:%.*]] = select <4 x i1> [[TMP40]], <4 x i8> [[TMP39]], <4 x i8> [[TMP37]]
-; SSE-NEXT:    [[TMP42:%.*]] = zext <4 x i8> [[TMP41]] to <4 x i32>
-; SSE-NEXT:    [[TMP43:%.*]] = mul <4 x i32> [[TMP42]], [[SHUFFLE2]]
-; SSE-NEXT:    [[TMP44:%.*]] = trunc <4 x i32> [[TMP43]] to <4 x i8>
-; SSE-NEXT:    [[TMP45:%.*]] = bitcast i8* [[ARRAYIDX104]] to <4 x i8>*
-; SSE-NEXT:    store <4 x i8> [[TMP44]], <4 x i8>* [[TMP45]], align 1
-; SSE-NEXT:    [[ARRAYIDX141:%.*]] = getelementptr inbounds i8, i8* [[C_ADDR_0352]], i64 12
-; SSE-NEXT:    [[ARRAYIDX143:%.*]] = getelementptr inbounds i8, i8* [[D_ADDR_0353]], i64 12
-; SSE-NEXT:    [[ARRAYIDX145:%.*]] = getelementptr inbounds i8, i8* [[A_ADDR_0355]], i64 12
-; SSE-NEXT:    [[ARRAYIDX148:%.*]] = getelementptr inbounds i8, i8* [[B_ADDR_0351]], i64 12
-; SSE-NEXT:    [[ARRAYIDX152:%.*]] = getelementptr inbounds i8, i8* [[E_ADDR_0354]], i64 12
-; SSE-NEXT:    [[TMP46:%.*]] = bitcast i8* [[ARRAYIDX141]] to <4 x i8>*
-; SSE-NEXT:    [[TMP47:%.*]] = load <4 x i8>, <4 x i8>* [[TMP46]], align 1
-; SSE-NEXT:    [[TMP48:%.*]] = bitcast i8* [[ARRAYIDX143]] to <4 x i8>*
-; SSE-NEXT:    [[TMP49:%.*]] = load <4 x i8>, <4 x i8>* [[TMP48]], align 1
-; SSE-NEXT:    [[TMP50:%.*]] = bitcast i8* [[ARRAYIDX145]] to <4 x i8>*
-; SSE-NEXT:    [[TMP51:%.*]] = load <4 x i8>, <4 x i8>* [[TMP50]], align 1
-; SSE-NEXT:    [[TMP52:%.*]] = bitcast i8* [[ARRAYIDX148]] to <4 x i8>*
-; SSE-NEXT:    [[TMP53:%.*]] = load <4 x i8>, <4 x i8>* [[TMP52]], align 1
-; SSE-NEXT:    [[TMP54:%.*]] = icmp ult <4 x i8> [[TMP47]], [[TMP49]]
-; SSE-NEXT:    [[TMP55:%.*]] = select <4 x i1> [[TMP54]], <4 x i8> [[TMP53]], <4 x i8> [[TMP51]]
-; SSE-NEXT:    [[TMP56:%.*]] = zext <4 x i8> [[TMP55]] to <4 x i32>
-; SSE-NEXT:    [[TMP57:%.*]] = mul <4 x i32> [[TMP56]], [[SHUFFLE3]]
-; SSE-NEXT:    [[TMP58:%.*]] = trunc <4 x i32> [[TMP57]] to <4 x i8>
-; SSE-NEXT:    [[TMP59:%.*]] = bitcast i8* [[ARRAYIDX152]] to <4 x i8>*
-; SSE-NEXT:    store <4 x i8> [[TMP58]], <4 x i8>* [[TMP59]], align 1
-; SSE-NEXT:    [[INC]] = add nuw nsw i32 [[I_0356]], 1
-; SSE-NEXT:    [[ADD_PTR]] = getelementptr inbounds i8, i8* [[A_ADDR_0355]], i64 16
-; SSE-NEXT:    [[ADD_PTR189]] = getelementptr inbounds i8, i8* [[B_ADDR_0351]], i64 16
-; SSE-NEXT:    [[ADD_PTR190]] = getelementptr inbounds i8, i8* [[C_ADDR_0352]], i64 16
-; SSE-NEXT:    [[ADD_PTR191]] = getelementptr inbounds i8, i8* [[D_ADDR_0353]], i64 16
-; SSE-NEXT:    [[ADD_PTR192]] = getelementptr inbounds i8, i8* [[E_ADDR_0354]], i64 16
-; SSE-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[INC]], 8
-; SSE-NEXT:    br i1 [[EXITCOND]], label [[FOR_END:%.*]], label [[FOR_BODY]]
-; SSE:       for.end:
-; SSE-NEXT:    ret void
-;
-; AVX512-LABEL: @bar(
-; AVX512-NEXT:  entry:
-; AVX512-NEXT:    [[TMP0:%.*]] = insertelement <16 x i32> poison, i32 [[W:%.*]], i32 0
-; AVX512-NEXT:    [[SHUFFLE:%.*]] = shufflevector <16 x i32> [[TMP0]], <16 x i32> poison, <16 x i32> zeroinitializer
-; AVX512-NEXT:    br label [[FOR_BODY:%.*]]
-; AVX512:       for.body:
-; AVX512-NEXT:    [[I_0356:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[INC:%.*]], [[FOR_BODY]] ]
-; AVX512-NEXT:    [[A_ADDR_0355:%.*]] = phi i8* [ [[A:%.*]], [[ENTRY]] ], [ [[ADD_PTR:%.*]], [[FOR_BODY]] ]
-; AVX512-NEXT:    [[E_ADDR_0354:%.*]] = phi i8* [ [[E:%.*]], [[ENTRY]] ], [ [[ADD_PTR192:%.*]], [[FOR_BODY]] ]
-; AVX512-NEXT:    [[D_ADDR_0353:%.*]] = phi i8* [ [[D:%.*]], [[ENTRY]] ], [ [[ADD_PTR191:%.*]], [[FOR_BODY]] ]
-; AVX512-NEXT:    [[C_ADDR_0352:%.*]] = phi i8* [ [[C:%.*]], [[ENTRY]] ], [ [[ADD_PTR190:%.*]], [[FOR_BODY]] ]
-; AVX512-NEXT:    [[B_ADDR_0351:%.*]] = phi i8* [ [[B:%.*]], [[ENTRY]] ], [ [[ADD_PTR189:%.*]], [[FOR_BODY]] ]
-; AVX512-NEXT:    [[TMP1:%.*]] = bitcast i8* [[C_ADDR_0352]] to <16 x i8>*
-; AVX512-NEXT:    [[TMP2:%.*]] = load <16 x i8>, <16 x i8>* [[TMP1]], align 1
-; AVX512-NEXT:    [[TMP3:%.*]] = bitcast i8* [[D_ADDR_0353]] to <16 x i8>*
-; AVX512-NEXT:    [[TMP4:%.*]] = load <16 x i8>, <16 x i8>* [[TMP3]], align 1
-; AVX512-NEXT:    [[TMP5:%.*]] = bitcast i8* [[A_ADDR_0355]] to <16 x i8>*
-; AVX512-NEXT:    [[TMP6:%.*]] = load <16 x i8>, <16 x i8>* [[TMP5]], align 1
-; AVX512-NEXT:    [[TMP7:%.*]] = bitcast i8* [[B_ADDR_0351]] to <16 x i8>*
-; AVX512-NEXT:    [[TMP8:%.*]] = load <16 x i8>, <16 x i8>* [[TMP7]], align 1
-; AVX512-NEXT:    [[TMP9:%.*]] = icmp ult <16 x i8> [[TMP2]], [[TMP4]]
-; AVX512-NEXT:    [[TMP10:%.*]] = select <16 x i1> [[TMP9]], <16 x i8> [[TMP8]], <16 x i8> [[TMP6]]
-; AVX512-NEXT:    [[TMP11:%.*]] = zext <16 x i8> [[TMP10]] to <16 x i32>
-; AVX512-NEXT:    [[TMP12:%.*]] = mul <16 x i32> [[TMP11]], [[SHUFFLE]]
-; AVX512-NEXT:    [[TMP13:%.*]] = trunc <16 x i32> [[TMP12]] to <16 x i8>
-; AVX512-NEXT:    [[TMP14:%.*]] = bitcast i8* [[E_ADDR_0354]] to <16 x i8>*
-; AVX512-NEXT:    store <16 x i8> [[TMP13]], <16 x i8>* [[TMP14]], align 1
-; AVX512-NEXT:    [[INC]] = add nuw nsw i32 [[I_0356]], 1
-; AVX512-NEXT:    [[ADD_PTR]] = getelementptr inbounds i8, i8* [[A_ADDR_0355]], i64 16
-; AVX512-NEXT:    [[ADD_PTR189]] = getelementptr inbounds i8, i8* [[B_ADDR_0351]], i64 16
-; AVX512-NEXT:    [[ADD_PTR190]] = getelementptr inbounds i8, i8* [[C_ADDR_0352]], i64 16
-; AVX512-NEXT:    [[ADD_PTR191]] = getelementptr inbounds i8, i8* [[D_ADDR_0353]], i64 16
-; AVX512-NEXT:    [[ADD_PTR192]] = getelementptr inbounds i8, i8* [[E_ADDR_0354]], i64 16
-; AVX512-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[INC]], 8
-; AVX512-NEXT:    br i1 [[EXITCOND]], label [[FOR_END:%.*]], label [[FOR_BODY]]
-; AVX512:       for.end:
-; AVX512-NEXT:    ret void
+; CHECK-LABEL: @bar(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[TMP0:%.*]] = insertelement <16 x i32> poison, i32 [[W:%.*]], i32 0
+; CHECK-NEXT:    [[SHUFFLE:%.*]] = shufflevector <16 x i32> [[TMP0]], <16 x i32> poison, <16 x i32> zeroinitializer
+; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
+; CHECK:       for.body:
+; CHECK-NEXT:    [[I_0356:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[INC:%.*]], [[FOR_BODY]] ]
+; CHECK-NEXT:    [[A_ADDR_0355:%.*]] = phi i8* [ [[A:%.*]], [[ENTRY]] ], [ [[ADD_PTR:%.*]], [[FOR_BODY]] ]
+; CHECK-NEXT:    [[E_ADDR_0354:%.*]] = phi i8* [ [[E:%.*]], [[ENTRY]] ], [ [[ADD_PTR192:%.*]], [[FOR_BODY]] ]
+; CHECK-NEXT:    [[D_ADDR_0353:%.*]] = phi i8* [ [[D:%.*]], [[ENTRY]] ], [ [[ADD_PTR191:%.*]], [[FOR_BODY]] ]
+; CHECK-NEXT:    [[C_ADDR_0352:%.*]] = phi i8* [ [[C:%.*]], [[ENTRY]] ], [ [[ADD_PTR190:%.*]], [[FOR_BODY]] ]
+; CHECK-NEXT:    [[B_ADDR_0351:%.*]] = phi i8* [ [[B:%.*]], [[ENTRY]] ], [ [[ADD_PTR189:%.*]], [[FOR_BODY]] ]
+; CHECK-NEXT:    [[TMP1:%.*]] = bitcast i8* [[C_ADDR_0352]] to <16 x i8>*
+; CHECK-NEXT:    [[TMP2:%.*]] = load <16 x i8>, <16 x i8>* [[TMP1]], align 1
+; CHECK-NEXT:    [[TMP3:%.*]] = bitcast i8* [[D_ADDR_0353]] to <16 x i8>*
+; CHECK-NEXT:    [[TMP4:%.*]] = load <16 x i8>, <16 x i8>* [[TMP3]], align 1
+; CHECK-NEXT:    [[TMP5:%.*]] = bitcast i8* [[A_ADDR_0355]] to <16 x i8>*
+; CHECK-NEXT:    [[TMP6:%.*]] = load <16 x i8>, <16 x i8>* [[TMP5]], align 1
+; CHECK-NEXT:    [[TMP7:%.*]] = bitcast i8* [[B_ADDR_0351]] to <16 x i8>*
+; CHECK-NEXT:    [[TMP8:%.*]] = load <16 x i8>, <16 x i8>* [[TMP7]], align 1
+; CHECK-NEXT:    [[TMP9:%.*]] = icmp ult <16 x i8> [[TMP2]], [[TMP4]]
+; CHECK-NEXT:    [[TMP10:%.*]] = select <16 x i1> [[TMP9]], <16 x i8> [[TMP8]], <16 x i8> [[TMP6]]
+; CHECK-NEXT:    [[TMP11:%.*]] = zext <16 x i8> [[TMP10]] to <16 x i32>
+; CHECK-NEXT:    [[TMP12:%.*]] = mul <16 x i32> [[TMP11]], [[SHUFFLE]]
+; CHECK-NEXT:    [[TMP13:%.*]] = trunc <16 x i32> [[TMP12]] to <16 x i8>
+; CHECK-NEXT:    [[TMP14:%.*]] = bitcast i8* [[E_ADDR_0354]] to <16 x i8>*
+; CHECK-NEXT:    store <16 x i8> [[TMP13]], <16 x i8>* [[TMP14]], align 1
+; CHECK-NEXT:    [[INC]] = add nuw nsw i32 [[I_0356]], 1
+; CHECK-NEXT:    [[ADD_PTR]] = getelementptr inbounds i8, i8* [[A_ADDR_0355]], i64 16
+; CHECK-NEXT:    [[ADD_PTR189]] = getelementptr inbounds i8, i8* [[B_ADDR_0351]], i64 16
+; CHECK-NEXT:    [[ADD_PTR190]] = getelementptr inbounds i8, i8* [[C_ADDR_0352]], i64 16
+; CHECK-NEXT:    [[ADD_PTR191]] = getelementptr inbounds i8, i8* [[D_ADDR_0353]], i64 16
+; CHECK-NEXT:    [[ADD_PTR192]] = getelementptr inbounds i8, i8* [[E_ADDR_0354]], i64 16
+; CHECK-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[INC]], 8
+; CHECK-NEXT:    br i1 [[EXITCOND]], label [[FOR_END:%.*]], label [[FOR_BODY]]
+; CHECK:       for.end:
+; CHECK-NEXT:    ret void
 ;
 entry:
   br label %for.body
 
 for.body:                                         ; preds = %for.body, %entry
   %i.0356 = phi i32 [ 0, %entry ], [ %inc, %for.body ]
   %a.addr.0355 = phi i8* [ %a, %entry ], [ %add.ptr, %for.body ]
   %e.addr.0354 = phi i8* [ %e, %entry ], [ %add.ptr192, %for.body ]

llvm/test/Transforms/SLPVectorizer/X86/memory-runtime-checks.ll

 ; CHECK-NEXT:    ret void
 ; CHECK:       then:
 ; CHECK-NEXT:    [[A_8:%.*]] = getelementptr inbounds float, float* [[A:%.*]], i64 8
 ; CHECK-NEXT:    store float 0.000000e+00, float* [[A_8]], align 4
 ; CHECK-NEXT:    [[L6:%.*]] = load float, float* [[B_14]], align 4
 ; CHECK-NEXT:    [[A_5:%.*]] = getelementptr inbounds float, float* [[A]], i64 5
 ; CHECK-NEXT:    store float [[L6]], float* [[A_5]], align 4
 ; CHECK-NEXT:    [[A_6:%.*]] = getelementptr inbounds float, float* [[A]], i64 6
-; CHECK-NEXT:    store float 0.000000e+00, float* [[A_6]], align 4
-; CHECK-NEXT:    [[A_7:%.*]] = getelementptr inbounds float, float* [[A]], i64 7
-; CHECK-NEXT:    store float 0.000000e+00, float* [[A_7]], align 4
+; CHECK-NEXT:    [[TMP0:%.*]] = bitcast float* [[A_6]] to <2 x float>*
+; CHECK-NEXT:    store <2 x float> zeroinitializer, <2 x float>* [[TMP0]], align 4
 ; CHECK-NEXT:    ret void
 ;
 entry:
   %b.10 = getelementptr inbounds float, float* %b, i64 10
   %b.14 = getelementptr inbounds float, float* %b, i64 14
   br i1 %c, label %then, label %else
 
 else:

 ; CHECK:       bb22:
 ; CHECK-NEXT:    [[TMP23:%.*]] = getelementptr inbounds float, float* [[ARG4:%.*]], i32 0
 ; CHECK-NEXT:    [[TMP24:%.*]] = getelementptr float, float* [[TMP23]], i64 7
 ; CHECK-NEXT:    br i1 [[C_2:%.*]], label [[BB25:%.*]], label [[BB22]]
 ; CHECK:       bb25:
 ; CHECK-NEXT:    [[TMP26:%.*]] = getelementptr float, float* [[TMP23]], i64 6
 ; CHECK-NEXT:    store float 0.000000e+00, float* [[TMP24]], align 4
 ; CHECK-NEXT:    [[TMP27:%.*]] = load float, float* [[ARG5:%.*]], align 4
-; CHECK-NEXT:    [[TMP28:%.*]] = getelementptr float, float* [[TMP23]], i64 5
 ; CHECK-NEXT:    [[TMP29:%.*]] = fadd float 0.000000e+00, 0.000000e+00
 ; CHECK-NEXT:    store float 0.000000e+00, float* [[TMP26]], align 4
 ; CHECK-NEXT:    [[TMP30:%.*]] = getelementptr float, float* [[TMP23]], i64 4
-; CHECK-NEXT:    store float 0.000000e+00, float* [[TMP28]], align 4
 ; CHECK-NEXT:    [[TMP31:%.*]] = fadd float 0.000000e+00, 0.000000e+00
-; CHECK-NEXT:    store float 0.000000e+00, float* [[TMP30]], align 4
+; CHECK-NEXT:    [[TMP5:%.*]] = bitcast float* [[TMP30]] to <2 x float>*
+; CHECK-NEXT:    store <2 x float> zeroinitializer, <2 x float>* [[TMP5]], align 4
 ; CHECK-NEXT:    [[TMP32:%.*]] = getelementptr inbounds float, float* [[ARG4]], i32 0
 ; CHECK-NEXT:    br label [[BB33:%.*]]
 ; CHECK:       bb33:
 ; CHECK-NEXT:    br label [[BB34:%.*]]
 ; CHECK:       bb34:
 ; CHECK-NEXT:    [[TMP35:%.*]] = getelementptr float, float* [[TMP32]], i64 3
-; CHECK-NEXT:    [[TMP36:%.*]] = getelementptr float, float* [[TMP32]], i64 2
 ; CHECK-NEXT:    [[TMP37:%.*]] = load float, float* [[TMP35]], align 4
 ; CHECK-NEXT:    [[TMP38:%.*]] = fadd float 0.000000e+00, [[TMP37]]
 ; CHECK-NEXT:    store float [[TMP38]], float* [[TMP35]], align 4
 ; CHECK-NEXT:    [[TMP39:%.*]] = getelementptr float, float* [[TMP32]], i64 1
-; CHECK-NEXT:    [[TMP40:%.*]] = load float, float* [[TMP36]], align 4
-; CHECK-NEXT:    [[TMP41:%.*]] = fadd float 0.000000e+00, [[TMP40]]
-; CHECK-NEXT:    store float [[TMP41]], float* [[TMP36]], align 4
-; CHECK-NEXT:    [[TMP42:%.*]] = load float, float* [[TMP39]], align 4
-; CHECK-NEXT:    [[TMP43:%.*]] = fadd float 0.000000e+00, [[TMP42]]
-; CHECK-NEXT:    store float [[TMP43]], float* [[TMP39]], align 4
+; CHECK-NEXT:    [[TMP6:%.*]] = bitcast float* [[TMP39]] to <2 x float>*
+; CHECK-NEXT:    [[TMP7:%.*]] = load <2 x float>, <2 x float>* [[TMP6]], align 4
+; CHECK-NEXT:    [[TMP8:%.*]] = fadd <2 x float> zeroinitializer, [[TMP7]]
+; CHECK-NEXT:    [[TMP9:%.*]] = bitcast float* [[TMP39]] to <2 x float>*
+; CHECK-NEXT:    store <2 x float> [[TMP8]], <2 x float>* [[TMP9]], align 4
 ; CHECK-NEXT:    [[TMP44:%.*]] = load float, float* [[ARG3:%.*]], align 4
 ; CHECK-NEXT:    [[TMP45:%.*]] = load float, float* [[TMP32]], align 4
 ; CHECK-NEXT:    [[TMP46:%.*]] = fadd float 0.000000e+00, [[TMP45]]
 ; CHECK-NEXT:    store float [[TMP46]], float* [[TMP32]], align 4
 ; CHECK-NEXT:    call void @quux()
 ; CHECK-NEXT:    br label [[BB47:%.*]]
 ; CHECK:       bb47:
 ; CHECK-NEXT:    br label [[BB17]]

llvm/test/Transforms/SLPVectorizer/X86/no_alternate_divrem.ll

 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
 ; RUN: opt < %s -slp-vectorizer -slp-threshold=-200 -mtriple=x86_64-unknown-linux -mcpu=core-avx2 -S | FileCheck %s
 
 define void @test_add_sdiv(i32 *%arr1, i32 *%arr2, i32 %a0, i32 %a1, i32 %a2, i32 %a3) {
 ; CHECK-LABEL: @test_add_sdiv(
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[GEP1_0:%.*]] = getelementptr i32, i32* [[ARR1:%.*]], i32 0
-; CHECK-NEXT:    [[GEP1_1:%.*]] = getelementptr i32, i32* [[ARR1]], i32 1
 ; CHECK-NEXT:    [[GEP1_2:%.*]] = getelementptr i32, i32* [[ARR1]], i32 2
 ; CHECK-NEXT:    [[GEP1_3:%.*]] = getelementptr i32, i32* [[ARR1]], i32 3
 ; CHECK-NEXT:    [[GEP2_0:%.*]] = getelementptr i32, i32* [[ARR2:%.*]], i32 0
-; CHECK-NEXT:    [[GEP2_1:%.*]] = getelementptr i32, i32* [[ARR2]], i32 1
 ; CHECK-NEXT:    [[GEP2_2:%.*]] = getelementptr i32, i32* [[ARR2]], i32 2
 ; CHECK-NEXT:    [[GEP2_3:%.*]] = getelementptr i32, i32* [[ARR2]], i32 3
-; CHECK-NEXT:    [[V0:%.*]] = load i32, i32* [[GEP1_0]], align 4
-; CHECK-NEXT:    [[V1:%.*]] = load i32, i32* [[GEP1_1]], align 4
 ; CHECK-NEXT:    [[V2:%.*]] = load i32, i32* [[GEP1_2]], align 4
 ; CHECK-NEXT:    [[V3:%.*]] = load i32, i32* [[GEP1_3]], align 4
-; CHECK-NEXT:    [[Y0:%.*]] = add nsw i32 [[A0:%.*]], 1146
-; CHECK-NEXT:    [[Y1:%.*]] = add nsw i32 [[A1:%.*]], 146
+; CHECK-NEXT:    [[TMP0:%.*]] = insertelement <2 x i32> poison, i32 [[A0:%.*]], i32 0
+; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <2 x i32> [[TMP0]], i32 [[A1:%.*]], i32 1
+; CHECK-NEXT:    [[TMP2:%.*]] = add nsw <2 x i32> [[TMP1]], <i32 1146, i32 146>
 ; CHECK-NEXT:    [[Y2:%.*]] = add nsw i32 [[A2:%.*]], 42
 ; CHECK-NEXT:    [[Y3:%.*]] = add nsw i32 [[A3:%.*]], 0
-; CHECK-NEXT:    [[RES0:%.*]] = add nsw i32 [[V0]], [[Y0]]
-; CHECK-NEXT:    [[RES1:%.*]] = add nsw i32 [[V1]], [[Y1]]
 ; CHECK-NEXT:    [[RES2:%.*]] = sdiv i32 [[V2]], [[Y2]]
 ; CHECK-NEXT:    [[RES3:%.*]] = add nsw i32 [[V3]], [[Y3]]
-; CHECK-NEXT:    store i32 [[RES0]], i32* [[GEP2_0]], align 4
-; CHECK-NEXT:    store i32 [[RES1]], i32* [[GEP2_1]], align 4
+; CHECK-NEXT:    [[TMP3:%.*]] = bitcast i32* [[GEP1_0]] to <2 x i32>*
+; CHECK-NEXT:    [[TMP4:%.*]] = load <2 x i32>, <2 x i32>* [[TMP3]], align 4
+; CHECK-NEXT:    [[TMP5:%.*]] = add nsw <2 x i32> [[TMP4]], [[TMP2]]
+; CHECK-NEXT:    [[TMP6:%.*]] = bitcast i32* [[GEP2_0]] to <2 x i32>*
+; CHECK-NEXT:    store <2 x i32> [[TMP5]], <2 x i32>* [[TMP6]], align 4
 ; CHECK-NEXT:    store i32 [[RES2]], i32* [[GEP2_2]], align 4
 ; CHECK-NEXT:    store i32 [[RES3]], i32* [[GEP2_3]], align 4
 ; CHECK-NEXT:    ret void
 ;
 entry:
   %gep1.0 = getelementptr i32, i32* %arr1, i32 0
   %gep1.1 = getelementptr i32, i32* %arr1, i32 1
   %gep1.2 = getelementptr i32, i32* %arr1, i32 2

llvm/test/Transforms/SLPVectorizer/X86/odd_store.ll

 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
 ; RUN: opt < %s -basic-aa -slp-vectorizer -dce -S -mtriple=x86_64-apple-macosx10.8.0 -mcpu=corei7-avx | FileCheck %s
 
 ;int foo(char * restrict A, float * restrict B, float T) {
 ;  A[0] = (T * B[10] + 4.0);
 ;  A[1] = (T * B[11] + 5.0);
 ;  A[2] = (T * B[12] + 6.0);
 ;}
 
 define i32 @foo(i8* noalias nocapture %A, float* noalias nocapture %B, float %T) {
 ; CHECK-LABEL: @foo(
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds float, float* [[B:%.*]], i64 10
-; CHECK-NEXT:    [[TMP2:%.*]] = load float, float* [[TMP1]], align 4
-; CHECK-NEXT:    [[TMP3:%.*]] = fmul float [[TMP2]], [[T:%.*]]
-; CHECK-NEXT:    [[TMP4:%.*]] = fpext float [[TMP3]] to double
-; CHECK-NEXT:    [[TMP5:%.*]] = fadd double [[TMP4]], 4.000000e+00
-; CHECK-NEXT:    [[TMP6:%.*]] = fptosi double [[TMP5]] to i8
-; CHECK-NEXT:    store i8 [[TMP6]], i8* [[A:%.*]], align 1
-; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds float, float* [[B]], i64 11
-; CHECK-NEXT:    [[TMP8:%.*]] = load float, float* [[TMP7]], align 4
-; CHECK-NEXT:    [[TMP9:%.*]] = fmul float [[TMP8]], [[T]]
-; CHECK-NEXT:    [[TMP10:%.*]] = fpext float [[TMP9]] to double
-; CHECK-NEXT:    [[TMP11:%.*]] = fadd double [[TMP10]], 5.000000e+00
-; CHECK-NEXT:    [[TMP12:%.*]] = fptosi double [[TMP11]] to i8
-; CHECK-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i8, i8* [[A]], i64 1
-; CHECK-NEXT:    store i8 [[TMP12]], i8* [[TMP13]], align 1
-; CHECK-NEXT:    [[TMP14:%.*]] = getelementptr inbounds float, float* [[B]], i64 12
-; CHECK-NEXT:    [[TMP15:%.*]] = load float, float* [[TMP14]], align 4
-; CHECK-NEXT:    [[TMP16:%.*]] = fmul float [[TMP15]], [[T]]
-; CHECK-NEXT:    [[TMP17:%.*]] = fpext float [[TMP16]] to double
-; CHECK-NEXT:    [[TMP18:%.*]] = fadd double [[TMP17]], 6.000000e+00
-; CHECK-NEXT:    [[TMP19:%.*]] = fptosi double [[TMP18]] to i8
-; CHECK-NEXT:    [[TMP20:%.*]] = getelementptr inbounds i8, i8* [[A]], i64 2
-; CHECK-NEXT:    store i8 [[TMP19]], i8* [[TMP20]], align 1
+; CHECK-NEXT:    [[TMP2:%.*]] = bitcast float* [[TMP1]] to <2 x float>*
+; CHECK-NEXT:    [[TMP3:%.*]] = load <2 x float>, <2 x float>* [[TMP2]], align 4
+; CHECK-NEXT:    [[TMP4:%.*]] = insertelement <2 x float> poison, float [[T:%.*]], i32 0
+; CHECK-NEXT:    [[TMP5:%.*]] = insertelement <2 x float> [[TMP4]], float [[T]], i32 1
+; CHECK-NEXT:    [[TMP6:%.*]] = fmul <2 x float> [[TMP3]], [[TMP5]]
+; CHECK-NEXT:    [[TMP7:%.*]] = fpext <2 x float> [[TMP6]] to <2 x double>
+; CHECK-NEXT:    [[TMP8:%.*]] = fadd <2 x double> [[TMP7]], <double 4.000000e+00, double 5.000000e+00>
+; CHECK-NEXT:    [[TMP9:%.*]] = fptosi <2 x double> [[TMP8]] to <2 x i8>
+; CHECK-NEXT:    [[TMP10:%.*]] = bitcast i8* [[A:%.*]] to <2 x i8>*
+; CHECK-NEXT:    store <2 x i8> [[TMP9]], <2 x i8>* [[TMP10]], align 1
+; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds float, float* [[B]], i64 12
+; CHECK-NEXT:    [[TMP12:%.*]] = load float, float* [[TMP11]], align 4
+; CHECK-NEXT:    [[TMP13:%.*]] = fmul float [[TMP12]], [[T]]
+; CHECK-NEXT:    [[TMP14:%.*]] = fpext float [[TMP13]] to double
+; CHECK-NEXT:    [[TMP15:%.*]] = fadd double [[TMP14]], 6.000000e+00
+; CHECK-NEXT:    [[TMP16:%.*]] = fptosi double [[TMP15]] to i8
+; CHECK-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i8, i8* [[A]], i64 2
+; CHECK-NEXT:    store i8 [[TMP16]], i8* [[TMP17]], align 1
 ; CHECK-NEXT:    ret i32 undef
 ;
   %1 = getelementptr inbounds float, float* %B, i64 10
   %2 = load float, float* %1, align 4
   %3 = fmul float %2, %T
   %4 = fpext float %3 to double
   %5 = fadd double %4, 4.000000e+00
   %6 = fptosi double %5 to i8

llvm/test/Transforms/SLPVectorizer/X86/pr49933.ll

 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
 ; RUN: opt < %s -slp-vectorizer -S -mtriple=x86_64-- -mcpu=skylake-avx512 | FileCheck %s
-; These code should be fully vectorized by D57059 patch
 
 define void @foo(i8* noalias nocapture %t0, i8* noalias nocapture readonly %t1) {
 ; CHECK-LABEL: @foo(
-; CHECK-NEXT:    [[T3:%.*]] = load i8, i8* [[T1:%.*]], align 1, !tbaa [[TBAA0:![0-9]+]]
-; CHECK-NEXT:    [[T4:%.*]] = icmp ult i8 [[T3]], 64
-; CHECK-NEXT:    [[T5:%.*]] = sub i8 0, [[T3]]
-; CHECK-NEXT:    [[T6:%.*]] = select i1 [[T4]], i8 [[T3]], i8 [[T5]]
-; CHECK-NEXT:    store i8 [[T6]], i8* [[T0:%.*]], align 1, !tbaa [[TBAA0]]
-; CHECK-NEXT:    [[T7:%.*]] = getelementptr inbounds i8, i8* [[T1]], i64 1
-; CHECK-NEXT:    [[T8:%.*]] = load i8, i8* [[T7]], align 1, !tbaa [[TBAA0]]
-; CHECK-NEXT:    [[T9:%.*]] = icmp ult i8 [[T8]], 64
-; CHECK-NEXT:    [[T10:%.*]] = sub i8 0, [[T8]]
-; CHECK-NEXT:    [[T11:%.*]] = select i1 [[T9]], i8 [[T8]], i8 [[T10]]
-; CHECK-NEXT:    [[T12:%.*]] = getelementptr inbounds i8, i8* [[T0]], i64 1
-; CHECK-NEXT:    store i8 [[T11]], i8* [[T12]], align 1, !tbaa [[TBAA0]]
-; CHECK-NEXT:    [[T13:%.*]] = getelementptr inbounds i8, i8* [[T1]], i64 2
-; CHECK-NEXT:    [[T14:%.*]] = load i8, i8* [[T13]], align 1, !tbaa [[TBAA0]]
-; CHECK-NEXT:    [[T15:%.*]] = icmp ult i8 [[T14]], 64
-; CHECK-NEXT:    [[T16:%.*]] = sub i8 0, [[T14]]
-; CHECK-NEXT:    [[T17:%.*]] = select i1 [[T15]], i8 [[T14]], i8 [[T16]]
-; CHECK-NEXT:    [[T18:%.*]] = getelementptr inbounds i8, i8* [[T0]], i64 2
-; CHECK-NEXT:    store i8 [[T17]], i8* [[T18]], align 1, !tbaa [[TBAA0]]
-; CHECK-NEXT:    [[T19:%.*]] = getelementptr inbounds i8, i8* [[T1]], i64 3
-; CHECK-NEXT:    [[T20:%.*]] = load i8, i8* [[T19]], align 1, !tbaa [[TBAA0]]
-; CHECK-NEXT:    [[T21:%.*]] = icmp ult i8 [[T20]], 64
-; CHECK-NEXT:    [[T22:%.*]] = sub i8 0, [[T20]]
-; CHECK-NEXT:    [[T23:%.*]] = select i1 [[T21]], i8 [[T20]], i8 [[T22]]
-; CHECK-NEXT:    [[T24:%.*]] = getelementptr inbounds i8, i8* [[T0]], i64 3
-; CHECK-NEXT:    store i8 [[T23]], i8* [[T24]], align 1, !tbaa [[TBAA0]]
-; CHECK-NEXT:    [[T25:%.*]] = getelementptr inbounds i8, i8* [[T1]], i64 4
-; CHECK-NEXT:    [[T26:%.*]] = load i8, i8* [[T25]], align 1, !tbaa [[TBAA0]]
-; CHECK-NEXT:    [[T27:%.*]] = icmp ult i8 [[T26]], 64
-; CHECK-NEXT:    [[T28:%.*]] = sub i8 0, [[T26]]
-; CHECK-NEXT:    [[T29:%.*]] = select i1 [[T27]], i8 [[T26]], i8 [[T28]]
-; CHECK-NEXT:    [[T30:%.*]] = getelementptr inbounds i8, i8* [[T0]], i64 4
-; CHECK-NEXT:    store i8 [[T29]], i8* [[T30]], align 1, !tbaa [[TBAA0]]
-; CHECK-NEXT:    [[T31:%.*]] = getelementptr inbounds i8, i8* [[T1]], i64 5
-; CHECK-NEXT:    [[T32:%.*]] = load i8, i8* [[T31]], align 1, !tbaa [[TBAA0]]
-; CHECK-NEXT:    [[T33:%.*]] = icmp ult i8 [[T32]], 64
-; CHECK-NEXT:    [[T34:%.*]] = sub i8 0, [[T32]]
-; CHECK-NEXT:    [[T35:%.*]] = select i1 [[T33]], i8 [[T32]], i8 [[T34]]
-; CHECK-NEXT:    [[T36:%.*]] = getelementptr inbounds i8, i8* [[T0]], i64 5
-; CHECK-NEXT:    store i8 [[T35]], i8* [[T36]], align 1, !tbaa [[TBAA0]]
-; CHECK-NEXT:    [[T37:%.*]] = getelementptr inbounds i8, i8* [[T1]], i64 6
-; CHECK-NEXT:    [[T38:%.*]] = load i8, i8* [[T37]], align 1, !tbaa [[TBAA0]]
-; CHECK-NEXT:    [[T39:%.*]] = icmp ult i8 [[T38]], 64
-; CHECK-NEXT:    [[T40:%.*]] = sub i8 0, [[T38]]
-; CHECK-NEXT:    [[T41:%.*]] = select i1 [[T39]], i8 [[T38]], i8 [[T40]]
-; CHECK-NEXT:    [[T42:%.*]] = getelementptr inbounds i8, i8* [[T0]], i64 6
-; CHECK-NEXT:    store i8 [[T41]], i8* [[T42]], align 1, !tbaa [[TBAA0]]
-; CHECK-NEXT:    [[T43:%.*]] = getelementptr inbounds i8, i8* [[T1]], i64 7
-; CHECK-NEXT:    [[T44:%.*]] = load i8, i8* [[T43]], align 1, !tbaa [[TBAA0]]
-; CHECK-NEXT:    [[T45:%.*]] = icmp ult i8 [[T44]], 64
-; CHECK-NEXT:    [[T46:%.*]] = sub i8 0, [[T44]]
-; CHECK-NEXT:    [[T47:%.*]] = select i1 [[T45]], i8 [[T44]], i8 [[T46]]
-; CHECK-NEXT:    [[T48:%.*]] = getelementptr inbounds i8, i8* [[T0]], i64 7
-; CHECK-NEXT:    store i8 [[T47]], i8* [[T48]], align 1, !tbaa [[TBAA0]]
+; CHECK-NEXT:    [[TMP1:%.*]] = bitcast i8* [[T1:%.*]] to <8 x i8>*
+; CHECK-NEXT:    [[TMP2:%.*]] = load <8 x i8>, <8 x i8>* [[TMP1]], align 1, !tbaa [[TBAA0:![0-9]+]]
+; CHECK-NEXT:    [[TMP3:%.*]] = icmp ult <8 x i8> [[TMP2]], <i8 64, i8 64, i8 64, i8 64, i8 64, i8 64, i8 64, i8 64>
+; CHECK-NEXT:    [[TMP4:%.*]] = sub <8 x i8> zeroinitializer, [[TMP2]]
+; CHECK-NEXT:    [[TMP5:%.*]] = select <8 x i1> [[TMP3]], <8 x i8> [[TMP2]], <8 x i8> [[TMP4]]
+; CHECK-NEXT:    [[TMP6:%.*]] = bitcast i8* [[T0:%.*]] to <8 x i8>*
+; CHECK-NEXT:    store <8 x i8> [[TMP5]], <8 x i8>* [[TMP6]], align 1, !tbaa [[TBAA0]]
 ; CHECK-NEXT:    ret void
 ;
   %t3 = load i8, i8* %t1, align 1, !tbaa !3
   %t4 = icmp ult i8 %t3, 64
   %t5 = sub i8 0, %t3
   %t6 = select i1 %t4, i8 %t3, i8 %t5
   store i8 %t6, i8* %t0, align 1, !tbaa !3
   %t7 = getelementptr inbounds i8, i8* %t1, i64 1

llvm/test/Transforms/SLPVectorizer/X86/remark_not_all_parts.ll

 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
 ; RUN: opt -S -mtriple=x86_64-pc-linux-gnu -mcpu=generic -slp-vectorizer -pass-remarks-output=%t < %s | FileCheck %s
 ; RUN: FileCheck --input-file=%t --check-prefix=YAML %s
 
 define i32 @foo(i32* nocapture readonly %diff) #0 {
 ; CHECK-LABEL: @foo(
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[M2:%.*]] = alloca [8 x [8 x i32]], align 16
 ; CHECK-NEXT:    [[TMP0:%.*]] = bitcast [8 x [8 x i32]]* [[M2]] to i8*
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK:       for.body:
 ; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
 ; CHECK-NEXT:    [[A_088:%.*]] = phi i32 [ 0, [[ENTRY]] ], [ [[ADD24:%.*]], [[FOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = shl i64 [[INDVARS_IV]], 3
 ; CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[DIFF:%.*]], i64 [[TMP1]]
-; CHECK-NEXT:    [[TMP2:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
-; CHECK-NEXT:    [[TMP3:%.*]] = or i64 [[TMP1]], 4
-; CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i32, i32* [[DIFF]], i64 [[TMP3]]
-; CHECK-NEXT:    [[TMP4:%.*]] = load i32, i32* [[ARRAYIDX2]], align 4
-; CHECK-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP4]], [[TMP2]]
+; CHECK-NEXT:    [[TMP2:%.*]] = or i64 [[TMP1]], 4
+; CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i32, i32* [[DIFF]], i64 [[TMP2]]
 ; CHECK-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [8 x [8 x i32]], [8 x [8 x i32]]* [[M2]], i64 0, i64 [[INDVARS_IV]], i64 0
-; CHECK-NEXT:    store i32 [[ADD3]], i32* [[ARRAYIDX6]], align 16
-; CHECK-NEXT:    [[ADD10:%.*]] = add nsw i32 [[ADD3]], [[A_088]]
-; CHECK-NEXT:    [[TMP5:%.*]] = or i64 [[TMP1]], 1
-; CHECK-NEXT:    [[ARRAYIDX13:%.*]] = getelementptr inbounds i32, i32* [[DIFF]], i64 [[TMP5]]
-; CHECK-NEXT:    [[TMP6:%.*]] = load i32, i32* [[ARRAYIDX13]], align 4
-; CHECK-NEXT:    [[TMP7:%.*]] = or i64 [[TMP1]], 5
-; CHECK-NEXT:    [[ARRAYIDX16:%.*]] = getelementptr inbounds i32, i32* [[DIFF]], i64 [[TMP7]]
-; CHECK-NEXT:    [[TMP8:%.*]] = load i32, i32* [[ARRAYIDX16]], align 4
-; CHECK-NEXT:    [[ADD17:%.*]] = add nsw i32 [[TMP8]], [[TMP6]]
-; CHECK-NEXT:    [[ARRAYIDX20:%.*]] = getelementptr inbounds [8 x [8 x i32]], [8 x [8 x i32]]* [[M2]], i64 0, i64 [[INDVARS_IV]], i64 1
-; CHECK-NEXT:    store i32 [[ADD17]], i32* [[ARRAYIDX20]], align 4
-; CHECK-NEXT:    [[ADD24]] = add nsw i32 [[ADD10]], [[ADD17]]
+; CHECK-NEXT:    [[TMP3:%.*]] = bitcast i32* [[ARRAYIDX]] to <2 x i32>*
+; CHECK-NEXT:    [[TMP4:%.*]] = load <2 x i32>, <2 x i32>* [[TMP3]], align 4
+; CHECK-NEXT:    [[TMP5:%.*]] = bitcast i32* [[ARRAYIDX2]] to <2 x i32>*
+; CHECK-NEXT:    [[TMP6:%.*]] = load <2 x i32>, <2 x i32>* [[TMP5]], align 4
+; CHECK-NEXT:    [[TMP7:%.*]] = add nsw <2 x i32> [[TMP6]], [[TMP4]]
+; CHECK-NEXT:    [[TMP8:%.*]] = extractelement <2 x i32> [[TMP7]], i32 0
+; CHECK-NEXT:    [[ADD10:%.*]] = add nsw i32 [[TMP8]], [[A_088]]
+; CHECK-NEXT:    [[TMP9:%.*]] = bitcast i32* [[ARRAYIDX6]] to <2 x i32>*
+; CHECK-NEXT:    store <2 x i32> [[TMP7]], <2 x i32>* [[TMP9]], align 16
+; CHECK-NEXT:    [[TMP10:%.*]] = extractelement <2 x i32> [[TMP7]], i32 1
+; CHECK-NEXT:    [[ADD24]] = add nsw i32 [[ADD10]], [[TMP10]]
 ; CHECK-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
 ; CHECK-NEXT:    [[EXITCOND:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], 8
 ; CHECK-NEXT:    br i1 [[EXITCOND]], label [[FOR_END:%.*]], label [[FOR_BODY]]
 ; CHECK:       for.end:
 ; CHECK-NEXT:    [[ARRAYDECAY:%.*]] = getelementptr inbounds [8 x [8 x i32]], [8 x [8 x i32]]* [[M2]], i64 0, i64 0
 ; CHECK-NEXT:    ret i32 [[ADD24]]
 ;
 entry:
for.body:                                         ; preds = %for.body, %entry
   %arrayidx16 = getelementptr inbounds i32, i32* %diff, i64 %7
   %8 = load i32, i32* %arrayidx16, align 4
   %add17 = add nsw i32 %8, %6
   %arrayidx20 = getelementptr inbounds [8 x [8 x i32]], [8 x [8 x i32]]* %m2, i64 0, i64 %indvars.iv, i64 1
   store i32 %add17, i32* %arrayidx20, align 4
   %add24 = add nsw i32 %add10, %add17
 
   ; YAML:      Pass:            slp-vectorizer
-  ; YAML-NEXT: Name:            NotPossible
+  ; YAML-NEXT: Name:            StoresVectorized
   ; YAML-NEXT: Function:        foo
   ; YAML-NEXT: Args:
-  ; YAML-NEXT:   - String:          'Cannot SLP vectorize list: vectorization was impossible'
-  ; YAML-NEXT:   - String:          ' with available vectorization factors'
+  ; YAML-NEXT:   - String:          'Stores SLP vectorized with cost '
+  ; YAML-NEXT:   - Cost:            '-1'
+  ; YAML-NEXT:   - String:          ' and with tree size '
+  ; YAML-NEXT:   - TreeSize:        '4'
+
+  ; YAML:      Pass:            slp-vectorizer
+  ; YAML-NEXT: Name:            UnsupportedType
+  ; YAML-NEXT: Function:        foo
+  ; YAML-NEXT: Args:
+  ; YAML-NEXT:   - String:          'Cannot SLP vectorize list: type '
+  ; YAML-NEXT:   - String:          '<2 x i32> is unsupported by vectorizer'
 
   %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
   %exitcond = icmp eq i64 %indvars.iv.next, 8
   br i1 %exitcond, label %for.end, label %for.body
 
 for.end:                                          ; preds = %for.body
   %arraydecay = getelementptr inbounds [8 x [8 x i32]], [8 x [8 x i32]]* %m2, i64 0, i64 0
   ret i32 %add24
 }
 

llvm/test/Transforms/SLPVectorizer/X86/reorder_phi.ll

 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
 ; RUN: opt < %s -basic-aa -slp-vectorizer  -S -mtriple=x86_64-unknown -mcpu=corei7-avx | FileCheck %s
 
 %struct.complex = type { float, float }
 
 define  void @foo (%struct.complex* %A, %struct.complex* %B, %struct.complex* %Result) {
 ; CHECK-LABEL: @foo(
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[TMP0:%.*]] = add i64 256, 0
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[TMP1:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[TMP20:%.*]], [[LOOP]] ]
-; CHECK-NEXT:    [[TMP2:%.*]] = phi float [ 0.000000e+00, [[ENTRY]] ], [ [[TMP19:%.*]], [[LOOP]] ]
-; CHECK-NEXT:    [[TMP3:%.*]] = phi float [ 0.000000e+00, [[ENTRY]] ], [ [[TMP18:%.*]], [[LOOP]] ]
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_COMPLEX:%.*]], %struct.complex* [[A:%.*]], i64 [[TMP1]], i32 0
+; CHECK-NEXT:    [[TMP2:%.*]] = phi <2 x float> [ zeroinitializer, [[ENTRY]] ], [ [[TMP19:%.*]], [[LOOP]] ]
+; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_COMPLEX:%.*]], %struct.complex* [[A:%.*]], i64 [[TMP1]], i32 0
+; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_COMPLEX]], %struct.complex* [[B:%.*]], i64 [[TMP1]], i32 0
 ; CHECK-NEXT:    [[TMP5:%.*]] = load float, float* [[TMP4]], align 4
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_COMPLEX]], %struct.complex* [[A]], i64 [[TMP1]], i32 1
+; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_COMPLEX]], %struct.complex* [[B]], i64 [[TMP1]], i32 1
 ; CHECK-NEXT:    [[TMP7:%.*]] = load float, float* [[TMP6]], align 4
-; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT_COMPLEX]], %struct.complex* [[B:%.*]], i64 [[TMP1]], i32 0
-; CHECK-NEXT:    [[TMP9:%.*]] = load float, float* [[TMP8]], align 4
-; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [[STRUCT_COMPLEX]], %struct.complex* [[B]], i64 [[TMP1]], i32 1
-; CHECK-NEXT:    [[TMP11:%.*]] = load float, float* [[TMP10]], align 4
-; CHECK-NEXT:    [[TMP12:%.*]] = fmul float [[TMP5]], [[TMP9]]
-; CHECK-NEXT:    [[TMP13:%.*]] = fmul float [[TMP7]], [[TMP11]]
-; CHECK-NEXT:    [[TMP14:%.*]] = fsub float [[TMP12]], [[TMP13]]
-; CHECK-NEXT:    [[TMP15:%.*]] = fmul float [[TMP7]], [[TMP9]]
-; CHECK-NEXT:    [[TMP16:%.*]] = fmul float [[TMP5]], [[TMP11]]
-; CHECK-NEXT:    [[TMP17:%.*]] = fadd float [[TMP15]], [[TMP16]]
-; CHECK-NEXT:    [[TMP18]] = fadd float [[TMP3]], [[TMP14]]
-; CHECK-NEXT:    [[TMP19]] = fadd float [[TMP2]], [[TMP17]]
+; CHECK-NEXT:    [[TMP8:%.*]] = bitcast float* [[TMP3]] to <2 x float>*
+; CHECK-NEXT:    [[TMP9:%.*]] = load <2 x float>, <2 x float>* [[TMP8]], align 4
+; CHECK-NEXT:    [[TMP10:%.*]] = insertelement <2 x float> poison, float [[TMP5]], i32 0
+; CHECK-NEXT:    [[TMP11:%.*]] = insertelement <2 x float> [[TMP10]], float [[TMP5]], i32 1
+; CHECK-NEXT:    [[TMP12:%.*]] = fmul <2 x float> [[TMP9]], [[TMP11]]
+; CHECK-NEXT:    [[TMP13:%.*]] = insertelement <2 x float> poison, float [[TMP7]], i32 0
+; CHECK-NEXT:    [[TMP14:%.*]] = insertelement <2 x float> [[TMP13]], float [[TMP7]], i32 1
+; CHECK-NEXT:    [[TMP15:%.*]] = fmul <2 x float> [[TMP9]], [[TMP14]]
+; CHECK-NEXT:    [[SHUFFLE:%.*]] = shufflevector <2 x float> [[TMP15]], <2 x float> poison, <2 x i32> <i32 1, i32 0>
+; CHECK-NEXT:    [[TMP16:%.*]] = fsub <2 x float> [[TMP12]], [[SHUFFLE]]
+; CHECK-NEXT:    [[TMP17:%.*]] = fadd <2 x float> [[TMP12]], [[SHUFFLE]]
+; CHECK-NEXT:    [[TMP18:%.*]] = shufflevector <2 x float> [[TMP16]], <2 x float> [[TMP17]], <2 x i32> <i32 0, i32 3>
+; CHECK-NEXT:    [[TMP19]] = fadd <2 x float> [[TMP2]], [[TMP18]]
 ; CHECK-NEXT:    [[TMP20]] = add nuw nsw i64 [[TMP1]], 1
 ; CHECK-NEXT:    [[TMP21:%.*]] = icmp eq i64 [[TMP20]], [[TMP0]]
 ; CHECK-NEXT:    br i1 [[TMP21]], label [[EXIT:%.*]], label [[LOOP]]
 ; CHECK:       exit:
 ; CHECK-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [[STRUCT_COMPLEX]], %struct.complex* [[RESULT:%.*]], i32 0, i32 0
-; CHECK-NEXT:    store float [[TMP18]], float* [[TMP22]], align 4
-; CHECK-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [[STRUCT_COMPLEX]], %struct.complex* [[RESULT]], i32 0, i32 1
-; CHECK-NEXT:    store float [[TMP19]], float* [[TMP23]], align 4
+; CHECK-NEXT:    [[TMP23:%.*]] = bitcast float* [[TMP22]] to <2 x float>*
+; CHECK-NEXT:    store <2 x float> [[TMP19]], <2 x float>* [[TMP23]], align 4
 ; CHECK-NEXT:    ret void
 ;
 entry:
   %0 = add i64 256, 0
   br label %loop
 
 loop:
   %1 = phi i64 [ 0, %entry ], [ %20, %loop ]

llvm/test/Transforms/SLPVectorizer/X86/saxpy.ll

 }
 
 ; Make sure we don't crash on this one.
 define void @SAXPY_crash(i32* noalias nocapture %x, i32* noalias nocapture %y, i64 %i) {
 ; CHECK-LABEL: @SAXPY_crash(
 ; CHECK-NEXT:    [[TMP1:%.*]] = add i64 [[I:%.*]], 1
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, i32* [[X:%.*]], i64 [[TMP1]]
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, i32* [[Y:%.*]], i64 [[TMP1]]
-; CHECK-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
-; CHECK-NEXT:    [[TMP5:%.*]] = add nsw i32 undef, [[TMP4]]
-; CHECK-NEXT:    store i32 [[TMP5]], i32* [[TMP2]], align 4
-; CHECK-NEXT:    [[TMP6:%.*]] = add i64 [[I]], 2
-; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, i32* [[X]], i64 [[TMP6]]
-; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, i32* [[Y]], i64 [[TMP6]]
-; CHECK-NEXT:    [[TMP9:%.*]] = load i32, i32* [[TMP8]], align 4
-; CHECK-NEXT:    [[TMP10:%.*]] = add nsw i32 undef, [[TMP9]]
-; CHECK-NEXT:    store i32 [[TMP10]], i32* [[TMP7]], align 4
+; CHECK-NEXT:    [[TMP4:%.*]] = bitcast i32* [[TMP3]] to <2 x i32>*
+; CHECK-NEXT:    [[TMP5:%.*]] = load <2 x i32>, <2 x i32>* [[TMP4]], align 4
+; CHECK-NEXT:    [[TMP6:%.*]] = add nsw <2 x i32> undef, [[TMP5]]
+; CHECK-NEXT:    [[TMP7:%.*]] = bitcast i32* [[TMP2]] to <2 x i32>*
+; CHECK-NEXT:    store <2 x i32> [[TMP6]], <2 x i32>* [[TMP7]], align 4
 ; CHECK-NEXT:    ret void
 ;
   %1 = add i64 %i, 1
   %2 = getelementptr inbounds i32, i32* %x, i64 %1
   %3 = getelementptr inbounds i32, i32* %y, i64 %1
   %4 = load i32, i32* %3, align 4
   %5 = add nsw i32 undef, %4
   store i32 %5, i32* %2, align 4
   %6 = add i64 %i, 2
   %7 = getelementptr inbounds i32, i32* %x, i64 %6
   %8 = getelementptr inbounds i32, i32* %y, i64 %6
   %9 = load i32, i32* %8, align 4
   %10 = add nsw i32 undef, %9
   store i32 %10, i32* %7, align 4
   ret void
 }

llvm/test/Transforms/SLPVectorizer/X86/schedule-bundle.ll

 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
 ; RUN: opt -S -slp-vectorizer -slp-vectorizer -mcpu=bdver1 < %s | FileCheck %s
 
 target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
 target triple = "x86_64-unknown-linux-gnu"
 
 @a = common local_unnamed_addr global [1 x i32] zeroinitializer, align 4
 @b = common local_unnamed_addr global [1 x i32] zeroinitializer, align 4
 
 define i32 @slp_schedule_bundle() local_unnamed_addr #0 {
 ; CHECK-LABEL: @slp_schedule_bundle(
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[TMP0:%.*]] = load <4 x i32>, <4 x i32>* bitcast ([1 x i32]* @b to <4 x i32>*), align 4
 ; CHECK-NEXT:    [[TMP1:%.*]] = lshr <4 x i32> [[TMP0]], <i32 31, i32 31, i32 31, i32 31>
 ; CHECK-NEXT:    [[TMP2:%.*]] = xor <4 x i32> [[TMP1]], <i32 1, i32 1, i32 1, i32 1>
 ; CHECK-NEXT:    store <4 x i32> [[TMP2]], <4 x i32>* bitcast ([1 x i32]* @a to <4 x i32>*), align 4
-; CHECK-NEXT:    [[TMP3:%.*]] = load i32, i32* getelementptr ([1 x i32], [1 x i32]* @b, i64 4, i64 0), align 4
-; CHECK-NEXT:    [[DOTLOBIT_4:%.*]] = lshr i32 [[TMP3]], 31
-; CHECK-NEXT:    [[DOTLOBIT_NOT_4:%.*]] = xor i32 [[DOTLOBIT_4]], 1
-; CHECK-NEXT:    store i32 [[DOTLOBIT_NOT_4]], i32* getelementptr ([1 x i32], [1 x i32]* @a, i64 4, i64 0), align 4
-; CHECK-NEXT:    [[TMP4:%.*]] = load i32, i32* getelementptr ([1 x i32], [1 x i32]* @b, i64 5, i64 0), align 4
-; CHECK-NEXT:    [[DOTLOBIT_5:%.*]] = lshr i32 [[TMP4]], 31
-; CHECK-NEXT:    [[DOTLOBIT_NOT_5:%.*]] = xor i32 [[DOTLOBIT_5]], 1
-; CHECK-NEXT:    store i32 [[DOTLOBIT_NOT_5]], i32* getelementptr ([1 x i32], [1 x i32]* @a, i64 5, i64 0), align 4
+; CHECK-NEXT:    [[TMP3:%.*]] = load <2 x i32>, <2 x i32>* bitcast (i32* getelementptr ([1 x i32], [1 x i32]* @b, i64 4, i64 0) to <2 x i32>*), align 4
+; CHECK-NEXT:    [[TMP4:%.*]] = lshr <2 x i32> [[TMP3]], <i32 31, i32 31>
+; CHECK-NEXT:    [[TMP5:%.*]] = xor <2 x i32> [[TMP4]], <i32 1, i32 1>
+; CHECK-NEXT:    store <2 x i32> [[TMP5]], <2 x i32>* bitcast (i32* getelementptr ([1 x i32], [1 x i32]* @a, i64 4, i64 0) to <2 x i32>*), align 4
 ; CHECK-NEXT:    ret i32 undef
 ;
 entry:
   %0 = load i32, i32* getelementptr inbounds ([1 x i32], [1 x i32]* @b, i64 0, i64 0), align 4
   %.lobit = lshr i32 %0, 31
   %.lobit.not = xor i32 %.lobit, 1
   store i32 %.lobit.not, i32* getelementptr inbounds ([1 x i32], [1 x i32]* @a, i64 0, i64 0), align 4
   %1 = load i32, i32* getelementptr inbounds ([1 x i32], [1 x i32]* @b, i64 1, i64 0), align 4

llvm/test/Transforms/SLPVectorizer/X86/simple-loop.ll

._crit_edge:                                      ; preds = %.lr.ph, %0
   ret i32 undef
 }
 
 define i32 @unrollable(i32* %in, i32* %out, i64 %n) nounwind ssp uwtable {
 ; CHECK-LABEL: @unrollable(
 ; CHECK-NEXT:    [[TMP1:%.*]] = icmp eq i64 [[N:%.*]], 0
 ; CHECK-NEXT:    br i1 [[TMP1]], label [[DOT_CRIT_EDGE:%.*]], label [[DOTLR_PH:%.*]]
 ; CHECK:       .lr.ph:
-; CHECK-NEXT:    [[I_019:%.*]] = phi i64 [ [[TMP26:%.*]], [[DOTLR_PH]] ], [ 0, [[TMP0:%.*]] ]
+; CHECK-NEXT:    [[I_019:%.*]] = phi i64 [ [[TMP18:%.*]], [[DOTLR_PH]] ], [ 0, [[TMP0:%.*]] ]
 ; CHECK-NEXT:    [[TMP2:%.*]] = shl i64 [[I_019]], 2
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, i32* [[IN:%.*]], i64 [[TMP2]]
-; CHECK-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
-; CHECK-NEXT:    [[TMP5:%.*]] = or i64 [[TMP2]], 1
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, i32* [[IN]], i64 [[TMP5]]
-; CHECK-NEXT:    [[TMP7:%.*]] = load i32, i32* [[TMP6]], align 4
-; CHECK-NEXT:    [[TMP8:%.*]] = or i64 [[TMP2]], 2
-; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i32, i32* [[IN]], i64 [[TMP8]]
-; CHECK-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
-; CHECK-NEXT:    [[TMP11:%.*]] = or i64 [[TMP2]], 3
-; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i32, i32* [[IN]], i64 [[TMP11]]
-; CHECK-NEXT:    [[TMP13:%.*]] = load i32, i32* [[TMP12]], align 4
-; CHECK-NEXT:    [[TMP14:%.*]] = mul i32 [[TMP4]], 7
-; CHECK-NEXT:    [[TMP15:%.*]] = add i32 [[TMP14]], 7
-; CHECK-NEXT:    [[TMP16:%.*]] = mul i32 [[TMP7]], 7
-; CHECK-NEXT:    [[TMP17:%.*]] = add i32 [[TMP16]], 14
-; CHECK-NEXT:    [[TMP18:%.*]] = mul i32 [[TMP10]], 7
-; CHECK-NEXT:    [[TMP19:%.*]] = add i32 [[TMP18]], 21
-; CHECK-NEXT:    [[TMP20:%.*]] = mul i32 [[TMP13]], 7
-; CHECK-NEXT:    [[TMP21:%.*]] = add i32 [[TMP20]], 28
-; CHECK-NEXT:    [[TMP22:%.*]] = getelementptr inbounds i32, i32* [[OUT:%.*]], i64 [[TMP2]]
-; CHECK-NEXT:    store i32 [[TMP15]], i32* [[TMP22]], align 4
-; CHECK-NEXT:    [[TMP23:%.*]] = getelementptr inbounds i32, i32* [[OUT]], i64 [[TMP5]]
-; CHECK-NEXT:    store i32 [[TMP17]], i32* [[TMP23]], align 4
+; CHECK-NEXT:    [[TMP4:%.*]] = or i64 [[TMP2]], 2
+; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, i32* [[IN]], i64 [[TMP4]]
+; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, i32* [[OUT:%.*]], i64 [[TMP2]]
+; CHECK-NEXT:    [[TMP7:%.*]] = bitcast i32* [[TMP3]] to <2 x i32>*
+; CHECK-NEXT:    [[TMP8:%.*]] = load <2 x i32>, <2 x i32>* [[TMP7]], align 4
+; CHECK-NEXT:    [[TMP9:%.*]] = mul <2 x i32> [[TMP8]], <i32 7, i32 7>
+; CHECK-NEXT:    [[TMP10:%.*]] = add <2 x i32> [[TMP9]], <i32 7, i32 14>
+; CHECK-NEXT:    [[TMP11:%.*]] = bitcast i32* [[TMP6]] to <2 x i32>*
+; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i32, i32* [[OUT]], i64 [[TMP4]]
+; CHECK-NEXT:    [[TMP13:%.*]] = bitcast i32* [[TMP5]] to <2 x i32>*
+; CHECK-NEXT:    [[TMP14:%.*]] = load <2 x i32>, <2 x i32>* [[TMP13]], align 4
+; CHECK-NEXT:    [[TMP15:%.*]] = mul <2 x i32> [[TMP14]], <i32 7, i32 7>
+; CHECK-NEXT:    [[TMP16:%.*]] = add <2 x i32> [[TMP15]], <i32 21, i32 28>
+; CHECK-NEXT:    store <2 x i32> [[TMP10]], <2 x i32>* [[TMP11]], align 4
 ; CHECK-NEXT:    [[BARRIER:%.*]] = call i32 @goo(i32 0)
-; CHECK-NEXT:    [[TMP24:%.*]] = getelementptr inbounds i32, i32* [[OUT]], i64 [[TMP8]]
-; CHECK-NEXT:    store i32 [[TMP19]], i32* [[TMP24]], align 4
-; CHECK-NEXT:    [[TMP25:%.*]] = getelementptr inbounds i32, i32* [[OUT]], i64 [[TMP11]]
-; CHECK-NEXT:    store i32 [[TMP21]], i32* [[TMP25]], align 4
-; CHECK-NEXT:    [[TMP26]] = add i64 [[I_019]], 1
-; CHECK-NEXT:    [[EXITCOND:%.*]] = icmp eq i64 [[TMP26]], [[N]]
+; CHECK-NEXT:    [[TMP17:%.*]] = bitcast i32* [[TMP12]] to <2 x i32>*
+; CHECK-NEXT:    store <2 x i32> [[TMP16]], <2 x i32>* [[TMP17]], align 4
+; CHECK-NEXT:    [[TMP18]] = add i64 [[I_019]], 1
+; CHECK-NEXT:    [[EXITCOND:%.*]] = icmp eq i64 [[TMP18]], [[N]]
 ; CHECK-NEXT:    br i1 [[EXITCOND]], label [[DOT_CRIT_EDGE]], label [[DOTLR_PH]]
 ; CHECK:       ._crit_edge:
 ; CHECK-NEXT:    ret i32 undef
 ;
   %1 = icmp eq i64 %n, 0
   br i1 %1, label %._crit_edge, label %.lr.ph
 
 .lr.ph:                                           ; preds = %0, %.lr.ph

llvm/test/Transforms/SLPVectorizer/X86/sitofp-inseltpoison.ll

;
   ret void
 }
 
 ;
 ; SITOFP to vXf32
 ;
 
 define void @sitofp_2i64_2f32() #0 {
-; CHECK-LABEL: @sitofp_2i64_2f32(
-; CHECK-NEXT:    [[LD0:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 0), align 64
-; CHECK-NEXT:    [[LD1:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 1), align 8
-; CHECK-NEXT:    [[CVT0:%.*]] = sitofp i64 [[LD0]] to float
-; CHECK-NEXT:    [[CVT1:%.*]] = sitofp i64 [[LD1]] to float
-; CHECK-NEXT:    store float [[CVT0]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 0), align 64
-; CHECK-NEXT:    store float [[CVT1]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 1), align 4
-; CHECK-NEXT:    ret void
+; SSE-LABEL: @sitofp_2i64_2f32(
+; SSE-NEXT:    [[LD0:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 0), align 64
+; SSE-NEXT:    [[LD1:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 1), align 8
+; SSE-NEXT:    [[CVT0:%.*]] = sitofp i64 [[LD0]] to float
+; SSE-NEXT:    [[CVT1:%.*]] = sitofp i64 [[LD1]] to float
+; SSE-NEXT:    store float [[CVT0]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 0), align 64
+; SSE-NEXT:    store float [[CVT1]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 1), align 4
+; SSE-NEXT:    ret void
+;
+; AVX256NODQ-LABEL: @sitofp_2i64_2f32(
+; AVX256NODQ-NEXT:    [[LD0:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 0), align 64
+; AVX256NODQ-NEXT:    [[LD1:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 1), align 8
+; AVX256NODQ-NEXT:    [[CVT0:%.*]] = sitofp i64 [[LD0]] to float
+; AVX256NODQ-NEXT:    [[CVT1:%.*]] = sitofp i64 [[LD1]] to float
+; AVX256NODQ-NEXT:    store float [[CVT0]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 0), align 64
+; AVX256NODQ-NEXT:    store float [[CVT1]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 1), align 4
+; AVX256NODQ-NEXT:    ret void
+;
+; AVX512-LABEL: @sitofp_2i64_2f32(
+; AVX512-NEXT:    [[TMP1:%.*]] = load <2 x i64>, <2 x i64>* bitcast ([8 x i64]* @src64 to <2 x i64>*), align 64
+; AVX512-NEXT:    [[TMP2:%.*]] = sitofp <2 x i64> [[TMP1]] to <2 x float>
+; AVX512-NEXT:    store <2 x float> [[TMP2]], <2 x float>* bitcast ([16 x float]* @dst32 to <2 x float>*), align 64
+; AVX512-NEXT:    ret void
+;
+; AVX256DQ-LABEL: @sitofp_2i64_2f32(
+; AVX256DQ-NEXT:    [[TMP1:%.*]] = load <2 x i64>, <2 x i64>* bitcast ([8 x i64]* @src64 to <2 x i64>*), align 64
+; AVX256DQ-NEXT:    [[TMP2:%.*]] = sitofp <2 x i64> [[TMP1]] to <2 x float>
+; AVX256DQ-NEXT:    store <2 x float> [[TMP2]], <2 x float>* bitcast ([16 x float]* @dst32 to <2 x float>*), align 64
+; AVX256DQ-NEXT:    ret void
 ;
   %ld0 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 0), align 64
   %ld1 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 1), align 8
   %cvt0 = sitofp i64 %ld0 to float
   %cvt1 = sitofp i64 %ld1 to float
   store float %cvt0, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 0), align 64
   store float %cvt1, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 1), align 4
   ret void

llvm/test/Transforms/SLPVectorizer/X86/sitofp.ll

;
   ret void
 }
 
 ;
 ; SITOFP to vXf32
 ;
 
 define void @sitofp_2i64_2f32() #0 {
-; CHECK-LABEL: @sitofp_2i64_2f32(
-; CHECK-NEXT:    [[LD0:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 0), align 64
-; CHECK-NEXT:    [[LD1:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 1), align 8
-; CHECK-NEXT:    [[CVT0:%.*]] = sitofp i64 [[LD0]] to float
-; CHECK-NEXT:    [[CVT1:%.*]] = sitofp i64 [[LD1]] to float
-; CHECK-NEXT:    store float [[CVT0]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 0), align 64
-; CHECK-NEXT:    store float [[CVT1]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 1), align 4
-; CHECK-NEXT:    ret void
+; SSE-LABEL: @sitofp_2i64_2f32(
+; SSE-NEXT:    [[LD0:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 0), align 64
+; SSE-NEXT:    [[LD1:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 1), align 8
+; SSE-NEXT:    [[CVT0:%.*]] = sitofp i64 [[LD0]] to float
+; SSE-NEXT:    [[CVT1:%.*]] = sitofp i64 [[LD1]] to float
+; SSE-NEXT:    store float [[CVT0]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 0), align 64
+; SSE-NEXT:    store float [[CVT1]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 1), align 4
+; SSE-NEXT:    ret void
+;
+; AVX256NODQ-LABEL: @sitofp_2i64_2f32(
+; AVX256NODQ-NEXT:    [[LD0:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 0), align 64
+; AVX256NODQ-NEXT:    [[LD1:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 1), align 8
+; AVX256NODQ-NEXT:    [[CVT0:%.*]] = sitofp i64 [[LD0]] to float
+; AVX256NODQ-NEXT:    [[CVT1:%.*]] = sitofp i64 [[LD1]] to float
+; AVX256NODQ-NEXT:    store float [[CVT0]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 0), align 64
+; AVX256NODQ-NEXT:    store float [[CVT1]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 1), align 4
+; AVX256NODQ-NEXT:    ret void
+;
+; AVX512-LABEL: @sitofp_2i64_2f32(
+; AVX512-NEXT:    [[TMP1:%.*]] = load <2 x i64>, <2 x i64>* bitcast ([8 x i64]* @src64 to <2 x i64>*), align 64
+; AVX512-NEXT:    [[TMP2:%.*]] = sitofp <2 x i64> [[TMP1]] to <2 x float>
+; AVX512-NEXT:    store <2 x float> [[TMP2]], <2 x float>* bitcast ([16 x float]* @dst32 to <2 x float>*), align 64
+; AVX512-NEXT:    ret void
+;
+; AVX256DQ-LABEL: @sitofp_2i64_2f32(
+; AVX256DQ-NEXT:    [[TMP1:%.*]] = load <2 x i64>, <2 x i64>* bitcast ([8 x i64]* @src64 to <2 x i64>*), align 64
+; AVX256DQ-NEXT:    [[TMP2:%.*]] = sitofp <2 x i64> [[TMP1]] to <2 x float>
+; AVX256DQ-NEXT:    store <2 x float> [[TMP2]], <2 x float>* bitcast ([16 x float]* @dst32 to <2 x float>*), align 64
+; AVX256DQ-NEXT:    ret void
 ;
   %ld0 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 0), align 64
   %ld1 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 1), align 8
   %cvt0 = sitofp i64 %ld0 to float
   %cvt1 = sitofp i64 %ld1 to float
   store float %cvt0, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 0), align 64
   store float %cvt1, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 1), align 4
   ret void

llvm/test/Transforms/SLPVectorizer/X86/uitofp.ll

;
   ret void
 }
 
 ;
 ; UITOFP to vXf32
 ;
 
 define void @uitofp_2i64_2f32() #0 {
-; CHECK-LABEL: @uitofp_2i64_2f32(
-; CHECK-NEXT:    [[LD0:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 0), align 64
-; CHECK-NEXT:    [[LD1:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 1), align 8
-; CHECK-NEXT:    [[CVT0:%.*]] = uitofp i64 [[LD0]] to float
-; CHECK-NEXT:    [[CVT1:%.*]] = uitofp i64 [[LD1]] to float
-; CHECK-NEXT:    store float [[CVT0]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 0), align 64
-; CHECK-NEXT:    store float [[CVT1]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 1), align 4
-; CHECK-NEXT:    ret void
+; SSE-LABEL: @uitofp_2i64_2f32(
+; SSE-NEXT:    [[LD0:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 0), align 64
+; SSE-NEXT:    [[LD1:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 1), align 8
+; SSE-NEXT:    [[CVT0:%.*]] = uitofp i64 [[LD0]] to float
+; SSE-NEXT:    [[CVT1:%.*]] = uitofp i64 [[LD1]] to float
+; SSE-NEXT:    store float [[CVT0]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 0), align 64
+; SSE-NEXT:    store float [[CVT1]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 1), align 4
+; SSE-NEXT:    ret void
+;
+; AVX1-LABEL: @uitofp_2i64_2f32(
+; AVX1-NEXT:    [[LD0:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 0), align 64
+; AVX1-NEXT:    [[LD1:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 1), align 8
+; AVX1-NEXT:    [[CVT0:%.*]] = uitofp i64 [[LD0]] to float
+; AVX1-NEXT:    [[CVT1:%.*]] = uitofp i64 [[LD1]] to float
+; AVX1-NEXT:    store float [[CVT0]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 0), align 64
+; AVX1-NEXT:    store float [[CVT1]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 1), align 4
+; AVX1-NEXT:    ret void
+;
+; AVX2-LABEL: @uitofp_2i64_2f32(
+; AVX2-NEXT:    [[LD0:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 0), align 64
+; AVX2-NEXT:    [[LD1:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 1), align 8
+; AVX2-NEXT:    [[CVT0:%.*]] = uitofp i64 [[LD0]] to float
+; AVX2-NEXT:    [[CVT1:%.*]] = uitofp i64 [[LD1]] to float
+; AVX2-NEXT:    store float [[CVT0]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 0), align 64
+; AVX2-NEXT:    store float [[CVT1]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 1), align 4
+; AVX2-NEXT:    ret void
+;
+; AVX512-LABEL: @uitofp_2i64_2f32(
+; AVX512-NEXT:    [[TMP1:%.*]] = load <2 x i64>, <2 x i64>* bitcast ([8 x i64]* @src64 to <2 x i64>*), align 64
+; AVX512-NEXT:    [[TMP2:%.*]] = uitofp <2 x i64> [[TMP1]] to <2 x float>
+; AVX512-NEXT:    store <2 x float> [[TMP2]], <2 x float>* bitcast ([16 x float]* @dst32 to <2 x float>*), align 64
+; AVX512-NEXT:    ret void
+;
+; AVX256DQ-LABEL: @uitofp_2i64_2f32(
+; AVX256DQ-NEXT:    [[TMP1:%.*]] = load <2 x i64>, <2 x i64>* bitcast ([8 x i64]* @src64 to <2 x i64>*), align 64
+; AVX256DQ-NEXT:    [[TMP2:%.*]] = uitofp <2 x i64> [[TMP1]] to <2 x float>
+; AVX256DQ-NEXT:    store <2 x float> [[TMP2]], <2 x float>* bitcast ([16 x float]* @dst32 to <2 x float>*), align 64
+; AVX256DQ-NEXT:    ret void
 ;
   %ld0 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 0), align 64
   %ld1 = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 1), align 8
   %cvt0 = uitofp i64 %ld0 to float
   %cvt1 = uitofp i64 %ld1 to float
   store float %cvt0, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 0), align 64
   store float %cvt1, float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 1), align 4
   ret void

llvm/test/Transforms/SLPVectorizer/X86/vect_copyable_in_binops.ll

 
 define void @add1(i32* noalias %dst, i32* noalias %src) {
 ; CHECK-LABEL: @add1(
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[INCDEC_PTR:%.*]] = getelementptr inbounds i32, i32* [[SRC:%.*]], i64 1
 ; CHECK-NEXT:    [[TMP0:%.*]] = load i32, i32* [[SRC]], align 4
 ; CHECK-NEXT:    [[INCDEC_PTR1:%.*]] = getelementptr inbounds i32, i32* [[DST:%.*]], i64 1
 ; CHECK-NEXT:    store i32 [[TMP0]], i32* [[DST]], align 4
-; CHECK-NEXT:    [[INCDEC_PTR2:%.*]] = getelementptr inbounds i32, i32* [[SRC]], i64 2
-; CHECK-NEXT:    [[TMP1:%.*]] = load i32, i32* [[INCDEC_PTR]], align 4
-; CHECK-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP1]], 1
-; CHECK-NEXT:    [[INCDEC_PTR4:%.*]] = getelementptr inbounds i32, i32* [[DST]], i64 2
-; CHECK-NEXT:    store i32 [[ADD3]], i32* [[INCDEC_PTR1]], align 4
 ; CHECK-NEXT:    [[INCDEC_PTR5:%.*]] = getelementptr inbounds i32, i32* [[SRC]], i64 3
-; CHECK-NEXT:    [[TMP2:%.*]] = load i32, i32* [[INCDEC_PTR2]], align 4
-; CHECK-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP2]], 2
 ; CHECK-NEXT:    [[INCDEC_PTR7:%.*]] = getelementptr inbounds i32, i32* [[DST]], i64 3
-; CHECK-NEXT:    store i32 [[ADD6]], i32* [[INCDEC_PTR4]], align 4
-; CHECK-NEXT:    [[TMP3:%.*]] = load i32, i32* [[INCDEC_PTR5]], align 4
-; CHECK-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP3]], 3
+; CHECK-NEXT:    [[TMP1:%.*]] = bitcast i32* [[INCDEC_PTR]] to <2 x i32>*
+; CHECK-NEXT:    [[TMP2:%.*]] = load <2 x i32>, <2 x i32>* [[TMP1]], align 4
+; CHECK-NEXT:    [[TMP3:%.*]] = add nsw <2 x i32> [[TMP2]], <i32 1, i32 2>
+; CHECK-NEXT:    [[TMP4:%.*]] = bitcast i32* [[INCDEC_PTR1]] to <2 x i32>*
+; CHECK-NEXT:    store <2 x i32> [[TMP3]], <2 x i32>* [[TMP4]], align 4
+; CHECK-NEXT:    [[TMP5:%.*]] = load i32, i32* [[INCDEC_PTR5]], align 4
+; CHECK-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP5]], 3
 ; CHECK-NEXT:    store i32 [[ADD9]], i32* [[INCDEC_PTR7]], align 4
 ; CHECK-NEXT:    ret void
 ;
 entry:
   %incdec.ptr = getelementptr inbounds i32, i32* %src, i64 1
   %0 = load i32, i32* %src, align 4
   %incdec.ptr1 = getelementptr inbounds i32, i32* %dst, i64 1
   store i32 %0, i32* %dst, align 4

 ; CHECK-NEXT:    [[TMP0:%.*]] = load i32, i32* [[SRC]], align 4
 ; CHECK-NEXT:    [[SUB:%.*]] = add nsw i32 [[TMP0]], -1
 ; CHECK-NEXT:    [[INCDEC_PTR1:%.*]] = getelementptr inbounds i32, i32* [[DST:%.*]], i64 1
 ; CHECK-NEXT:    store i32 [[SUB]], i32* [[DST]], align 4
 ; CHECK-NEXT:    [[INCDEC_PTR2:%.*]] = getelementptr inbounds i32, i32* [[SRC]], i64 2
 ; CHECK-NEXT:    [[TMP1:%.*]] = load i32, i32* [[INCDEC_PTR]], align 4
 ; CHECK-NEXT:    [[INCDEC_PTR3:%.*]] = getelementptr inbounds i32, i32* [[DST]], i64 2
 ; CHECK-NEXT:    store i32 [[TMP1]], i32* [[INCDEC_PTR1]], align 4
-; CHECK-NEXT:    [[INCDEC_PTR4:%.*]] = getelementptr inbounds i32, i32* [[SRC]], i64 3
-; CHECK-NEXT:    [[TMP2:%.*]] = load i32, i32* [[INCDEC_PTR2]], align 4
-; CHECK-NEXT:    [[SUB5:%.*]] = add nsw i32 [[TMP2]], -2
-; CHECK-NEXT:    [[INCDEC_PTR6:%.*]] = getelementptr inbounds i32, i32* [[DST]], i64 3
-; CHECK-NEXT:    store i32 [[SUB5]], i32* [[INCDEC_PTR3]], align 4
-; CHECK-NEXT:    [[TMP3:%.*]] = load i32, i32* [[INCDEC_PTR4]], align 4
-; CHECK-NEXT:    [[SUB8:%.*]] = add nsw i32 [[TMP3]], -3
-; CHECK-NEXT:    store i32 [[SUB8]], i32* [[INCDEC_PTR6]], align 4
+; CHECK-NEXT:    [[TMP2:%.*]] = bitcast i32* [[INCDEC_PTR2]] to <2 x i32>*
+; CHECK-NEXT:    [[TMP3:%.*]] = load <2 x i32>, <2 x i32>* [[TMP2]], align 4
+; CHECK-NEXT:    [[TMP4:%.*]] = add nsw <2 x i32> [[TMP3]], <i32 -2, i32 -3>
+; CHECK-NEXT:    [[TMP5:%.*]] = bitcast i32* [[INCDEC_PTR3]] to <2 x i32>*
+; CHECK-NEXT:    store <2 x i32> [[TMP4]], <2 x i32>* [[TMP5]], align 4
 ; CHECK-NEXT:    ret void
 ;
 entry:
   %incdec.ptr = getelementptr inbounds i32, i32* %src, i64 1
   %0 = load i32, i32* %src, align 4
   %sub = add nsw i32 %0, -1
   %incdec.ptr1 = getelementptr inbounds i32, i32* %dst, i64 1
   store i32 %sub, i32* %dst, align 4

 ; CHECK-NEXT:    [[TMP0:%.*]] = load i32, i32* [[SRC]], align 4
 ; CHECK-NEXT:    [[SUB:%.*]] = add nsw i32 [[TMP0]], -1
 ; CHECK-NEXT:    [[INCDEC_PTR1:%.*]] = getelementptr inbounds i32, i32* [[DST:%.*]], i64 1
 ; CHECK-NEXT:    store i32 [[SUB]], i32* [[DST]], align 4
 ; CHECK-NEXT:    [[INCDEC_PTR2:%.*]] = getelementptr inbounds i32, i32* [[SRC]], i64 2
 ; CHECK-NEXT:    [[TMP1:%.*]] = load i32, i32* [[INCDEC_PTR]], align 4
 ; CHECK-NEXT:    [[INCDEC_PTR3:%.*]] = getelementptr inbounds i32, i32* [[DST]], i64 2
 ; CHECK-NEXT:    store i32 [[TMP1]], i32* [[INCDEC_PTR1]], align 4
-; CHECK-NEXT:    [[INCDEC_PTR4:%.*]] = getelementptr inbounds i32, i32* [[SRC]], i64 3
-; CHECK-NEXT:    [[TMP2:%.*]] = load i32, i32* [[INCDEC_PTR2]], align 4
-; CHECK-NEXT:    [[SUB5:%.*]] = add nsw i32 [[TMP2]], -2
-; CHECK-NEXT:    [[INCDEC_PTR6:%.*]] = getelementptr inbounds i32, i32* [[DST]], i64 3
-; CHECK-NEXT:    store i32 [[SUB5]], i32* [[INCDEC_PTR3]], align 4
-; CHECK-NEXT:    [[TMP3:%.*]] = load i32, i32* [[INCDEC_PTR4]], align 4
-; CHECK-NEXT:    [[SUB8:%.*]] = sub nsw i32 [[TMP3]], -3
-; CHECK-NEXT:    store i32 [[SUB8]], i32* [[INCDEC_PTR6]], align 4
+; CHECK-NEXT:    [[TMP2:%.*]] = bitcast i32* [[INCDEC_PTR2]] to <2 x i32>*
+; CHECK-NEXT:    [[TMP3:%.*]] = load <2 x i32>, <2 x i32>* [[TMP2]], align 4
+; CHECK-NEXT:    [[TMP4:%.*]] = add nsw <2 x i32> [[TMP3]], <i32 -2, i32 -3>
+; CHECK-NEXT:    [[TMP5:%.*]] = sub nsw <2 x i32> [[TMP3]], <i32 -2, i32 -3>
+; CHECK-NEXT:    [[TMP6:%.*]] = shufflevector <2 x i32> [[TMP4]], <2 x i32> [[TMP5]], <2 x i32> <i32 0, i32 3>
+; CHECK-NEXT:    [[TMP7:%.*]] = bitcast i32* [[INCDEC_PTR3]] to <2 x i32>*
+; CHECK-NEXT:    store <2 x i32> [[TMP6]], <2 x i32>* [[TMP7]], align 4
 ; CHECK-NEXT:    ret void
 ;
 entry:
   %incdec.ptr = getelementptr inbounds i32, i32* %src, i64 1
   %0 = load i32, i32* %src, align 4
   %sub = add nsw i32 %0, -1
   %incdec.ptr1 = getelementptr inbounds i32, i32* %dst, i64 1
   store i32 %sub, i32* %dst, align 4
entry:
   %sub8 = sub nsw i32 %3, -3
   store i32 %sub8, i32* %incdec.ptr6, align 4
   ret void
 }
 
 define void @addsub1(i32* noalias %dst, i32* noalias %src) {
 ; CHECK-LABEL: @addsub1(
 ; CHECK-NEXT:  entry:
-; CHECK-NEXT:    [[INCDEC_PTR:%.*]] = getelementptr inbounds i32, i32* [[SRC:%.*]], i64 1
-; CHECK-NEXT:    [[TMP0:%.*]] = load i32, i32* [[SRC]], align 4
-; CHECK-NEXT:    [[SUB:%.*]] = add nsw i32 [[TMP0]], -1
-; CHECK-NEXT:    [[INCDEC_PTR1:%.*]] = getelementptr inbounds i32, i32* [[DST:%.*]], i64 1
-; CHECK-NEXT:    store i32 [[SUB]], i32* [[DST]], align 4
-; CHECK-NEXT:    [[INCDEC_PTR2:%.*]] = getelementptr inbounds i32, i32* [[SRC]], i64 2
-; CHECK-NEXT:    [[TMP1:%.*]] = load i32, i32* [[INCDEC_PTR]], align 4
-; CHECK-NEXT:    [[SUB1:%.*]] = sub nsw i32 [[TMP1]], -1
-; CHECK-NEXT:    [[INCDEC_PTR3:%.*]] = getelementptr inbounds i32, i32* [[DST]], i64 2
-; CHECK-NEXT:    store i32 [[SUB1]], i32* [[INCDEC_PTR1]], align 4
+; CHECK-NEXT:    [[INCDEC_PTR2:%.*]] = getelementptr inbounds i32, i32* [[SRC:%.*]], i64 2
+; CHECK-NEXT:    [[INCDEC_PTR3:%.*]] = getelementptr inbounds i32, i32* [[DST:%.*]], i64 2
+; CHECK-NEXT:    [[TMP0:%.*]] = bitcast i32* [[SRC]] to <2 x i32>*
+; CHECK-NEXT:    [[TMP1:%.*]] = load <2 x i32>, <2 x i32>* [[TMP0]], align 4
+; CHECK-NEXT:    [[TMP2:%.*]] = add nsw <2 x i32> [[TMP1]], <i32 -1, i32 -1>
+; CHECK-NEXT:    [[TMP3:%.*]] = sub nsw <2 x i32> [[TMP1]], <i32 -1, i32 -1>
+; CHECK-NEXT:    [[TMP4:%.*]] = shufflevector <2 x i32> [[TMP2]], <2 x i32> [[TMP3]], <2 x i32> <i32 0, i32 3>
+; CHECK-NEXT:    [[TMP5:%.*]] = bitcast i32* [[DST]] to <2 x i32>*
+; CHECK-NEXT:    store <2 x i32> [[TMP4]], <2 x i32>* [[TMP5]], align 4
 ; CHECK-NEXT:    [[INCDEC_PTR4:%.*]] = getelementptr inbounds i32, i32* [[SRC]], i64 3
-; CHECK-NEXT:    [[TMP2:%.*]] = load i32, i32* [[INCDEC_PTR2]], align 4
+; CHECK-NEXT:    [[TMP6:%.*]] = load i32, i32* [[INCDEC_PTR2]], align 4
 ; CHECK-NEXT:    [[INCDEC_PTR6:%.*]] = getelementptr inbounds i32, i32* [[DST]], i64 3
-; CHECK-NEXT:    store i32 [[TMP2]], i32* [[INCDEC_PTR3]], align 4
-; CHECK-NEXT:    [[TMP3:%.*]] = load i32, i32* [[INCDEC_PTR4]], align 4
-; CHECK-NEXT:    [[SUB8:%.*]] = sub nsw i32 [[TMP3]], -3
+; CHECK-NEXT:    store i32 [[TMP6]], i32* [[INCDEC_PTR3]], align 4
+; CHECK-NEXT:    [[TMP7:%.*]] = load i32, i32* [[INCDEC_PTR4]], align 4
+; CHECK-NEXT:    [[SUB8:%.*]] = sub nsw i32 [[TMP7]], -3
 ; CHECK-NEXT:    store i32 [[SUB8]], i32* [[INCDEC_PTR6]], align 4
 ; CHECK-NEXT:    ret void
 ;
 entry:
   %incdec.ptr = getelementptr inbounds i32, i32* %src, i64 1
   %0 = load i32, i32* %src, align 4
   %sub = add nsw i32 %0, -1
   %incdec.ptr1 = getelementptr inbounds i32, i32* %dst, i64 1
entry:
   %sub8 = sub nsw i32 %3, -3
   store i32 %sub8, i32* %incdec.ptr6, align 4
   ret void
 }
 
 define void @mul(i32* noalias %dst, i32* noalias %src) {
 ; CHECK-LABEL: @mul(
 ; CHECK-NEXT:  entry:
-; CHECK-NEXT:    [[INCDEC_PTR:%.*]] = getelementptr inbounds i32, i32* [[SRC:%.*]], i64 1
-; CHECK-NEXT:    [[TMP0:%.*]] = load i32, i32* [[SRC]], align 4
-; CHECK-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP0]], 257
-; CHECK-NEXT:    [[INCDEC_PTR1:%.*]] = getelementptr inbounds i32, i32* [[DST:%.*]], i64 1
-; CHECK-NEXT:    store i32 [[MUL]], i32* [[DST]], align 4
-; CHECK-NEXT:    [[INCDEC_PTR2:%.*]] = getelementptr inbounds i32, i32* [[SRC]], i64 2
-; CHECK-NEXT:    [[TMP1:%.*]] = load i32, i32* [[INCDEC_PTR]], align 4
-; CHECK-NEXT:    [[MUL3:%.*]] = mul nsw i32 [[TMP1]], -3
-; CHECK-NEXT:    [[INCDEC_PTR4:%.*]] = getelementptr inbounds i32, i32* [[DST]], i64 2
-; CHECK-NEXT:    store i32 [[MUL3]], i32* [[INCDEC_PTR1]], align 4
+; CHECK-NEXT:    [[INCDEC_PTR2:%.*]] = getelementptr inbounds i32, i32* [[SRC:%.*]], i64 2
+; CHECK-NEXT:    [[INCDEC_PTR4:%.*]] = getelementptr inbounds i32, i32* [[DST:%.*]], i64 2
+; CHECK-NEXT:    [[TMP0:%.*]] = bitcast i32* [[SRC]] to <2 x i32>*
+; CHECK-NEXT:    [[TMP1:%.*]] = load <2 x i32>, <2 x i32>* [[TMP0]], align 4
+; CHECK-NEXT:    [[TMP2:%.*]] = mul nsw <2 x i32> [[TMP1]], <i32 257, i32 -3>
+; CHECK-NEXT:    [[TMP3:%.*]] = bitcast i32* [[DST]] to <2 x i32>*
+; CHECK-NEXT:    store <2 x i32> [[TMP2]], <2 x i32>* [[TMP3]], align 4
 ; CHECK-NEXT:    [[INCDEC_PTR5:%.*]] = getelementptr inbounds i32, i32* [[SRC]], i64 3
-; CHECK-NEXT:    [[TMP2:%.*]] = load i32, i32* [[INCDEC_PTR2]], align 4
+; CHECK-NEXT:    [[TMP4:%.*]] = load i32, i32* [[INCDEC_PTR2]], align 4
 ; CHECK-NEXT:    [[INCDEC_PTR7:%.*]] = getelementptr inbounds i32, i32* [[DST]], i64 3
-; CHECK-NEXT:    store i32 [[TMP2]], i32* [[INCDEC_PTR4]], align 4
-; CHECK-NEXT:    [[TMP3:%.*]] = load i32, i32* [[INCDEC_PTR5]], align 4
-; CHECK-NEXT:    [[MUL9:%.*]] = mul nsw i32 [[TMP3]], -9
+; CHECK-NEXT:    store i32 [[TMP4]], i32* [[INCDEC_PTR4]], align 4
+; CHECK-NEXT:    [[TMP5:%.*]] = load i32, i32* [[INCDEC_PTR5]], align 4
+; CHECK-NEXT:    [[MUL9:%.*]] = mul nsw i32 [[TMP5]], -9
 ; CHECK-NEXT:    store i32 [[MUL9]], i32* [[INCDEC_PTR7]], align 4
 ; CHECK-NEXT:    ret void
 ;
 entry:
   %incdec.ptr = getelementptr inbounds i32, i32* %src, i64 1
   %0 = load i32, i32* %src, align 4
   %mul = mul nsw i32 %0, 257
   %incdec.ptr1 = getelementptr inbounds i32, i32* %dst, i64 1

 
 define void @shl0(i32* noalias %dst, i32* noalias %src) {
 ; CHECK-LABEL: @shl0(
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[INCDEC_PTR:%.*]] = getelementptr inbounds i32, i32* [[SRC:%.*]], i64 1
 ; CHECK-NEXT:    [[TMP0:%.*]] = load i32, i32* [[SRC]], align 4
 ; CHECK-NEXT:    [[INCDEC_PTR1:%.*]] = getelementptr inbounds i32, i32* [[DST:%.*]], i64 1
 ; CHECK-NEXT:    store i32 [[TMP0]], i32* [[DST]], align 4
-; CHECK-NEXT:    [[INCDEC_PTR2:%.*]] = getelementptr inbounds i32, i32* [[SRC]], i64 2
-; CHECK-NEXT:    [[TMP1:%.*]] = load i32, i32* [[INCDEC_PTR]], align 4
-; CHECK-NEXT:    [[SHL:%.*]] = shl i32 [[TMP1]], 1
-; CHECK-NEXT:    [[INCDEC_PTR3:%.*]] = getelementptr inbounds i32, i32* [[DST]], i64 2
-; CHECK-NEXT:    store i32 [[SHL]], i32* [[INCDEC_PTR1]], align 4
 ; CHECK-NEXT:    [[INCDEC_PTR4:%.*]] = getelementptr inbounds i32, i32* [[SRC]], i64 3
-; CHECK-NEXT:    [[TMP2:%.*]] = load i32, i32* [[INCDEC_PTR2]], align 4
-; CHECK-NEXT:    [[SHL5:%.*]] = shl i32 [[TMP2]], 2
 ; CHECK-NEXT:    [[INCDEC_PTR6:%.*]] = getelementptr inbounds i32, i32* [[DST]], i64 3
-; CHECK-NEXT:    store i32 [[SHL5]], i32* [[INCDEC_PTR3]], align 4
-; CHECK-NEXT:    [[TMP3:%.*]] = load i32, i32* [[INCDEC_PTR4]], align 4
-; CHECK-NEXT:    [[SHL8:%.*]] = shl i32 [[TMP3]], 3
+; CHECK-NEXT:    [[TMP1:%.*]] = bitcast i32* [[INCDEC_PTR]] to <2 x i32>*
+; CHECK-NEXT:    [[TMP2:%.*]] = load <2 x i32>, <2 x i32>* [[TMP1]], align 4
+; CHECK-NEXT:    [[TMP3:%.*]] = shl <2 x i32> [[TMP2]], <i32 1, i32 2>
+; CHECK-NEXT:    [[TMP4:%.*]] = bitcast i32* [[INCDEC_PTR1]] to <2 x i32>*
+; CHECK-NEXT:    store <2 x i32> [[TMP3]], <2 x i32>* [[TMP4]], align 4
+; CHECK-NEXT:    [[TMP5:%.*]] = load i32, i32* [[INCDEC_PTR4]], align 4
+; CHECK-NEXT:    [[SHL8:%.*]] = shl i32 [[TMP5]], 3
 ; CHECK-NEXT:    store i32 [[SHL8]], i32* [[INCDEC_PTR6]], align 4
 ; CHECK-NEXT:    ret void
 ;
 entry:
   %incdec.ptr = getelementptr inbounds i32, i32* %src, i64 1
   %0 = load i32, i32* %src, align 4
   %incdec.ptr1 = getelementptr inbounds i32, i32* %dst, i64 1
   store i32 %0, i32* %dst, align 4

 
 define void @add1f(float* noalias %dst, float* noalias %src) {
 ; CHECK-LABEL: @add1f(
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[INCDEC_PTR:%.*]] = getelementptr inbounds float, float* [[SRC:%.*]], i64 1
 ; CHECK-NEXT:    [[TMP0:%.*]] = load float, float* [[SRC]], align 4
 ; CHECK-NEXT:    [[INCDEC_PTR1:%.*]] = getelementptr inbounds float, float* [[DST:%.*]], i64 1
 ; CHECK-NEXT:    store float [[TMP0]], float* [[DST]], align 4
-; CHECK-NEXT:    [[INCDEC_PTR2:%.*]] = getelementptr inbounds float, float* [[SRC]], i64 2
-; CHECK-NEXT:    [[TMP1:%.*]] = load float, float* [[INCDEC_PTR]], align 4
-; CHECK-NEXT:    [[ADD3:%.*]] = fadd fast float [[TMP1]], 1.000000e+00
-; CHECK-NEXT:    [[INCDEC_PTR4:%.*]] = getelementptr inbounds float, float* [[DST]], i64 2
-; CHECK-NEXT:    store float [[ADD3]], float* [[INCDEC_PTR1]], align 4
 ; CHECK-NEXT:    [[INCDEC_PTR5:%.*]] = getelementptr inbounds float, float* [[SRC]], i64 3
-; CHECK-NEXT:    [[TMP2:%.*]] = load float, float* [[INCDEC_PTR2]], align 4
-; CHECK-NEXT:    [[ADD6:%.*]] = fadd fast float [[TMP2]], 2.000000e+00
 ; CHECK-NEXT:    [[INCDEC_PTR7:%.*]] = getelementptr inbounds float, float* [[DST]], i64 3
-; CHECK-NEXT:    store float [[ADD6]], float* [[INCDEC_PTR4]], align 4
-; CHECK-NEXT:    [[TMP3:%.*]] = load float, float* [[INCDEC_PTR5]], align 4
-; CHECK-NEXT:    [[ADD9:%.*]] = fadd fast float [[TMP3]], 3.000000e+00
+; CHECK-NEXT:    [[TMP1:%.*]] = bitcast float* [[INCDEC_PTR]] to <2 x float>*
+; CHECK-NEXT:    [[TMP2:%.*]] = load <2 x float>, <2 x float>* [[TMP1]], align 4
+; CHECK-NEXT:    [[TMP3:%.*]] = fadd fast <2 x float> [[TMP2]], <float 1.000000e+00, float 2.000000e+00>
+; CHECK-NEXT:    [[TMP4:%.*]] = bitcast float* [[INCDEC_PTR1]] to <2 x float>*
+; CHECK-NEXT:    store <2 x float> [[TMP3]], <2 x float>* [[TMP4]], align 4
+; CHECK-NEXT:    [[TMP5:%.*]] = load float, float* [[INCDEC_PTR5]], align 4
+; CHECK-NEXT:    [[ADD9:%.*]] = fadd fast float [[TMP5]], 3.000000e+00
 ; CHECK-NEXT:    store float [[ADD9]], float* [[INCDEC_PTR7]], align 4
 ; CHECK-NEXT:    ret void
 ;
 entry:
   %incdec.ptr = getelementptr inbounds float, float* %src, i64 1
   %0 = load float, float* %src, align 4
   %incdec.ptr1 = getelementptr inbounds float, float* %dst, i64 1
   store float %0, float* %dst, align 4

 ; CHECK-NEXT:    [[TMP0:%.*]] = load float, float* [[SRC]], align 4
 ; CHECK-NEXT:    [[ADD:%.*]] = fadd fast float [[TMP0]], -1.000000e+00
 ; CHECK-NEXT:    [[INCDEC_PTR1:%.*]] = getelementptr inbounds float, float* [[DST:%.*]], i64 1
 ; CHECK-NEXT:    store float [[ADD]], float* [[DST]], align 4
 ; CHECK-NEXT:    [[INCDEC_PTR2:%.*]] = getelementptr inbounds float, float* [[SRC]], i64 2
 ; CHECK-NEXT:    [[TMP1:%.*]] = load float, float* [[INCDEC_PTR]], align 4
 ; CHECK-NEXT:    [[INCDEC_PTR4:%.*]] = getelementptr inbounds float, float* [[DST]], i64 2
 ; CHECK-NEXT:    store float [[TMP1]], float* [[INCDEC_PTR1]], align 4
-; CHECK-NEXT:    [[INCDEC_PTR5:%.*]] = getelementptr inbounds float, float* [[SRC]], i64 3
-; CHECK-NEXT:    [[TMP2:%.*]] = load float, float* [[INCDEC_PTR2]], align 4
-; CHECK-NEXT:    [[ADD6:%.*]] = fadd fast float [[TMP2]], -2.000000e+00
-; CHECK-NEXT:    [[INCDEC_PTR7:%.*]] = getelementptr inbounds float, float* [[DST]], i64 3
-; CHECK-NEXT:    store float [[ADD6]], float* [[INCDEC_PTR4]], align 4
-; CHECK-NEXT:    [[TMP3:%.*]] = load float, float* [[INCDEC_PTR5]], align 4
-; CHECK-NEXT:    [[ADD9:%.*]] = fadd fast float [[TMP3]], -3.000000e+00
-; CHECK-NEXT:    store float [[ADD9]], float* [[INCDEC_PTR7]], align 4
+; CHECK-NEXT:    [[TMP2:%.*]] = bitcast float* [[INCDEC_PTR2]] to <2 x float>*
+; CHECK-NEXT:    [[TMP3:%.*]] = load <2 x float>, <2 x float>* [[TMP2]], align 4
+; CHECK-NEXT:    [[TMP4:%.*]] = fadd fast <2 x float> [[TMP3]], <float -2.000000e+00, float -3.000000e+00>
+; CHECK-NEXT:    [[TMP5:%.*]] = bitcast float* [[INCDEC_PTR4]] to <2 x float>*
+; CHECK-NEXT:    store <2 x float> [[TMP4]], <2 x float>* [[TMP5]], align 4
 ; CHECK-NEXT:    ret void
 ;
 entry:
   %incdec.ptr = getelementptr inbounds float, float* %src, i64 1
   %0 = load float, float* %src, align 4
   %add = fadd fast float %0, -1.000000e+00
   %incdec.ptr1 = getelementptr inbounds float, float* %dst, i64 1
   store float %add, float* %dst, align 4

 ; CHECK-NEXT:    [[TMP0:%.*]] = load float, float* [[SRC]], align 4
 ; CHECK-NEXT:    [[SUB:%.*]] = fadd fast float [[TMP0]], -1.000000e+00
 ; CHECK-NEXT:    [[INCDEC_PTR1:%.*]] = getelementptr inbounds float, float* [[DST:%.*]], i64 1
 ; CHECK-NEXT:    store float [[SUB]], float* [[DST]], align 4
 ; CHECK-NEXT:    [[INCDEC_PTR2:%.*]] = getelementptr inbounds float, float* [[SRC]], i64 2
 ; CHECK-NEXT:    [[TMP1:%.*]] = load float, float* [[INCDEC_PTR]], align 4
 ; CHECK-NEXT:    [[INCDEC_PTR3:%.*]] = getelementptr inbounds float, float* [[DST]], i64 2
 ; CHECK-NEXT:    store float [[TMP1]], float* [[INCDEC_PTR1]], align 4
-; CHECK-NEXT:    [[INCDEC_PTR4:%.*]] = getelementptr inbounds float, float* [[SRC]], i64 3
-; CHECK-NEXT:    [[TMP2:%.*]] = load float, float* [[INCDEC_PTR2]], align 4
-; CHECK-NEXT:    [[SUB5:%.*]] = fadd fast float [[TMP2]], -2.000000e+00
-; CHECK-NEXT:    [[INCDEC_PTR6:%.*]] = getelementptr inbounds float, float* [[DST]], i64 3
-; CHECK-NEXT:    store float [[SUB5]], float* [[INCDEC_PTR3]], align 4
-; CHECK-NEXT:    [[TMP3:%.*]] = load float, float* [[INCDEC_PTR4]], align 4
-; CHECK-NEXT:    [[SUB8:%.*]] = fsub fast float [[TMP3]], -3.000000e+00
-; CHECK-NEXT:    store float [[SUB8]], float* [[INCDEC_PTR6]], align 4
+; CHECK-NEXT:    [[TMP2:%.*]] = bitcast float* [[INCDEC_PTR2]] to <2 x float>*
+; CHECK-NEXT:    [[TMP3:%.*]] = load <2 x float>, <2 x float>* [[TMP2]], align 4
+; CHECK-NEXT:    [[TMP4:%.*]] = fadd fast <2 x float> [[TMP3]], <float -2.000000e+00, float -3.000000e+00>
+; CHECK-NEXT:    [[TMP5:%.*]] = fsub fast <2 x float> [[TMP3]], <float -2.000000e+00, float -3.000000e+00>
+; CHECK-NEXT:    [[TMP6:%.*]] = shufflevector <2 x float> [[TMP4]], <2 x float> [[TMP5]], <2 x i32> <i32 0, i32 3>
+; CHECK-NEXT:    [[TMP7:%.*]] = bitcast float* [[INCDEC_PTR3]] to <2 x float>*
+; CHECK-NEXT:    store <2 x float> [[TMP6]], <2 x float>* [[TMP7]], align 4
 ; CHECK-NEXT:    ret void
 ;
 entry:
   %incdec.ptr = getelementptr inbounds float, float* %src, i64 1
   %0 = load float, float* %src, align 4
   %sub = fadd fast float %0, -1.000000e+00
   %incdec.ptr1 = getelementptr inbounds float, float* %dst, i64 1
   store float %sub, float* %dst, align 4
entry:
   %sub8 = fsub fast float %3, -3.000000e+00
   store float %sub8, float* %incdec.ptr6, align 4
   ret void
 }
 
 define void @addsub1f(float* noalias %dst, float* noalias %src) {
 ; CHECK-LABEL: @addsub1f(
 ; CHECK-NEXT:  entry:
-; CHECK-NEXT:    [[INCDEC_PTR:%.*]] = getelementptr inbounds float, float* [[SRC:%.*]], i64 1
-; CHECK-NEXT:    [[TMP0:%.*]] = load float, float* [[SRC]], align 4
-; CHECK-NEXT:    [[SUB:%.*]] = fadd fast float [[TMP0]], -1.000000e+00
-; CHECK-NEXT:    [[INCDEC_PTR1:%.*]] = getelementptr inbounds float, float* [[DST:%.*]], i64 1
-; CHECK-NEXT:    store float [[SUB]], float* [[DST]], align 4
-; CHECK-NEXT:    [[INCDEC_PTR2:%.*]] = getelementptr inbounds float, float* [[SRC]], i64 2
-; CHECK-NEXT:    [[TMP1:%.*]] = load float, float* [[INCDEC_PTR]], align 4
-; CHECK-NEXT:    [[SUB1:%.*]] = fsub fast float [[TMP1]], -1.000000e+00
-; CHECK-NEXT:    [[INCDEC_PTR3:%.*]] = getelementptr inbounds float, float* [[DST]], i64 2
-; CHECK-NEXT:    store float [[SUB1]], float* [[INCDEC_PTR1]], align 4
+; CHECK-NEXT:    [[INCDEC_PTR2:%.*]] = getelementptr inbounds float, float* [[SRC:%.*]], i64 2
+; CHECK-NEXT:    [[INCDEC_PTR3:%.*]] = getelementptr inbounds float, float* [[DST:%.*]], i64 2
+; CHECK-NEXT:    [[TMP0:%.*]] = bitcast float* [[SRC]] to <2 x float>*
+; CHECK-NEXT:    [[TMP1:%.*]] = load <2 x float>, <2 x float>* [[TMP0]], align 4
+; CHECK-NEXT:    [[TMP2:%.*]] = fadd fast <2 x float> [[TMP1]], <float -1.000000e+00, float -1.000000e+00>
+; CHECK-NEXT:    [[TMP3:%.*]] = fsub fast <2 x float> [[TMP1]], <float -1.000000e+00, float -1.000000e+00>
+; CHECK-NEXT:    [[TMP4:%.*]] = shufflevector <2 x float> [[TMP2]], <2 x float> [[TMP3]], <2 x i32> <i32 0, i32 3>
+; CHECK-NEXT:    [[TMP5:%.*]] = bitcast float* [[DST]] to <2 x float>*
+; CHECK-NEXT:    store <2 x float> [[TMP4]], <2 x float>* [[TMP5]], align 4
 ; CHECK-NEXT:    [[INCDEC_PTR4:%.*]] = getelementptr inbounds float, float* [[SRC]], i64 3
-; CHECK-NEXT:    [[TMP2:%.*]] = load float, float* [[INCDEC_PTR2]], align 4
+; CHECK-NEXT:    [[TMP6:%.*]] = load float, float* [[INCDEC_PTR2]], align 4
 ; CHECK-NEXT:    [[INCDEC_PTR6:%.*]] = getelementptr inbounds float, float* [[DST]], i64 3
-; CHECK-NEXT:    store float [[TMP2]], float* [[INCDEC_PTR3]], align 4
-; CHECK-NEXT:    [[TMP3:%.*]] = load float, float* [[INCDEC_PTR4]], align 4
-; CHECK-NEXT:    [[SUB8:%.*]] = fsub fast float [[TMP3]], -3.000000e+00
+; CHECK-NEXT:    store float [[TMP6]], float* [[INCDEC_PTR3]], align 4
+; CHECK-NEXT:    [[TMP7:%.*]] = load float, float* [[INCDEC_PTR4]], align 4
+; CHECK-NEXT:    [[SUB8:%.*]] = fsub fast float [[TMP7]], -3.000000e+00
 ; CHECK-NEXT:    store float [[SUB8]], float* [[INCDEC_PTR6]], align 4
 ; CHECK-NEXT:    ret void
 ;
 entry:
   %incdec.ptr = getelementptr inbounds float, float* %src, i64 1
   %0 = load float, float* %src, align 4
   %sub = fadd fast float %0, -1.000000e+00
   %incdec.ptr1 = getelementptr inbounds float, float* %dst, i64 1
entry:
   %sub8 = fsub fast float %3, -3.000000e+00
   store float %sub8, float* %incdec.ptr6, align 4
   ret void
 }
 
 define void @mulf(float* noalias %dst, float* noalias %src) {
 ; CHECK-LABEL: @mulf(
 ; CHECK-NEXT:  entry:
-; CHECK-NEXT:    [[INCDEC_PTR:%.*]] = getelementptr inbounds float, float* [[SRC:%.*]], i64 1
-; CHECK-NEXT:    [[TMP0:%.*]] = load float, float* [[SRC]], align 4
-; CHECK-NEXT:    [[SUB:%.*]] = fmul fast float [[TMP0]], 2.570000e+02
-; CHECK-NEXT:    [[INCDEC_PTR1:%.*]] = getelementptr inbounds float, float* [[DST:%.*]], i64 1
-; CHECK-NEXT:    store float [[SUB]], float* [[DST]], align 4
-; CHECK-NEXT:    [[INCDEC_PTR2:%.*]] = getelementptr inbounds float, float* [[SRC]], i64 2
-; CHECK-NEXT:    [[TMP1:%.*]] = load float, float* [[INCDEC_PTR]], align 4
-; CHECK-NEXT:    [[SUB3:%.*]] = fmul fast float [[TMP1]], -3.000000e+00
-; CHECK-NEXT:    [[INCDEC_PTR4:%.*]] = getelementptr inbounds float, float* [[DST]], i64 2
-; CHECK-NEXT:    store float [[SUB3]], float* [[INCDEC_PTR1]], align 4
+; CHECK-NEXT:    [[INCDEC_PTR2:%.*]] = getelementptr inbounds float, float* [[SRC:%.*]], i64 2
+; CHECK-NEXT:    [[INCDEC_PTR4:%.*]] = getelementptr inbounds float, float* [[DST:%.*]], i64 2
+; CHECK-NEXT:    [[TMP0:%.*]] = bitcast float* [[SRC]] to <2 x float>*
+; CHECK-NEXT:    [[TMP1:%.*]] = load <2 x float>, <2 x float>* [[TMP0]], align 4
+; CHECK-NEXT:    [[TMP2:%.*]] = fmul fast <2 x float> [[TMP1]], <float 2.570000e+02, float -3.000000e+00>
+; CHECK-NEXT:    [[TMP3:%.*]] = bitcast float* [[DST]] to <2 x float>*
+; CHECK-NEXT:    store <2 x float> [[TMP2]], <2 x float>* [[TMP3]], align 4
 ; CHECK-NEXT:    [[INCDEC_PTR5:%.*]] = getelementptr inbounds float, float* [[SRC]], i64 3
-; CHECK-NEXT:    [[TMP2:%.*]] = load float, float* [[INCDEC_PTR2]], align 4
+; CHECK-NEXT:    [[TMP4:%.*]] = load float, float* [[INCDEC_PTR2]], align 4
 ; CHECK-NEXT:    [[INCDEC_PTR7:%.*]] = getelementptr inbounds float, float* [[DST]], i64 3
-; CHECK-NEXT:    store float [[TMP2]], float* [[INCDEC_PTR4]], align 4
-; CHECK-NEXT:    [[TMP3:%.*]] = load float, float* [[INCDEC_PTR5]], align 4
-; CHECK-NEXT:    [[SUB9:%.*]] = fmul fast float [[TMP3]], -9.000000e+00
+; CHECK-NEXT:    store float [[TMP4]], float* [[INCDEC_PTR4]], align 4
+; CHECK-NEXT:    [[TMP5:%.*]] = load float, float* [[INCDEC_PTR5]], align 4
+; CHECK-NEXT:    [[SUB9:%.*]] = fmul fast float [[TMP5]], -9.000000e+00
 ; CHECK-NEXT:    store float [[SUB9]], float* [[INCDEC_PTR7]], align 4
 ; CHECK-NEXT:    ret void
 ;
 entry:
   %incdec.ptr = getelementptr inbounds float, float* %src, i64 1
   %0 = load float, float* %src, align 4
   %sub = fmul fast float %0, 2.570000e+02
   %incdec.ptr1 = getelementptr inbounds float, float* %dst, i64 1

 
 define void @add1fn(float* noalias %dst, float* noalias %src) {
 ; CHECK-LABEL: @add1fn(
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[INCDEC_PTR:%.*]] = getelementptr inbounds float, float* [[SRC:%.*]], i64 1
 ; CHECK-NEXT:    [[TMP0:%.*]] = load float, float* [[SRC]], align 4
 ; CHECK-NEXT:    [[INCDEC_PTR1:%.*]] = getelementptr inbounds float, float* [[DST:%.*]], i64 1
 ; CHECK-NEXT:    store float [[TMP0]], float* [[DST]], align 4
-; CHECK-NEXT:    [[INCDEC_PTR2:%.*]] = getelementptr inbounds float, float* [[SRC]], i64 2
-; CHECK-NEXT:    [[TMP1:%.*]] = load float, float* [[INCDEC_PTR]], align 4
-; CHECK-NEXT:    [[ADD3:%.*]] = fadd float [[TMP1]], 1.000000e+00
-; CHECK-NEXT:    [[INCDEC_PTR4:%.*]] = getelementptr inbounds float, float* [[DST]], i64 2
-; CHECK-NEXT:    store float [[ADD3]], float* [[INCDEC_PTR1]], align 4
 ; CHECK-NEXT:    [[INCDEC_PTR5:%.*]] = getelementptr inbounds float, float* [[SRC]], i64 3
-; CHECK-NEXT:    [[TMP2:%.*]] = load float, float* [[INCDEC_PTR2]], align 4
-; CHECK-NEXT:    [[ADD6:%.*]] = fadd float [[TMP2]], 2.000000e+00
 ; CHECK-NEXT:    [[INCDEC_PTR7:%.*]] = getelementptr inbounds float, float* [[DST]], i64 3
-; CHECK-NEXT:    store float [[ADD6]], float* [[INCDEC_PTR4]], align 4
-; CHECK-NEXT:    [[TMP3:%.*]] = load float, float* [[INCDEC_PTR5]], align 4
-; CHECK-NEXT:    [[ADD9:%.*]] = fadd float [[TMP3]], 3.000000e+00
+; CHECK-NEXT:    [[TMP1:%.*]] = bitcast float* [[INCDEC_PTR]] to <2 x float>*
+; CHECK-NEXT:    [[TMP2:%.*]] = load <2 x float>, <2 x float>* [[TMP1]], align 4
+; CHECK-NEXT:    [[TMP3:%.*]] = fadd <2 x float> [[TMP2]], <float 1.000000e+00, float 2.000000e+00>
+; CHECK-NEXT:    [[TMP4:%.*]] = bitcast float* [[INCDEC_PTR1]] to <2 x float>*
+; CHECK-NEXT:    store <2 x float> [[TMP3]], <2 x float>* [[TMP4]], align 4
+; CHECK-NEXT:    [[TMP5:%.*]] = load float, float* [[INCDEC_PTR5]], align 4
+; CHECK-NEXT:    [[ADD9:%.*]] = fadd float [[TMP5]], 3.000000e+00
 ; CHECK-NEXT:    store float [[ADD9]], float* [[INCDEC_PTR7]], align 4
 ; CHECK-NEXT:    ret void
 ;
 entry:
   %incdec.ptr = getelementptr inbounds float, float* %src, i64 1
   %0 = load float, float* %src, align 4
   %incdec.ptr1 = getelementptr inbounds float, float* %dst, i64 1
   store float %0, float* %dst, align 4

 ; CHECK-NEXT:    [[TMP0:%.*]] = load float, float* [[SRC]], align 4
 ; CHECK-NEXT:    [[ADD:%.*]] = fadd fast float [[TMP0]], -1.000000e+00
 ; CHECK-NEXT:    [[INCDEC_PTR1:%.*]] = getelementptr inbounds float, float* [[DST:%.*]], i64 1
 ; CHECK-NEXT:    store float [[ADD]], float* [[DST]], align 4
 ; CHECK-NEXT:    [[INCDEC_PTR2:%.*]] = getelementptr inbounds float, float* [[SRC]], i64 2
 ; CHECK-NEXT:    [[TMP1:%.*]] = load float, float* [[INCDEC_PTR]], align 4
 ; CHECK-NEXT:    [[INCDEC_PTR4:%.*]] = getelementptr inbounds float, float* [[DST]], i64 2
 ; CHECK-NEXT:    store float [[TMP1]], float* [[INCDEC_PTR1]], align 4
-; CHECK-NEXT:    [[INCDEC_PTR5:%.*]] = getelementptr inbounds float, float* [[SRC]], i64 3
-; CHECK-NEXT:    [[TMP2:%.*]] = load float, float* [[INCDEC_PTR2]], align 4
-; CHECK-NEXT:    [[ADD6:%.*]] = fadd float [[TMP2]], -2.000000e+00
-; CHECK-NEXT:    [[INCDEC_PTR7:%.*]] = getelementptr inbounds float, float* [[DST]], i64 3
-; CHECK-NEXT:    store float [[ADD6]], float* [[INCDEC_PTR4]], align 4
-; CHECK-NEXT:    [[TMP3:%.*]] = load float, float* [[INCDEC_PTR5]], align 4
-; CHECK-NEXT:    [[ADD9:%.*]] = fadd float [[TMP3]], -3.000000e+00
-; CHECK-NEXT:    store float [[ADD9]], float* [[INCDEC_PTR7]], align 4
+; CHECK-NEXT:    [[TMP2:%.*]] = bitcast float* [[INCDEC_PTR2]] to <2 x float>*
+; CHECK-NEXT:    [[TMP3:%.*]] = load <2 x float>, <2 x float>* [[TMP2]], align 4
+; CHECK-NEXT:    [[TMP4:%.*]] = fadd <2 x float> [[TMP3]], <float -2.000000e+00, float -3.000000e+00>
+; CHECK-NEXT:    [[TMP5:%.*]] = bitcast float* [[INCDEC_PTR4]] to <2 x float>*
+; CHECK-NEXT:    store <2 x float> [[TMP4]], <2 x float>* [[TMP5]], align 4
 ; CHECK-NEXT:    ret void
 ;
 entry:
   %incdec.ptr = getelementptr inbounds float, float* %src, i64 1
   %0 = load float, float* %src, align 4
   %add = fadd fast float %0, -1.000000e+00
   %incdec.ptr1 = getelementptr inbounds float, float* %dst, i64 1
   store float %add, float* %dst, align 4
entry:
   %sub9 = fadd float %3, -3.000000e+00
   store float %sub9, float* %incdec.ptr7, align 4
   ret void
 }
 
 define void @mulfn(float* noalias %dst, float* noalias %src) {
 ; CHECK-LABEL: @mulfn(
 ; CHECK-NEXT:  entry:
-; CHECK-NEXT:    [[INCDEC_PTR:%.*]] = getelementptr inbounds float, float* [[SRC:%.*]], i64 1
-; CHECK-NEXT:    [[TMP0:%.*]] = load float, float* [[SRC]], align 4
-; CHECK-NEXT:    [[SUB:%.*]] = fmul float [[TMP0]], 2.570000e+02
-; CHECK-NEXT:    [[INCDEC_PTR1:%.*]] = getelementptr inbounds float, float* [[DST:%.*]], i64 1
-; CHECK-NEXT:    store float [[SUB]], float* [[DST]], align 4
-; CHECK-NEXT:    [[INCDEC_PTR2:%.*]] = getelementptr inbounds float, float* [[SRC]], i64 2
-; CHECK-NEXT:    [[TMP1:%.*]] = load float, float* [[INCDEC_PTR]], align 4
-; CHECK-NEXT:    [[SUB3:%.*]] = fmul float [[TMP1]], -3.000000e+00
-; CHECK-NEXT:    [[INCDEC_PTR4:%.*]] = getelementptr inbounds float, float* [[DST]], i64 2
-; CHECK-NEXT:    store float [[SUB3]], float* [[INCDEC_PTR1]], align 4
+; CHECK-NEXT:    [[INCDEC_PTR2:%.*]] = getelementptr inbounds float, float* [[SRC:%.*]], i64 2
+; CHECK-NEXT:    [[INCDEC_PTR4:%.*]] = getelementptr inbounds float, float* [[DST:%.*]], i64 2
+; CHECK-NEXT:    [[TMP0:%.*]] = bitcast float* [[SRC]] to <2 x float>*
+; CHECK-NEXT:    [[TMP1:%.*]] = load <2 x float>, <2 x float>* [[TMP0]], align 4
+; CHECK-NEXT:    [[TMP2:%.*]] = fmul <2 x float> [[TMP1]], <float 2.570000e+02, float -3.000000e+00>
+; CHECK-NEXT:    [[TMP3:%.*]] = bitcast float* [[DST]] to <2 x float>*
+; CHECK-NEXT:    store <2 x float> [[TMP2]], <2 x float>* [[TMP3]], align 4
 ; CHECK-NEXT:    [[INCDEC_PTR5:%.*]] = getelementptr inbounds float, float* [[SRC]], i64 3
-; CHECK-NEXT:    [[TMP2:%.*]] = load float, float* [[INCDEC_PTR2]], align 4
+; CHECK-NEXT:    [[TMP4:%.*]] = load float, float* [[INCDEC_PTR2]], align 4
 ; CHECK-NEXT:    [[INCDEC_PTR7:%.*]] = getelementptr inbounds float, float* [[DST]], i64 3
-; CHECK-NEXT:    store float [[TMP2]], float* [[INCDEC_PTR4]], align 4
-; CHECK-NEXT:    [[TMP3:%.*]] = load float, float* [[INCDEC_PTR5]], align 4
-; CHECK-NEXT:    [[SUB9:%.*]] = fmul fast float [[TMP3]], -9.000000e+00
+; CHECK-NEXT:    store float [[TMP4]], float* [[INCDEC_PTR4]], align 4
+; CHECK-NEXT:    [[TMP5:%.*]] = load float, float* [[INCDEC_PTR5]], align 4
+; CHECK-NEXT:    [[SUB9:%.*]] = fmul fast float [[TMP5]], -9.000000e+00
 ; CHECK-NEXT:    store float [[SUB9]], float* [[INCDEC_PTR7]], align 4
 ; CHECK-NEXT:    ret void
 ;
 entry:
   %incdec.ptr = getelementptr inbounds float, float* %src, i64 1
   %0 = load float, float* %src, align 4
   %sub = fmul float %0, 2.570000e+02
   %incdec.ptr1 = getelementptr inbounds float, float* %dst, i64 1

llvm/test/Transforms/SLPVectorizer/slp-max-reg-size.ll

 ; CHECK-VF2-128-NEXT:    store <2 x i32> <i32 7, i32 8>, <2 x i32>* bitcast (i32* getelementptr inbounds ([8 x i32], [8 x i32]* @X, i16 0, i16 6) to <2 x i32>*), align 1
 ; CHECK-VF2-128-NEXT:    ret void
 ;
 ; CHECK-VF8-256-LABEL: @foo(
 ; CHECK-VF8-256-NEXT:    store <8 x i32> <i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8>, <8 x i32>* bitcast ([8 x i32]* @X to <8 x i32>*), align 1
 ; CHECK-VF8-256-NEXT:    ret void
 ;
 ; CHECK-VF2-128-128-LABEL: @foo(
-; CHECK-VF2-128-128-NEXT:    store i32 1, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @X, i16 0, i16 0), align 1
-; CHECK-VF2-128-128-NEXT:    store i32 2, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @X, i16 0, i16 1), align 1
-; CHECK-VF2-128-128-NEXT:    store i32 3, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @X, i16 0, i16 2), align 1
-; CHECK-VF2-128-128-NEXT:    store i32 4, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @X, i16 0, i16 3), align 1
-; CHECK-VF2-128-128-NEXT:    store i32 5, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @X, i16 0, i16 4), align 1
-; CHECK-VF2-128-128-NEXT:    store i32 6, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @X, i16 0, i16 5), align 1
-; CHECK-VF2-128-128-NEXT:    store i32 7, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @X, i16 0, i16 6), align 1
-; CHECK-VF2-128-128-NEXT:    store i32 8, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @X, i16 0, i16 7), align 1
+; CHECK-VF2-128-128-NEXT:    store <2 x i32> <i32 1, i32 2>, <2 x i32>* bitcast ([8 x i32]* @X to <2 x i32>*), align 1
+; CHECK-VF2-128-128-NEXT:    store <2 x i32> <i32 3, i32 4>, <2 x i32>* bitcast (i32* getelementptr inbounds ([8 x i32], [8 x i32]* @X, i16 0, i16 2) to <2 x i32>*), align 1
+; CHECK-VF2-128-128-NEXT:    store <2 x i32> <i32 5, i32 6>, <2 x i32>* bitcast (i32* getelementptr inbounds ([8 x i32], [8 x i32]* @X, i16 0, i16 4) to <2 x i32>*), align 1
+; CHECK-VF2-128-128-NEXT:    store <2 x i32> <i32 7, i32 8>, <2 x i32>* bitcast (i32* getelementptr inbounds ([8 x i32], [8 x i32]* @X, i16 0, i16 6) to <2 x i32>*), align 1
 ; CHECK-VF2-128-128-NEXT:    ret void
 ;
   store i32 1, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @X, i16 0, i16 0), align 1
   store i32 2, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @X, i16 0, i16 1), align 1
   store i32 3, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @X, i16 0, i16 2), align 1
   store i32 4, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @X, i16 0, i16 3), align 1
   store i32 5, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @X, i16 0, i16 4), align 1
   store i32 6, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @X, i16 0, i16 5), align 1
   store i32 7, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @X, i16 0, i16 6), align 1
   store i32 8, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @X, i16 0, i16 7), align 1
   ret void
 }