This patch seems to break
   libgomp.fortran/examples-4/simd-3.f90
at run time as the calculated value is way off: "diff = sum - sum_ref" is -12576 while abs(diff) < 1e-16 is the check. It works with -O0 or without -mavx; (Hence, -Os or -O1 + -mavx cause the fail.)

Please have a look.

Tobias

PS: I also see other failures, namely, libgomp.fortran/nestedfn5.f90 and libgomp.fortran/simd1.f90 – but I have not checked whether the reason is the same or due to my tree or … For libgomp.fortran/examples-4/simd-3.f90 I did a proper bisecting.


On 10/30/19 7:16 PM, Andre Vieira (lists) wrote:
Hi,

In this patch I turn epilogue vectorization on by default for all targets. After some discussions I decided to take the following testing approach:

1) I have disabled epilogue vectorization for all tests that failed due to scan-tree-dump failures inside:
  - gcc.dg/vect
  - gcc.target/i386
  - gcc.target/aarch6
  - gfortran.dg/vect

2) Added the Bugzilla's reported testcase but xfail the scan for EPILOGUE VECTORIZED for arm*-*-* as that target does not vectorize the epilogue.


I have only been able to test this for aarch64, arm and x86_64 (with avx512).  Other targets may also want to disable epilogue vectorization like I did for gcc.target and gcc.dg/vect tests or xfail the gcc.dg/vect/vect-epilogues.c test. I am relying on others to test this for their available targets, thanks in advance!

Is this OK for trunk?


Cheers,
Andre

gcc/ChangeLog:

2019-10-30  Andre Vieira  <andre.simoesdiasvie...@arm.com>

        * params.def (PARAM_VECT_EPILOGUES_NOMASK): Enable by default.

gcc/testsuite/ChangeLog:

2019-10-30  Andre Vieira  <andre.simoesdiasvie...@arm.com>

        * gcc.dg/vect/vect-epilogues.c: New test.
        * gcc.dg/vect/fast-math-vect-call-1.c: Disable for epilogue
        vectorization.
        * gcc.dg/vect/no-fast-math-vect16.c: Likewise.
        * gcc.dg/vect/no-scevccp-noreassoc-slp-reduc-7.c: Likewise.
        * gcc.dg/vect/no-scevccp-vect-iv-3.c: Likewise.
        * gcc.dg/vect/no-section-anchors-vect-31.c: Likewise.
        * gcc.dg/vect/no-section-anchors-vect-64.c: Likewise.
        * gcc.dg/vect/no-section-anchors-vect-66.c: Likewise.
        * gcc.dg/vect/no-section-anchors-vect-68.c: Likewise.
        * gcc.dg/vect/no-vfa-vect-dv-2.c: Likewise.
        * gcc.dg/vect/pr33804.c: Likewise.
        * gcc.dg/vect/pr53773.c: Likewise.
        * gcc.dg/vect/pr65947-1.c: Likewise.
        * gcc.dg/vect/pr65947-13.c: Likewise.
        * gcc.dg/vect/pr65947-14.c: Likewise.
        * gcc.dg/vect/pr65947-4.c: Likewise.
        * gcc.dg/vect/pr80631-1.c: Likewise.
        * gcc.dg/vect/pr80631-2.c: Likewise.
        * gcc.dg/vect/slp-23.c: Likewise.
        * gcc.dg/vect/slp-25.c: Likewise.
        * gcc.dg/vect/slp-reduc-2.c: Likewise.
        * gcc.dg/vect/slp-reduc-5.c: Likewise.
        * gcc.dg/vect/slp-reduc-6.c: Likewise.
        * gcc.dg/vect/slp-reduc-sad-2.c: Likewise.
        * gcc.dg/vect/slp-widen-mult-half.c: Likewise.
        * gcc.dg/vect/trapv-vect-reduc-4.c: Likewise.
        * gcc.dg/vect/vect-103.c: Likewise.
        * gcc.dg/vect/vect-109.c: Likewise.
        * gcc.dg/vect/vect-119.c: Likewise.
        * gcc.dg/vect/vect-24.c: Likewise.
        * gcc.dg/vect/vect-26.c: Likewise.
        * gcc.dg/vect/vect-27.c: Likewise.
        * gcc.dg/vect/vect-29.c: Likewise.
        * gcc.dg/vect/vect-42.c: Likewise.
        * gcc.dg/vect/vect-44.c: Likewise.
        * gcc.dg/vect/vect-48.c: Likewise.
        * gcc.dg/vect/vect-50.c: Likewise.
        * gcc.dg/vect/vect-52.c: Likewise.
        * gcc.dg/vect/vect-54.c: Likewise.
        * gcc.dg/vect/vect-56.c: Likewise.
        * gcc.dg/vect/vect-58.c: Likewise.
        * gcc.dg/vect/vect-60.c: Likewise.
        * gcc.dg/vect/vect-72.c: Likewise.
        * gcc.dg/vect/vect-75-big-array.c: Likewise.
        * gcc.dg/vect/vect-75.c: Likewise.
        * gcc.dg/vect/vect-77-alignchecks.c: Likewise.
        * gcc.dg/vect/vect-77-global.c: Likewise.
        * gcc.dg/vect/vect-78-alignchecks.c: Likewise.
        * gcc.dg/vect/vect-78-global.c: Likewise.
        * gcc.dg/vect/vect-89-big-array.c: Likewise.
        * gcc.dg/vect/vect-89.c: Likewise.
        * gcc.dg/vect/vect-91.c: Likewise.
        * gcc.dg/vect/vect-92.c: Likewise.
        * gcc.dg/vect/vect-96.c: Likewise.
        * gcc.dg/vect/vect-cond-reduc-3.c: Likewise.
        * gcc.dg/vect/vect-cond-reduc-4.c: Likewise.
        * gcc.dg/vect/vect-live-1.c: Likewise.
        * gcc.dg/vect/vect-live-2.c: Likewise.
        * gcc.dg/vect/vect-live-3.c: Likewise.
        * gcc.dg/vect/vect-live-4.c: Likewise.
        * gcc.dg/vect/vect-live-slp-1.c: Likewise.
        * gcc.dg/vect/vect-live-slp-2.c: Likewise.
        * gcc.dg/vect/vect-live-slp-3.c: Likewise.
        * gcc.dg/vect/vect-multitypes-3.c: Likewise.
        * gcc.dg/vect/vect-multitypes-4.c: Likewise.
        * gcc.dg/vect/vect-multitypes-6.c: Likewise.
        * gcc.dg/vect/vect-peel-1-epilogue.c: Likewise. New test.
        * gcc.dg/vect/vect-peel-1-src.c: Likewise. New test.
        * gcc.dg/vect/vect-peel-1.c: Likewise.
        * gcc.dg/vect/vect-peel-3-epilogue.c: Likewise. New test.
        * gcc.dg/vect/vect-peel-3-src.c: Likewise. New test.
        * gcc.dg/vect/vect-peel-3.c: Likewise.
        * gcc.dg/vect/vect-peel-4-epilogue.c: Likewise. New test.
        * gcc.dg/vect/vect-peel-4-src.c: Likewise. New test.
        * gcc.dg/vect/vect-peel-4.c: Likewise.
        * gcc.dg/vect/vect-reduc-6.c: Likewise.
        * gcc.dg/vect/vect-reduc-dot-s16a.c: Likewise.
        * gcc.dg/vect/vect-reduc-dot-s8a.c: Likewise.
        * gcc.dg/vect/vect-reduc-dot-s8b.c: Likewise.
        * gcc.dg/vect/vect-reduc-dot-u16b.c: Likewise.
        * gcc.dg/vect/vect-reduc-dot-u8a.c: Likewise.
        * gcc.dg/vect/vect-reduc-dot-u8b.c: Likewise.
        * gcc.dg/vect/vect-reduc-pattern-1a.c: Likewise.
        * gcc.dg/vect/vect-reduc-pattern-1b-big-array.c: Likewise.
        * gcc.dg/vect/vect-reduc-pattern-1c-big-array.c: Likewise.
        * gcc.dg/vect/vect-reduc-pattern-2a.c: Likewise.
        * gcc.dg/vect/vect-reduc-pattern-2b-big-array.c: Likewise.
        * gcc.dg/vect/vect-reduc-sad.c: Likewise.
        * gcc.dg/vect/vect-widen-mult-const-s16.c: Likewise.
        * gcc.dg/vect/vect-widen-mult-const-u16.c: Likewise.
        * gcc.dg/vect/vect-widen-mult-half-u8.c: Likewise.
        * gcc.dg/vect/vect-widen-mult-half.c: Likewise.
        * gcc.dg/vect/vect-widen-mult-u16.c: Likewise.
        * gcc.dg/vect/vect-widen-mult-u8-s16-s32.c: Likewise.
        * gcc.dg/vect/wrapv-vect-reduc-dot-s8b.c: Likewise.
        * gcc.target/aarch64/sve/reduc_strict_3.c: Likewise.
        * gcc.target/i386/avx-vzeroupper-19.c: Likewise.
        * gcc.target/i386/avx2-vect-mask-store-move1.c: Likewise.
        * gcc.target/i386/avx512f-gather-2.c: Likewise.
        * gcc.target/i386/avx512f-gather-5.c: Likewise.
        * gcc.target/i386/l_fma_double_1.c: Likewise.
        * gcc.target/i386/l_fma_double_2.c: Likewise.
        * gcc.target/i386/l_fma_double_3.c: Likewise.
        * gcc.target/i386/l_fma_double_4.c: Likewise.
        * gcc.target/i386/l_fma_double_5.c: Likewise.
        * gcc.target/i386/l_fma_double_6.c: Likewise.
        * gcc.target/i386/l_fma_float_1.c: Likewise.
        * gcc.target/i386/l_fma_float_2.c: Likewise.
        * gcc.target/i386/l_fma_float_3.c: Likewise.
        * gcc.target/i386/l_fma_float_4.c: Likewise.
        * gcc.target/i386/l_fma_float_5.c: Likewise.
        * gcc.target/i386/l_fma_float_6.c: Likewise.
        * gcc.target/i386/mask-pack.c: Likewise.
        * gcc.target/i386/mask-unpack.c: Likewise.
        * gfortran.dg/vect/vect-4.f90: Likewise.

Reply via email to