[Bug target/109807] New: [14 Regression] sse2-mmx-pmaddwd.c met ICE after commit gcc-14-666-g608e7f3ab47 with march=cascadelake

[Bug target/109807] New: [14 Regression] sse2-mmx-pmaddwd.c met ICE after commit gcc-14-666-g608e7f3ab47 with march=cascadelake

[haochen.jiang at intel dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109807

            Bug ID: 109807
           Summary: [14 Regression] sse2-mmx-pmaddwd.c met ICE after
                    commit gcc-14-666-g608e7f3ab47 with march=cascadelake
           Product: gcc
           Version: 14.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: target
          Assignee: unassigned at gcc dot gnu.org
          Reporter: haochen.jiang at intel dot com
  Target Milestone: ---

After that commit in the testcase with command:

/export/users/haochenj/env/build_no_bootstrap_master/gcc/xgcc
-B/export/users/haochenj/env/build_no_bootstrap_master/gcc/
/export/users/haochenj/src/gcc/master/gcc/testsuite/gcc.
target/i386/sse2-mmx-pmaddwd.c -m64 -march=cascadelake
-fdiagnostics-plain-output -O2 -fno-strict-aliasing -msse2 -mno-mmx -lm -o
./sse2-mmx-pmaddwd.exe

We met an ICE:

In file included from
/export/users/haochenj/src/gcc/master/gcc/testsuite/gcc.target/i386/sse2-mmx-pmaddwd.c:5:
/export/users/haochenj/src/gcc/master/gcc/testsuite/gcc.target/i386/sse2-check.h:
In function ‘do_test’:                                                         
                                            
/export/users/haochenj/src/gcc/master/gcc/testsuite/gcc.target/i386/sse2-check.h:10:1:
internal compiler error: in ix86_widen_mult_cost, at config/i386/i386.cc:20442
0x1a57e23 ix86_widen_mult_cost                                                 
                                                                               
                                                      
/export/users/haochenj/src/gcc/master/./gcc/config/i386/i386.cc:20442
0x1a62835 ix86_vector_costs::add_stmt_cost(int, vect_cost_for_stmt,
_stmt_vec_info*, _slp_tree*, tree_node*, int, vect_cost_model_location)
        /export/users/haochenj/src/gcc/master/./gcc/config/i386/i386.cc:23479
0x18dc317 add_stmt_cost(vector_costs*, int, vect_cost_for_stmt,
_stmt_vec_info*, _slp_tree*, tree_node*, int, vect_cost_model_location)
        /export/users/haochenj/src/gcc/master/./gcc/tree-vectorizer.h:1779
0x190fa68 add_stmt_cost(vector_costs*, stmt_info_for_cost*)
        /export/users/haochenj/src/gcc/master/./gcc/tree-vectorizer.h:1801
0x190567f vect_bb_vectorization_profitable_p
        /export/users/haochenj/src/gcc/master/./gcc/tree-vect-slp.cc:6928
0x1907bec vect_slp_region
        /export/users/haochenj/src/gcc/master/./gcc/tree-vect-slp.cc:7441
0x19087dc vect_slp_bbs
        /export/users/haochenj/src/gcc/master/./gcc/tree-vect-slp.cc:7611
0x1908d6c vect_slp_function(function*)
        /export/users/haochenj/src/gcc/master/./gcc/tree-vect-slp.cc:7712
0x192d9b5 execute
        /export/users/haochenj/src/gcc/master/./gcc/tree-vectorizer.cc:1529
Please submit a full bug report, with preprocessed source (by using
-freport-bug).
Please include the complete backtrace with any bug report.
See <https://gcc.gnu.org/bugs/> for instructions.

I took a quick look, it seems that V2SImode is not handled in function
ix86_widen_mult_cost.

[Bug target/109807] [14 Regression] sse2-mmx-pmaddwd.c met ICE after commit r14-666-g608e7f3ab47 with march=cascadelake

[haochen.jiang at intel dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109807

--- Comment #1 from Haochen Jiang <haochen.jiang at intel dot com> ---
I further checked the reason, V2SI should never dropped into that function
because we have no pattern under V2SI.

I suppose it is because -march=cascadelake will open SSE4.1, with the new
pattern, it wrongly dropped into that.

-m32 will not ICE since TARGET_MMX_WITH_SSE need 64 bit and won't enable the
new pattern.

[Bug target/109807] [14 Regression] sse2-mmx-pmaddwd.c met ICE after commit r14-666-g608e7f3ab47 with march=cascadelake

[ubizjak at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109807

--- Comment #2 from Uroš Bizjak <ubizjak at gmail dot com> ---
(In reply to Haochen Jiang from comment #1)
> I further checked the reason, V2SI should never dropped into that function
> because we have no pattern under V2SI.
> 
> I suppose it is because -march=cascadelake will open SSE4.1, with the new
> pattern, it wrongly dropped into that.
> 
> -m32 will not ICE since TARGET_MMX_WITH_SSE need 64 bit and won't enable the
> new pattern.

V2SI mul was introduced in r14-493 (AKA partial fix for PR109690).

[Bug target/109807] [14 Regression] sse2-mmx-pmaddwd.c met ICE after commit r14-666-g608e7f3ab47 with march=cascadelake

[dcb314 at hotmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109807

David Binderman <dcb314 at hotmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |dcb314 at hotmail dot com

--- Comment #3 from David Binderman <dcb314 at hotmail dot com> ---
I see the same for this C code on znver1:

typedef struct MonitorInfo MonitorInfo;
char *decode_edid_edid;
struct MonitorInfo {
  int width_mm;
  int height_mm
} decode_edid() {
  MonitorInfo info;
  MonitorInfo *__trans_tmp_3 = &info;
  {
    MonitorInfo *info = __trans_tmp_3;
    info->width_mm = 10 * decode_edid_edid[5];
    info->height_mm = 10 * decode_edid_edid[6];
  }
  decode_headerdecode_vendor_and_product_identificationdecode_edid_version(
      info);
}

$ ~/gcc/results.20230511.asan.ubsan/bin/gcc -c -O2 -march=znver1 -w bug919.c
during GIMPLE pass: slp
bug919.c: In function ‘decode_edid’:
bug919.c:6:3: internal compiler error: in ix86_widen_mult_cost, at
config/i386/i386.cc:20442
    6 | } decode_edid() {
      |   ^~~~~~~~~~~

It was ok yesterday:

$ ~/gcc/results.20230510.asan.ubsan/bin/gcc -c -O2 -march=znver1 -w bug919.c
$

[Bug target/109807] [14 Regression] sse2-mmx-pmaddwd.c met ICE after commit r14-666-g608e7f3ab47 with march=cascadelake

[haochen.jiang at intel dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109807

--- Comment #4 from Haochen Jiang <haochen.jiang at intel dot com> ---
(In reply to Uroš Bizjak from comment #2)
> (In reply to Haochen Jiang from comment #1)
> > I further checked the reason, V2SI should never dropped into that function
> > because we have no pattern under V2SI.
> > 
> > I suppose it is because -march=cascadelake will open SSE4.1, with the new
> > pattern, it wrongly dropped into that.
> > 
> > -m32 will not ICE since TARGET_MMX_WITH_SSE need 64 bit and won't enable the
> > new pattern.
> 
> V2SI mul was introduced in r14-493 (AKA partial fix for PR109690).

I see. So we might need to add cost for that right?

[Bug target/109807] [14 Regression] sse2-mmx-pmaddwd.c met ICE after commit r14-666-g608e7f3ab47 with march=cascadelake

[ubizjak at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109807

--- Comment #5 from Uroš Bizjak <ubizjak at gmail dot com> ---
(In reply to Haochen Jiang from comment #4)
> (In reply to Uroš Bizjak from comment #2)
> > (In reply to Haochen Jiang from comment #1)
> > > I further checked the reason, V2SI should never dropped into that function
> > > because we have no pattern under V2SI.
> > > 
> > > I suppose it is because -march=cascadelake will open SSE4.1, with the new
> > > pattern, it wrongly dropped into that.
> > > 
> > > -m32 will not ICE since TARGET_MMX_WITH_SSE need 64 bit and won't enable the
> > > new pattern.
> > 
> > V2SI mul was introduced in r14-493 (AKA partial fix for PR109690).
> 
> I see. So we might need to add cost for that right?

Adding "case V2SImode:" above "case V4SImode:" in ix86_widen_mult_cost will fix
this issue. Probably V4HImode is also needed above V8HImode, we have had a
mulv4hi3 pattern for some time now.

[Bug target/109807] [14 Regression] sse2-mmx-pmaddwd.c met ICE after commit r14-666-g608e7f3ab47 with march=cascadelake

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109807

Richard Biener <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |14.0

[Bug target/109807] [14 Regression] sse2-mmx-pmaddwd.c met ICE after commit r14-666-g608e7f3ab47 with march=cascadelake

[ubizjak at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109807

Uroš Bizjak <ubizjak at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Last reconfirmed|                            |2023-05-11
     Ever confirmed|0                           |1
             Status|UNCONFIRMED                 |ASSIGNED
           Assignee|unassigned at gcc dot gnu.org      |ubizjak at gmail dot com

--- Comment #6 from Uroš Bizjak <ubizjak at gmail dot com> ---
Created attachment 55053
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=55053&action=edit
Proposed patch.

Patch in testing.

[Bug target/109807] [14 Regression] sse2-mmx-pmaddwd.c met ICE after commit r14-666-g608e7f3ab47 with march=cascadelake

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109807

--- Comment #7 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by Uros Bizjak <uros@gcc.gnu.org>:

https://gcc.gnu.org/g:10a0d98356680a387b69f8659099450f7af86d9f

commit r14-713-g10a0d98356680a387b69f8659099450f7af86d9f
Author: Uros Bizjak <ubizjak@gmail.com>
Date:   Thu May 11 20:54:24 2023 +0200

    i386: Handle V4HI and V2SImode in ix86_widen_mult_cost [PR109807]

    Do not crash when asking ix86_widen_mult_cost for the cost of
    a widening mul operation to V4HI or V2SImode.

    gcc/ChangeLog:

            PR target/109807
            * config/i386/i386.cc (ix86_widen_mult_cost):
            Handle V4HImode and V2SImode.

    gcc/testsuite/ChangeLog:

            PR target/109807
            * gcc.target/i386/pr109807.c: New test.

[Bug target/109807] [14 Regression] sse2-mmx-pmaddwd.c met ICE after commit r14-666-g608e7f3ab47 with march=cascadelake

[ubizjak at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109807

Uroš Bizjak <ubizjak at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         Resolution|---                         |FIXED
             Status|ASSIGNED                    |RESOLVED
             Target|                            |x86_64

--- Comment #8 from Uroš Bizjak <ubizjak at gmail dot com> ---
Fixed.

[Bug target/109807] [14 Regression] sse2-mmx-pmaddwd.c met ICE after commit r14-666-g608e7f3ab47 with march=cascadelake

[ubizjak at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109807

Uroš Bizjak <ubizjak at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         Resolution|FIXED                       |DUPLICATE

--- Comment #9 from Uroš Bizjak <ubizjak at gmail dot com> ---


*** This bug has been marked as a duplicate of bug 109825 ***

[Bug target/109807] [14 Regression] sse2-mmx-pmaddwd.c met ICE after commit r14-666-g608e7f3ab47 with march=cascadelake

[dcb314 at hotmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109807

--- Comment #10 from David Binderman <dcb314 at hotmail dot com> ---
(In reply to Uroš Bizjak from comment #8)
> Fixed.

I don't think so. The code I gave seems still to crash the compiler:

$ ~/gcc/results/bin/gcc -v
Using built-in specs.
COLLECT_GCC=/home/dcb38/gcc/results/bin/gcc
COLLECT_LTO_WRAPPER=/home/dcb38/gcc/results.20230513.asan.ubsan/libexec/gcc/x86_64-pc-linux-gnu/14.0.0/lto-wrapper
Target: x86_64-pc-linux-gnu
Configured with: ../trunk.year/configure
--prefix=/home/dcb38/gcc/results.20230513.asan.ubsan --disable-multilib
--disable-bootstrap --with-build-config=bootstrap-asan
--with-build-config=bootstrap-ubsan --with-pkgversion=1d339ce8d002920f
--enable-checking=df,extra,fold,rtl,yes --enable-languages=c,c++,fortran
Thread model: posix
Supported LTO compression algorithms: zlib zstd
gcc version 14.0.0 20230513 (experimental) (1d339ce8d002920f)


$ ~/gcc/results/bin/gcc -c -O2 -march=znver1 bug919.c
during GIMPLE pass: slp
edid-parse.c: In function ‘decode_edid’:
edid-parse.c:521:14: internal compiler error: in ix86_widen_mult_cost, at
config
/i386/i386.cc:20444
0x11f3ab1 ix86_widen_mult_cost(processor_costs const*, machine_mode, bool)
        ../../trunk.year/gcc/config/i386/i386.cc:20444

I thought at first I hadn't picked up Uros's change, but:

$ git log | fgrep V4HI
fgrep: warning: fgrep is obsolescent; using grep -F
    i386: Handle V4HI and V2SImode in ix86_widen_mult_cost [PR109807]
    a widening mul operation to V4HI or V2SImode.
            Handle V4HImode and V2SImode.
    The operation ADDHN on V4SI, for example, is represented as (truncate:V4HI
((src1:V4SI + src2:V4SI) >> 16))
    and RADDHN as (truncate:V4HI ((src1:V4SI + src2:V4SI + (1 << 15)) >> 16)).
$

[Bug target/109807] [14 Regression] sse2-mmx-pmaddwd.c met ICE after commit r14-666-g608e7f3ab47 with march=cascadelake

[ubizjak at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109807

--- Comment #11 from Uroš Bizjak <ubizjak at gmail dot com> ---
(In reply to David Binderman from comment #10)
> (In reply to Uroš Bizjak from comment #8)
> > Fixed.
> 
> I don't think so. The code I gave seems still to crash the compiler:

Yes, the cost function is ICEing on unhandled modes (that is *not* a good
approach). Please look at the PR109825 how it will be solved.

[Bug target/109807] [14 Regression] sse2-mmx-pmaddwd.c met ICE after commit r14-666-g608e7f3ab47 with march=cascadelake

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109807

--- Comment #12 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by Uros Bizjak <uros@gcc.gnu.org>:

https://gcc.gnu.org/g:aed51e2051b24a6a2127c6626f451641557a571a

commit r14-812-gaed51e2051b24a6a2127c6626f451641557a571a
Author: Uros Bizjak <ubizjak@gmail.com>
Date:   Sun May 14 21:53:17 2023 +0200

    i386: Handle unsupported modes from ix86_widen_mult_cost [PR109807]

    Revert my previous change that faked handling of V4HI and V2SImodes
    in ix86_widen_mult_cost and rather return arbitrary high value
    for unsupported modes. This should prevent cost estimator from
    selecting non-existent vector widen multiply operation.

    gcc/ChangeLog:

            PR target/109807
            * config/i386/i386.cc: Revert the 2023-05-11 change.
            (ix86_widen_mult_cost): Return high value instead of
            ICEing for unsupported modes.

    gcc/testsuite/ChangeLog:

            PR target/109807
            * gcc.target/i386/pr109825.c: New test.

[Bug target/109807] [14 Regression] sse2-mmx-pmaddwd.c met ICE after commit r14-666-g608e7f3ab47 with march=cascadelake

[ubizjak at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109807

Uroš Bizjak <ubizjak at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         Resolution|DUPLICATE                   |FIXED

--- Comment #13 from Uroš Bizjak <ubizjak at gmail dot com> ---
Fixed.

[Bug target/109807] [14 Regression] sse2-mmx-pmaddwd.c met ICE after commit r14-666-g608e7f3ab47 with march=cascadelake

[ubizjak at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109807

Uroš Bizjak <ubizjak at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |slyfox at gcc dot gnu.org

--- Comment #14 from Uroš Bizjak <ubizjak at gmail dot com> ---
*** Bug 109838 has been marked as a duplicate of this bug. ***

[Bug target/109807] [14 Regression] sse2-mmx-pmaddwd.c met ICE after commit r14-666-g608e7f3ab47 with march=cascadelake

[ubizjak at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109807

Uroš Bizjak <ubizjak at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |jamborm at gcc dot gnu.org

--- Comment #15 from Uroš Bizjak <ubizjak at gmail dot com> ---
*** Bug 109825 has been marked as a duplicate of this bug. ***