[Bug testsuite/68112] New: [6 Regression] FAIL: gcc.target/i386/avx512ifma-vpmaddhuq-2.c (test for excess errors)

[Bug testsuite/68112] New: [6 Regression] FAIL: gcc.target/i386/avx512ifma-vpmaddhuq-2.c (test for excess errors)

[hjl.tools at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=68112

            Bug ID: 68112
           Summary: [6 Regression] FAIL:
                    gcc.target/i386/avx512ifma-vpmaddhuq-2.c (test for
                    excess errors)
           Product: gcc
           Version: 6.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: testsuite
          Assignee: unassigned at gcc dot gnu.org
          Reporter: hjl.tools at gmail dot com
                CC: kirill.yukhin at intel dot com
  Target Milestone: ---

On Linux/x86, r229437 gave

spawn -ignore SIGHUP /export/gnu/import/git/gcc-test-ia32corei7/bld/gcc/xgcc
-B/export/gnu/import/git/gcc-test-ia32corei7/bld/gcc/
/export/gnu/import/git/gcc-test-ia32corei7/src-trunk/gcc/testsuite/gcc.target/i386/avx512ifma-vpmaddhuq-2.c
-B/export/gnu/import/git/gcc-test-ia32corei7/bld/i686-linux/./libmpx/
-B/export/gnu/import/git/gcc-test-ia32corei7/bld/i686-linux/./libmpx/mpxrt
-L/export/gnu/import/git/gcc-test-ia32corei7/bld/i686-linux/./libmpx/mpxrt/.libs
-B/export/gnu/import/git/gcc-test-ia32corei7/bld/i686-linux/./libmpx/
-B/export/gnu/import/git/gcc-test-ia32corei7/bld/i686-linux/./libmpx/mpxwrap
-L/export/gnu/import/git/gcc-test-ia32corei7/bld/i686-linux/./libmpx/mpxwrap/.libs
-fno-diagnostics-show-caret -fdiagnostics-color=never -O2 -mavx512ifma -lm -o
./avx512ifma-vpmaddhuq-2.exe^M
/export/gnu/import/git/gcc-test-ia32corei7/src-trunk/gcc/testsuite/gcc.target/i386/avx512ifma-vpmaddhuq-2.c:
In function 'test_512':^M
/export/gnu/import/git/gcc-test-ia32corei7/src-trunk/gcc/testsuite/gcc.target/i386/avx512ifma-vpmaddhuq-2.c:41:29:
warning: iteration 4 invokes undefined behavior
[-Waggressive-loop-optimizations]^M
/export/gnu/import/git/gcc-test-ia32corei7/src-trunk/gcc/testsuite/gcc.target/i386/avx512ifma-vpmaddhuq-2.c:36:3:
note: within this loop^M
output is:
/export/gnu/import/git/gcc-test-ia32corei7/src-trunk/gcc/testsuite/gcc.target/i386/avx512ifma-vpmaddhuq-2.c:
In function 'test_512':^M
/export/gnu/import/git/gcc-test-ia32corei7/src-trunk/gcc/testsuite/gcc.target/i386/avx512ifma-vpmaddhuq-2.c:41:29:
warning: iteration 4 invokes undefined behavior
[-Waggressive-loop-optimizations]^M
/export/gnu/import/git/gcc-test-ia32corei7/src-trunk/gcc/testsuite/gcc.target/i386/avx512ifma-vpmaddhuq-2.c:36:3:
note: within this loop^M

FAIL: gcc.target/i386/avx512ifma-vpmaddhuq-2.c (test for excess errors)
Excess errors:
/export/gnu/import/git/gcc-test-ia32corei7/src-trunk/gcc/testsuite/gcc.target/i386/avx512ifma-vpmaddhuq-2.c:41:29:
warning: iteration 4 invokes undefined behavior
[-Waggressive-loop-optimizations]

r229395 is OK.

[Bug middle-end/68112] [6 Regression] FAIL: gcc.target/i386/avx512ifma-vpmaddhuq-2.c (test for excess errors)

[hjl.tools at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=68112

H.J. Lu <hjl.tools at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2015-10-27
                 CC|                            |alalaw01 at gcc dot gnu.org
          Component|testsuite                   |middle-end
   Target Milestone|---                         |6.0
     Ever confirmed|0                           |1

--- Comment #1 from H.J. Lu <hjl.tools at gmail dot com> ---
It was caused by r229437.

[Bug middle-end/68112] [6 Regression] FAIL: gcc.target/i386/avx512ifma-vpmaddhuq-2.c (test for excess errors)

[alalaw01 at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=68112

--- Comment #2 from alalaw01 at gcc dot gnu.org ---
So (a << CONSTANT) is not equivalent to a * (1<<CONSTANT), as the former is
well-defined, whereas the latter invokes UB if bits would have been shifted off
the far end.

[Bug middle-end/68112] [6 Regression] FAIL: gcc.target/i386/avx512ifma-vpmaddhuq-2.c (test for excess errors)

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=68112

--- Comment #3 from Richard Biener <rguenth at gcc dot gnu.org> ---
Not exactly sure but at least GCC doesn't take advantage of any undefinedness
in
left-shifts (apart from the shift count being in-range for the type).

[Bug middle-end/68112] [6 Regression] FAIL: gcc.target/i386/avx512ifma-vpmaddhuq-2.c (test for excess errors)

[alalaw01 at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=68112

--- Comment #4 from alalaw01 at gcc dot gnu.org ---
Sure, but gcc exploits undefinedness of multiply, so rewriting shift to
multiply is not equivalent in the general case :(.

One way forward might be to make definedness of overflow a bit finer-grained
(either on types, i.e. TYPE_OVERFLOW_DEFINED, or maybe as a property of
chrecs?)