Re: [PATCH v2] MATCH: fold signbit comparison and conditional negate to copysign [PR109843]

On Fri, Jul 31, 2026 at 2:33 AM Jakub Jelinek <[email protected]> wrote:

> On Fri, Jul 24, 2026 at 05:14:11PM -0600, Jeffrey Law wrote:
> >
> >
> > On 7/23/2026 1:11 AM, Eikansh Gupta wrote:
> > > Fold (signbit (x) cmp1 0) cmp (signbit (y) cmp2 0) ? y : -y
> > > to copysign (y, +-x).  The result keeps the magnitude of Y and takes
> its
> > > sign from X (or -X).  Emitted as IFN_COPYSIGN when the target supports
> it.
> > >
> > >     PR tree-optimization/109843
> > >
> > > gcc/ChangeLog:
> > >
> > >     * match.pd ((signbit (x) cmp 0) cmp (signbit (y) cmp 0) ? y : -y):
> > >     New simplification to copysign (y, +-x).
> > >
> > > gcc/testsuite/ChangeLog:
> > >
> > >     * gcc.dg/tree-ssa/pr109843.c: New test.
> > >
> > > Signed-off-by: Eikansh Gupta <[email protected]>
> > OK.  Please commit to the trunk.
>
> This test FAILs for me, both on i686-linux and on x86_64-linux too,
> on the latter with
> RUNTESTFLAGS='--target_board=unix\{-m32,-m64\} tree-ssa.exp=pr109843.c'
>
> PASS: gcc.dg/tree-ssa/pr112472.c (test for excess errors)
> PASS: gcc.dg/tree-ssa/pr112472.c scan-tree-dump-not optimized "=
> __builtin_copysign"
> FAIL: gcc.dg/tree-ssa/pr112472.c scan-tree-dump-not optimized " double "
> gcc.dg/tree-ssa/pr112472.c: pattern found 0 times
> FAIL: gcc.dg/tree-ssa/pr112472.c scan-tree-dump-times optimized
> ".COPYSIGN" 1
> gcc.dg/tree-ssa/pr112472.c: pattern found 0 times
> FAIL: gcc.dg/tree-ssa/pr112472.c scan-tree-dump-times optimized
> "-1.0e\\+0" 1
> gcc.dg/tree-ssa/pr112472.c: pattern found 2 times
> FAIL: gcc.dg/tree-ssa/pr112472.c scan-tree-dump-times optimized " ABS_EXPR
> " 1
>
> during the lower pass __builtin_copysign is lowered to -ABS_EXPR etc.
>
>         Jakub
>
> I have pushed a fix for both the test cases.

Regards,
Eikansh