On Wed, Aug 26, 2026 at 4:38 PM Richard Biener <[email protected]> wrote: > > The following removes the scan for a addr32 vgather, but keeps the > negative scan for vgather. There is no reason for using a gather > in the first place and after hoisting the invariant load of the > ncost pointer load from the unswitched loop body we don't use one. > > Tested on x86_64-unknown-linux-gnu. > > OK? > > PR testsuite/127048 > * gcc.target/i386/pr89523-1a.c: Remove scan for addr32 vgather. > * gcc.target/i386/pr89523-1b.c: Likewise. > --- > gcc/testsuite/gcc.target/i386/pr89523-1a.c | 1 - > gcc/testsuite/gcc.target/i386/pr89523-1b.c | 1 - > 2 files changed, 2 deletions(-) > > diff --git a/gcc/testsuite/gcc.target/i386/pr89523-1a.c > b/gcc/testsuite/gcc.target/i386/pr89523-1a.c > index 0d0edab0363..98f8270f850 100644 > --- a/gcc/testsuite/gcc.target/i386/pr89523-1a.c > +++ b/gcc/testsuite/gcc.target/i386/pr89523-1a.c > @@ -2,7 +2,6 @@ > /* { dg-require-effective-target maybe_x32 } */ > /* { dg-options "-maddress-mode=short -mx32 -Ofast -funroll-loops > -march=haswell" } */ > /* { dg-final { scan-assembler-not "\tvgather" } } */ > -/* { dg-final { scan-assembler "addr32 vgather" } } */ > > void foo (void); > > diff --git a/gcc/testsuite/gcc.target/i386/pr89523-1b.c > b/gcc/testsuite/gcc.target/i386/pr89523-1b.c > index 6a5c1d43625..146566c3846 100644 > --- a/gcc/testsuite/gcc.target/i386/pr89523-1b.c > +++ b/gcc/testsuite/gcc.target/i386/pr89523-1b.c > @@ -2,6 +2,5 @@ > /* { dg-require-effective-target maybe_x32 } */ > /* { dg-options "-maddress-mode=long -mx32 -Ofast -funroll-loops > -march=haswell" } */ > /* { dg-final { scan-assembler-not "\tvgather" } } */ > -/* { dg-final { scan-assembler "addr32 vgather" } } */ > > #include "pr89523-1a.c" > -- > 2.51.0 >
LGTM. Thanks. -- H.J.
