On 2025-12-16 13:50, Rainer Orth wrote:
Hi Andi,
diff --git a/gcc/testsuite/gcc.target/i386/shift-gf2p8affine-2.c
b/gcc/testsuite/gcc.target/i386/shift-gf2p8affine-2.c
--- a/gcc/testsuite/gcc.target/i386/shift-gf2p8affine-2.c
+++ b/gcc/testsuite/gcc.target/i386/shift-gf2p8affine-2.c
@@ -1,5 +1,6 @@
/* { dg-do run } */
/* { dg-options "-mgfni -mavx512vl -mavx512bw -mavx512f -O3
-Wno-shift-count-negative -march=x86-64 -mtune=generic" } */
+/* { dg-skip-if "Assembler support missing" { *-*-solaris2.* && { !
gas } } } */
It seems it would be better to introduce some generic equivalent of
dg-effective-target for compile only tests instead of all these
special
cases.
good point: we already have all of gfni, avx512vl, avx512bw, and
avxv512f effective-target keywords. This test this should require all
of them, I guess.
effective-target is for what can be run. This is a compile test, so it
would overconstrain it.
But something similar for compile would be useful.
-Andi
