On 5/17/18 3:31 PM, Segher Boessenkool wrote: > On Wed, May 16, 2018 at 12:53:13PM -0700, Carl Love wrote: >> @@ -27,21 +27,21 @@ >> /* { dg-final { scan-assembler-times "vmulosb" 1 } } */ >> >> // For LE platforms P9 and later, we generate the lxv insn instead of >> lxvd2x. >> -/* { dg-final { scan-assembler-times {\mlxvd2x\M} 0 { target { { >> powerpc64*le-*-* } && { p9vector_hw } } } } } */ >> -/* { dg-final { scan-assembler-times {\mlxv\M} 36 { target { { >> powerpc64*le-*-* } && { p9vector_hw } } } } } */ >> +/* { dg-final { scan-assembler-times {lxvd2x} 0 { target { { >> powerpc64*le-*-* } && { p9vector_hw } } } } } */ >> +/* { dg-final { scan-assembler-times {lxv} 36 { target { { >> powerpc64*le-*-* } && { p9vector_hw } } } } } */ > > This {lxv} matches {lxvd2x} as well. \m\M in Tcl are like \b\b in Perl, > or \<\> in many other regex dialects.
The target triplet of powerpc64*le-*-* isn't modified by the patch, but the '*' in powerpc64*le seems superfluous, so can we just remove it? Peter