Re: PR target/70243: Do not generate fmaddfp and fnmsubfp

2023-04-07 Thread Segher Boessenkool
Hi! On Fri, Apr 07, 2023 at 02:32:04AM -0400, Michael Meissner wrote: > On Thu, Apr 06, 2023 at 03:37:59PM -0500, Segher Boessenkool wrote: > > > This patch eliminates the generation of the Altivec fmaddfp and fnmsubfp > > > instructions as alternatives in the VSX instruction insn support, and in

Re: PR target/70243: Do not generate fmaddfp and fnmsubfp

2023-04-07 Thread Michael Meissner via Gcc-patches
On Thu, Apr 06, 2023 at 03:37:59PM -0500, Segher Boessenkool wrote: > Hi! > > On Thu, Apr 06, 2023 at 11:12:11AM -0400, Michael Meissner wrote: > > The Altivec instructions fmaddfp and fnmsubfp have different rounding > > behaviors > > Those are not existing instructions. You mean "vmaddfp"

Re: PR target/70243: Do not generate fmaddfp and fnmsubfp

2023-04-06 Thread Segher Boessenkool
Hi! On Thu, Apr 06, 2023 at 11:12:11AM -0400, Michael Meissner wrote: > The Altivec instructions fmaddfp and fnmsubfp have different rounding > behaviors Those are not existing instructions. You mean "vmaddfp" etc. > than the VSX xvmaddsp and xvnmsubsp instructions. In particular, generating

PR target/70243: Do not generate fmaddfp and fnmsubfp

2023-04-06 Thread Michael Meissner via Gcc-patches
The Altivec instructions fmaddfp and fnmsubfp have different rounding behaviors than the VSX xvmaddsp and xvnmsubsp instructions. In particular, generating these instructions seems to break Eigen. GCC has generated the Altivec fmaddfp and fnmsubfp instructions on VSX systems as an alternative to