Hi Carl, On Wed, Jun 10, 2020 at 09:05:36AM -0700, Carl Love wrote: > I committed this patch to mainline and backported to GCC 9. > > I have looked at GCC 8. The functional issue is there, i.e. the > vcmpnez is used instead of vcmpne. However the test case > builtins-8-p9-runnable.c does not exist in GCC 8. The patch consists > of the functional fix: > > --- a/gcc/config/rs6000/vsx.md > +++ b/gcc/config/rs6000/vsx.md > @@ -4803,8 +4803,8 @@ > rtx cmp_result = gen_reg_rtx (<MODE>mode); > rtx not_result = gen_reg_rtx (<MODE>mode); > > - emit_insn (gen_vcmpnez<VSX_EXTRACT_WIDTH> (cmp_result, operands[1], > - operands[2])); > + emit_insn (gen_vcmpne<VSX_EXTRACT_WIDTH> (cmp_result, operands[1], > + operands[2])); > emit_insn (gen_one_cmpl<mode>2 (not_result, cmp_result)); > > sh = GET_MODE_SIZE (GET_MODE_INNER (<MODE>mode)) / 2; > > So, I am a bit unsure how to proceed. I think we need the functional > change. But without applying the test cases fixes I don't feel that I > am really backporting the patch as approved for backporting. > > I think we could reference the mainline commit, as we always do when > backporting, but then note the test case fix was not included since the > testcase does not exist. Would that be OK?
Yes, that is fine, this fix is obvious enough :-) You could also include the whole testcase from trunk (or 9), but that can be quite a bit of (testing) work, and is it worth it at all. > Please let me know how you would like me to handle this issue. Either way is okay (I'd go for just the 2-line patch). Thanks, Segher
