On Mon, Aug 4, 2025 at 11:05 PM H.J. Lu <hjl.to...@gmail.com> wrote: > > On Mon, Aug 4, 2025 at 8:50 AM Richard Sandiford > <richard.sandif...@arm.com> wrote: > > > > Uros Bizjak <ubiz...@gmail.com> writes: > > > On Sat, Aug 2, 2025 at 8:56 PM H.J. Lu <hjl.to...@gmail.com> wrote: > > >> > > >> On Fri, Aug 1, 2025 at 10:32 PM Uros Bizjak <ubiz...@gmail.com> wrote: > > >> > > > >> > On Sat, Aug 2, 2025 at 3:22 AM H.J. Lu <hjl.to...@gmail.com> wrote: > > >> > > > > >> > > After > > >> > > > > >> > > commit 965564eafb721f8000013a3112f1bba8d8fae32b > > >> > > Author: Richard Sandiford <richard.sandif...@arm.com> > > >> > > Date: Tue Jul 29 15:58:34 2025 +0100 > > >> > > > > >> > > simplify-rtx: Simplify subregs of logic ops > > >> > > > > >> > > combine generates > > >> > > > > >> > > (set (zero_extract:SI (reg/v:SI 101 [ a ]) > > >> > > (const_int 8 [0x8]) > > >> > > (const_int 8 [0x8])) > > >> > > (not:SI (sign_extract:SI (reg:SI 107 [ b ]) > > >> > > (const_int 8 [0x8]) > > >> > > (const_int 8 [0x8])))) > > >> > > > > >> > > instead of > > >> > > > > >> > > (set (zero_extract:SI (reg/v:SI 101 [ a ]) > > >> > > (const_int 8 [0x8]) > > >> > > (const_int 8 [0x8])) > > >> > > (subreg:SI (not:QI (subreg:QI (sign_extract:SI (reg:SI 107 [ b ]) > > >> > > (const_int 8 [0x8]) > > >> > > (const_int 8 [0x8])) 0)) 0)) > > >> > > > > >> > > Add *one_cmplqi_ext<mode>_2 to support the new pattern. > > >> > > > > >> > > PR target/121306 > > >> > > * config/i386/i386.md (*one_cmplqi_ext<mode>_2): New. > > >> > > > >> > Why not just change the old pattern? I'd expect that the old form is > > >> > now obsolete. > > >> > > > >> > > >> *one_cmplqi_ext<mode>_1 is still needed. Otherwise combine will > > >> fail to match this instruction: > > >> > > >> (set (zero_extract:SI (reg/v:SI 102 [ a ]) > > >> (const_int 8 [0x8]) > > >> (const_int 8 [0x8])) > > >> (subreg:SI (not:QI (subreg:QI (reg:SI 105 [ _2 ]) 0)) 0)) > > > > > > Do you perhaps have a testcase that still shows this combination? > > > Perhaps the author of the simplification (CC'd) would be interested in > > > this missing case. > > > > > > I was under the impression that the new simplification should always > > > trigger, there is no point in having it if the backend still has to > > > provide simplified and non-simplified patterns. > > > > > > There are also other similar logic patterns in i386.md that would have > > > to be amended. > > > > Sorry, I hadn't realised that there were still unfixed regressions > > from that patch. I suppose if we wanted to avoid two patterns here, > > we'd need to extend the pre-existing word_mode folds to support > > subword modes too (for !WORD_REGISTER_OPERATIONS). The attached > > untested patch does that, but I expect it would have similar > > knock-on effects. I'll give it a spin overnight on x86 anyway > > just to see what happens. > > Yes, it fixes: > > FAIL: gcc.target/i386/pr82524.c scan-assembler-not mov[sz]bl > FAIL: gcc.target/i386/pr82524.c scan-assembler [ \t]notb > > together with the enclosed patch.
+ (match_op_dup 2 + [(match_dup 0) + (const_int 8) + (const_int 8)])))] IMO, it is better to leave arguments to match_op_dup together in one line, they fit into line character limit. OK for the x86 part. Thanks, Uros.
