On Wed, Feb 05, 2020 at 10:58:04AM +0100, Jakub Jelinek wrote: > Attached are 3 possible incremental patches for recog.c, all of them fix > all the above regressions, but haven't fully bootstrapped/regtested any of > them yet. My preference would be the last one, which for -O0 and x86 > disables split2 and enables split3, as it doesn't add any extra passes. > The first one just enables split3 for -O0 on x86, the second one enables > split5 for -O0 on x86.
The -da -O0 changes of the last patch are: test.c.275r.split1 test.c.277r.dfinit test.c.278r.mode_sw test.c.279r.asmcons test.c.284r.ira test.c.285r.reload -test.c.289r.split2 test.c.292r.pro_and_epilogue test.c.295r.jump2 +test.c.306r.split3 test.c.307r.stack test.c.308r.alignments test.c.310r.mach test.c.311r.barriers test.c.316r.shorten test.c.317r.nothrow test.c.318r.dwarf2 test.c.319r.final test.c.320r.dfinish while the first one just adds test.c.306r.split3 and doesn't remove test.c.289r.split2 and the second one just adds test.c.313r.split5 before test.c.316r.shorten. Jakub