On 10/11/18 3:05 PM, Peter Bergner wrote: > On 10/11/18 2:40 PM, Jeff Law wrote: >> On 10/11/18 1:23 PM, Peter Bergner wrote: >>> On 10/11/18 1:18 PM, Peter Bergner wrote: >>>> Ok, after working in gdb, I see that the PA-RISC port still uses reload >>>> and not LRA, but it too seems to have the same issue of reusing input >>>> regs that have REG_DEAD notes, so the question still stands. It's just >>>> that whatever fix we come up with will have to be to both LRA and reload. >>> >>> On second thought, I'm thinking we should just leave reload alone and >>> only fix this in LRA. That means we'd have to disable the reg copy >>> handling when not using LRA though, which might be another reason to >>> get targets to move to LRA? I've verified the following patch gets >>> the PA-RISC test case to pass again. Thoughts? >>> >>> If ok, I still have to dig into the fails we're seeing on LRA targets. >> Hmmm. Interesting. I wonder if all the failing targets were reload >> targets..... If so, this may be the way forward -- I certainly don't >> want to spend much, if any, time fixing reload. >> >> I'm in the middle of something, but will try to look at each of the >> failing targets and confirm they use reload by default. > > These are the easy ones (they default to reload): > > bergner@pike:~/gcc/gcc-fsf-mainline/gcc/config$ grep -r TARGET_LRA_P | grep > false | sort > alpha/alpha.c:#define TARGET_LRA_P hook_bool_void_false > avr/avr.c:#define TARGET_LRA_P hook_bool_void_false > bfin/bfin.c:#define TARGET_LRA_P hook_bool_void_false > c6x/c6x.c:#define TARGET_LRA_P hook_bool_void_false > cr16/cr16.c:#define TARGET_LRA_P hook_bool_void_false > cris/cris.c:#define TARGET_LRA_P hook_bool_void_false > epiphany/epiphany.c:#define TARGET_LRA_P hook_bool_void_false > fr30/fr30.c:#define TARGET_LRA_P hook_bool_void_false > frv/frv.c:#define TARGET_LRA_P hook_bool_void_false > h8300/h8300.c:#define TARGET_LRA_P hook_bool_void_false > ia64/ia64.c:#define TARGET_LRA_P hook_bool_void_false > iq2000/iq2000.c:#define TARGET_LRA_P hook_bool_void_false > lm32/lm32.c:#define TARGET_LRA_P hook_bool_void_false > m32c/m32c.c:#define TARGET_LRA_P hook_bool_void_false > m32r/m32r.c:#define TARGET_LRA_P hook_bool_void_false > m68k/m68k.c:#define TARGET_LRA_P hook_bool_void_false > mcore/mcore.c:#define TARGET_LRA_P hook_bool_void_false > microblaze/microblaze.c:#define TARGET_LRA_P hook_bool_void_false > mmix/mmix.c:#define TARGET_LRA_P hook_bool_void_false > mn10300/mn10300.c:#define TARGET_LRA_P hook_bool_void_false > moxie/moxie.c:#define TARGET_LRA_P hook_bool_void_false > msp430/msp430.c:#define TARGET_LRA_P hook_bool_void_false > nvptx/nvptx.c:#define TARGET_LRA_P hook_bool_void_false > pa/pa.c:#define TARGET_LRA_P hook_bool_void_false > rl78/rl78.c:#define TARGET_LRA_P hook_bool_void_false > spu/spu.c:#define TARGET_LRA_P hook_bool_void_false > stormy16/stormy16.c:#define TARGET_LRA_P hook_bool_void_false > tilegx/tilegx.c:#define TARGET_LRA_P hook_bool_void_false > tilepro/tilepro.c:#define TARGET_LRA_P hook_bool_void_false > vax/vax.c:#define TARGET_LRA_P hook_bool_void_false > visium/visium.c:#define TARGET_LRA_P hook_bool_void_false > xtensa/xtensa.c:#define TARGET_LRA_P hook_bool_void_false > > These are harder since they support -mlra: > > arc/arc.c:#define TARGET_LRA_P arc_lra_p > ft32/ft32.c:#define TARGET_LRA_P ft32_lra_p > mips/mips.c:#define TARGET_LRA_P mips_lra_p > pdp11/pdp11.c:#define TARGET_LRA_P pdp11_lra_p > powerpcspe/powerpcspe.c:#define TARGET_LRA_P rs6000_lra_p > rx/rx.c:#define TARGET_LRA_P rx_enable_lra > s390/s390.c:#define TARGET_LRA_P s390_lra_p > sh/sh.c:#define TARGET_LRA_P sh_lra_p > sparc/sparc.c:#define TARGET_LRA_P sparc_lra_p > > Quickly looking into their *.opt files, the follwoing default to LRA: > mips, s390 So the failing targets were aarch64, alpha, arm, sh4, s390, alpha and hppa. In theory your patch has a reasonable chance of fixing sh4, alpha and hppa. So I suspect we're still going to have the aarch64, arm and s390 issues.
I've had my tester turned off while we sorted this out. I'll put your patch to disable the conflict pruning for non-LRA targets and see what we get overnight. jeff
