| Message ID | 20240711142919.1328689-1-stefansf@gcc.gnu.org |
|---|---|
| State | New |
| Headers | show |
| Series | s390: Align *cjump_64 and *icjump_64 | expand |
On Thu, Jul 11, 2024 at 04:29:19PM +0200, Stefan Schulze Frielinghaus wrote: > During machine reorg we optimize backward jumps and transform insns as > e.g. > > (jump_insn 118 117 119 (set (pc) > (if_then_else (ne (reg:CCRAW 33 %cc) > (const_int 8 [0x8])) > (label_ref 134) > (pc))) "dec_math_1.f90":204:8 discrim 1 2161 {*cjump_64} > (expr_list:REG_DEAD (reg:CCRAW 33 %cc) > (int_list:REG_BR_PROB 719407028 (nil))) > -> 134) > > into > > (jump_insn 118 117 432 (set (pc) > (if_then_else (ne (reg:CCRAW 33 %cc) > (const_int 8 [0x8])) > (pc) > (label_ref 433))) "dec_math_1.f90":204:8 discrim 1 -1 > (expr_list:REG_DEAD (reg:CCRAW 33 %cc) > (int_list:REG_BR_PROB 719407028 (nil))) > -> 433) > > The latter is not recognized anymore since *icjump_64 only matches > CC_REGNUM against zero. Fixed by aligning *cjump_64 and *icjump_64. > > gcc/ChangeLog: > > * config/s390/s390.md (*icjump_64): Allow raw CC comparisons, > i.e., any constant integer between 0 and 15 for CC comparisons. > --- > Bootstrap and regtest or still running. Assuming no regressions, ok > for {mainline,11,12,13,14}? Would be great to see this in 14.2 RC :) I didn't have the schedule for 11.5 RC in mind which is tomorrow and the release a week afterwards. I hope this is still appropriate for 11.5? Cheers, Stefan > > gcc/config/s390/s390.md | 3 ++- > 1 file changed, 2 insertions(+), 1 deletion(-) > > diff --git a/gcc/config/s390/s390.md b/gcc/config/s390/s390.md > index f5d7003dfad..d3931b09417 100644 > --- a/gcc/config/s390/s390.md > +++ b/gcc/config/s390/s390.md > @@ -9556,7 +9556,8 @@ > (define_insn "*icjump_64" > [(set (pc) > (if_then_else > - (match_operator 1 "s390_comparison" [(reg CC_REGNUM) (const_int 0)]) > + (match_operator 1 "s390_comparison" [(reg CC_REGNUM) > + (match_operand 2 "const_int_operand" "")]) > (pc) > (label_ref (match_operand 0 "" ""))))] > "" > -- > 2.45.2 >
On Thu, Jul 11, 2024 at 05:09:41PM +0200, Stefan Schulze Frielinghaus wrote: > I didn't have the schedule for 11.5 RC in mind which is tomorrow and the > release a week afterwards. I hope this is still appropriate for 11.5? From my side, if Andreas or somebody else approves it, it is tested on 11 branch and committed by tomorrow, it can be added. But I'd like to know what patches I should wait for tomorrow and approximate ETA (and ideally before end of working day in Europe). Once rc1 is done, only severe blockers will be possible. Jakub
On 7/11/24 16:29, Stefan Schulze Frielinghaus wrote: > During machine reorg we optimize backward jumps and transform insns as > e.g. > > (jump_insn 118 117 119 (set (pc) > (if_then_else (ne (reg:CCRAW 33 %cc) > (const_int 8 [0x8])) > (label_ref 134) > (pc))) "dec_math_1.f90":204:8 discrim 1 2161 {*cjump_64} > (expr_list:REG_DEAD (reg:CCRAW 33 %cc) > (int_list:REG_BR_PROB 719407028 (nil))) > -> 134) > > into > > (jump_insn 118 117 432 (set (pc) > (if_then_else (ne (reg:CCRAW 33 %cc) > (const_int 8 [0x8])) > (pc) > (label_ref 433))) "dec_math_1.f90":204:8 discrim 1 -1 > (expr_list:REG_DEAD (reg:CCRAW 33 %cc) > (int_list:REG_BR_PROB 719407028 (nil))) > -> 433) > > The latter is not recognized anymore since *icjump_64 only matches > CC_REGNUM against zero. Fixed by aligning *cjump_64 and *icjump_64. > > gcc/ChangeLog: > > * config/s390/s390.md (*icjump_64): Allow raw CC comparisons, > i.e., any constant integer between 0 and 15 for CC comparisons. Ok. Thanks! Andreas
On Thu, Jul 11, 2024 at 05:14:58PM +0200, Jakub Jelinek wrote: > On Thu, Jul 11, 2024 at 05:09:41PM +0200, Stefan Schulze Frielinghaus wrote: > > I didn't have the schedule for 11.5 RC in mind which is tomorrow and the > > release a week afterwards. I hope this is still appropriate for 11.5? > > From my side, if Andreas or somebody else approves it, it is tested on 11 > branch and committed by tomorrow, it can be added. > But I'd like to know what patches I should wait for tomorrow and approximate > ETA (and ideally before end of working day in Europe). Once rc1 is done, only > severe blockers will be possible. The tester is running over night and will finish around 7 AM CEST. I will let you know once it has finished. If anything goes wrong we can skip this patch of course. Cheers, Stefan
On Thu, Jul 11, 2024 at 07:32:17PM +0200, Stefan Schulze Frielinghaus wrote: > On Thu, Jul 11, 2024 at 05:14:58PM +0200, Jakub Jelinek wrote: > > On Thu, Jul 11, 2024 at 05:09:41PM +0200, Stefan Schulze Frielinghaus wrote: > > > I didn't have the schedule for 11.5 RC in mind which is tomorrow and the > > > release a week afterwards. I hope this is still appropriate for 11.5? > > > > From my side, if Andreas or somebody else approves it, it is tested on 11 > > branch and committed by tomorrow, it can be added. > > But I'd like to know what patches I should wait for tomorrow and approximate > > ETA (and ideally before end of working day in Europe). Once rc1 is done, only > > severe blockers will be possible. > > The tester is running over night and will finish around 7 AM CEST. I > will let you know once it has finished. If anything goes wrong we can > skip this patch of course. The tester was extremely slow this time and still didn't finish. I don't wanna rush it risking to introduce late time problems for the 11.5 release. Since I'm testing for three different architectures and the first one hasn't finished, let's drop this patch for 11.5. Sorry for the noise, Stefan
On Fri, Jul 12, 2024 at 07:09:37AM +0200, Stefan Schulze Frielinghaus wrote: > On Thu, Jul 11, 2024 at 07:32:17PM +0200, Stefan Schulze Frielinghaus wrote: > > On Thu, Jul 11, 2024 at 05:14:58PM +0200, Jakub Jelinek wrote: > > > On Thu, Jul 11, 2024 at 05:09:41PM +0200, Stefan Schulze Frielinghaus wrote: > > > > I didn't have the schedule for 11.5 RC in mind which is tomorrow and the > > > > release a week afterwards. I hope this is still appropriate for 11.5? > > > > > > From my side, if Andreas or somebody else approves it, it is tested on 11 > > > branch and committed by tomorrow, it can be added. > > > But I'd like to know what patches I should wait for tomorrow and approximate > > > ETA (and ideally before end of working day in Europe). Once rc1 is done, only > > > severe blockers will be possible. > > > > The tester is running over night and will finish around 7 AM CEST. I > > will let you know once it has finished. If anything goes wrong we can > > skip this patch of course. > > The tester was extremely slow this time and still didn't finish. I > don't wanna rush it risking to introduce late time problems for the > 11.5 release. Since I'm testing for three different architectures and > the first one hasn't finished, let's drop this patch for 11.5. We haven't started the RC yet, so if it finished testing by now, feel free to commit within the next 15 minutes or so. Jakub
diff --git a/gcc/config/s390/s390.md b/gcc/config/s390/s390.md index f5d7003dfad..d3931b09417 100644 --- a/gcc/config/s390/s390.md +++ b/gcc/config/s390/s390.md @@ -9556,7 +9556,8 @@ (define_insn "*icjump_64" [(set (pc) (if_then_else - (match_operator 1 "s390_comparison" [(reg CC_REGNUM) (const_int 0)]) + (match_operator 1 "s390_comparison" [(reg CC_REGNUM) + (match_operand 2 "const_int_operand" "")]) (pc) (label_ref (match_operand 0 "" ""))))] ""
During machine reorg we optimize backward jumps and transform insns as e.g. (jump_insn 118 117 119 (set (pc) (if_then_else (ne (reg:CCRAW 33 %cc) (const_int 8 [0x8])) (label_ref 134) (pc))) "dec_math_1.f90":204:8 discrim 1 2161 {*cjump_64} (expr_list:REG_DEAD (reg:CCRAW 33 %cc) (int_list:REG_BR_PROB 719407028 (nil))) -> 134) into (jump_insn 118 117 432 (set (pc) (if_then_else (ne (reg:CCRAW 33 %cc) (const_int 8 [0x8])) (pc) (label_ref 433))) "dec_math_1.f90":204:8 discrim 1 -1 (expr_list:REG_DEAD (reg:CCRAW 33 %cc) (int_list:REG_BR_PROB 719407028 (nil))) -> 433) The latter is not recognized anymore since *icjump_64 only matches CC_REGNUM against zero. Fixed by aligning *cjump_64 and *icjump_64. gcc/ChangeLog: * config/s390/s390.md (*icjump_64): Allow raw CC comparisons, i.e., any constant integer between 0 and 15 for CC comparisons. --- Bootstrap and regtest or still running. Assuming no regressions, ok for {mainline,11,12,13,14}? Would be great to see this in 14.2 RC :) gcc/config/s390/s390.md | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)