Re: [RFA] Minor improvement to coremark, avoid unconditional jump to return

[Jeff Law <jlaw@ventanamicro.com>]

On 10/7/22 04:51, Franz Sirl wrote:
> Am 2022-09-25 um 18:28 schrieb Jeff Law:
>> This is a minor improvement for the core_list_find routine in coremark.
>>
>>
>> Basically for riscv, and likely other targets, we can end up with an 
>> unconditional jump to a return statement.    This is a result of 
>> compensation code created by bb-reorder, and no jump optimization 
>> pass runs after bb-reorder to clean this stuff up.
>>
>> This patch utilizes preexisting code to identify suitable branch 
>> targets as well as preexisting code to emit a suitable return, so 
>> it's pretty simple.  Note that when we arrange to do this 
>> optimization, the original return block may become unreachable. So we 
>> conditionally call delete_unreachable_blocks to fix that up.
>>
>> This triggers ~160 times during an x86_64 bootstrap.  Naturally it 
>> bootstraps and regression tests on x86_64.
>>
>> I've also bootstrapped this on riscv64, regression testing with qemu 
>> shows some regressions, but AFAICT they're actually qemu bugs with 
>> signal handling/delivery -- qemu user mode emulation is not 
>> consistently calling user defined signal handlers.  Given the same 
>> binary, sometimes they'll get called and the test passes, other times 
>> the handler isn't called and the test (of course) fails. I'll 
>> probably spend some time to try and chase this down for the sake of 
>> making testing easier.
>>
>>
>> OK for the trunk?
>
> Hello Jeff,
>
> I've bisected this change to break a "profiledbootstrap" on x86_64 
> like that:
>
> make[3]: Entering directory 
> '/home/fsirl/rpmbuild/BUILD/gcc-13.0.0+gitr13+2871/obj-x86_64-suse-linux/gcc'
> /home/fsirl/rpmbuild/BUILD/gcc-13.0.0+gitr13+2871/obj-x86_64-suse-linux/./prev-gcc/xg++ 
> -B/home/fsirl/rpmbuild/BUILD/gcc-13.0.0+gitr13+2871/obj-x86_64-suse-linux/./prev-gcc/ 
> -B/usr/x86_64-suse-linux/
> bin/ -nostdinc++ 
> -B/home/fsirl/rpmbuild/BUILD/gcc-13.0.0+gitr13+2871/obj-x86_64-suse-linux/prev-x86_64-suse-linux/libstdc++-v3/src/.libs 
> -B/home/fsirl/rpmbuild/BUILD/gcc-13.0.0+gitr13+2871/obj-x86_64
> -suse-linux/prev-x86_64-suse-linux/libstdc++-v3/libsupc++/.libs 
> -I/home/fsirl/rpmbuild/BUILD/gcc-13.0.0+gitr13+2871/obj-x86_64-suse-linux/prev-x86_64-suse-linux/libstdc++-v3/include/x86_64-suse-linu
> x 
> -I/home/fsirl/rpmbuild/BUILD/gcc-13.0.0+gitr13+2871/obj-x86_64-suse-linux/prev-x86_64-suse-linux/libstdc++-v3/include 
>
> -I/home/fsirl/rpmbuild/BUILD/gcc-13.0.0+gitr13+2871/libstdc++-v3/libsupc++ 
> -L
> /home/fsirl/rpmbuild/BUILD/gcc-13.0.0+gitr13+2871/obj-x86_64-suse-linux/prev-x86_64-suse-linux/libstdc++-v3/src/.libs 
> -L/home/fsirl/rpmbuild/BUILD/gcc-13.0.0+gitr13+2871/obj-x86_64-suse-linux/prev-x8 
>
> 6_64-suse-linux/libstdc++-v3/libsupc++/.libs  -fno-PIE -c   -O2 -g 
> -fmessage-length=0 -D_FORTIFY_SOURCE=2 -funwind-tables 
> -fasynchronous-unwind-tables -U_FORTIFY_SOURCE -fprofile-use 
> -fprofile-reprod
> ucible=parallel-runs -DIN_GCC     -fno-exceptions -fno-rtti 
> -fasynchronous-unwind-tables -W -Wall -Wno-narrowing -Wwrite-strings 
> -Wcast-qual -Wmissing-format-attribute -Woverloaded-virtual -pedantic
> -Wno-long-long -Wno-variadic-macros -Wno-overlength-strings 
> -fno-common -DHAVE_CONFIG_H -I. -I. -I../../gcc -I../../gcc/. 
> -I../../gcc/../include -I../../gcc/../libcpp/include 
> -I../../gcc/../libcody
>  -I../../gcc/../libdecnumber -I../../gcc/../libdecnumber/bid 
> -I../libdecnumber -I../../gcc/../libbacktrace   -o cgraph.o -MT 
> cgraph.o -MMD -MP -MF ./.deps/cgraph.TPo ../../gcc/cgraph.cc
> ../../gcc/cgraph.cc: In member function 'cgraph_edge* 
> cgraph_edge::first_speculative_call_target()':
> ../../gcc/cgraph.cc:1166:1: error: EDGE_CROSSING incorrectly set 
> across same section
>  1166 | }
>       | ^
> ../../gcc/cgraph.cc:1166:1: error: No region crossing jump at section 
> boundary in bb 19
> during RTL pass: bbro
> ../../gcc/cgraph.cc:1166:1: internal compiler error: verify_flow_info 
> failed
> 0xa7116e verify_flow_info()
>         ../../gcc/cfghooks.cc:284
> 0x1c64958 execute
>         ../../gcc/bb-reorder.cc:2663
>
> In such a case, what do you need to reproduce it? I'm a mere user of 
> the Suse RPM builds here, no idea if profiling needs any extra data to 
> reproduce to bug.

Franz, it's been a long time.  Good to hear from you.


I'll take a look.  I'll start with a profiledbootstrap and if that 
doesn't reproduce I'll get in contact


jeff