[Bug libgcc/111372] New: libgcc: RISCV C++ exception handling stack usage grew in 13.1

[Bug libgcc/111372] New: libgcc: RISCV C++ exception handling stack usage grew in 13.1

[alexey.lapshin at espressif dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111372

            Bug ID: 111372
           Summary: libgcc: RISCV C++ exception handling stack usage grew
                    in 13.1
           Product: gcc
           Version: 13.2.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: libgcc
          Assignee: unassigned at gcc dot gnu.org
          Reporter: alexey.lapshin at espressif dot com
                CC: juzhe.zhong at rivai dot ai
  Target Milestone: ---

Created attachment 55877
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=55877&action=edit
0001-libgcc-reduce-DWARF_FRAME_REGISTERS-to-use-only-base-registers.patch

After vector registers were added (commit
31380d4b7925912b9c34afe8cdb8dffe6cd50b94) c++ exception handling stack
consumption grew too large.
(GCC 12.2) ~1700 bytes vs (GCC 13.2) ~2800 bytes of stack required. This issue
may be critical for embedded projects that use C++.

As a temporary solution, I have patched gcc (see attached file).


It seems better to define macros (FIRST_PSEUDO_REGISTER, FIXED_REGISTERS...)
depending on "vector" extension is set or not. This would also prevent future
bugs.

[Bug libgcc/111372] libgcc: RISCV C++ exception handling stack usage grew in 13.1

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111372

--- Comment #1 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
(In reply to Alexey from comment #0) 
> It seems better to define macros (FIRST_PSEUDO_REGISTER, FIXED_REGISTERS...)
> depending on "vector" extension is set or not. This would also prevent
> future bugs.

Except that is broken if you mix code that uses and does not use vectors.
Especially when the vector extension does not need to be compiled in libgcc ...

[Bug libgcc/111372] libgcc: RISCV C++ exception handling stack usage grew in 13.1

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111372

--- Comment #2 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
(In reply to Alexey from comment #0)
> It seems better to define macros (FIRST_PSEUDO_REGISTER, FIXED_REGISTERS...)
> depending on "vector" extension is set or not. This would also prevent
> future bugs.

Also that especially does not work with LTO where some functions can have
vectors enabled and some don't and a lot of the RTL code does not really get
reinitialized with those kind of differences ...

[Bug target/111372] libgcc: RISCV C++ exception handling stack usage grew in 13.1

[alexey.lapshin at espressif dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111372

--- Comment #3 from Alexey <alexey.lapshin at espressif dot com> ---
Ok, but it's better to have configure option or something else just for
toolchains that definitely do not use vector extension

[Bug target/111372] libgcc: RISCV C++ exception handling stack usage grew in 13.1

[juzhe.zhong at rivai dot ai]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111372

--- Comment #4 from JuzheZhong <juzhe.zhong at rivai dot ai> ---
I am not familiar with this stuff.
I would like to see whether kito has good suggestions.

[Bug target/111372] libgcc: RISCV C++ exception handling stack usage grew in 13.1

[kito at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111372

--- Comment #5 from Kito Cheng <kito at gcc dot gnu.org> ---
> Ok, but it's better to have configure option or something else just
> for toolchains that definitely do not use vector extension

I can understand that there would be such a demand in the embedded world, but
that's not critical issue, so this won't get high priority to most RISC-V GCC
developer, it would be appreciate if you could send a patch for that.