public inbox for gcc-bugs@sourceware.org
help / color / mirror / Atom feed
* [Bug target/114906] New: aarch64 locally_streaming ICE in aarch64_expand_prologue
@ 2024-05-01 13:59 matmal01 at gcc dot gnu.org
  2024-06-15 23:40 ` [Bug target/114906] aarch64 locally_streaming ICE with -fstack-clash-protection " pinskia at gcc dot gnu.org
  0 siblings, 1 reply; 2+ messages in thread
From: matmal01 at gcc dot gnu.org @ 2024-05-01 13:59 UTC (permalink / raw)
  To: gcc-bugs

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114906

            Bug ID: 114906
           Summary: aarch64 locally_streaming ICE in
                    aarch64_expand_prologue
           Product: gcc
           Version: unknown
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: target
          Assignee: unassigned at gcc dot gnu.org
          Reporter: matmal01 at gcc dot gnu.org
  Target Milestone: ---

Bug (testcase + ICE) below.
I believe this is because:
1) We save `r20` below `VG_REGNUM` in `aarch64_layout_frame` (and above the
point that `bytes_below_hard_fp` describes).
2) Despite that save of `r20` causing us to also set
`frame.wb_push_candidate1`, because we have a poly-int sized frame (due to the
-O0 in this case, but I don't think has to be -O0) we still end up in the
"General case" in `aarch64_layout_frame`.
3) Hence we end up with `initial_adjust` non zero, `sve_callee_adjust`
non-zero, and the `VG_REGNUM` not pointing to the same place as
`bytes_below_hard_fp` because there is that r20 saved in between.

My initial guess would be that we should simply change the assertion that
failed to check that VG_REGNUM is *greater than or equal to* `bytes_below_sp`.
To be honest I'm not entirely sure what this assertion is there for so would
not like to actually make that suggestion.  The commit message of ad4df8cd080c
seems to say the assertion is there to ensure that the allocation of VG_REGNUM
is not folded into the initial_allocation, but I don't 100% follow what's going
on.


vshcmd: > cat ../streaming-prologues.c                                          
[[arm::locally_streaming]] void                                                 
with_callee_saved_regs (__SVBool_t x)                                           
{                                                                               
  asm ("" : : : "r20");                                                         
}                                                                               
testing [14:47:20] $                                                            
vshcmd: > ${install_dir}/aarch64-none-linux-gnu-gcc \                           
vshcmd: >       ../streaming-prologues.c \                                      
vshcmd: >       -fdiagnostics-plain-output -O0 -fstack-clash-protection \       
vshcmd: >       -march=armv9-a+sme -mtune=generic -moverride=tune=none \        
vshcmd: >       -S -o prologues-with-streaming-1.s                              
> > > > during RTL pass: late_pro_and_epilogue                                                                                                                                                                                                                                                                                 
../streaming-prologues.c: In function ‘with_callee_saved_regs’:                 
../streaming-prologues.c:5:1: internal compiler error: in
aarch64_expand_prologue, at config/aarch64/aarch64.cc:9705                      
0x142e3af aarch64_expand_prologue()                                             
       
/workspace/GNU-toolchain/fsf-trunk/src/gcc/gcc/config/aarch64/aarch64.cc:9701   
0x1a7eee7 gen_prologue()                                                        
       
/workspace/GNU-toolchain/fsf-trunk/src/gcc/gcc/config/aarch64/aarch64.md:1008   
0x140219f target_gen_prologue                                                   
       
/workspace/GNU-toolchain/fsf-trunk/src/gcc/gcc/config/aarch64/aarch64.md:8121   
0xb8d242 make_prologue_seq                                                      
        /workspace/GNU-toolchain/fsf-trunk/src/gcc/gcc/function.cc:5818         
0xb8d3aa thread_prologue_and_epilogue_insns()                                   
        /workspace/GNU-toolchain/fsf-trunk/src/gcc/gcc/function.cc:6053         
0xb8dc4e rest_of_handle_thread_prologue_and_epilogue                            
        /workspace/GNU-toolchain/fsf-trunk/src/gcc/gcc/function.cc:6567         
0xb8dcbf execute                                                                
        /workspace/GNU-toolchain/fsf-trunk/src/gcc/gcc/function.cc:6692         
Please submit a full bug report, with preprocessed source (by using
-freport-bug).                                                                  
Please include the complete backtrace with any bug report.                      
See <https://gcc.gnu.org/bugs/> for instructions.                               
testing [14:47:22] $

^ permalink raw reply	[flat|nested] 2+ messages in thread

* [Bug target/114906] aarch64 locally_streaming ICE with -fstack-clash-protection in aarch64_expand_prologue
  2024-05-01 13:59 [Bug target/114906] New: aarch64 locally_streaming ICE in aarch64_expand_prologue matmal01 at gcc dot gnu.org
@ 2024-06-15 23:40 ` pinskia at gcc dot gnu.org
  0 siblings, 0 replies; 2+ messages in thread
From: pinskia at gcc dot gnu.org @ 2024-06-15 23:40 UTC (permalink / raw)
  To: gcc-bugs

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114906

Andrew Pinski <pinskia at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
     Ever confirmed|0                           |1
   Last reconfirmed|                            |2024-06-15
            Summary|aarch64 locally_streaming   |aarch64 locally_streaming
                   |ICE in                      |ICE with
                   |aarch64_expand_prologue     |-fstack-clash-protection in
                   |                            |aarch64_expand_prologue
             Status|UNCONFIRMED                 |NEW

--- Comment #1 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Confirmed.

^ permalink raw reply	[flat|nested] 2+ messages in thread

end of thread, other threads:[~2024-06-15 23:40 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2024-05-01 13:59 [Bug target/114906] New: aarch64 locally_streaming ICE in aarch64_expand_prologue matmal01 at gcc dot gnu.org
2024-06-15 23:40 ` [Bug target/114906] aarch64 locally_streaming ICE with -fstack-clash-protection " pinskia at gcc dot gnu.org

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).