[Bug target/103396] New: [GCN][BUILD] ICE RTL check: access of elt 4 of vector with last elt 3 in move_callee_saved_registers, at config/gcn/gcn.c:2821

[Bug target/103396] New: [GCN][BUILD] ICE RTL check: access of elt 4 of vector with last elt 3 in move_callee_saved_registers, at config/gcn/gcn.c:2821

[burnus at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103396

            Bug ID: 103396
           Summary: [GCN][BUILD] ICE RTL check: access of elt 4 of vector
                    with last elt 3 in move_callee_saved_registers, at
                    config/gcn/gcn.c:2821
           Product: gcc
           Version: 12.0
            Status: UNCONFIRMED
          Keywords: build, ice-on-valid-code
          Severity: normal
          Priority: P3
         Component: target
          Assignee: unassigned at gcc dot gnu.org
          Reporter: burnus at gcc dot gnu.org
                CC: ams at gcc dot gnu.org, rdapp at gcc dot gnu.org, sayle at gcc dot gnu.org
  Target Milestone: ---
              Host: x86_64-gnu-linux
            Target: amdgcn-amdhsa

Bisecting points at r12-5436-gdc915b361bbc99da83fc53db7f7e0e28d0ce12c8
Date:   Sun Nov 21 11:40:08 2021 +0000

    Tweak tree-ssa-math-opts.c to solve PR target/102117.
...
    2021-11-21  Roger Sayle  <roger@nextmovesoftware.com>
                Robin Dapp  <rdapp@linux.ibm.com>

    gcc/ChangeLog
            PR target/102117
            * tree-ssa-math-opts.c (convert_mult_to_widen): Recognize
            signed WIDEN_MULT_EXPR if the target supports umul_widen_optab.
...

Compiling libgfortran/runtime/environ.c during bootstrap with -march=gfx900 -O2
fails with:

during RTL pass: pro_and_epilogue

libgfortran/runtime/environ.c: In function 'search_unit':
libgfortran/runtime/environ.c:319:1: internal compiler error: RTL check: access
of elt 4 of vector with last elt 3 in move_callee_saved_registers, at
config/gcn/gcn.c:2821
  319 | }
      | ^
0x908087 rtvec_check_failed_bounds(rtvec_def const*, int, char const*, int,
char const*)
        gcc-mainline/gcc/rtl.c:975
0x9aa041 move_callee_saved_registers
        gcc-mainline/gcc/config/gcn/gcn.c:2821
0x14679cf gcn_expand_prologue()
        gcc-mainline/gcc/config/gcn/gcn.c:2939
0x17d331a gen_prologue()
        gcc-mainline/gcc/config/gcn/gcn.md:714
0x1460758 target_gen_prologue
        ./insn-target-def.h:433
0xdb3537 make_prologue_seq
        gcc-mainline/gcc/function.c:5823
0xdb36c2 thread_prologue_and_epilogue_insns()

[Bug target/103396] [GCN][BUILD] ICE RTL check: access of elt 4 of vector with last elt 3 in move_callee_saved_registers, at config/gcn/gcn.c:2821

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103396

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |tschwinge at gcc dot gnu.org

--- Comment #1 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
*** Bug 103397 has been marked as a duplicate of this bug. ***

[Bug target/103396] [GCN][BUILD] ICE RTL check: access of elt 4 of vector with last elt 3 in move_callee_saved_registers, at config/gcn/gcn.c:2821

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103396

--- Comment #2 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Is the cross compiler built with this GCC that includes this patch or this
patch just exposing a bug in the backend?

[Bug target/103396] [GCN][BUILD] ICE RTL check: access of elt 4 of vector with last elt 3 in move_callee_saved_registers, at config/gcn/gcn.c:2821

[burnus at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103396

--- Comment #3 from Tobias Burnus <burnus at gcc dot gnu.org> ---
Created attachment 51864
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=51864&action=edit
Reduced testcase – compile with -O2 -march=gfx900 / gfx906 / gfx908

Note: No ICE with -match=fiji

ICE at:

Breakpoint 1, move_callee_saved_registers (sp=0x7ffff73090f0,
offsets=0x7ffff7402e00, prologue=true)
    at gcc-mainline/gcc/config/gcn/gcn.c:2821
2821                        RTVEC_ELT (seq, count) = set;

with

(gdb) p debug_rtx(sp)
(reg/f:DI 16 s16)

(gdb) p debug_rtx(set)
(set/f (mem:DI (plus:DI (reg/f:DI 16 s16)
            (const_int 8 [0x8])) [0  S8 A64])
    (reg:DI 420 ?dwlr))
$18 = void

(gdb) p count
$8 = 4

(gdb) p *seq
$10 = {num_elem = 4, elem = {0x7ffff742ef30}}
(gdb) p *seq.elem 
$11 = (rtx) 0x7ffff742ef30
(gdb) p seq.elem[0] 
$12 = (rtx) 0x7ffff742ef30
(gdb) p debug_rtx(seq.elem[0])
(set/f (mem:SI (plus:DI (reg/f:DI 16 s16)
            (const_int 0 [0])) [0  S4 A32])
    (reg:SI 14 s14))
$13 = void
(gdb) p debug_rtx(seq.elem[1])
(set/f (mem:SI (plus:DI (reg/f:DI 16 s16)
            (const_int 4 [0x4])) [0  S4 A32])
    (reg:SI 15 s15))
$14 = void
(gdb) p debug_rtx(seq.elem[2])
(set/f (mem:SI (plus:DI (reg/f:DI 16 s16)
            (const_int 8 [0x8])) [0  S4 A32])
    (reg:SI 18 s18))
$15 = void
(gdb) p debug_rtx(seq.elem[3])
(set/f (mem:SI (plus:DI (reg/f:DI 16 s16)
            (const_int 12 [0xc])) [0  S4 A32])
    (reg:SI 19 s19))
$16 = void

[Bug target/103396] [12 Regression][GCN][BUILD] ICE RTL check: access of elt 4 of vector with last elt 3 in move_callee_saved_registers, at config/gcn/gcn.c:2821

[burnus at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103396

Tobias Burnus <burnus at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|[GCN][BUILD] ICE RTL check: |[12 Regression][GCN][BUILD]
                   |access of elt 4 of vector   |ICE RTL check: access of
                   |with last elt 3 in          |elt 4 of vector with last
                   |move_callee_saved_registers |elt 3 in
                   |, at config/gcn/gcn.c:2821  |move_callee_saved_registers
                   |                            |, at config/gcn/gcn.c:2821
   Target Milestone|---                         |12.0

[Bug target/103396] [12 Regression][GCN][BUILD] ICE RTL check: access of elt 4 of vector with last elt 3 in move_callee_saved_registers, at config/gcn/gcn.c:2821

[ams at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103396

Andrew Stubbs <ams at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Last reconfirmed|                            |2021-11-24
             Status|UNCONFIRMED                 |ASSIGNED
           Assignee|unassigned at gcc dot gnu.org      |ams at gcc dot gnu.org
     Ever confirmed|0                           |1

--- Comment #4 from Andrew Stubbs <ams at gcc dot gnu.org> ---
I think I have a fix for this. It happens when the link register has to be
saved because it is used implicitly by a function call, but the register is
never explicitly mentioned anywhere else in the function. I don't know why this
hasn't been a problem before now?

[Bug target/103396] [12 Regression][GCN][BUILD] ICE RTL check: access of elt 4 of vector with last elt 3 in move_callee_saved_registers, at config/gcn/gcn.c:2821

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103396

--- Comment #5 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by Andrew Stubbs <ams@gcc.gnu.org>:

https://gcc.gnu.org/g:58d50a5dd6344179eebaeb6fd2f895e59463cf74

commit r12-5525-g58d50a5dd6344179eebaeb6fd2f895e59463cf74
Author: Andrew Stubbs <ams@codesourcery.com>
Date:   Thu Nov 25 15:59:20 2021 +0000

    amdgcn: Fix ICE generating CFI [PR103396]

    gcc/ChangeLog:

            PR target/103396
            * config/gcn/gcn.c (move_callee_saved_registers): Ensure that the
            number of spilled registers is counted correctly.

[Bug target/103396] [12 Regression][GCN][BUILD] ICE RTL check: access of elt 4 of vector with last elt 3 in move_callee_saved_registers, at config/gcn/gcn.c:2821

[ams at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103396

Andrew Stubbs <ams at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         Resolution|---                         |FIXED
             Status|ASSIGNED                    |RESOLVED

--- Comment #6 from Andrew Stubbs <ams at gcc dot gnu.org> ---
This problem should be fixed now.