From: Diego Novillo <dnovillo@redhat.com>
To: gcc@gcc.gnu.org
Cc: Franz.Sirl-kernel@lauterbach.com
Subject: SPEC95 failure on 3.0 branch
Date: Wed, 10 Oct 2001 06:53:00 -0000 [thread overview]
Message-ID: <20011010095234.A6528@tornado.cygnus.com> (raw)
This failure started showing up on Oct09. We fail to compile
124.m88ksim. The compiler ICEs in unroll.c:
---------------------------------------------------------------------------
dpath.c: In function `execute':
dpath.c:863: Internal compiler error in loop_iterations, at unroll.c:3510
Please submit a full bug report,
with preprocessed source if appropriate.
See <URL: http://www.gnu.org/software/gcc/bugs.html > for instructions.
specmake: *** [dpath.o] Error 1
---------------------------------------------------------------------------
Looking at the ChangeLog diffs it looks as if Franz's patch
could've caused this:
+ 2001-10-08 Neil Booth <neil@daikokuya.demon.co.uk>
+
+ * doc/cpp.texi: Update.
+
+ 2001-10-08 Neil Booth <neil@daikokuya.demon.co.uk>
+
+ * cppmacro.c (_cpp_create_definition): Leave comments off.
+ * cpplex.c (_cpp_lex_token): Don't save comments if skipping.
+
+ * doc/cpp.texi: Update.
+
+ 2001-10-08 Franz Sirl <Franz.Sirl-kernel@lauterbach.com>
+
+ * unroll.c (loop_iterations): Extend check for multiple back edges.
+
+ 2001-10-08 Jeffrey A Law <law@cygnus.com>
+
+ * sibcall.c (optimize_sibling_and_tail_recursive_calls): Call
+ purge_mem_unchanging_flag on all instructions, not just on those
+ before NOTE_INSN_FUNCTION_BEG.
+
This is a debugging session of this failure. I'm not too sure
what I'm looking for. Hopefully this will give some context:
---------------------------------------------------------------------------
$ gdb /home/dnovillo/perf/sandbox-gcc-3_0-branch/i686/inst/lib/gcc-lib/i686-pc-linux-gnu/3.0.2/cc1
[ ... ]
(gdb) b unroll.c:3510
Breakpoint 1 at 0x81505b0: file /home/dnovillo/perf/sandbox-gcc-3_0-branch/src/gcc/unroll.c, line 3510.
(gdb) run -fpreprocessed dpath.i -quiet -dumpbase dpath.c -O2 -version -o dpath.s
Starting program: /home/dnovillo/perf/sandbox-gcc-3_0-branch/i686/inst/lib/gcc-lib/i686-pc-linux-gnu/3.0.2/cc1 -fpreprocessed dpath.i -quiet -dumpbase dpath.c -O2 -version -o dpath.s
During symbol reading, register number 1769234796 too large (max 32) in symbol buf.
GNU CPP version 3.0.2 20011010 (prerelease) (cpplib) (i386 Linux/ELF)
GNU C version 3.0.2 20011010 (prerelease) (i686-pc-linux-gnu)
compiled by GNU C version 3.0.2 20011010 (prerelease).
/home/dnovillo/SPEC95/benchspec/CINT95/124.m88ksim/src/dpath.c: In function `execute':
/home/dnovillo/SPEC95/benchspec/CINT95/124.m88ksim/src/dpath.c:863: Internal compiler error in loop_iterations, at unroll.c:3510
Please submit a full bug report,
with preprocessed source if appropriate.
See <URL: http://www.gnu.org/software/gcc/bugs.html > for instructions.
Breakpoint 1, loop_iterations (loop=0x83a6418)
at /home/dnovillo/perf/sandbox-gcc-3_0-branch/src/gcc/unroll.c:3511
3511 && INSN_LUID (JUMP_LABEL (temp)) < INSN_LUID (loop->cont))
(gdb) l 3511
3506
3507 do
3508 {
3509 if (GET_CODE (temp) == JUMP_INSN
3510 && INSN_LUID (JUMP_LABEL (temp)) > INSN_LUID (loop->top)
3511 && INSN_LUID (JUMP_LABEL (temp)) < INSN_LUID (loop->cont))
3512 {
3513 if (loop_dump_stream)
3514 fprintf (loop_dump_stream,
3515 "Loop iterations: Loop has multiple back edges.\n");
(gdb) bt
#0 loop_iterations (loop=0x83a6418)
at /home/dnovillo/perf/sandbox-gcc-3_0-branch/src/gcc/unroll.c:3511
#1 0x08144314 in strength_reduce (loop=0x83a6418, flags=0)
at /home/dnovillo/perf/sandbox-gcc-3_0-branch/src/gcc/loop.c:4276
#2 0x0813f4ad in scan_loop (loop=0x83a6418, flags=0)
at /home/dnovillo/perf/sandbox-gcc-3_0-branch/src/gcc/loop.c:1029
#3 0x0813f0a7 in loop_optimize (f=0x4030c380, dumpfile=0x0, flags=0)
at /home/dnovillo/perf/sandbox-gcc-3_0-branch/src/gcc/loop.c:460
#4 0x0808016c in rest_of_compilation (decl=0x4025eea0)
at /home/dnovillo/perf/sandbox-gcc-3_0-branch/src/gcc/toplev.c:3177
#5 0x08058b09 in c_expand_body (fndecl=0x4025eea0, nested_p=0)
at /home/dnovillo/perf/sandbox-gcc-3_0-branch/src/gcc/c-decl.c:6781
#6 0x0805897d in finish_function (nested=0)
at /home/dnovillo/perf/sandbox-gcc-3_0-branch/src/gcc/c-decl.c:6697
#7 0x0804a06d in yyparse_1 () at /usr/lib/bison.simple:324
#8 0x0807f048 in compile_file (name=0xbffff4c2 "dpath.i")
at /home/dnovillo/perf/sandbox-gcc-3_0-branch/src/gcc/toplev.c:2369
#9 0x08082daa in main (argc=10, argv=0xbffff2b4)
at /home/dnovillo/perf/sandbox-gcc-3_0-branch/src/gcc/toplev.c:4990
#10 0x40047177 in __libc_start_main (main=0x8082890 <main>, argc=10,
ubp_av=0xbffff2b4, init=0x80490d4 <_init>, fini=0x823b9e0 <_fini>,
rtld_fini=0x4000e184 <_dl_fini>, stack_end=0xbffff2ac)
at ../sysdeps/generic/libc-start.c:129
(gdb) pr loop->top
(code_label 3489 3487 3488 311 "" "" [1 uses])
(gdb) pr loop->cont
(note 2037 2019 3485 NOTE_INSN_LOOP_CONT 0)
(gdb) pr temp
(jump_insn 2030 2029 2056 (set (pc)
(if_then_else (ne (reg:CCZ 17 flags)
(const_int 0 [0x0]))
(label_ref 3489)
(pc))) -1 (nil)
(nil))
---------------------------------------------------------------------------
Diego.
next reply other threads:[~2001-10-10 6:53 UTC|newest]
Thread overview: 3+ messages / expand[flat|nested] mbox.gz Atom feed top
2001-10-10 6:53 Diego Novillo [this message]
2001-10-10 7:01 ` Franz Sirl
2001-10-11 8:43 ` Diego Novillo
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20011010095234.A6528@tornado.cygnus.com \
--to=dnovillo@redhat.com \
--cc=Franz.Sirl-kernel@lauterbach.com \
--cc=gcc@gcc.gnu.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
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).