* [lno] compute_immediate_uses in vectorizer causes ICE in pass_redundant_phi
@ 2004-03-18 15:32 Dorit Naishlos
2004-03-18 21:28 ` Zdenek Dvorak
0 siblings, 1 reply; 2+ messages in thread
From: Dorit Naishlos @ 2004-03-18 15:32 UTC (permalink / raw)
To: gcc
I added a call to compute_immediate_uses() in the vectorizer, and I get
ICE's all over SPEC during the subsequent redundant-phi-removal pass (even
if I immediately return after the call to compute_immediate_uses(), doing
no vectorization). If I disable the redundant-phi-removal pass after
vectorization, I can compile SPEC successfully. Are there any restrictions
on calling compute_immediate_uses(), or any cleanups that need to be done
after it? For example, this is the error that I get compiling SPEC's gzip:
getopt.c: In function `_getopt_internal':
getopt.c:334: internal compiler error: Bus error
(gdb)
Program received signal EXC_BAD_ACCESS, Could not access memory.
replace_immediate_uses (var=0x0, repl=0x681900) at
../../gcc/gcc/tree-flow-inline.h:3814
Line number 3814 out of range; ../../gcc/gcc/tree-flow-inline.h has 601
lines.
(gdb) backtrace
#0 replace_immediate_uses (var=0x0, repl=0x681900) at
../../gcc/gcc/tree-flow-inline.h:3814
#1 0x001be760 in kill_redundant_phi_nodes () at
../../gcc/gcc/tree-ssa.c:4021
#2 0x00200234 in execute_one_pass (pass=0x5916a4) at
../../gcc/gcc/tree-optimize.c:416
#3 0x002002d0 in execute_pass_list (pass=0x411006f0) at
../../gcc/gcc/tree-optimize.c:445
#4 0x002002e8 in execute_pass_list (pass=0x571d28) at
../../gcc/gcc/tree-optimize.c:446
#5 0x002005f8 in tree_rest_of_compilation (fndecl=0x613000,
nested_p=false) at ../../gcc/gcc/tree-optimize.c:539
#6 0x0001cac8 in c_expand_body_1 (fndecl=0x613000, nested_p=0) at
../../gcc/gcc/c-decl.c:6204
#7 0x001f73e4 in cgraph_expand_function (node=0x0) at
../../gcc/gcc/cgraphunit.c:782
#8 0x001f923c in cgraph_expand_all_functions () at
../../gcc/gcc/cgraphunit.c:1664
#9 0x001f95a4 in cgraph_optimize () at ../../gcc/gcc/cgraphunit.c:1754
#10 0x0005c768 in c_objc_common_finish_file () at
../../gcc/gcc/c-objc-common.c:243
#11 0x0009e880 in compile_file () at ../../gcc/gcc/toplev.c:1879
#12 0x000a2e4c in do_compile () at ../../gcc/gcc/toplev.c:4581
#13 0x000a2f04 in toplev_main (argc=18, argv=0x58c974) at
../../gcc/gcc/toplev.c:4621
#14 0x000020ec in _start (argc=18, argv=0x58c974, envp=0x574228) at
/SourceCache/Csu/Csu-46/crt.c:267
#15 0x00001f60 in start ()
thanks,
dorit
^ permalink raw reply [flat|nested] 2+ messages in thread
* Re: [lno] compute_immediate_uses in vectorizer causes ICE in pass_redundant_phi
2004-03-18 15:32 [lno] compute_immediate_uses in vectorizer causes ICE in pass_redundant_phi Dorit Naishlos
@ 2004-03-18 21:28 ` Zdenek Dvorak
0 siblings, 0 replies; 2+ messages in thread
From: Zdenek Dvorak @ 2004-03-18 21:28 UTC (permalink / raw)
To: Dorit Naishlos; +Cc: gcc
Hello,
> I added a call to compute_immediate_uses() in the vectorizer, and I get
> ICE's all over SPEC during the subsequent redundant-phi-removal pass (even
> if I immediately return after the call to compute_immediate_uses(), doing
> no vectorization). If I disable the redundant-phi-removal pass after
> vectorization, I can compile SPEC successfully. Are there any restrictions
> on calling compute_immediate_uses(), or any cleanups that need to be done
> after it?
you must call free_df () afterwards.
Zdenek
^ permalink raw reply [flat|nested] 2+ messages in thread
end of thread, other threads:[~2004-03-18 21:11 UTC | newest]
Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2004-03-18 15:32 [lno] compute_immediate_uses in vectorizer causes ICE in pass_redundant_phi Dorit Naishlos
2004-03-18 21:28 ` Zdenek Dvorak
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).