[Bug bootstrap/26361] [4.2 regression] bootstrap failure on Alpha: xgcc runs out of memory compiling libiberty/md5.c
-- pinskia at gcc dot gnu dot org changed: What|Removed |Added Keywords||build, ice-on-valid-code, ||memory-hog Summary|bootstrap failure on Alpha: |[4.2 regression] bootstrap |xgcc runs out of memory |failure on Alpha: xgcc runs |compiling libiberty/md5.c |out of memory compiling ||libiberty/md5.c Target Milestone|--- |4.2.0 http://gcc.gnu.org/bugzilla/show_bug.cgi?id=26361
[Bug bootstrap/26361] [4.2 regression] bootstrap failure on Alpha: xgcc runs out of memory compiling libiberty/md5.c
--- Comment #2 from pinskia at gcc dot gnu dot org 2006-02-19 15:38 --- I wonder if this is related to at all PR 26348. -- pinskia at gcc dot gnu dot org changed: What|Removed |Added BugsThisDependsOn||26348 http://gcc.gnu.org/bugzilla/show_bug.cgi?id=26361
[Bug bootstrap/26361] [4.2 regression] bootstrap failure on Alpha: xgcc runs out of memory compiling libiberty/md5.c
--- Comment #3 from falk at debian dot org 2006-02-19 15:57 --- (In reply to comment #2) I wonder if this is related to at all PR 26348. Probably not, because it already happens without any options (forgot to mention that). -- falk at debian dot org changed: What|Removed |Added Target Milestone|4.2.0 |--- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=26361
[Bug bootstrap/26361] [4.2 regression] bootstrap failure on Alpha: xgcc runs out of memory compiling libiberty/md5.c
--- Comment #4 from pinskia at gcc dot gnu dot org 2006-02-19 16:30 ---
[11:22] mellum The Alpha bootstrap bug seems to be due to
alpha_expand_prologue being miscompiled
[11:23] mellum there's a loop
[11:24] mellum for (probed = 4096; probed frame_size; probed += 8192) {
[11:24] mellum fprintf(stderr, probed = %d, frame_size = %ld\n, probed,
frame_size);
[11:24] mellum and this gives
[11:24] mellum probed = 4096, frame_size = 4112
[11:24] mellum probed = 12288, frame_size = 4112
[11:24] mellum probed = 20480, frame_size = 4112
[11:24] mellum etc so the loop exit isn't taken
[11:24] mellum does this look like a known problem?
From that description, this looks like PR 25985.
--
pinskia at gcc dot gnu dot org changed:
What|Removed |Added
BugsThisDependsOn|26348 |25985
Target Milestone|--- |4.2.0
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=26361
[Bug bootstrap/26361] [4.2 regression] bootstrap failure on Alpha: xgcc runs out of memory compiling libiberty/md5.c
--- Comment #5 from falk at debian dot org 2006-02-19 18:19 ---
The problem apparently comes from using negation on an induction variable,
in a context where widening is needed:
[EMAIL PROTECTED]:/tmp% cat alpha.c
void abort(void);
int printf(const char *format, ...);
__attribute__((noinline))
void gen_rtx_CONST_INT(long x) {
if (-x 10)
abort();
}
__attribute__((noinline))
int alpha_expand_prologue(long frame_size)
{
int probed;
for (probed = 1; probed frame_size; probed += 2) {
printf(probed = %d, frame_size = %ld\n,
probed, frame_size);
gen_rtx_CONST_INT (-probed);
}
}
int main(void) {
alpha_expand_prologue(10);
return 0;
}
[EMAIL PROTECTED]:/tmp% /src/gcc-2006.02.19/build/prev-gcc/xgcc
-B/src/gcc-2006.02.19/build/prev-gcc/ -g -O2 alpha.c ./a.out
probed = 1, frame_size = 10
probed = 3, frame_size = 10
probed = 5, frame_size = 10
probed = 7, frame_size = 10
probed = 9, frame_size = 10
probed = 11, frame_size = 10
zsh: abort (core dumped) ./a.out
[EMAIL PROTECTED]:/tmp% /src/gcc-2006.02.19/build/prev-gcc/xgcc
-B/src/gcc-2006.02.19/build/prev-gcc/ -g -O2 -fno-ivopts alpha.c ./a.out
probed = 1, frame_size = 10
probed = 3, frame_size = 10
probed = 5, frame_size = 10
probed = 7, frame_size = 10
probed = 9, frame_size = 10
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=26361
[Bug bootstrap/26361] [4.2 regression] bootstrap failure on Alpha: xgcc runs out of memory compiling libiberty/md5.c
--- Comment #6 from steven at gcc dot gnu dot org 2006-02-19 18:56 --- Also fails on AMD64. -- steven at gcc dot gnu dot org changed: What|Removed |Added Status|UNCONFIRMED |NEW Ever Confirmed|0 |1 Last reconfirmed|-00-00 00:00:00 |2006-02-19 18:56:19 date|| http://gcc.gnu.org/bugzilla/show_bug.cgi?id=26361
[Bug bootstrap/26361] [4.2 regression] bootstrap failure on Alpha: xgcc runs out of memory compiling libiberty/md5.c
--- Comment #7 from falk at debian dot org 2006-02-19 18:58 --- not really ice-on-valid-code nor memory-hog, those were only due to miscompiled xgcc -- falk at debian dot org changed: What|Removed |Added Keywords|ice-on-valid-code, memory- | |hog | http://gcc.gnu.org/bugzilla/show_bug.cgi?id=26361
[Bug bootstrap/26361] [4.2 regression] bootstrap failure on Alpha: xgcc runs out of memory compiling libiberty/md5.c
--- Comment #8 from pinskia at gcc dot gnu dot org 2006-02-19 19:36 --- This looks more related to PR 26304. -- pinskia at gcc dot gnu dot org changed: What|Removed |Added BugsThisDependsOn||26304 http://gcc.gnu.org/bugzilla/show_bug.cgi?id=26361
[Bug bootstrap/26361] [4.2 regression] bootstrap failure on Alpha: xgcc runs out of memory compiling libiberty/md5.c
--- Comment #9 from falk at debian dot org 2006-02-19 22:46 --- This bug was introduced with: r75 | law | 2006-02-17 05:15:32 +0100 (Fri, 17 Feb 2006) | 33 lines * tree-vrp.c (set_value_range_to_nonnegative): New function. (vrp_expr_computes_nonnegative, ssa_name_nonnegative_p): Likewise. (ssa_name_nonzero_p): Likewise. (get_value_range): Return NULL if VRP is not running. (extract_range_from_expr): Fallback to tree_expr_XXX_p if VRP routines do not discover a range. (vrp_finalize): Clear VR_VALUE to indicate VRP is not running. * tree.h (ssa_name_nonzero_p, ssa_name_nonnegative_p): Prototype. * fold-const.c (tree_expr_nonzero_p): For SSA_NAMEs, query VRP. (tree_expr_nonnegative_p): Similarly. * tree-ssa-dom.c (nonzero_vars, nonzero_vars_stack): Remove. (restore_nonzero_vars_to_original_value): Remove. (unsafe_associative_fp_binop): Remove. (tree_ssa_dominator_optimize): Remove initialization and finalization of nonzero_vars and nonzero_vars_stack. (dom_opt_initialize_block): No longer push marker on nonzero_vars_stack. (dom_opt_finalize_block): No longer call restore_nonzero_vars_to_original_value. (record_equivalences_from_phis): No longer look for nonzero PHI arguments. (cprop_into_successor_phis): No longer propagate nonzero property into PHI arguments. Lose unused argument. Caller updated. (record_equivalences_from_stmt): No longer record nonzero values for SSA_NAMEs. (lookup_avail_expr): No longer use nonzero_vars. * gcc.dg/tree-ssa/vrp24.c: Update expected output. * gcc.dg/tree-ssa/vrp26.c: New test. -- falk at debian dot org changed: What|Removed |Added CC||law at gcc dot gnu dot org http://gcc.gnu.org/bugzilla/show_bug.cgi?id=26361
[Bug bootstrap/26361] [4.2 regression] bootstrap failure on Alpha: xgcc runs out of memory compiling libiberty/md5.c
--- Comment #10 from roger at eyesopen dot com 2006-02-20 03:11 --- Analyzing the code in comment #5, this looks like a bad interaction between ivopts and vrp. Either of -fno-ivopts or -fno-tree-vrp cures the problem. But it looks like the output of .084t.ivopts is reasonable. -- roger at eyesopen dot com changed: What|Removed |Added CC||roger at eyesopen dot com http://gcc.gnu.org/bugzilla/show_bug.cgi?id=26361
[Bug bootstrap/26361] [4.2 regression] bootstrap failure on Alpha: xgcc runs out of memory compiling libiberty/md5.c
--- Comment #11 from roger at eyesopen dot com 2006-02-20 03:52 ---
Hmmm. Looks like the problem is in .088t.vrp2
We have
unsigned int D.1981;
unsigned int D.1982;
D.1982_9 = -D.1981_1;
D.1981_1: [0, +INF] EQUIVALENCES: { } (0 elements)
D.1982_9: [0, 1] EQUIVALENCES: { } (0 elements)
Seems like a bug in the interval arithmetic of NEGATE_EXPR for unsigned types.
- [0, (unsigned)-1] seems to be converted to [0, 1]!!? I hope this helps.
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=26361
[Bug bootstrap/26361] [4.2 regression] bootstrap failure on Alpha: xgcc runs out of memory compiling libiberty/md5.c
--- Comment #12 from pinskia at gcc dot gnu dot org 2006-02-20 04:58
---
Here is a testcase independent of IV-OPTs:
void abort(void);
int printf(const char *format, ...);
__attribute__((noinline))
void gen_rtx_CONST_INT(long long x) {
if (-x 10)
abort();
}
__attribute__((noinline))
int alpha_expand_prologue(long frame_size)
{
unsigned long long a;
int probed;
if (frame_size = 1) return;
unsigned long long b = -2;
a = -2;
do {
int a1 = a;
probed = -a1;
printf(probed = %d, frame_size = %ld\n,
probed, frame_size);
gen_rtx_CONST_INT (a1);
a -= 2;
a1 = -a;
probed = a1;
} while (probed frame_size);
}
int main(void) {
alpha_expand_prologue(10);
return 0;
}
-
I tried this on 4.1.0 and could not reproduce it there, though my 4.1.0 is from
20060208.
--
pinskia at gcc dot gnu dot org changed:
What|Removed |Added
CC||pinskia at gcc dot gnu dot
||org
BugsThisDependsOn|25985, 26304|
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=26361
