[Bug tree-optimization/38051] [4.4 Regression] Miscompilation of glibc's memcmp
--- Comment #16 from rguenth at gcc dot gnu dot org 2008-12-10 18:43 --- *** Bug 38478 has been marked as a duplicate of this bug. *** -- rguenth at gcc dot gnu dot org changed: What|Removed |Added CC||bjg at gnu dot org http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38051
[Bug tree-optimization/38051] [4.4 Regression] Miscompilation of glibc's memcmp
--- Comment #15 from rguenth at gcc dot gnu dot org 2008-11-20 12:15 --- *** Bug 38169 has been marked as a duplicate of this bug. *** -- rguenth at gcc dot gnu dot org changed: What|Removed |Added CC||holger dot hopp at sap dot ||com http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38051
[Bug tree-optimization/38051] [4.4 Regression] Miscompilation of glibc's memcmp
--- Comment #11 from dominiq at lps dot ens dot fr 2008-11-18 13:38 --- The test fails on powerpc-apple-darwin9 (revision 141945, 32 and 64 bit modes): FAIL: gcc.c-torture/execute/pr38051.c execution, -O0 FAIL: gcc.c-torture/execute/pr38051.c execution, -O1 FAIL: gcc.c-torture/execute/pr38051.c execution, -O2 FAIL: gcc.c-torture/execute/pr38051.c execution, -O3 -fomit-frame-pointer FAIL: gcc.c-torture/execute/pr38051.c execution, -O3 -fomit-frame-pointer -funroll-loops FAIL: gcc.c-torture/execute/pr38051.c execution, -O3 -fomit-frame-pointer -funroll-all-loops -finline-functions FAIL: gcc.c-torture/execute/pr38051.c execution, -O3 -g FAIL: gcc.c-torture/execute/pr38051.c execution, -Os -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38051
[Bug tree-optimization/38051] [4.4 Regression] Miscompilation of glibc's memcmp
--- Comment #12 from sje at cup dot hp dot com 2008-11-18 17:31 --- The new test is also failing on IA64 HP-UX and PA HP-UX (32 and 64 bits). It is not failing on IA64 Linux. Could the test have a big-endian/little-endian issue? -- sje at cup dot hp dot com changed: What|Removed |Added CC||sje at cup dot hp dot com http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38051
[Bug tree-optimization/38051] [4.4 Regression] Miscompilation of glibc's memcmp
--- Comment #13 from jakub at gcc dot gnu dot org 2008-11-18 18:19 --- See http://gcc.gnu.org/ml/gcc-patches/2008-11/msg00905.html -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38051
[Bug tree-optimization/38051] [4.4 Regression] Miscompilation of glibc's memcmp
--- Comment #14 from jakub at gcc dot gnu dot org 2008-11-18 23:03 --- Subject: Bug 38051 Author: jakub Date: Tue Nov 18 23:01:35 2008 New Revision: 141983 URL: http://gcc.gnu.org/viewcvs?root=gccview=revrev=141983 Log: PR tree-optimization/38051 * gcc.c-torture/execute/pr38051.c (buf): Remove aligned attribute. (buf2): Removed. (main): Only run on little endian targets with sizeof (long) == sizeof (void *). Use just one buffer, align the pointers at runtime. Modified: trunk/gcc/testsuite/ChangeLog trunk/gcc/testsuite/gcc.c-torture/execute/pr38051.c -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38051
[Bug tree-optimization/38051] [4.4 Regression] Miscompilation of glibc's memcmp
-- jakub at gcc dot gnu dot org changed: What|Removed |Added Priority|P3 |P1 http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38051
[Bug tree-optimization/38051] [4.4 Regression] Miscompilation of glibc's memcmp
--- Comment #9 from rguenth at gcc dot gnu dot org 2008-11-15 15:39 --- Subject: Bug 38051 Author: rguenth Date: Sat Nov 15 15:37:57 2008 New Revision: 141887 URL: http://gcc.gnu.org/viewcvs?root=gccview=revrev=141887 Log: 2008-11-15 Richard Guenther [EMAIL PROTECTED] PR tree-optimization/38051 * tree-ssa-alias.c (update_alias_info_1): Manually find written variables. * gcc.c-torture/execute/pr38051.c: New testcase. Added: trunk/gcc/testsuite/gcc.c-torture/execute/pr38051.c Modified: trunk/gcc/ChangeLog trunk/gcc/testsuite/ChangeLog trunk/gcc/tree-ssa-alias.c -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38051
[Bug tree-optimization/38051] [4.4 Regression] Miscompilation of glibc's memcmp
--- Comment #10 from rguenth at gcc dot gnu dot org 2008-11-15 15:42 --- Fixed. -- rguenth at gcc dot gnu dot org changed: What|Removed |Added Status|ASSIGNED|RESOLVED Resolution||FIXED http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38051
[Bug tree-optimization/38051] [4.4 Regression] Miscompilation of glibc's memcmp
--- Comment #1 from jakub at gcc dot gnu dot org 2008-11-07 14:57 --- -fno-tree-pre fixes this. -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38051
[Bug tree-optimization/38051] [4.4 Regression] Miscompilation of glibc's memcmp
--- Comment #2 from rguenth at gcc dot gnu dot org 2008-11-07 15:29 --- Confirmed. PRE doesn't do anything interesting though, neither are the differences in the optimized dump interesting from a first look. -- rguenth at gcc dot gnu dot org changed: What|Removed |Added CC||rguenth at gcc dot gnu dot ||org Status|UNCONFIRMED |NEW Ever Confirmed|0 |1 Last reconfirmed|-00-00 00:00:00 |2008-11-07 15:29:03 date|| http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38051
[Bug tree-optimization/38051] [4.4 Regression] Miscompilation of glibc's memcmp
--- Comment #4 from jakub at gcc dot gnu dot org 2008-11-07 16:32 ---
*.copyprop4 still contains the needed stores:
[gg5.c : 151] D.2147_293 = a3D.2068_292 shlD.2074_195;
[gg5.c : 151] D.2148_295 = a0D.2065_294 shrD.2075_198;
[gg5.c : 151] xD.2073_296 = D.2148_295 | D.2147_293;
[gg5.c : 152] if (xD.2073_296 != b0D.2069_297)
goto bb 51;
else
goto bb 55 (do2);
# SUCC: 51 [72.0%] (true,exec) 55 [28.0%] (irreducible,false,exec)
# BLOCK 51 freq:370, starting at line 0
# PRED: 50 [72.0%] (true,exec)
# aD.2099_621 = VDEF aD.2099_620(D) { aD.2099 }
aD.2099 = xD.2073_296;
# bD.2098_623 = VDEF bD.2098_622(D) { bD.2098 }
bD.2098 = b0D.2069_297;
[gg5.c : 8] srcp1D.2100_298 = (long intD.2) [gg5.c : 8] aD.2099;
[gg5.c : 9] srcp2D.2101_299 = (long intD.2) [gg5.c : 9] bD.2098;
# BLOCK 52 freq:2640
# PRED: 51 [100.0%] (fallthru,exec) 53 [100.0%] (fallthru,exec)
# srcp1D.2100_300 = PHI srcp1D.2100_298(51), srcp1D.2100_308(53)
# srcp2D.2101_304 = PHI srcp2D.2101_299(51), srcp2D.2101_309(53)
[gg5.c : 13] srcp1.0D.2149_301 = (unsigned charD.10 *) srcp1D.2100_300;
[gg5.c : 13] # VUSE aD.1965_413(D), bD.1970_414(D), aD.1971_415(D),
bD.1976_416(D), aD.1977_417(D), bD.1982_418(D), aD.1983_41
9(D), SMT.25D.2200_420(D) { aD.1965 bD.1970 aD.1971 bD.1976 aD.1977 bD.1982
aD.1983 SMT.25D.2200 }
D.2150_302 = *srcp1.0D.2149_301;
[gg5.c : 13] a0D.2102_303 = (long unsigned intD.4) D.2150_302;
[gg5.c : 14] srcp2.1D.2151_305 = (unsigned charD.10 *) srcp2D.2101_304;
[gg5.c : 14] # VUSE aD.1965_413(D), bD.1970_414(D), aD.1971_415(D),
bD.1976_416(D), aD.1977_417(D), bD.1982_418(D), aD.1983_41
9(D), SMT.25D.2200_420(D) { aD.1965 bD.1970 aD.1971 bD.1976 aD.1977 bD.1982
aD.1983 SMT.25D.2200 }
D.2152_306 = *srcp2.1D.2151_305;
(a = x and b = b0) before taking their addresses, but *.dceloop1 nukes them,
supposedly because VUSE on the *srcp1 and *srcp2 reads is wrong. This is with
-O2 -fno-strict-aliasing, but even with strict aliasing taking an address of an
unsigned long variable, casting it to unsigned char * and reading it a single
byte at a time through unsigned char * pointer is valid.
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38051
[Bug tree-optimization/38051] [4.4 Regression] Miscompilation of glibc's memcmp
--- Comment #3 from jakub at gcc dot gnu dot org 2008-11-07 16:19 --- In assembly, it seems the a and b arguments to mymemcmp1 aren't stored into memory at all: .loc 1 128 0 movq(%rax), %r12#* srcp2.272, b0 .LVL54: .loc 1 126 0 movq(%rdi), %rsi#* srcp1, a3 .LVL55: .loc 1 127 0 movq8(%rdi), %rbx #, a0 .LVL56: .loc 1 129 0 addq$8, %rdi#, srcp1.270 .LVL57: .L28: .loc 1 152 0 movl%r9d, %ecx #, .LVL58: movq%rbx, %r10 # a0, tmp154 .loc 1 149 0 movq8(%rdi), %r11 #, a1 .LVL59: .loc 1 152 0 salq%cl, %r10 #, tmp154 movl%r8d, %ecx #, .loc 1 150 0 movq8(%rax), %rbp #, b1 .LVL60: .loc 1 152 0 shrq%cl, %rsi #, tmp155 .LVL61: orq %rsi, %r10 # tmp155, tmp156 cmpq%r12, %r10 # b0, tmp156 je .L26#, .LBB112: .LBB113: .loc 1 8 0 leaq-8(%rsp), %rcx #, srcp1 .LVL62: .loc 1 9 0 leaq-16(%rsp), %rdx #, srcp2 .LVL63: .p2align 4,,10 .p2align 3 .L31: .loc 1 13 0 movzbl (%rcx), %eax#* srcp1, a0 .LVL64: .loc 1 14 0 movzbl (%rdx), %edi#* srcp2, b0 ... At je .L26 insn %r12 register contains correct value of b0 variable and %r10 of x (shifted + ored together value). But then suddenly the code assumes it is in memory without actually storing it there. -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38051
[Bug tree-optimization/38051] [4.4 Regression] Miscompilation of glibc's memcmp
--- Comment #6 from rguenth at gcc dot gnu dot org 2008-11-07 17:35 --- I will have a look. -- rguenth at gcc dot gnu dot org changed: What|Removed |Added AssignedTo|unassigned at gcc dot gnu |rguenth at gcc dot gnu dot |dot org |org Status|NEW |ASSIGNED Keywords||alias Last reconfirmed|2008-11-07 15:29:03 |2008-11-07 17:35:59 date|| http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38051
[Bug tree-optimization/38051] [4.4 Regression] Miscompilation of glibc's memcmp
-- jakub at gcc dot gnu dot org changed: What|Removed |Added Target Milestone|--- |4.4.0 http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38051
[Bug tree-optimization/38051] [4.4 Regression] Miscompilation of glibc's memcmp
--- Comment #5 from jakub at gcc dot gnu dot org 2008-11-07 16:46 ---
At *.crited there is still a virtual dependency between a = x; (and b = b0;)
and
*srcp1 + *srcp2 reads:
# MPT.26D.2201_438 = VDEF MPT.26D.2201_421(D) { MPT.26D.2201 }
aD.2099 = xD.2073_296;
# MPT.26D.2201_439 = VDEF MPT.26D.2201_438 { MPT.26D.2201 }
bD.2098 = b0D.2069_297;
...
[gg5.c : 13] # VUSE aD.1965_413(D), bD.1970_414(D), aD.1971_415(D),
bD.1976_416(D), aD.1977_417(D), bD.1982_418(D), aD.1983_41
9(D), SMT.25D.2200_420(D), MPT.26D.2201_439 { aD.1965 bD.1970 aD.1971 bD.1976
aD.1977 bD.1982 aD.1983 SMT.25D.2200 MPT.26D.2201
}
D.2150_302 = *srcp1.0D.2149_301;
...
[gg5.c : 14] # VUSE aD.1965_413(D), bD.1970_414(D), aD.1971_415(D),
bD.1976_416(D), aD.1977_417(D), bD.1982_418(D), aD.1983_41
9(D), SMT.25D.2200_420(D), MPT.26D.2201_439 { aD.1965 bD.1970 aD.1971 bD.1976
aD.1977 bD.1982 aD.1983 SMT.25D.2200 MPT.26D.2201
}
D.2152_306 = *srcp2.1D.2151_305;
through MPT.26. But in *.pre that's gone and there is no virtual dependency
between them anymore.
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38051
[Bug tree-optimization/38051] [4.4 Regression] Miscompilation of glibc's memcmp
--- Comment #7 from rguenth at gcc dot gnu dot org 2008-11-07 18:43 --- During update_alias_info we collect the variables that are written to, but because we use gimple_stored_syms to get at them we miss those that are in memory partitions. If the new partitioning moves them out of a partition we then miss conflicts for them. Oops. I have a patch. -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38051
[Bug tree-optimization/38051] [4.4 Regression] Miscompilation of glibc's memcmp
--- Comment #8 from rguenth at gcc dot gnu dot org 2008-11-07 19:33 --- Created an attachment (id=16633) -- (http://gcc.gnu.org/bugzilla/attachment.cgi?id=16633action=view) prototype patch Prototype patch. It needs some work as appearantly we cannot use MPT_SYMBOLS reliably there - I get random segfaults / referenced_var_lookup ICEs. As we are only interested in symbols we can just look at the statement manually. But at least for asms that will need some work. -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38051
