[Bug tree-optimization/27659] New: ICE on autovect-branch
with preprocessed source if appropriate. See http://gcc.gnu.org/bugs.html> for instructions. I tried to separate covar2corrcoef but that doesn't ICE. This does not ICE on trunk, rev. 113884 -- Summary: ICE on autovect-branch Product: gcc Version: 4.2.0 Status: UNCONFIRMED Severity: normal Priority: P3 Component: tree-optimization AssignedTo: unassigned at gcc dot gnu dot org ReportedBy: gcc at pdoerfler dot com GCC build triplet: i686-pc-linux-gnu GCC host triplet: i686-pc-linux-gnu GCC target triplet: i686-pc-linux-gnu http://gcc.gnu.org/bugzilla/show_bug.cgi?id=27659
[Bug tree-optimization/27659] ICE on autovect-branch
--- Comment #1 from gcc at pdoerfler dot com 2006-05-18 14:56 --- Created an attachment (id=11485) --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=11485&action=view) gziped preprocessed source that produces the ICE gziped because it's too large. Sorry, don't know how to reduce this. -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=27659
[Bug tree-optimization/27659] ICE on autovect-branch
--- Comment #2 from gcc at pdoerfler dot com 2006-05-22 09:27 ---
I figured out the reduction. Here is a much smaller testcase:
=
extern int rows_;
extern int cols_;
template
class secondOrderStatistics {
public:
void covar2corrcoef(U** &cv) const {
U* k=*cv;
for (int row=0; row;
=
/usr/local/vect/libexec/gcc/i686-pc-linux-gnu/4.2.0-autovect/cc1plus
-fpreprocessed reduced2.ii -quiet -march=pentium-m -ansi -O3 -version
-ftree-vectorize -o reduced2.s
GNU C++ version 4.2.0-autovect 20060518 (experimental) (i686-pc-linux-gnu)
compiled by GNU C version 4.2.0-autovect 20060518 (experimental).
GGC heuristics: --param ggc-min-expand=30 --param ggc-min-heapsize=4096
Compiler executable checksum: ce12e1a5e2bda479d797d2bc5f9a8466
reduced2.ii: In member function void
secondOrderStatistics::covar2corrcoef(U**&) const [with U = float]:
reduced2.ii:7: internal compiler error: Segmentation fault
Please submit a full bug report,
with preprocessed source if appropriate.
See http://gcc.gnu.org/bugs.html> for instructions.
gdb says:
Program received signal SIGSEGV, Segmentation fault.
remove_all_phi_uses (var=0x4022da28) at tree-if-conv.c:180
180 if (TREE_CODE (usestmt) == PHI_NODE)
(gdb) print usestmt
$1 = 0x0
(gdb) bt
#0 remove_all_phi_uses (var=0x4022da28) at tree-if-conv.c:180
#1 0x0857fe48 in remove_all_phi_uses (var=0x4022f64c) at tree-if-conv.c:184
#2 0x0857fe48 in remove_all_phi_uses (var=0x4022da28) at tree-if-conv.c:184
#3 0x0857fe48 in remove_all_phi_uses (var=0x4022d9c0) at tree-if-conv.c:184
#4 0x0857fe48 in remove_all_phi_uses (var=0x4022db60) at tree-if-conv.c:184
#5 0x0857ff97 in remove_store (bsi=0xbfd512dc) at tree-if-conv.c:206
#6 0x08582763 in main_tree_if_conversion () at tree-if-conv.c:248
#7 0x084dad64 in execute_one_pass (pass=0x86f0900) at passes.c:874
#8 0x084daea7 in execute_pass_list (pass=0x86f0900) at passes.c:906
#9 0x084daeba in execute_pass_list (pass=0x86edca0) at passes.c:907
#10 0x084daeba in execute_pass_list (pass=0x86ed660) at passes.c:907
#11 0x081bacc6 in tree_rest_of_compilation (fndecl=0x40223c80) at
tree-optimize.c:419
#12 0x081294fd in expand_body (fn=0x40223c80) at semantics.c:3005
#13 0x08526f12 in cgraph_expand_function (node=0x402270d0) at cgraphunit.c:1055
#14 0x08529395 in cgraph_optimize () at cgraphunit.c:1121
#15 0x080d2bf5 in cp_finish_file () at decl2.c:3115
#16 0x0818d79f in c_common_parse_file (set_yydebug=0) at c-opts.c:1136
#17 0x084a90e0 in toplev_main (argc=11, argv=0xbfd515b4) at toplev.c:990
#18 0x0819ae6f in main (argc=Cannot access memory at address 0x0
) at main.c:35
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=27659
[Bug tree-optimization/27742] New: ICE with -ftree-vectorizer-verbose
/usr/local/4.2/bin/g++4.2.0 -v
Using built-in specs.
Target: i686-pc-linux-gnu
Configured with: ../gcc-4.2-svn/configure --enable-languages=c,c++
--prefix=/usr/local/4.2 --program-suffix=4.2.0
Thread model: posix
gcc version 4.2.0 20060523 (experimental)
Revision: 114016
/usr/local/4.2/libexec/gcc/i686-pc-linux-gnu/4.2.0/cc1plus -quiet -v
-I/home/doerfler/ltilib-2/42/linux -I../src/basics -I../src/types -I../src/math
-I../src/imgProc -I../src/ioBasics -I../src/io -I../src/draw -I../src/viewer
-I../src/system -I../src/classifiers -Iworkbench -I../tester
-I/usr/include/cairo -I/usr/include/freetype2 -I/usr/X11R6/include
-I/usr/include/libpng12 -I/opt/gnome/include/gtk-2.0
-I/opt/gnome/lib/gtk-2.0/include -I/opt/gnome/include/atk-1.0
-I/opt/gnome/include/pango-1.0 -I/opt/gnome/include/glib-2.0
-I/opt/gnome/lib/glib-2.0/include -D_GNU_SOURCE -D_REENTRANT -D_BSD_SOURCE
-D_XOPEN_SOURCE=500 -D_REENTRANT -DNDEBUG -D_USE_FIRE_WIRE_DCAM
../src/imgProc/ltiLocalColorDescriptorExtractor.cpp -quiet -dumpbase
ltiLocalColorDescriptorExtractor.cpp -march=pentium-m -msse -msse2 -ansi
-auxbase-strip ../obj/release/ltiLocalColorDescriptorExtractor.o -O3 -Wundef
-Wwrite-strings -Wall -ansi -version -ftree-vectorize
-ftree-vectorizer-verbose=5 -fpic -o /tmp/ccrEtWY9.s
...
../src/imgProc/ltiGridSampling_template.h:227: note: vectorized 0 loops in
function.
../src/imgProc/ltiSliceSampling_template.h: In function void __tcf_15(void*):
../src/imgProc/ltiSliceSampling_template.h:82: internal compiler error:
Segmentation fault
Please submit a full bug report,
with preprocessed source if appropriate.
See http://gcc.gnu.org/bugs.html> for instructions.
Unfortunately, if I use -save-temps the ICE does not occur.
It also goes away without -ftree-vectorizer-verbose
Here is what gdb tells me:
Program received signal SIGSEGV, Segmentation fault.
0x40099d9b in strlen () from /lib/tls/libc.so.6
(gdb) bt
#0 0x40099d9b in strlen () from /lib/tls/libc.so.6
#1 0x4006e3d8 in vfprintf () from /lib/tls/libc.so.6
#2 0x08694581 in __FUNCTION__.28397 ()
#3 0x0001 in ?? ()
#4 0x4072a1b8 in ?? ()
#5 0x in ?? ()
#6 0x4060a898 in ?? ()
#7 0x0813df36 in gt_ggc_mx_lang_tree_node (x_p=0x0) at gt-cp-tree.h:89
#8 0x0813df36 in gt_ggc_mx_lang_tree_node (x_p=0x40494b2c) at gt-cp-tree.h:89
#9 0x in ?? ()
#10 0x0002 in ?? ()
#11 0x in ?? ()
#12 0x in ?? ()
#13 0x0813df36 in gt_ggc_mx_lang_tree_node (x_p=0x8694582) at gt-cp-tree.h:89
#14 0x in ?? ()
(gdb) cont
Continuing.
../src/imgProc/ltiSliceSampling_template.h: In function void __tcf_15(void*):
../src/imgProc/ltiSliceSampling_template.h:82: internal compiler error:
Segmentation fault
The source this happens on is a large library (ltilib-2). Here's a snippet
-
template
bool sliceSampling::apply(Acc& acc,
const matrix& img,
const location& loc,
dvector& destv) {
const parameters& par = getParameters();
// interpolation
static const bilinearInterpolation bilin;
//create accumulators
static std::vector accuVec;
-
the last line is the one reported as causing the ICE.
Acc is a template struct with a few member functions and a C array.
When compiled without -ftree-vectorizer-verbose __tcf_15 looks like this:
--
.align 2
.p2align 4,,15
.type __tcf_15, @function
__tcf_15:
.LFB3101:
pushl %ebp
.LCFI428:
movl%esp, %ebp
.LCFI429:
pushl %ebx
.LCFI430:
call__i686.get_pc_thunk.bx
addl$_GLOBAL_OFFSET_TABLE_, %ebx
subl$4, %esp
.LCFI431:
movl
[EMAIL PROTECTED](%ebx)
, %eax
movl(%eax), %eax
testl %eax, %eax
je .L1085
movl%eax, (%esp)
call[EMAIL PROTECTED]
.L1085:
popl%eax
popl%ebx
popl%ebp
ret
.LFE3101:
.size __tcf_15, .-__tcf_15
I can post instructions for downloading and reproduction of the ICE if this is
desired.
I can also perform further analysis if somebody can give instructions what else
I can do.
--
Summary: ICE with -ftree-vectorizer-verbose
Product: gcc
Version: 4.2.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: tree-optimization
AssignedTo: unassigned at gcc dot gnu dot org
ReportedBy: gcc at pdoerfler dot com
GCC host triplet: i686-pc-linux-gnu
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=27742
[Bug tree-optimization/27742] ICE with -ftree-vectorizer-verbose
--- Comment #2 from gcc at pdoerfler dot com 2006-05-23 15:38 --- Created an attachment (id=11501) --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=11501&action=view) gzipped prepocessed source code Please note again that the preprocessed source code does *not* trigger the ICE. -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=27742
[Bug tree-optimization/27753] New: ICE in make_ssa_name on autovect-branch
/usr/local/vect/bin/g++ -v
Using built-in specs.
Target: i686-pc-linux-gnu
Configured with: ../gcc-vect/configure --enable-languages=c,c++
--prefix=/usr/local/vect
Thread model: posix
gcc version 4.2.0-autovect 20060518 (experimental)
Reduced testcase below ICEs:
/usr/local/vect/bin/g++ -O3 -ftree-vectorize -march=pentium-m -c vecreduced.cpp
vecreduced.cpp: In member function vector& vector::add(T) [with T =
ppoint]:
vecreduced.cpp:14: internal compiler error: in make_ssa_name, at
tree-ssanames.c:129
template
struct ppoint {
int x, y;
inline ppoint& operator+=(const ppoint& p) {
x+=p.x;
y+=p.y;
return *this;
}
};
template
class vector {
public:
vector& add(T cst) {
for (int i=0; i >;
--
Summary: ICE in make_ssa_name on autovect-branch
Product: gcc
Version: 4.2.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: tree-optimization
AssignedTo: unassigned at gcc dot gnu dot org
ReportedBy: gcc at pdoerfler dot com
GCC host triplet: i686-pc-linux-gnu
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=27753
[Bug tree-optimization/27753] ICE in make_ssa_name on autovect-branch
--- Comment #1 from gcc at pdoerfler dot com 2006-05-24 11:41 ---
This is probably related to or a duplicate of PR26362.
Anyways, this is now a C++ testcase instead of fortran.
The patch by rth posted in that PR makes the ICE go away.
Here is an updated version that applies cleanly:
Index: targhooks.c
===
--- targhooks.c (revision 114040)
+++ targhooks.c (working copy)
@@ -455,7 +455,8 @@
tree type;
enum machine_mode mode;
block_stmt_iterator bsi;
- tree th, tl, result, x;
+ tree t1, t2, x;
+ int i, n;
/* If the first argument is a type, just check if support
is available. Return a non NULL value if supported, NULL_TREE otherwise.
@@ -479,31 +480,36 @@
return NULL;
bsi = bsi_for_stmt (stmt);
-
- th = make_rename_temp (type, NULL);
- x = build2 (VEC_INTERLEAVE_HIGH_EXPR, type, vec1, vec2);
- x = build2 (MODIFY_EXPR, type, th, x);
- th = make_ssa_name (th, x);
- TREE_OPERAND (x, 0) = th;
- bsi_insert_before (&bsi, x, BSI_SAME_STMT);
+ n = exact_log2 (GET_MODE_NUNITS (mode)) - 1;
+ for (i = 0; i < n; ++i)
+{
+ t1 = create_tmp_var (type, NULL);
+ add_referenced_tmp_var (t1);
+ x = build2 (VEC_INTERLEAVE_HIGH_EXPR, type, vec1, vec2);
+ x = build2 (MODIFY_EXPR, type, t1, x);
+ t1 = make_ssa_name (t1, x);
+ TREE_OPERAND (x, 0) = t1;
+ bsi_insert_before (&bsi, x, BSI_SAME_STMT);
- tl = make_rename_temp (type, NULL);
- x = build2 (VEC_INTERLEAVE_LOW_EXPR, type, vec1, vec2);
- x = build2 (MODIFY_EXPR, type, tl, x);
- tl = make_ssa_name (tl, x);
- TREE_OPERAND (x, 0) = tl;
- bsi_insert_before (&bsi, x, BSI_SAME_STMT);
+ t2 = create_tmp_var (type, NULL);
+ add_referenced_tmp_var (t2);
+ x = build2 (VEC_INTERLEAVE_LOW_EXPR, type, vec1, vec2);
+ x = build2 (MODIFY_EXPR, type, t2, x);
+ t2 = make_ssa_name (t2, x);
+ TREE_OPERAND (x, 0) = t2;
+ bsi_insert_before (&bsi, x, BSI_SAME_STMT);
- result = make_rename_temp (type, NULL);
- /* ??? Endianness issues? */
+ if (BYTES_BIG_ENDIAN)
+vec1 = t1, vec2 = t2;
+ else
+vec1 = t2, vec2 = t1;
+}
+
x = build2 (odd_p ? VEC_INTERLEAVE_HIGH_EXPR : VEC_INTERLEAVE_LOW_EXPR,
- type, th, tl);
- x = build2 (MODIFY_EXPR, type, result, x);
- result = make_ssa_name (result, x);
- TREE_OPERAND (x, 0) = result;
- bsi_insert_before (&bsi, x, BSI_SAME_STMT);
+ type, vec1, vec2);
+ x = build2 (MODIFY_EXPR, type, dest, x);
- return result;
+ return x;
}
tree
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=27753
[Bug other/22313] [4.2 Regression] profiledbootstrap is broken on the mainline
--- Comment #35 from gcc at pdoerfler dot com 2006-05-29 09:39 --- FWIW profiledbootstrap still fails with -march=pentium4 and fortran enabled because of PR26449. Last tested with 20060520. -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=22313
[Bug tree-optimization/27742] [4.2 regression] ICE with -ftree-vectorizer-verbose
--- Comment #4 from gcc at pdoerfler dot com 2006-06-03 12:23 --- This does not fail with gcc-4_1-branch, Revision: 114325, BTW. So I'm marking it a 4.2 regression. -- gcc at pdoerfler dot com changed: What|Removed |Added Known to fail||4.2.0 Known to work||4.1.1 Summary|ICE with -ftree-vectorizer- |[4.2 regression] ICE with - |verbose |ftree-vectorizer-verbose http://gcc.gnu.org/bugzilla/show_bug.cgi?id=27742
[Bug rtl-optimization/21676] [4.0/4.1/4.2 Regression] Optimizer regression: SciMark sparse matrix benchmark
--- Comment #3 from gcc at pdoerfler dot com 2006-06-06 11:22 --- I get the following with -O3 -march=pentium4 -fomit-frame-pointer on a pentium4 gentoo machine: gcc-3.4.6 gcc-4.0.2 gcc-4.1.1 2.69s 4.14s 3.26s These are all with gentoo's patches. Also, current mainline is the same as gcc-4.1.1 I can confirm that the difference without -fomit-frame-pointer is much smaller. In fact, 3.4.6 and 4.1.1 are almost the same without it. -- gcc at pdoerfler dot com changed: What|Removed |Added CC| |gcc at pdoerfler dot com http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21676
[Bug c/27950] New: [4.2 regression] undefined reference when compiling valgrind 3.2.0
if /usr/local/4.2/bin/gcc4.2.0 -DHAVE_CONFIG_H -I. -I. -I.. -I../include
-I../VEX/pub -DVGA_x86=1 -DVGO_linux=1 -DVGP_x86_linux=1 -m32
-mpreferred-stack-boundary=2 -O -g -Wmissing-prototypes -Winline -Wall -Wshadow
-Wpointer-arith -Wstrict-prototypes -Wmissing-declarations -O2 -Wno-long-long
-Wno-pointer-sign -Wdeclaration-after-statement -save-temps -v -MT
memcheck_x86_linux-mc_main.o -MD -MP -MF ".deps/memcheck_x86_linux-mc_main.Tpo"
-c -o memcheck_x86_linux-mc_main.o `test -f 'mc_main.c' || echo './'`mc_main.c;
\
then mv -f ".deps/memcheck_x86_linux-mc_main.Tpo"
".deps/memcheck_x86_linux-mc_main.Po"; else rm -f
".deps/memcheck_x86_linux-mc_main.Tpo"; exit 1; fi
Using built-in specs.
Target: i686-pc-linux-gnu
Configured with: ../gcc-4.2-svn/configure --enable-languages=c,c++
--prefix=/usr/local/4.2 --program-suffix=4.2.0
Thread model: posix
gcc version 4.2.0 20060603 (experimental)
/usr/local/4.2/libexec/gcc/i686-pc-linux-gnu/4.2.0/cc1 -E -quiet -v -I. -I.
-I.. -I../include -I../VEX/pub -MD memcheck_x86_linux-mc_main.d -MF
.deps/memcheck_x86_linux-mc_main.Tpo -MP -MT memcheck_x86_linux-mc_main.o -MQ
memcheck_x86_linux-mc_main.o -DHAVE_CONFIG_H -DVGA_x86=1 -DVGO_linux=1
-DVGP_x86_linux=1 mc_main.c -m32 -mpreferred-stack-boundary=2 -mtune=generic
-Wmissing-prototypes -Winline -Wall -Wshadow -Wpointer-arith
-Wstrict-prototypes -Wmissing-declarations -Wno-long-long -Wno-pointer-sign
-Wdeclaration-after-statement -fworking-directory -O -O2 -fpch-preprocess -o
mc_main.i
ignoring nonexistent directory
"/usr/local/4.2/lib/gcc/i686-pc-linux-gnu/4.2.0/../../../../i686-pc-linux-gnu/include"
ignoring duplicate directory "."
#include "..." search starts here:
#include <...> search starts here:
.
..
../include
../VEX/pub
/usr/local/include
/usr/local/4.2/include
/usr/local/4.2/lib/gcc/i686-pc-linux-gnu/4.2.0/include
/usr/include
End of search list.
/usr/local/4.2/libexec/gcc/i686-pc-linux-gnu/4.2.0/cc1 -fpreprocessed
mc_main.i -quiet -dumpbase mc_main.c -m32 -mpreferred-stack-boundary=2
-mtune=generic -auxbase-strip memcheck_x86_linux-mc_main.o -g -O -O2
-Wmissing-prototypes -Winline -Wall -Wshadow -Wpointer-arith
-Wstrict-prototypes -Wmissing-declarations -Wno-long-long -Wno-pointer-sign
-Wdeclaration-after-statement -version -o mc_main.s
GNU C version 4.2.0 20060603 (experimental) (i686-pc-linux-gnu)
compiled by GNU C version 4.2.0 20060603 (experimental).
GGC heuristics: --param ggc-min-expand=30 --param ggc-min-heapsize=4096
Compiler executable checksum: a42b76558fd6d6863486421b7e5252eb
as -V -Qy -o memcheck_x86_linux-mc_main.o mc_main.s
GNU assembler version 2.16.91.0.2 (i586-suse-linux) using BFD version
2.16.91.0.2 20050720 (SuSE Linux)
/usr/local/4.2/bin/gcc4.2.0 -Wno-long-long -Wno-pointer-sign
-Wdeclaration-after-statement -save-temps -v -o memcheck-x86-linux -static
-Wl,-defsym,valt_load_address=0x3800 -nodefaultlibs -nostartfiles -u _start
-m32 -Wl,-T,../valt_load_address_x86_linux.lds
memcheck_x86_linux-mc_leakcheck.o memcheck_x86_linux-mc_malloc_wrappers.o
memcheck_x86_linux-mc_main.o memcheck_x86_linux-mc_translate.o
../coregrind/libcoregrind_x86_linux.a ../VEX/libvex_x86_linux.a -lgcc
Using built-in specs.
Target: i686-pc-linux-gnu
Configured with: ../gcc-4.2-svn/configure --enable-languages=c,c++
--prefix=/usr/local/4.2 --program-suffix=4.2.0
Thread model: posix
gcc version 4.2.0 20060603 (experimental)
/usr/local/4.2/libexec/gcc/i686-pc-linux-gnu/4.2.0/collect2 -m elf_i386
-static -o memcheck-x86-linux -u _start
-L/usr/local/4.2/lib/gcc/i686-pc-linux-gnu/4.2.0
-L/usr/local/4.2/lib/gcc/i686-pc-linux-gnu/4.2.0/../../.. -defsym
valt_load_address=0x3800 -T ../valt_load_address_x86_linux.lds
memcheck_x86_linux-mc_leakcheck.o memcheck_x86_linux-mc_malloc_wrappers.o
memcheck_x86_linux-mc_main.o memcheck_x86_linux-mc_translate.o
../coregrind/libcoregrind_x86_linux.a ../VEX/libvex_x86_linux.a -lgcc
memcheck_x86_linux-mc_main.o:(.debug_info+0x9d50): undefined reference to
`hacky_auxmaps'
collect2: ld returned 1 exit status
--
Summary: [4.2 regression] undefined reference when compiling
valgrind 3.2.0
Product: gcc
Version: 4.2.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: c
AssignedTo: unassigned at gcc dot gnu dot org
ReportedBy: gcc at pdoerfler dot com
GCC target triplet: i686-pc-linux-gnu
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=27950
[Bug c/27950] [4.2 regression] undefined reference when compiling valgrind 3.2.0
--- Comment #1 from gcc at pdoerfler dot com 2006-06-08 11:00 --- Created an attachment (id=11633) --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=11633&action=view) preprocessed source Preprocessed source of the file that contains declaration of hacky_auxmaps. Probably, this is not enough to reproduce the failure. valgrind-3.2.0 is available from valgrind.org ./configure make to reproduce at least on x86 -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=27950
[Bug tree-optimization/28029] New: [4.2 regression] missed optimization with -ftree-vectorize
0
vectorizer.cpp:16: note: skip.
vectorizer.cpp:16: note: ==> examining statement: D.2381_12 = pretmp.24_1 +
D.2380_11
vectorizer.cpp:16: note: skip.
vectorizer.cpp:16: note: ==> examining statement: D.2382_17 = *D.2381_12
vectorizer.cpp:16: note: vectype: vector float
vectorizer.cpp:16: note: nunits = 4
vectorizer.cpp:16: note: ==> examining statement: D.2383_19 = (const float *)
D.2379_10
vectorizer.cpp:16: note: skip.
vectorizer.cpp:16: note: ==> examining statement: D.2384_20 = D.2383_19 + op_3
vectorizer.cpp:16: note: skip.
vectorizer.cpp:16: note: ==> examining statement: D.2385_21 = *D.2384_20
vectorizer.cpp:16: note: vectype: const vector float
vectorizer.cpp:16: note: nunits = 4
vectorizer.cpp:16: note: ==> examining statement: D.2386_22 = D.2382_17 *
D.2385_21
vectorizer.cpp:16: note: get vectype for scalar type: float
vectorizer.cpp:16: note: get vectype with 4 units of type float
vectorizer.cpp:16: note: vectype: vector float
vectorizer.cpp:16: note: vectype: vector float
vectorizer.cpp:16: note: nunits = 4
vectorizer.cpp:16: note: ==> examining statement: *D.2381_12 = D.2386_22
vectorizer.cpp:16: note: vectype: vector float
vectorizer.cpp:16: note: nunits = 4
vectorizer.cpp:16: note: ==> examining statement: i_23 = i_4 + 1
vectorizer.cpp:16: note: skip.
vectorizer.cpp:16: note: ==> examining statement: if (D.2376_16 > i_23) goto
; else goto ;
vectorizer.cpp:16: note: skip.
vectorizer.cpp:16: note: ==> examining statement: :
vectorizer.cpp:16: note: skip.
vectorizer.cpp:16: note: === vect_analyze_dependences ===
vectorizer.cpp:16: note: dependence distance = 0.
vectorizer.cpp:16: note: accesses have the same alignment.
vectorizer.cpp:16: note: dependence distance modulo vf == 0 between *D.2381_12
and *D.2381_12
vectorizer.cpp:16: note: not vectorized: can't determine dependence between
*D.2384_20 and *D.2381_12
vectorizer.cpp:16: note: bad data dependence.
vectorizer.cpp:16: note: vectorized 0 loops in function.
The workaround with "op" is not needed with the current autovect-branch BTW.
--
Summary: [4.2 regression] missed optimization with -ftree-
vectorize
Product: gcc
Version: tree-ssa
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: tree-optimization
AssignedTo: unassigned at gcc dot gnu dot org
ReportedBy: gcc at pdoerfler dot com
GCC build triplet: i686-pc-linux-gnu
GCC host triplet: i686-pc-linux-gnu
GCC target triplet: i686-pc-linux-gnu
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=28029
[Bug tree-optimization/28030] New: missed optimization with -ftree-vectorize and templates
The following code only gets vectorized with explicitly copying the "size_"
member to the local "sz" variable for 4.1.1, 4.2.0 rev. 114610 and autovect
branch.
=
template
class vec
{
public:
vec& multiply(const vec& other)
{
// do something to make sure restrict is valid...
const T* __restrict__ op = other.data_;
T* __restrict__ tp = data_;
unsigned int sz = size_; // NEEDED!
for (unsigned int i=0; i;
=
Without the local variable I get the following output:
g++ -O3 -ftree-vectorize -ftree-vectorizer-verbose=7 -march=pentium-m -c
vectorizer.cpp
vectorizer.cpp:16: note: = analyze_loop_nest =
vectorizer.cpp:16: note: === vect_analyze_loop_form ===
vectorizer.cpp:16: note: split exit edge.
vectorizer.cpp:16: note: === get_loop_niters ===
vectorizer.cpp:16: note: not vectorized: number of iterations cannot be
computed.
vectorizer.cpp:16: note: bad loop form.
vectorizer.cpp:16: note: vectorized 0 loops in function.
It works fine for T in [float, double, long int]
Using
typedef int aint __attribute__ ((__aligned__(16)));
as in the examples doesn't help.
Replacing the template declaration with int and using the class in main() leads
to vectorization of the loop without needing the local variable.
--
Summary: missed optimization with -ftree-vectorize and templates
Product: gcc
Version: tree-ssa
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: tree-optimization
AssignedTo: unassigned at gcc dot gnu dot org
ReportedBy: gcc at pdoerfler dot com
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=28030
[Bug c++/29123] New: accepts typedef as elaborated type specifier in friend declaration
In the following testcase wrap::type is accepted as elaborate type specifier
which AFAIU is not permitted by the current standard.
-
template
struct wrap
{
typedef C type;
};
template
class foo
{
friend class wrap::type;
// T t;
};
template
class bar
{
// bar();
};
foo > f;
ICC9.1 rejects this code with:
testTypedefFriend.cpp(10): error: typedef "type" may not be used in an
elaborated type specifier
friend class wrap::type;
^
detected during instantiation of class "foo [with T=bar]" at
line 20
This code will probably be valid in C++0x if I understand
http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2005/n1791.pdf correctly.
But gcc doesn't implement that behavior either as can be seen by uncommenting
the two lines in the snippet above.
Possibly related to: PR22047
/usr/local/4.2/bin/g++4.2.0 -v -c testTypedefFriend.cpp
Using built-in specs.
Target: i686-pc-linux-gnu
Configured with: ../gcc-4.2-svn/configure --prefix=/usr/local/4.2
--program-suffix=4.2.0 --enable-languages=c,c++,fortran
Thread model: posix
gcc version 4.2.0 20060918 (experimental)
/usr/local/4.2/libexec/gcc/i686-pc-linux-gnu/4.2.0/cc1plus -quiet -v
-D_GNU_SOURCE testTypedefFriend.cpp -quiet -dumpbase testTypedefFriend.cpp
-mtune=generic -auxbase testTypedefFriend -version -o /tmp/cc79A4Nh.s
ignoring nonexistent directory
"/usr/local/4.2/lib/gcc/i686-pc-linux-gnu/4.2.0/../../../../i686-pc-linux-gnu/include"
#include "..." search starts here:
#include <...> search starts here:
/usr/local/4.2/lib/gcc/i686-pc-linux-gnu/4.2.0/../../../../include/c++/4.2.0
/usr/local/4.2/lib/gcc/i686-pc-linux-gnu/4.2.0/../../../../include/c++/4.2.0/i686-pc-linux-gnu
/usr/local/4.2/lib/gcc/i686-pc-linux-gnu/4.2.0/../../../../include/c++/4.2.0/backward
/usr/local/include
/usr/local/4.2/include
/usr/local/4.2/lib/gcc/i686-pc-linux-gnu/4.2.0/include
/usr/include
End of search list.
GNU C++ version 4.2.0 20060918 (experimental) (i686-pc-linux-gnu)
compiled by GNU C version 4.2.0 20060918 (experimental).
GGC heuristics: --param ggc-min-expand=30 --param ggc-min-heapsize=4096
Compiler executable checksum: febaaa122a095acbe60b52a3bc2a2eca
as -V -Qy -o testTypedefFriend.o /tmp/cc79A4Nh.s
GNU assembler version 2.16.91.0.5 (i586-suse-linux) using BFD version
2.16.91.0.5 20051219 (SUSE Linux)
--
Summary: accepts typedef as elaborated type specifier in friend
declaration
Product: gcc
Version: 3.3.6
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: c++
AssignedTo: unassigned at gcc dot gnu dot org
ReportedBy: gcc at pdoerfler dot com
GCC target triplet: i686-pc-linux-gnu
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=29123
