Re: [HACKERS] GIN FailedAssertions on Itanium2 with Intel
I suggest you test for the Intel compiler, and if it is there, make the static variable volatile, and add a comment about why that is being done. --- Sergey E. Koposov wrote: On Sat, 2 Sep 2006, Bruce Momjian wrote: Teodor Sigaev wrote: What does that option do? Is it practical to enable it for the entire backend? From docs: Disables inline expansion of standard library or intrinsic functions. And isn't this a straightforward compiler bug they should be notified about? What's a choice? Now I see 3: 1) -O1 2) volatile 3) -nolib_inline IMHO, only -O1 is guarantee for other possible places... But I'm not familiar enough with such kinds of bugs. My guess is that the compiler writers saw you calling a libc function, and assumed that library could not modify the file static variable, forgetting that the libc function can call back into the original file. Can you detect the Itanium compiler and optimization levels via preprocessor symbols, and test for that, and throw an #error? No, it's impossible. Unfortunately the __OPTIMIZE__ preproc. symbol of icc doesn't allow to distinguish between different optimization levels. (only between -O0 and anything else). Regards, Sergey *** Sergey E. Koposov Max Planck Institute for Astronomy/Sternberg Astronomical Institute Tel: +49-6221-528-349 Web: http://lnfm1.sai.msu.ru/~math E-mail: [EMAIL PROTECTED] ---(end of broadcast)--- TIP 4: Have you searched our list archives? http://archives.postgresql.org -- Bruce Momjian [EMAIL PROTECTED] EnterpriseDBhttp://www.enterprisedb.com + If your life is a hard drive, Christ can be your backup. + ---(end of broadcast)--- TIP 3: Have you checked our extensive FAQ? http://www.postgresql.org/docs/faq
Re: [HACKERS] GIN FailedAssertions on Itanium2 with Intel compiler
On Sat, 2 Sep 2006, Bruce Momjian wrote: Teodor Sigaev wrote: What does that option do? Is it practical to enable it for the entire backend? From docs: Disables inline expansion of standard library or intrinsic functions. And isn't this a straightforward compiler bug they should be notified about? What's a choice? Now I see 3: 1) -O1 2) volatile 3) -nolib_inline IMHO, only -O1 is guarantee for other possible places... But I'm not familiar enough with such kinds of bugs. My guess is that the compiler writers saw you calling a libc function, and assumed that library could not modify the file static variable, forgetting that the libc function can call back into the original file. Can you detect the Itanium compiler and optimization levels via preprocessor symbols, and test for that, and throw an #error? No, it's impossible. Unfortunately the __OPTIMIZE__ preproc. symbol of icc doesn't allow to distinguish between different optimization levels. (only between -O0 and anything else). Regards, Sergey *** Sergey E. Koposov Max Planck Institute for Astronomy/Sternberg Astronomical Institute Tel: +49-6221-528-349 Web: http://lnfm1.sai.msu.ru/~math E-mail: [EMAIL PROTECTED] ---(end of broadcast)--- TIP 4: Have you searched our list archives? http://archives.postgresql.org
Re: [HACKERS] GIN FailedAssertions on Itanium2 with Intel
Teodor Sigaev wrote: What does that option do? Is it practical to enable it for the entire backend? From docs: Disables inline expansion of standard library or intrinsic functions. And isn't this a straightforward compiler bug they should be notified about? What's a choice? Now I see 3: 1) -O1 2) volatile 3) -nolib_inline IMHO, only -O1 is guarantee for other possible places... But I'm not familiar enough with such kinds of bugs. My guess is that the compiler writers saw you calling a libc function, and assumed that library could not modify the file static variable, forgetting that the libc function can call back into the original file. Can you detect the Itanium compiler and optimization levels via preprocessor symbols, and test for that, and throw an #error? -- Bruce Momjian [EMAIL PROTECTED] EnterpriseDBhttp://www.enterprisedb.com + If your life is a hard drive, Christ can be your backup. + ---(end of broadcast)--- TIP 6: explain analyze is your friend
Re: [HACKERS] GIN FailedAssertions on Itanium2 with Intel compiler
Huh, it's a over-optimization by icc on Itanium. With -00 or -02 there is no any problem, only -O2 produces such effect. The problem is in code Sorry, right With -00 or -01 there is no any problem -- Teodor Sigaev E-mail: [EMAIL PROTECTED] WWW: http://www.sigaev.ru/ ---(end of broadcast)--- TIP 6: explain analyze is your friend
Re: [HACKERS] GIN FailedAssertions on Itanium2 with Intel compiler
% icc -O2 -o 1 1.c ./1 CALL cmp SET SI = 1 BUG: SI==0 % icc -O1 -o 1 1.c ./1 CALL cmp SET SI = 1 OK BTW, this example works correct with -nolib-inline (http://www.intel.com/software/products/compilers/clin/docs/main_cls/mergedprojects/copts_cls/ccpp_options/option_nolib_inline.htm): % icc -O2 -nolib_inline -o 1 1.c ./1 CALL cmp SET SI = 1 OK PS Have anybody any thoughts about workaround? -- Teodor Sigaev E-mail: [EMAIL PROTECTED] WWW: http://www.sigaev.ru/ ---(end of broadcast)--- TIP 9: In versions below 8.0, the planner will ignore your desire to choose an index scan if your joining column's datatypes do not match
Re: [HACKERS] GIN FailedAssertions on Itanium2 with Intel compiler
On Fri, Sep 01, 2006 at 03:10:44PM +0400, Teodor Sigaev wrote: BTW, this example works correct with -nolib-inline (http://www.intel.com/software/products/compilers/clin/docs/main_cls/mergedprojects/copts_cls/ccpp_options/option_nolib_inline.htm): % icc -O2 -nolib_inline -o 1 1.c ./1 CALL cmp SET SI = 1 OK PS Have anybody any thoughts about workaround? What does that option do? Is it practical to enable it for the entire backend? And isn't this a straightforward compiler bug they should be notified about? Sounds like the -fno-strict-aliasing option for gcc, the compiler is making some assumptions about the code that arn't true, so we tell the compiler not to do that. Have a nice day, -- Martijn van Oosterhout kleptog@svana.org http://svana.org/kleptog/ From each according to his ability. To each according to his ability to litigate. signature.asc Description: Digital signature
Re: [HACKERS] GIN FailedAssertions on Itanium2 with Intel compiler
What does that option do? Is it practical to enable it for the entire backend? From docs: Disables inline expansion of standard library or intrinsic functions. And isn't this a straightforward compiler bug they should be notified about? What's a choice? Now I see 3: 1) -O1 2) volatile 3) -nolib_inline IMHO, only -O1 is guarantee for other possible places... But I'm not familiar enough with such kinds of bugs. -- Teodor Sigaev E-mail: [EMAIL PROTECTED] WWW: http://www.sigaev.ru/ ---(end of broadcast)--- TIP 9: In versions below 8.0, the planner will ignore your desire to choose an index scan if your joining column's datatypes do not match
Re: [HACKERS] GIN FailedAssertions on Itanium2 with Intel compiler
Huh, it's a over-optimization by icc on Itanium. With -00 or -02 there is no any
problem, only -O2 produces such effect. The problem is in code at lines 125-172
in ginutils.c:
static bool needUnique = false;
int cmpFunc(...) {
...
if (...) needUnique = true;
...
}
...
needUnique = false;
qsort(, cmpFunc);
if (needUnique)
And, needUnique was setted to true in last call of cmpFunc (by accident, in
fact), so between last call and checking of needUnique there isn't any call of
function. Insertion after qsort() any call (elog, for example) solves the problem.
If needUnique is marked as volatile, all is ok too. But this way doesn't seem
to me as reasonable, because there is a lot of places with potentially the same
problem... and thats may cause unpredictable failures. May be, better way is
limiting optimization level on Itanium with icc.
--
Teodor Sigaev E-mail: [EMAIL PROTECTED]
WWW: http://www.sigaev.ru/
---(end of broadcast)---
TIP 5: don't forget to increase your free space map settings
Re: [HACKERS] GIN FailedAssertions on Itanium2 with Intel compiler
Simple illustration:
#include stdio.h
#include stdlib.h
static char SI = 0;
static int
cmp(const void *a, const void *b) {
puts(CALL cmp);
if ( *(int*)a == *(int*)b ) {
puts(SET SI = 1);
SI = 1;
return 0;
}
return ( *(int*)a *(int*)b ) ? 1 : -1;
}
int
main(int argn, char *argv[]) {
int a[]={43,43};
SI = 0;
qsort(a, sizeof(a)/sizeof(int), sizeof(int), cmp);
if ( SI )
puts(OK);
else
puts(BUG: SI==0);
return 0;
}
% icc -O2 -o 1 1.c ./1
CALL cmp
SET SI = 1
BUG: SI==0
% icc -O1 -o 1 1.c ./1
CALL cmp
SET SI = 1
OK
--
Teodor Sigaev E-mail: [EMAIL PROTECTED]
WWW: http://www.sigaev.ru/
---(end of broadcast)---
TIP 6: explain analyze is your friend
[HACKERS] GIN FailedAssertions on Itanium2 with Intel compiler
Hello -hackers, I see a make check failures on Itanium2 platform with Intel Compiler with CVS HEAD. A failure is coming for GIN. The problem is coming at the assertion at ginbulk.c:62 TRAP: FailedAssertion(!(res != 0), File: ginbulk.c, Line: 62) during the index creation from the regr. tests CREATE INDEX ii on array_index_op_test using gin(i); The backtraces before the assertion (unfortunately from intel debugger...): (idb) bt #0 0x4014d370 in ginChooseElem (accum=0x3, heapptr=0x182, entries=0x120012, nentry=16384, low=2, high=16, offset=8192) at ginbulk.c:62 #1 0x4014cd00 in ginChooseElem (accum=0x3, heapptr=0x182, entries=0x120012, nentry=16384, low=2, high=16, offset=8192) at ginbulk.c:168 #2 0x4014c790 in ginInsertRecordBA (accum=0x3, heapptr=0x182, entries=0x120012, nentry=16384) at ginbulk.c:192 #3 0x40123380 in ginBuildCallback (index=0x3, htup=0x182, values=0x120012, isnull=0x4000 no value, tupleIsAlive=2 '\002', state=0x180010) at gininsert.c:203 #4 0x40174ee0 in IndexBuildHeapScan (heapRelation=0x3, indexRelation=0x182, indexInfo=0x120012, callback=0x4000, callback_state=0x182) at index.c:1534 #5 0x40124ea0 in ginbuild (fcinfo=0x3) at gininsert.c:311 #6 0x406eaf00 in OidFunctionCall3 (functionId=3, arg1=1649267441666, arg2=77309411346, arg3=16384) at fmgr.c:1460 #7 0x40177390 in reindex_relation (relid=3, toast_too=2 '\002') at index.c:1282 #8 0x40172700 in index_create (heapRelationId=3, indexRelationName=0x182 no value, indexRelationId=18, indexInfo=0x4000, accessMethodObjectId=2, tableSpaceId=16, classObjectId=0x2000, reloptions=1649267441666, isprimary=0 '\000', isconstraint=0 '\000', allow_system_table_mods=0 '\000', skip_build=0 '\000', concurrent=0 '\000') at index.c:782 ---Type return to continue, or q return to quit--- #9 0x40270df0 in DefineIndex (heapRelation=0x3, indexRelationName=0x182 no value, indexRelationId=18, accessMethodName=0x4000 no value, tableSpaceName=0x182 no value, attributeList=0x180010, predicate=0x2000, rangetable=0x182, options=0x0, unique=0 '\000', primary=0 '\000', isconstraint=0 '\000', is_alter_table=0 '\000', check_rights=1 '\001', skip_build=0 '\000', quiet=0 '\000', concurrent=0 '\000') at indexcmds.c:440 #10 0x4051b3b0 in ProcessUtility (parsetree=0x3, params=0x182, dest=0x120012, completionTag=0x4000 no value) at utility.c:789 #11 0x40518100 in PortalRunMulti (portal=0x3, dest=0x182, altdest=0x120012, completionTag=0x4000 no value) at pquery.c:1062 #12 0x40516f90 in PortalRun (portal=0x3, count=1649267441666, dest=0x120012, altdest=0x4000, completionTag=0x182 no value) at pquery.c:700 #13 0x4050e770 in exec_simple_query (query_string=0x3 no value) at postgres.c:1003 #14 0x4050bc20 in PostgresMain (argc=3, argv=0x182, username=0x120012 no value) at postgres.c:3263 #15 0x4045fc20 in BackendRun (port=0x3) at postmaster.c:2848 #16 0x4045ea80 in ServerLoop () at postmaster.c:2485 #17 0x4045c310 in PostmasterMain (argc=3, argv=0x182) at postmaster.c:950 #18 0x4035a0b0 in main (argc=3, argv=0x182) at main.c:187 (idb) bt full #0 0x4014d370 in ginChooseElem (accum=0x3, heapptr=0x182, entries=0x120012, nentry=16384, low=2, high=16, offset=8192) at ginbulk.c:62 middle=Info: symbol middle is defined but not allocated (optimized away) no value pos=Info: symbol pos is defined but not allocated (optimized away) no value #1 0x4014cd00 in ginChooseElem (accum=0x3, heapptr=0x182, entries=0x120012, nentry=16384, low=2, high=16, offset=8192) at ginbulk.c:168 middle=Info: symbol middle is defined but not allocated (optimized away) no value pos=Info: symbol pos is defined but not allocated (optimized away) no value #2 0x4014c790 in ginInsertRecordBA (accum=0x3, heapptr=0x182, entries=0x120012, nentry=16384) at ginbulk.c:192 i=Info: symbol i is defined but not allocated (optimized away) no value nbit=Info: symbol nbit is defined but not allocated (optimized away) no value offset=Info: symbol offset is defined but not allocated (optimized away) no value #3 0x40123380 in ginBuildCallback (index=0x3, htup=0x182, values=0x120012, isnull=0x4000 no value, tupleIsAlive=2 '\002', state=0x180010) at gininsert.c:203 ---Type return to continue, or q return to quit--- buildstate=Info: symbol buildstate is defined but not allocated (optimized away) no value oldCtx=Info: symbol oldCtx is defined but not allocated (optimized away) no value #4 0x40174ee0 in IndexBuildHeapScan (heapRelation=0x3, indexRelation=0x182, indexInfo=0x120012,
