http://llvm.org/bugs/show_bug.cgi?id=3560
Summary: fpmath code generation performance issue
Product: tools
Version: trunk
Platform: PC
OS/Version: Linux
Status: NEW
Severity: normal
Priority: P2
Component: llvm-gcc
AssignedTo: [email protected]
ReportedBy: [email protected]
CC: [email protected]
Experiencing a 2x slowdown with nested loops in PHP source.
1) grab the php source (http://snaps.php.net)
2) tar xjvf php5.3[...].tar.bz2
3) cd php5.3[...]
4) ./configure --enable-debug --enable-cli --disable-all --without-sqlite3
5) make && make install && make test
ZEND_POST_INC_SPEC_CV_HANDLER is poorly generated, llvm-gcc generates x86 asm
which uses fldl/fstpl vs gcc which does not (and consequently does not
experience the slowdown).
The test in question is nestedloop(12), performs better/on par with gcc with
>=msse2.
--
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.
_______________________________________________
LLVMbugs mailing list
[email protected]
http://lists.cs.uiuc.edu/mailman/listinfo/llvmbugs
