This is the mail archive of the gcc@gcc.gnu.org mailing list for the GCC project.
Index Nav: | [Date Index] [Subject Index] [Author Index] [Thread Index] | |
---|---|---|
Message Nav: | [Date Prev] [Date Next] | [Thread Prev] [Thread Next] |
MSVC and gcc don't have much in common, other than both being C and/or C++ compilers with differing standards compliance goals, but my floating point performance comparison this week shows gcc-3.1 code running typically 15% faster than MSVC6 on P4 (20% if P4 specific code and abrupt underflow is selected). gcc.3.0.1 is not so consistently faster. In order to get this performance from gcc on Windows, I must use a non-standard configuration of binutils to get more frequently favorable data alignments, as well as working out mathinline.h. Don't tell BG, because my employer doesn't want to buck him on the issue of running free software on Windows. Most people will be running gcc for its compatibility with environments other than those where MSVC is workable. Too many factors involved for any direct comparison. In order to view the code generation conveniently, as opposed to make a working .s file, make your .o with -g on, and run objdump -S against it, and don't rely 100% on correctness of the display. ----- Original Message ----- From: "akbara" <akbar_a@acm.org> To: "Mailing List - GCC" <gcc@gcc.gnu.org> Sent: Thursday, September 06, 2001 2:06 PM Subject: gcc versus cl > hey, > i was wondering, but do you guys know of any articles where the author > compares > gcc release 3.0.1 hopefully against the newest release of microsoft's > compiler. > be intersting to read.. > > cause i always get windows developers telling me what's so special about gcc > and what it can do better then cl... > i'm not an advanced user of gcc or anything, but usually i just tell them > that usually we can get builds that generate faster code in our floating > point intensive apps... > although gcc takes a while to compile though... and iirc, the stl > implemention with gcc is supposed to be a lot more robust/complete then > microsoft's take at it.. > > btw, is there a way to get gcc -S to dump with our own variable names that > we used when we wrote the orignal code (instead of having gcc dump with the > way it handles everythign internally..)? > > anyways, > l8rz > -akbara. >
Index Nav: | [Date Index] [Subject Index] [Author Index] [Thread Index] | |
---|---|---|
Message Nav: | [Date Prev] [Date Next] | [Thread Prev] [Thread Next] |