maybe they were a bit modest about revealing the full story.....
IBM's latest XLF compilers released for free for G5 show a 210% improvment on floating point scaler over GCC optimized -03 code, 70% improvement on vector with altivec over GCC optimized -03 and about 30 -40% better on integer over the GCC3.3 optimized -O3
http://arstechnica.infopop.net/Open...=4250963285&p=1
Fixed peterh's formatting (XL improvement over Absoft and GCC 3.3):
Type of Code G4 G5
Scalar +70% +210%
Vector +40% +70%
If this holds that places his 2.0GHz G5 (single)=254 at 787
Note also that the P4 2.66 is 255.
Inicdentally this is an improvement ONLY from recompiling the code! :O
Hmm.... I can't wait till a newly compiled SPEC comes out compiled with this processor. It seams the doubters have been proven wrong, the performance we have seen so far is only the tip of the iceberg.
Incidentally all apple did when compiling for the SPEC marks was compile with GCC 3.3 with -O3. They themselves said that the GCC didn't do any favours what so ever for the G5 and that it wasn't really producing code that took full advantage of the chip. Those IBM compiler results would seam to definately prove them right.
For those of you that will wonder how apple will squeeze 200% + performance improvement out of OSX 10.4 , my money is on them recompiling OS X with IBM's compiler!
IBM's latest XLF compilers released for free for G5 show a 210% improvment on floating point scaler over GCC optimized -03 code, 70% improvement on vector with altivec over GCC optimized -03 and about 30 -40% better on integer over the GCC3.3 optimized -O3
http://arstechnica.infopop.net/Open...=4250963285&p=1
Fixed peterh's formatting (XL improvement over Absoft and GCC 3.3):
Type of Code G4 G5
Scalar +70% +210%
Vector +40% +70%
If this holds that places his 2.0GHz G5 (single)=254 at 787
Note also that the P4 2.66 is 255.
Inicdentally this is an improvement ONLY from recompiling the code! :O
Hmm.... I can't wait till a newly compiled SPEC comes out compiled with this processor. It seams the doubters have been proven wrong, the performance we have seen so far is only the tip of the iceberg.
Incidentally all apple did when compiling for the SPEC marks was compile with GCC 3.3 with -O3. They themselves said that the GCC didn't do any favours what so ever for the G5 and that it wasn't really producing code that took full advantage of the chip. Those IBM compiler results would seam to definately prove them right.
For those of you that will wonder how apple will squeeze 200% + performance improvement out of OSX 10.4 , my money is on them recompiling OS X with IBM's compiler!