This is the mail archive of the java@gcc.gnu.org mailing list for the Java project.
Index Nav: | [Date Index] [Subject Index] [Author Index] [Thread Index] | |
---|---|---|
Message Nav: | [Date Prev] [Date Next] | [Thread Prev] [Thread Next] |
Other format: | [Raw text] |
Following the advice of the list, I've got new test results (with 263 iterations instead of 100 iterations, so all result times can be compared to those of Mr. Gabriele (err, John:). The difference in the C code execution time I chalk up to noise. The Java code execution is about 5% faster. I *also* happen to have a copy of the java-gui-branch of GCJ v3.5 from CVS. In the last set of runs, it didn't do so hot so I dropped it, but look at it with the --fast-math flas! I'm thinking that the bounds check isn't as important an optimization as the fast math functions. GCJ v3.5 CVS (java-gui-branch) (gcj Main.java --main=Main -O3 -save-temps --fast-math -fno-bounds-check -o java3.5.cvs-fast.exe; time ./java3.5.cvs-fast.exe) real 0m4.298s user 0m4.126s sys 0m0.016s GCJ v3.5 CVS (java-gui-branch) (gcj Main.java --main=Main -O3 --fast-math -o java3.5.cvs-fast-wb.exe; time ./java3.5.cvs-fast-wb.exe) real 0m4.323s user 0m4.246s sys 0m0.009s GCC v3.3.4 (gcc -o c-fast.exe -lm -ffast-math benchmark.c; time ./c-fast.exe) real 0m8.549s user 0m8.477s sys 0m0.003s GCJ v3.3.4 (gcj Main.java --main=Main -O3 -save-temps --fast-math -fno-bounds-check -o java3.3.4-fast.exe; time ./java3.3.4-fast.exe) real 0m15.811s user 0m15.466s sys 0m0.021s GCJ v3.5 CVS (java-gui-branch) (gcj Main.java --main=Main -O3 -o java3.5.cvs-slow.exe; time ./java3.5.cvs-slow.exe) real 0m35.503s user 0m35.135s sys 0m0.015s -- James Damour (Suvarov454) <suvarov454@users.sourceforge.net>
Attachment:
signature.asc
Description: This is a digitally signed message part
Index Nav: | [Date Index] [Subject Index] [Author Index] [Thread Index] | |
---|---|---|
Message Nav: | [Date Prev] [Date Next] | [Thread Prev] [Thread Next] |