Testing OpenWatcom

[staalmannen]

Hi all. I have been playing around with the OpenWatcom (V2-fork) compiler lately and it is pretty cool (especially as a multi-target cross compiler, see for example: https://wiki.archlinux.org/index.php/Open_Watcom). It also has some "nostalgia value" for those of us who used to play DOS4GW games

I got curious to see how well it performs on Linux compared to GCC and looked for relevant tests in this article:

http://www.phoronix.com/scan.php?page=article&item=gcc_49_optimizations&num=5

I am especially looking at the tests that do not have external dependencies (looking at the compiler options under the graphs to make sure that nothing gets explicitly linked). simple stuff like libz would work to use with Watcom too but stuff like OpenMP are most likely doomed.

The first one I did was the "himeno" test since it was pretty fast

http://openbenchmarking.org/result/1410225-KH-HIMENOOWC10

OpenWatcom does pretty poorly vs GCC when optimization is activated. I do not have this feeling when running binaries built with it, but it might be true. I also assume that the similar -O0 results indicate that the Watcom libc and glibc do not have inherently different performance in this test?

I am now trying the "hint" benchmark, but it is a much slower process ... but the trends look even worse I think.

So... what are your thoughts about relevant benchmarks?
Ideas and feedback welcome.

[staalmannen]

some temporary updates on the "hint" test with -O3 optimization flag. My poor computer is however very slow in these tests

FLOAT:
owcc: 29431983.16
gcc: 82335706.39

so GCC some 3-4 X faster

DOUBLE
owcc: 64415960.69
gcc: 149858304.83

also about 3X more for GCC