Mail Archives: pgcc/1998/05/25/09:34:06
Well, there was a discussion about the size of programs compiled by pgcc
on this list.
I made a little testing on my program and...
gcc 2.7.2.3, -m386 -O2 -malign-jumps=0 -malign-loops=0 -malign-functions=0
-ffast-math
file size: 44672
-fno-strength-reduce added:
file size: 44336
pgcc 19980508, haifa enabled, -mpentium -malign-double -O3 -malign-jumps=0
-malign-loops=0 -malign-functions=0 -fno-inline-functions -ffast-math
file size: 46668
pgcc 19980508, haifa disabled, -mpentium -malign-double -O3 -malign-jumps=0
-malign-loops=0 -malign-functions=0 -fno-inline-functions -ffast-math
file size: 46636
pgcc 19980508, haifa enabled, -mpentium -malign-double -Os -malign-jumps=0
-malign-loops=0 -malign-functions=0 -ffast-math
file size: 46732
pgcc 19980508, haifa disabled, -mpentium -malign-double -Os -malign-jumps=0
-malign-loops=0 -malign-functions=0 -ffast-math
file size: 46700
Old pgcc (made against gcc 2.7.2), -mpentium -O3 -malign-double -malign-jumps=0
-malign-loops=0 -malign-functions=0 -fno-omit-frame-pointer -ffast-math
-fno-inline-functions
file size: 44608
without -malign-double: 44592
The Haifa scheduler gives slightly larger code. I must check the speed
differences. The "-O3 -fno-inline-functions" gives more compact code than -Os.
Sorry guys, but pgcc gives (in)significantly larger code than gcc 2.7.2.3.
Everything has been linked with the same libgcc.a (from gcc 2.7.2.3).
The reason of the code bloat must be in egcs, because the old pgcc gives
smaller programs.
Krzysztof Strasburger
- Raw text -