horrible disk thorughput on itanium
Jakub Jelinek
jakub@redhat.com
Sat Dec 8 07:50:00 GMT 2001
I have repeated the tests on a couple of other arches,
and although TEST 4 is noticably slower than inlined putc_unlocked on IA-64,
it is still on all arches (with exception of SPARC) slower than non-inlined
putc_unlocked (even on SPARC such inlined putc is about 4 times faster than
current glibc putc).
IA-64, gcc-3.1-current
real 0m9.479s
user 0m9.469s
sys 0m0.004s
real 0m2.521s
user 0m2.515s
sys 0m0.005s
real 0m5.773s
user 0m5.765s
sys 0m0.003s
real 0m6.900s
user 0m6.886s
sys 0m0.010s
real 0m4.404s
user 0m4.395s
sys 0m0.007s
Alpha, gcc-3.1-current, -mtune=ev67
real 0m6.386s
user 0m6.380s
sys 0m0.008s
real 0m1.385s
user 0m1.378s
sys 0m0.007s
real 0m2.759s
user 0m2.754s
sys 0m0.005s
real 0m3.030s
user 0m3.021s
sys 0m0.010s
real 0m2.127s
user 0m2.124s
sys 0m0.003s
i686, gcc-3.1-current, -mcpu=i686
real 0m8.465s
user 0m8.460s
sys 0m0.010s
real 0m1.744s
user 0m1.740s
sys 0m0.000s
real 0m4.306s
user 0m4.280s
sys 0m0.020s
real 0m3.643s
user 0m3.630s
sys 0m0.010s
real 0m1.749s
user 0m1.730s
sys 0m0.010s
i686, gcc-2.96-RH, -mcpu=i686
real 0m9.559s
user 0m9.530s
sys 0m0.030s
real 0m1.876s
user 0m1.860s
sys 0m0.020s
real 0m4.706s
user 0m4.680s
sys 0m0.030s
real 0m4.315s
user 0m4.320s
sys 0m0.000s
real 0m1.883s
user 0m1.840s
sys 0m0.040s
Sparc, gcc-2.96-RH, -mcpu=ultrasparc
real 0m23.424s
user 0m23.410s
sys 0m0.010s
real 0m4.223s
user 0m4.220s
sys 0m0.010s
real 0m5.593s
user 0m5.580s
sys 0m0.010s
real 0m15.611s
user 0m15.560s
sys 0m0.050s
real 0m5.893s
user 0m5.880s
sys 0m0.010s
Jakub
More information about the Libc-alpha
mailing list