This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: Ping! Re: [patch, libgfortran] PR37754 [4.4 Regression] READ I/O Performance regression from 4.3 to 4.4/4.5
Jerry,
FYI, I ran the Polyhedron 2005 benchmarks on gcc 4.4 branch
from 20090524 and 20090528 on x86_64-apple-darwin10 to see what
the effect was of the libgfortran io fixes. The results were...
20090524
================================================================================
Date & Time : 28 May 2009 19:20:27
Test Name : gfortran_lin_p4
Compile Command : gfortran -ffast-math -funroll-loops -msse3 -O3 %n.f90 -o %n
Benchmarks : ac aermod air capacita channel doduc fatigue gas_dyn induct linpk mdbx nf protein rnflow test_fpu tfft
Maximum Times : 2000.0
Target Error % : 0.100
Minimum Repeats : 10
Maximum Repeats : 100
Benchmark Compile Executable Ave Run Number Estim
Name (secs) (bytes) (secs) Repeats Err %
--------- ------- ---------- ------- ------- ------
ac 2.51 10000 11.84 10 0.0592
aermod 48.53 10000 28.21 10 0.0294
air 2.80 10000 7.82 10 0.0647
capacita 1.74 10000 43.84 10 0.0143
channel 0.99 10000 2.12 12 0.0891
doduc 5.55 10000 39.42 10 0.0182
fatigue 2.18 10000 10.78 10 0.0383
gas_dyn 3.11 10000 7.84 15 0.0907
induct 5.18 10000 32.12 10 0.0179
linpk 0.96 10000 30.17 10 0.0142
mdbx 1.77 10000 13.51 10 0.0177
nf 3.95 10000 38.89 14 0.0931
protein 4.49 10000 43.27 10 0.0713
rnflow 5.81 10000 33.17 10 0.0238
test_fpu 5.41 10000 11.33 10 0.0191
tfft 0.57 10000 2.63 23 0.0953
Geometric Mean Execution Time = 15.95 seconds
================================================================================
20090528
================================================================================
Date & Time : 28 May 2009 22:13:35
Test Name : gfortran_lin_p4
Compile Command : gfortran -ffast-math -funroll-loops -msse3 -O3 %n.f90 -o %n
Benchmarks : ac aermod air capacita channel doduc fatigue gas_dyn induct linpk mdbx nf protein rnflow test_fpu tfft
Maximum Times : 2000.0
Target Error % : 0.100
Minimum Repeats : 10
Maximum Repeats : 100
Benchmark Compile Executable Ave Run Number Estim
Name (secs) (bytes) (secs) Repeats Err %
--------- ------- ---------- ------- ------- ------
ac 1.44 10000 11.83 10 0.0190
aermod 48.64 10000 28.21 10 0.0139
air 2.81 10000 7.80 12 0.0671
capacita 1.73 10000 43.47 10 0.0122
channel 0.98 10000 2.11 12 0.0349
doduc 5.56 10000 39.41 10 0.0101
fatigue 2.36 10000 10.56 10 0.0836
gas_dyn 3.30 10000 7.72 15 0.0750
induct 4.87 10000 32.11 10 0.0781
linpk 0.92 10000 26.61 10 0.0401
mdbx 1.77 10000 13.50 10 0.0245
nf 3.93 10000 38.26 18 0.0989
protein 4.44 10000 42.85 10 0.0291
rnflow 5.80 10000 32.68 10 0.0135
test_fpu 5.42 10000 11.29 10 0.0114
tfft 0.57 10000 2.60 14 0.0822
Geometric Mean Execution Time = 15.72 seconds
================================================================================
So there appears to be no speed regressions and the best improvement
is about a 12% speedup for the linpk benchmark.
Jack