* graphite optimizations
@ 2009-08-16 13:51 Jack Howarth
0 siblings, 0 replies; only message in thread
From: Jack Howarth @ 2009-08-16 13:51 UTC (permalink / raw)
To: gcc, spop
Sebastian,
With the current r150790 gcc trunk, I am not seeing any
particular improvements in the polyhedron benchmarks with the
available graphite loop optimizations...
================================================================================
Date & Time : 15 Aug 2009 13:41:47
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 1.60 10000 9.73 10 0.0136
aermod 57.99 10000 21.16 10 0.0062
air 4.16 10000 5.71 10 0.0263
capacita 2.67 10000 33.22 10 0.0168
channel 0.99 10000 1.83 10 0.0206
doduc 8.18 10000 27.29 10 0.0126
fatigue 3.08 10000 8.11 10 0.0847
gas_dyn 4.97 10000 4.50 14 0.0927
induct 6.74 10000 12.42 10 0.0155
linpk 0.99 10000 15.41 10 0.0409
mdbx 2.39 10000 11.44 10 0.0222
nf 2.65 10000 28.07 16 0.0953
protein 7.85 10000 36.55 10 0.0144
rnflow 8.30 10000 23.89 10 0.0072
test_fpu 6.80 10000 9.93 10 0.0161
tfft 0.73 10000 1.87 10 0.0534
Geometric Mean Execution Time = 11.31 seconds
================================================================================
================================================================================
Date & Time : 15 Aug 2009 14:30:46
Test Name : gfortran_lin_p4_graphite_interchange
Compile Command : gfortran -ffast-math -funroll-loops -msse3 -O3 -fgraphite-identity -floop-interchange %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 6.04 10000 9.39 10 0.0308
aermod 88.35 10000 20.92 10 0.0048
air 7.76 10000 0.18 17 0.0939
capacita 3.21 10000 33.29 10 0.0098
channel 2.25 10000 1.83 12 0.0200
doduc 8.92 10000 27.38 10 0.0131
fatigue 5.42 10000 8.33 10 0.0464
gas_dyn 5.67 10000 4.57 17 0.0977
induct 26.51 10000 7.42 10 0.0056
linpk 15.67 10000 15.53 10 0.0228
mdbx 2.80 10000 11.41 10 0.0088
nf 4.27 10000 27.97 15 0.0932
protein 8.25 10000 36.18 10 0.0316
rnflow 11.00 10000 24.19 10 0.0127
test_fpu 9.62 10000 10.69 10 0.0287
tfft 2.32 10000 1.91 10 0.0651
Geometric Mean Execution Time = 8.86 seconds
================================================================================
================================================================================
Date & Time : 15 Aug 2009 11:43:28
Test Name : gfortran_lin_p4_graphite_stripmine
Compile Command : gfortran -ffast-math -funroll-loops -msse3 -O3 -fgraphite-identity -floop-strip-mine %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 5.92 10000 9.37 10 0.0178
aermod 88.31 10000 21.01 10 0.0172
air 7.57 10000 0.18 34 0.0914
capacita 3.22 10000 33.30 10 0.0111
channel 2.49 10000 2.02 11 0.0955
doduc 8.94 10000 27.39 10 0.0204
fatigue 5.43 10000 8.33 10 0.0571
gas_dyn 5.66 10000 4.57 16 0.0885
induct 25.21 10000 20.18 10 0.0040
linpk 15.73 10000 15.53 10 0.0402
mdbx 2.84 10000 11.41 10 0.0094
nf 4.28 10000 27.95 10 0.0892
protein 8.26 10000 36.15 10 0.0192
rnflow 11.29 10000 23.95 10 0.0143
test_fpu 9.65 10000 10.68 10 0.0289
tfft 2.32 10000 1.91 10 0.0737
Geometric Mean Execution Time = 9.49 seconds
================================================================================
================================================================================
Date & Time : 15 Aug 2009 12:39:32
Test Name : gfortran_lin_p4_graphite_stripmine_interchange
Compile Command : gfortran -ffast-math -funroll-loops -msse3 -O3 -fgraphite-identity -floop-interchange -floop-strip-mine %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 6.07 10000 9.37 10 0.0156
aermod 88.54 10000 21.01 10 0.0191
air 7.79 10000 0.18 23 0.0748
capacita 3.20 10000 33.28 10 0.0112
channel 2.54 10000 2.02 14 0.0811
doduc 8.92 10000 27.39 10 0.0196
fatigue 5.42 10000 8.34 10 0.0990
gas_dyn 5.65 10000 4.57 13 0.0908
induct 26.53 10000 7.42 10 0.0085
linpk 15.77 10000 15.55 10 0.0350
mdbx 2.83 10000 11.41 10 0.0104
nf 4.27 10000 27.95 15 0.0837
protein 8.24 10000 36.16 10 0.0193
rnflow 11.29 10000 24.00 10 0.0115
test_fpu 9.60 10000 10.68 10 0.0140
tfft 2.31 10000 1.90 10 0.0455
Geometric Mean Execution Time = 8.92 seconds
================================================================================
on x86_64-apple-darwin10 (ignoring the air benchmark which is miscompiled).
Are you seeing the same thing or am I using other compiler options here which
might interfere with the graphite optimizations? I thought we should expect
some of these to show a few precent improvement in run time, no?
Jack
^ permalink raw reply [flat|nested] only message in thread
only message in thread, other threads:[~2009-08-15 23:25 UTC | newest]
Thread overview: (only message) (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2009-08-16 13:51 graphite optimizations Jack Howarth
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).