[LLVMdev] darwin dragon-egg build issues

Jack Howarth howarth at bromo.med.uc.edu
Thu Apr 15 16:21:16 PDT 2010


Duncan,
    Do a quick check here on x86_64-apple-darwin10
with svn llvm and svn dragon-egg against release gcc 4.5.0,
the results from the himenoBMTxpa benchmark compiled at -O3
look pretty good. With stock gcc-4.5.0, we get...

Grid-size = M

mimax = 128 mjmax = 128 mkmax = 256
imax = 127 jmax = 127 kmax =255
 Start rehearsal measurement process.
 Measure the performance in 3 times.

 MFLOPS: 156.560865 time(s): 2.575476 1.733593e-03

 Now, start the actual measurement process.
 The loop will be excuted in 69 times
 This will take about one minute.
 Wait for a while

 Loop executed for 69 times
 Gosa : 1.449961e-03 
 MFLOPS measured : 156.771134	cpu : 59.156498
 Score based on Pentium III 600MHz using Fortran 77: 1.911843

and for the same through the plugin we get...

Grid-size = M

mimax = 128 mjmax = 128 mkmax = 256
imax = 127 jmax = 127 kmax =255
 Start rehearsal measurement process.
 Measure the performance in 3 times.

 MFLOPS: 151.130276 time(s): 2.668021 1.733593e-03

 Now, start the actual measurement process.
 The loop will be excuted in 67 times
 This will take about one minute.
 Wait for a while

 Loop executed for 67 times
 Gosa : 1.452966e-03 
 MFLOPS measured : 151.208295	cpu : 59.555058
 Score based on Pentium III 600MHz using Fortran 77: 1.844004

So dragon-egg is less than 5% slower.
            Jack



More information about the llvm-dev mailing list