This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [gfortran] patch for pr 21130 - 38822 lines of Fortran 90 takes more than 10 minutes to compile on a dual 3GHz P4 Linux box with lots of RAM


Bud,

A last PS on this!

My Athlon1700 takes 25 minutes 47 seconds to compile the testcase, whout the patch. With the patch, it takes 2 minutes 46 seconds. ifort, on the other hand, takes 12 seconds to do the same job.

We still have some work to do, in spite of the very remarkable improvement that you have made. I believe that module reading has to be broken up into two stages: (i) To read the module file and create a namespace with all the symbols on it; and (ii) To read the symbols from that namespace, according to the requirements of the USE statement, and to populate the current namespace with them. That way, the file IO and the processing is only done once for each module and each compilation.

Paul


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]