Google performance tools – part trois
Earlier this week I talked about the google performance tools.
The alternative memory allocator was the most interesting to me and I had a chance to test it out last night. The test I ran was very simple and unscientific. I have an application which allocated lots of little bits of memory so I thought it would be a good candidate for this test. I ran it both with the memory allocator in the glibc library and with the google alternative memory allocator.
The results were quite disappointing. The application did not run faster which is probably due to memory allocation being such a small part of the overall runtime that reducing it made no difference. And the application had a marginally smaller memory footprint, about 1.5% which could easily be attributed to randomness.
From what I can see, it is not worth the effort using at this point.
Updated June 3rd, 2007 – maybe it is worth the effort, I have to confess that my test ran a single thread only. I will have to set up a multi-threaded test and report on findings.
IM: fschiettecatte
YES…. it will NOT show a performance boost until you’re running more threads than CPUs…….
Great, that is good to know, I will set up a test tomorrow and blog the outcome, thanks for pointing that out, I came across your blog quite serendipitously over the weekend, very worthwhile already, thanks.