[Gretl-devel] Re: Speed of matrix "block" operations

Am 14.10.19 um 20:02 schrieb Allin Cottrell: > On Sun, 13 Oct 2019, Marcin Błażejowski wrote: > >> On 13.10.2019 02:15, Allin Cottrell wrote: >>> Thanks, Jack. So, given the options I posited, our machines agree on a >>> best chunk size of (25 * 5000 * 8) bytes (=~ 1 MB) for use with >>> memcpy. (5000 being the number of rows in the matrix to be copied, and >>> 8 the number of bytes to represent a double-precision floating point >>> value.) >>> >>> Now, to optimize libgretl's copying of contiguous data, we just have >>> to figure out how that relates to the size of L1 or L2 cache, or >>> whatever is truly the relevant hardware parameter here! >> >> But Allin, isn't it something that could/should(?) be done by AVX >> extensions? But I have to admit that I no idea how to use such low-level >> optimisation for scripting language. > > The AVX extensions are facilities that a compiler may or may not end > up using as part of its optimization efforts. In the matrix-copying > context the relevant question would be how the compiler and C-library > jointly interpret and implement calls to memcpy (e.g. does memcpy > translate to some AVX-optimized variant?). > > From the timings we're looking at (thanks for sending yours, Marcin) > it's clear that whatever optimizations are employed by gcc and the > relevant C-libraries, they are not automatically dividing a big call > to memcpy into smaller chunks whenever that would speed up the total > copy. Hence the idea that we may want to try some dividing up at the > libgretl level. I suspect that things are slowing when we try to copy > more data than fits into L2 cache in one go. By the way, these are results for an Intel(R) Core(TM) i5-6600 CPU 3.30GHz, L1=265kb, L2=1MB, L3=6MB <output: ROW=5000, COL=500, LOOP=600> 1 columns per chunk: 1.5847s 2 columns per chunk: 0.7398s 5 columns per chunk: 0.2226s 10 columns per chunk: 0.0837s 20 columns per chunk: 0.0663s 21 columns per chunk: 0.0700s 22 columns per chunk: 0.0356s 23 columns per chunk: 0.0368s 24 columns per chunk: 0.0389s 25 columns per chunk: 0.0406s 35 columns per chunk: 0.0617s 45 columns per chunk: 0.0880s 50 columns per chunk: 0.1195s 100 columns per chunk: 0.3994s 125 columns per chunk: 0.3895s 500 columns per chunk: 1.7884s </> <output: ROW=10000, COL=500, LOOP=600> 1 columns per chunk: 3.2237s 2 columns per chunk: 1.6661s 5 columns per chunk: 0.6233s 10 columns per chunk: 0.2529s 20 columns per chunk: 0.1990s 21 columns per chunk: 0.2190s 22 columns per chunk: 0.0884s 23 columns per chunk: 0.0938s 24 columns per chunk: 0.1050s 25 columns per chunk: 0.1095s 35 columns per chunk: 0.2152s 45 columns per chunk: 0.3409s 50 columns per chunk: 0.3980s 100 columns per chunk: 0.6861s 125 columns per chunk: 0.8611s 500 columns per chunk: 11.8921s </>