Post

Andres Freund
@anarazel.de

Sometimes I hate performance stuff. How to quickly zero large amounts of memory differs rather vastly between cpu generations. On both systems core and memory are bound to node 0 and the same mmap flags are used (MAP_SHARED | MAP_ANONYMOUS | MAP_HUGETLB | MAP_POPULATE)

Andres Freund
@anarazel.de

2x Xeon Gold 5215: .. memzero_rep_movsb(): 6.171 GB/s 1x_mm256_store_si256(): 7.100 GB/s 2x_mm256_store_si256(): 10.685 GB/s 1x_mm256_stream_si256(): 6.988 GB/s 2x_mm256_stream_si256(): 6.994 GB/s ...

Andres Freund
@anarazel.de

2x Xeon Gold 6442Y: memzero_rep_movsb(): 11.155 GB/s 1x_mm256_store_si256(): 28.981 GB/s 2x_mm256_store_si256(): 10.034 GB/s 1x_mm256_stream_si256(): 29.154 GB/s 2x_mm256_stream_si256(): 29.155 GB/s

Andres Freund
@anarazel.de

I.e on Cascade Lake non-temporal stores suck and it's important to use more than one store per-loop. On Sapphire Rapids it's the other way round. On both ERMSB is considerably worse than the alternatives.

0 reposts0 likes

Andres Freund
@anarazel.de

Cascade lake has the added fun that using non-temporal stores on memory on another node is actually a lot faster than what's achievable on the local node. nodebind: 0, membind: 0 1x_mm_stream_si128(): 6.926 GB/s nodebind: 0, membind: 1 1x_mm_stream_si128(): 20.693 GB/s