define flops events/metrics for AMD Zen4

facebookincubator / dynolog

Dynolog is a telemetry daemon for performance monitoring and tracing. It exports metrics from different components in the system like the linux kernel, CPU, disks, Intel PT, GPUs etc. Dynolog also integrates with pytorch and can trigger traces for distributed training applications.

MIT License

260 stars 38 forks source link

define flops events/metrics for AMD Zen4 #217

Closed Alston-Tang closed 8 months ago

Alston-Tang commented 8 months ago

Summary: define two new events zen4::fp_ret_x87_fp_ops.all and zen4::fp_ret_sse_avx_ops.all to count scalar/vector fp ops.

also define a new metric fp_ops_all that will use two events above on zen4 hosts and fallback to intel events on other hosts.

Differential Revision: D52861377

facebook-github-bot commented 8 months ago

This pull request was exported from Phabricator. Differential Revision: D52861377

facebook-github-bot commented 8 months ago

This pull request was exported from Phabricator. Differential Revision: D52861377

facebook-github-bot commented 8 months ago

This pull request has been merged in facebookincubator/dynolog@18f46eb6f2afc5c92dbb530f908ed5ad020782ea.