Can I monitor AMX and AVX-512 usage with pcm?

[gcecchi]

I have some processes and also some virtual machines potentially using VAX-512 and/or AMX, depending on cpu architecture. Can I verify usage with pcm tools? How?
Thanks,
Gianluca

[rdementi]

Hi Gianluca,

sorry for delayed response. First you need to check if the VM exposes perfmon counters:
https://github.com/intel/pcm/blob/master/doc/FAQ.md#q3

Then there are specific events which one can collect with pcm-raw utility:
EXE.AMX_BUSY
512B events: search the event DB https://github.com/intel/perfmon/blob/main/SPR/events/sapphirerapids_core.json

Hope this helps,

Roman

[gcecchi]

Thanks Roman for the reply, for the event name and for the event DB link! My current hw is a 5th Gen intel Scalable Processor (EMR). It was ok to run pcm-raw on the host against the pid of the qemu-system-x86_64 process representing the VM. I'm testing a demo inside the VM, with stable diffusion, based on a Hugging Face model using OpenVINO acceleration. With the command "pcm-raw 1 -pid pid_nr -e EXE.AMX_BUSY" I'm now able to verify that when the VM runs with "-cpu host" the values are not zero for the 64 cores/threads representing the VM vpcus configured on it. At the same time if I run the same VM with "-cpu host,-amx-bf16,-amx-tile,-amx-int8" the event instead remains at zero (and for the image generation 16 seconds are necessary vs 6 seconds when AMX is used). Are the AVX-512 events these three ones: FP_ARITH_INST_RET, FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE, FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE? Can I distinguish between SSE and AVX usage?

[rdementi]

yes.

SSE/AVX: AFAIK, and as you likely already found out it is only possible to distinguish between the different width of operations with 128B and 256B events.

A different tool based on sampling (https://github.com/aayasin/perf-tools) can report instruction-mix: "./do.py profile -pm 100"