avx-turbo

avx-turbo copied to clipboard

Add CPU count override

2

Useful when some cores are isolated with `isolcpus`, setting maximum number of threads gives `WARNING: can't run the requested number of threads (4) because there are only 1 available logical...

vient

Resolves https://github.com/travisdowns/avx-turbo/issues/21

vient

Currently, we barf a huge amount of output to the screen (especially with many threads) but don't actually summarize the most important information: the apparent CPU speeds for different instruction...

travisdowns

Things like turbo boost 3.0 mean that what core you pin to is important: some cores can clock higher than others. Currently we assign threads always starting from thread ID...

travisdowns

To minimize the chance of thermal limit related throttling, one should be able to specify a sleep period or sleep "duty cycle" which will rest between each test (or possibly...

travisdowns

Results for AMD Zen4 7950X

1

Results for AMD Zen4 7950X : https://gist.github.com/ErwanAliasr1/f50cc41d6c08ba6da7ee88980960ea0f

ErwanAliasr1

Not really an issue, and not sure if this is of any interest (to update the blog post? General musing?) But here are the results for: * [tiger-lake i7-1165G7](https://gist.github.com/damageboy/c2600bc6b7650cd80d9d0ab352465c58) (4-cores,...

damageboy

 get by ``` bash sudo ./avx-turbo --warmup-ms=1000 ``` ``` bash $ lscpu Architecture: x86_64 CPU op-mode(s): 32-bit, 64-bit Address sizes: 46 bits physical, 57 bits virtual Byte Order: Little...

Reihold-Chen

> A/M This is the ratio of the APERF and MPERF ratios exposed in an MSR. For details, see the [Intel SDM Vol 3](https://software.intel.com/en-us/download/intel-64-and-ia-32-architectures-sdm-combined-volumes-3a-3b-3c-and-3d-system-programming-guide), but basically APERF is a free...

Reihold-Chen

"Mops" can hit 5 columns wide for add and xor causing poor alignment. "A/M ratio" will never hit 10.0, so reclaim that extra column of space. Before:  After: 

austindonisan