VectorizationBase seems to wrongly detect the number of the physical cores

Elrod · January 4, 2023, 11:33am

CPUSummary.num_cores is wrong.

Once upon a time, it was critically important that I supported running under WINE.
using Hwloc.jl would segfault Julia under WINE, so after CPUSummary 1.8, I dropped Hwloc.

github.com

JuliaSIMD/CPUSummary.jl/blob/v0.1.8/src/CPUSummary.jl

module CPUSummary

using Static, Hwloc
using Static: Zero, One, gt, lt
using IfElse: ifelse

export cache_size, cache_linesize, cache_associativity, cache_type,
  cache_inclusive, num_cache, num_cores, num_threads

include("topology.jl")
const BASELINE_CORES = Int(num_cores()) * ((Sys.ARCH === :aarch64) && Sys.isapple() ? 2 : 1)
function __init__()
  Sys.isapple() && Sys.ARCH === :aarch64 && return # detect M1
  ccall(:jl_generating_output, Cint, ()) == 1 && return
  safe_topology_load!()
  if count_attr(:Core) ≢ BASELINE_CORES
    redefine_attr_count()
    foreach(redefine_cache, 1:4)
  end
  redefine_num_threads()

This file has been truncated. show original

I haven’t heard anything about WINE for a while, but I don’t think supporting it is still necessary.
so the solution might be to basically revert CPUSummary to its 1.8 version.

A PR to that effect would be welcome.
There are probably a few updates in behavior needed. Three that come to mind:

Delete CPUSummary.num_threads that exists in 1.8 but not in the latest version.
The meaning of the cache_size functions may also have changed since then.
Maybe hwloc supports the M1 now? Does it detect big vs little cores? Either way, at the time, Julia reported the 4 bit/4 small M1 as 8 threads, but now it is 4. The 8 big core version was also released since then, so my hack at the time is no longer appropriate.

Topic		Replies	Views
How to set up number of threads appropriately based on Hardware? New to Julia question	18	3242	July 1, 2021
JULIA_NUM_THREADS and physical cores New to Julia	3	3449	November 1, 2019
Why julia is not using all my CPU? General Usage	18	3897	April 25, 2020
Does Julia detect the maximum number of threads availible and if so how? Performance threads	29	6276	November 19, 2021
Decrease in performance using Threads.@threads in Linux Julia at Scale	16	2047	July 23, 2019

VectorizationBase seems to wrongly detect the number of the physical cores

Related topics