Ecosystem/xmrig
Ecosystem/xmrig
1
0
Fork 0
mirror of https://github.com/xmrig/xmrig.git synced 2024-10-29 20:47:50 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

Merge pull request #3254 from SChernykh/dev

Browse source 
Tweaked auto-tuning for Intel CPUs
This commit is contained in:
xmrig 2023-04-19 12:29:58 +07:00 committed by GitHub
commit 7b51e23aa0
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
1 changed files with 7 additions and 2 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

9
src/backend/cpu/platform/HwlocCpuInfo.cpp
View file

@ -298,8 +298,10 @@ void xmrig::HwlocCpuInfo::processTopLevelCache(hwloc_obj_t cache, const Algorith
cores.reserve(m_cores);
findByType(cache, HWLOC_OBJ_CORE, [&cores](hwloc_obj_t found) { cores.emplace_back(found); });
const bool L3_exclusive = isCacheExclusive(cache);
# ifdef XMRIG_ALGO_GHOSTRIDER
if ((algorithm == Algorithm::GHOSTRIDER_RTM) && (PUs > cores.size()) && (PUs < cores.size() * 2)) {
if ((algorithm == Algorithm::GHOSTRIDER_RTM) && L3_exclusive && (PUs > cores.size()) && (PUs < cores.size() * 2)) {
// Don't use E-cores on Alder Lake
cores.erase(std::remove_if(cores.begin(), cores.end(), [](hwloc_obj_t c) { return hwloc_bitmap_weight(c->cpuset) == 1; }), cores.end());
@ -311,7 +313,6 @@ void xmrig::HwlocCpuInfo::processTopLevelCache(hwloc_obj_t cache, const Algorith
# endif
size_t L3 = cache->attr->cache.size;
const bool L3_exclusive = isCacheExclusive(cache);
size_t L2 = 0;
int L2_associativity = 0;
size_t extra = 0;
@ -349,6 +350,10 @@ void xmrig::HwlocCpuInfo::processTopLevelCache(hwloc_obj_t cache, const Algorith
}
# ifdef XMRIG_ALGO_RANDOMX
if ((algorithm.family() == Algorithm::RANDOM_X) && L3_exclusive && (PUs > cores.size()) && (PUs < cores.size() * 2)) {
// Use all L3+L2 on latest Intel CPUs with P-cores, E-cores and exclusive L3 cache
cacheHashes = (L3 + L2) / scratchpad;
}
if (extra == 0 && algorithm.l2() > 0) {
cacheHashes = std::min<size_t>(std::max<size_t>(L2 / algorithm.l2(), cores.size()), cacheHashes);
}