leonarski_fandClaude Opus 5.5
3ebf085e4c
get_host_total_memory (sysconf / hw.memsize / GlobalMemoryStatusEx), DefaultHostMemoryBudget at 75% of the total - decided from the total, never from what is free, as the GPU plans are - and HostWorkerPool, a process-wide pool (a third of the budget) that workers reserve their large buffers against before making them, so how many run at once is decided by bytes rather than threads. Only scheduling reads any of it. [HostMemory] tests. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01PXc1zD1yxHqaR1DGdCTHL4
78 lines
2.0 KiB
C++
78 lines
2.0 KiB
C++
// SPDX-FileCopyrightText: 2026 Filip Leonarski, Paul Scherrer Institute <filip.leonarski@psi.ch>
|
|
// SPDX-License-Identifier: GPL-3.0-only
|
|
|
|
#include "HostMemory.h"
|
|
|
|
#if defined(_WIN32)
|
|
#ifndef NOMINMAX
|
|
#define NOMINMAX
|
|
#endif
|
|
#include <windows.h>
|
|
#elif defined(__APPLE__)
|
|
#include <cstdint>
|
|
#include <sys/types.h>
|
|
#include <sys/sysctl.h>
|
|
#else
|
|
#include <unistd.h>
|
|
#endif
|
|
|
|
size_t get_host_total_memory() {
|
|
#if defined(_WIN32)
|
|
MEMORYSTATUSEX status{};
|
|
status.dwLength = sizeof(status);
|
|
if (!GlobalMemoryStatusEx(&status))
|
|
return 0;
|
|
return static_cast<size_t>(status.ullTotalPhys);
|
|
#elif defined(__APPLE__)
|
|
uint64_t bytes = 0;
|
|
size_t len = sizeof(bytes);
|
|
if (sysctlbyname("hw.memsize", &bytes, &len, nullptr, 0) != 0)
|
|
return 0;
|
|
return static_cast<size_t>(bytes);
|
|
#else
|
|
const long pages = sysconf(_SC_PHYS_PAGES), page = sysconf(_SC_PAGE_SIZE);
|
|
if (pages <= 0 || page <= 0)
|
|
return 0;
|
|
return static_cast<size_t>(pages) * static_cast<size_t>(page);
|
|
#endif
|
|
}
|
|
|
|
size_t DefaultHostMemoryBudget() {
|
|
return static_cast<size_t>(HOST_MEMORY_BUDGET_FRACTION * static_cast<double>(get_host_total_memory()));
|
|
}
|
|
|
|
HostWorkerPool &HostWorkerPool::Instance() {
|
|
static HostWorkerPool pool;
|
|
return pool;
|
|
}
|
|
|
|
HostWorkerPool::HostWorkerPool() {
|
|
SetSize(0);
|
|
}
|
|
|
|
void HostWorkerPool::SetSize(size_t bytes) {
|
|
std::lock_guard lock(m);
|
|
size = bytes > 0 ? bytes
|
|
: static_cast<size_t>(HOST_WORKER_POOL_FRACTION * static_cast<double>(DefaultHostMemoryBudget()));
|
|
cv.notify_all();
|
|
}
|
|
|
|
size_t HostWorkerPool::Size() {
|
|
std::lock_guard lock(m);
|
|
return size;
|
|
}
|
|
|
|
HostWorkerPool::Reservation::Reservation(HostWorkerPool &pool, size_t bytes) : pool(pool), bytes(bytes) {
|
|
std::unique_lock lock(pool.m);
|
|
pool.cv.wait(lock, [&] { return pool.reserved == 0 || pool.reserved + bytes <= pool.size; });
|
|
pool.reserved += bytes;
|
|
}
|
|
|
|
HostWorkerPool::Reservation::~Reservation() {
|
|
{
|
|
std::lock_guard lock(pool.m);
|
|
pool.reserved -= bytes;
|
|
}
|
|
pool.cv.notify_all();
|
|
}
|