vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

ggml : style / formatting

Browse source
This commit is contained in:
Georgi Gerganov 2023-06-26 20:37:55 +03:00
parent 875a1e111e
commit 4a555b4539
No known key found for this signature in database
GPG key ID: 449E073F9DC10735
1 changed files with 21 additions and 7 deletions
Download patch file Download diff file Expand all files Collapse all files

28
ggml.c
View file

@ -16626,39 +16626,50 @@ typedef pthread_t ggml_thread_t;
#endif #endif
#ifdef __linux__ #ifdef __linux__
void set_numa_thread_affinity(int thread_n, int n_threads) void set_numa_thread_affinity(int thread_n, int n_threads) {
{ if (!ggml_is_numa()) {
if (!ggml_is_numa()) { return; } return;
}
// run thread on node_num thread_n / (threads per node) // run thread on node_num thread_n / (threads per node)
int node_num = thread_n / ((n_threads + g_state.numa.n_nodes - 1) / g_state.numa.n_nodes); const int node_num = thread_n / ((n_threads + g_state.numa.n_nodes - 1) / g_state.numa.n_nodes);
struct ggml_numa_node * node = &g_state.numa.nodes[node_num]; struct ggml_numa_node * node = &g_state.numa.nodes[node_num];
size_t setsize = CPU_ALLOC_SIZE(g_state.numa.total_cpus); size_t setsize = CPU_ALLOC_SIZE(g_state.numa.total_cpus);
cpu_set_t * cpus = CPU_ALLOC(g_state.numa.total_cpus); cpu_set_t * cpus = CPU_ALLOC(g_state.numa.total_cpus);
CPU_ZERO_S(setsize, cpus); CPU_ZERO_S(setsize, cpus);
for (size_t i = 0; i < node->n_cpus; ++i) { for (size_t i = 0; i < node->n_cpus; ++i) {
CPU_SET_S(node->cpus[i], setsize, cpus); CPU_SET_S(node->cpus[i], setsize, cpus);
} }
int rv = pthread_setaffinity_np(pthread_self(), setsize, cpus); int rv = pthread_setaffinity_np(pthread_self(), setsize, cpus);
if (rv) { if (rv) {
fprintf(stderr, "warning: pthread_setaffinity_np() failed: %s\n", fprintf(stderr, "warning: pthread_setaffinity_np() failed: %s\n",
strerror(rv)); strerror(rv));
} }
CPU_FREE(cpus); CPU_FREE(cpus);
} }
void clear_numa_thread_affinity(void)
{ void clear_numa_thread_affinity(void) {
if (!ggml_is_numa()) { return; } if (!ggml_is_numa()) {
return;
}
size_t setsize = CPU_ALLOC_SIZE(g_state.numa.total_cpus); size_t setsize = CPU_ALLOC_SIZE(g_state.numa.total_cpus);
cpu_set_t * cpus = CPU_ALLOC(g_state.numa.total_cpus); cpu_set_t * cpus = CPU_ALLOC(g_state.numa.total_cpus);
CPU_ZERO_S(setsize, cpus); CPU_ZERO_S(setsize, cpus);
for (unsigned i = 0; i < g_state.numa.total_cpus; ++i) { for (unsigned i = 0; i < g_state.numa.total_cpus; ++i) {
CPU_SET_S(i, setsize, cpus); CPU_SET_S(i, setsize, cpus);
} }
int rv = pthread_setaffinity_np(pthread_self(), setsize, cpus); int rv = pthread_setaffinity_np(pthread_self(), setsize, cpus);
if (rv) { if (rv) {
fprintf(stderr, "warning: pthread_setaffinity_np() failed: %s\n", fprintf(stderr, "warning: pthread_setaffinity_np() failed: %s\n",
strerror(rv)); strerror(rv));
} }
CPU_FREE(cpus); CPU_FREE(cpus);
} }
#else #else
@ -16699,10 +16710,12 @@ static void ggml_graph_compute_perf_stats_node(struct ggml_tensor * node, const
static thread_ret_t ggml_graph_compute_thread(void * data) { static thread_ret_t ggml_graph_compute_thread(void * data) {
struct ggml_compute_state * state = (struct ggml_compute_state *) data; struct ggml_compute_state * state = (struct ggml_compute_state *) data;
struct ggml_cgraph * cgraph = state->shared->cgraph; struct ggml_cgraph * cgraph = state->shared->cgraph;
const int n_threads = state->shared->n_threads; const int n_threads = state->shared->n_threads;
set_numa_thread_affinity(state->ith, n_threads); set_numa_thread_affinity(state->ith, n_threads);
int node_n = -1; int node_n = -1;
while (true) { while (true) {
if (atomic_fetch_sub(&state->shared->n_active, 1) == 1) { if (atomic_fetch_sub(&state->shared->n_active, 1) == 1) {
// all other threads are finished and spinning // all other threads are finished and spinning
@ -17165,6 +17178,7 @@ void ggml_graph_compute(struct ggml_context * ctx, struct ggml_cgraph * cgraph)
// this is a work thread too // this is a work thread too
ggml_graph_compute_thread(&workers[0]); ggml_graph_compute_thread(&workers[0]);
// don't leave affinity set on the main thread // don't leave affinity set on the main thread
clear_numa_thread_affinity(); clear_numa_thread_affinity();