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

Direct I/O and Transparent HugePages

Browse source 
--direct-io for bypassing page cache (and using THP on Linux)

Up to 3-6x faster uncached loading, fewer pageouts, no page cache pollution.
This commit is contained in:
Pavel Fatin 2024-05-20 21:55:33 +02:00
parent 917dc8cfa6
commit 1b17ed7ab6
10 changed files with 297 additions and 30 deletions
Download patch file Download diff file Expand all files Collapse all files

9
common/common.cpp
View file

@ -1072,6 +1072,10 @@ bool gpt_params_find_arg(int argc, char ** argv, const std::string & arg, gpt_pa
params.use_mmap = false;
return true;
}
if (arg == "--direct-io") {
params.use_direct_io = true;
return true;
}
if (arg == "--numa") {
if (++i >= argc) {
invalid_param = true;
@ -1544,6 +1548,9 @@ void gpt_print_usage(int /*argc*/, char ** argv, const gpt_params & params) {
if (llama_supports_mmap()) {
printf(" --no-mmap do not memory-map model (slower load but may reduce pageouts if not using mlock)\n");
}
if (llama_supports_direct_io()) {
printf(" --direct-io use direct I/O (potentially faster uncached loading, fewer pageouts, no page cache pollution)\n");
}
printf(" --numa TYPE attempt optimizations that help on some NUMA systems\n");
printf(" - distribute: spread execution evenly over all nodes\n");
printf(" - isolate: only spawn threads on CPUs on the node that execution started on\n");
@ -1844,6 +1851,7 @@ struct llama_model_params llama_model_params_from_gpt_params(const gpt_params &
mparams.split_mode = params.split_mode;
mparams.tensor_split = params.tensor_split;
mparams.use_mmap = params.use_mmap;
mparams.use_direct_io = params.use_direct_io;
mparams.use_mlock = params.use_mlock;
mparams.check_tensors = params.check_tensors;
if (params.kv_overrides.empty()) {
@ -2706,6 +2714,7 @@ void dump_non_result_info_yaml(FILE * stream, const gpt_params & params, const l
fprintf(stream, "n_predict: %d # default: -1 (unlimited)\n", params.n_predict);
fprintf(stream, "n_probs: %d # only used by server binary, default: 0\n", sparams.n_probs);
fprintf(stream, "no_mmap: %s # default: false\n", !params.use_mmap ? "true" : "false");
fprintf(stream, "direct-io: %s # default: false\n", params.use_direct_io ? "true" : "false");
fprintf(stream, "penalize_nl: %s # default: false\n", sparams.penalize_nl ? "true" : "false");
fprintf(stream, "ppl_output_type: %d # default: 0\n", params.ppl_output_type);
fprintf(stream, "ppl_stride: %d # default: 0\n", params.ppl_stride);