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

Clean up thread blocks with spares calculation a bit

Browse source
This commit is contained in:
KerfuffleV2 2023-06-05 13:36:35 -06:00
parent fe9ed7d36f
commit 7ed5aca9ca
1 changed files with 1 additions and 1 deletions
Download patch file Download diff file Expand all files Collapse all files

2
llama.cpp
View file

@ -2160,7 +2160,7 @@ static void llama_convert_tensor_internal(const llama_load_tensor & tensor, llam
std::vector<std::thread> workers; std::vector<std::thread> workers;
for (auto tnum = 0, in_buff_offs = 0, out_buff_offs = 0; tnum < nthread; tnum++) { for (auto tnum = 0, in_buff_offs = 0, out_buff_offs = 0; tnum < nthread; tnum++) {
auto thr_blocks = blocks_per_thread + (tnum == nthread - 1 && spare_blocks ? spare_blocks : 0); // num blocks for this thread auto thr_blocks = blocks_per_thread + (tnum == nthread - 1 ? spare_blocks : 0); // num blocks for this thread
auto thr_elems = thr_blocks * block_size; // number of elements for this thread auto thr_elems = thr_blocks * block_size; // number of elements for this thread
auto thr_block_bytes = thr_blocks * block_size_bytes; // number of input bytes for this thread auto thr_block_bytes = thr_blocks * block_size_bytes; // number of input bytes for this thread