p/gallery
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Limit concurrency to number of hardware threads

Browse Source
This commit is contained in:
Přemysl Eric Janouch 2024-01-18 01:24:15 +01:00
parent 36f6612603
commit 819d2d80e0
Signed by: p
GPG Key ID: A0420B94F92B9493
1 changed files with 7 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

10
deeptagger/deeptagger.cpp
View File

@ -598,7 +598,11 @@ infer(Ort::Env &env, const char *path, const std::vector<std::string> &images)
Thumbnailing ctx;
for (const auto &path : images)
ctx.input.push(path);
for (auto i = g.batch; i--; )
auto workers = g.batch;
if (auto threads = std::thread::hardware_concurrency())
workers = std::min(workers, long(threads));
for (auto i = workers; i--; )
std::thread(thumbnail, std::ref(config), *width, *height,
std::ref(ctx)).detach();
@ -610,7 +614,7 @@ infer(Ort::Env &env, const char *path, const std::vector<std::string> &images)
std::unique_lock<std::mutex> output_lock(ctx.output_mutex);
ctx.output_cv.wait(output_lock,
[&]{ return ctx.output.size() == g.batch || ctx.done == g.batch; });
[&]{ return ctx.output.size() == g.batch || ctx.done == workers; });
// It would be possible to add dummy entries to the batch,
// so that the model doesn't need to be rebuilt.
@ -618,7 +622,7 @@ infer(Ort::Env &env, const char *path, const std::vector<std::string> &images)
run(ctx.output, config, session, shape);
ctx.output.clear();
}
if (ctx.done == g.batch)
if (ctx.done == workers)
break;
}
}