Use new threadpool namespace for all of //executorch/... (#5851)

Use new threadpool namespace for all of //executorch/... (#5826)

Summary:
Pull Request resolved: #5826

Migrate all `//executorch/...` code to use the new threadpool namespace.

Reviewed By: mergennachin

Differential Revision: D63783748

fbshipit-source-id: 19e55a47689a765f58cc4b6cdbb1907cc0d24e5c
(cherry picked from commit 9263d88)

Co-authored-by: Dave Bort <[email protected]>

Author: pytorchbot

backends/xnnpack/runtime/XNNCompiler.cpp

@@ -1776,14 +1776,14 @@ ET_NODISCARD Error XNNCompiler::compileModel(
       subgraph.get(),
       /*weight_cache=*/nullptr, // TODO - support weight cache
       workspace,
-      torch::executorch::threadpool::get_pthreadpool(),
+      ::executorch::extension::threadpool::get_pthreadpool(),
       runtime_flags,
       &runtime_ptr);
 #else
   status = xnn_create_runtime_v3(
       subgraph.get(),
       /*weight_cache=*/nullptr, // TODO - support weight cache
-      torch::executorch::threadpool::get_pthreadpool(),
+      ::executorch::extension::threadpool::get_pthreadpool(),
       runtime_flags,
       &runtime_ptr);
 #endif

examples/models/llama2/main.cpp

@@ -59,13 +59,13 @@ int32_t main(int32_t argc, char** argv) {
 
 #if defined(ET_USE_THREADPOOL)
   uint32_t num_performant_cores = cpu_threads == -1
-      ? torch::executorch::cpuinfo::get_num_performant_cores()
+      ? ::executorch::extension::cpuinfo::get_num_performant_cores()
       : static_cast<uint32_t>(cpu_threads);
   ET_LOG(
       Info, "Resetting threadpool with num threads = %d", num_performant_cores);
   if (num_performant_cores > 0) {
-    torch::executorch::threadpool::get_threadpool()->_unsafe_reset_threadpool(
-        num_performant_cores);
+    ::executorch::extension::threadpool::get_threadpool()
+        ->_unsafe_reset_threadpool(num_performant_cores);
   }
 #endif
   // create llama runner

examples/models/llava/main.cpp

@@ -72,13 +72,13 @@ int32_t main(int32_t argc, char** argv) {
 
 #if defined(ET_USE_THREADPOOL)
   uint32_t num_performant_cores = cpu_threads == -1
-      ? torch::executorch::cpuinfo::get_num_performant_cores()
+      ? ::executorch::extension::cpuinfo::get_num_performant_cores()
       : static_cast<uint32_t>(cpu_threads);
   ET_LOG(
       Info, "Resetting threadpool with num threads = %d", num_performant_cores);
   if (num_performant_cores > 0) {
-    torch::executorch::threadpool::get_threadpool()->_unsafe_reset_threadpool(
-        num_performant_cores);
+    ::executorch::extension::threadpool::get_threadpool()
+        ->_unsafe_reset_threadpool(num_performant_cores);
   }
 #endif
   // create llama runner

extension/android/jni/jni_layer_llama.cpp

@@ -95,11 +95,11 @@ class ExecuTorchLlamaJni
 #if defined(ET_USE_THREADPOOL)
     // Reserve 1 thread for the main thread.
     uint32_t num_performant_cores =
-        torch::executorch::cpuinfo::get_num_performant_cores() - 1;
+        ::executorch::extension::cpuinfo::get_num_performant_cores() - 1;
     if (num_performant_cores > 0) {
       ET_LOG(Info, "Resetting threadpool to %d threads", num_performant_cores);
-      torch::executorch::threadpool::get_threadpool()->_unsafe_reset_threadpool(
-          num_performant_cores);
+      ::executorch::extension::threadpool::get_threadpool()
+          ->_unsafe_reset_threadpool(num_performant_cores);
     }
 #endif
 

extension/llm/custom_ops/op_sdpa.cpp

@@ -367,7 +367,7 @@ void cpu_flash_attention(
   int64_t qSlice = (qSize - 1) / qSplitSize + 1;
 #ifdef ET_USE_THREADPOOL
   int64_t num_thread =
-      torch::executorch::threadpool::get_threadpool()->get_thread_count();
+      ::executorch::extension::threadpool::get_threadpool()->get_thread_count();
 #else
   int64_t num_thread = 1;
 #endif

extension/parallel/thread_parallel.cpp

@@ -20,7 +20,7 @@ namespace {
 thread_local int64_t thread_num_ = 0;
 }
 
-using namespace torch::executorch::threadpool;
+using namespace ::executorch::extension::threadpool;
 
 inline int64_t divup(int64_t x, int64_t y) {
   return (x + y - 1) / y;