From 0c363de882e8cb40a45948b65f2e1cb28c4d6bf8 Mon Sep 17 00:00:00 2001
From: piotrekobi <48731682+piotrekobi@users.noreply.github.com>
Date: Thu, 25 Aug 2022 16:07:48 +0200
Subject: [PATCH] Enable OMP multithreading in lookup_table_v2 (#45249)

* Add omp parallel for directives

* Revert "Add omp parallel for directives"

This reverts commit f4e4f8ddb12454018d9c1e49c074af2543659de6.

* Add #pragma omp parallel for to correct file

* Add check for _OPENMP definition

* Disable omp on gpu

* Trigger CI

* Readd check for _OPENMP definition

* Change macro disabling changes on GPU

* Improve macro readability
---
 paddle/phi/kernels/cpu/embedding_kernel.cc | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/paddle/phi/kernels/cpu/embedding_kernel.cc b/paddle/phi/kernels/cpu/embedding_kernel.cc
index 0430f7a005..7accdffe13 100644
--- a/paddle/phi/kernels/cpu/embedding_kernel.cc
+++ b/paddle/phi/kernels/cpu/embedding_kernel.cc
@@ -48,6 +48,10 @@ struct EmbeddingCPUFunctor {
     dev_ctx_.template Alloc<T>(out_);
     auto* output = out_->data<T>();
 
+#if defined(_OPENMP) && !defined(PADDLE_WITH_CUDA)
+#pragma omp parallel for
+#endif
+
     for (int64_t i = 0; i < ids_numel; ++i) {
       if (padding_idx_ != kNoPadding && ids[i] == padding_idx_) {
         memset(output + i * row_width, 0, row_width * sizeof(T));
-- 
GitLab