From 472e2c62c9ba5c76f1dc6bbe55686d85976aa28f Mon Sep 17 00:00:00 2001
From: Olivier BICHLER <olivier.bichler@cea.fr>
Date: Mon, 7 Apr 2025 14:23:28 +0200
Subject: [PATCH] Optimized Fc Impl
---
.../backend/cpu/operator/FCImpl_kernels.hpp | 19 +++++++------------
1 file changed, 7 insertions(+), 12 deletions(-)
diff --git a/include/aidge/backend/cpu/operator/FCImpl_kernels.hpp b/include/aidge/backend/cpu/operator/FCImpl_kernels.hpp
index b77f749f..ca4d5def 100644
--- a/include/aidge/backend/cpu/operator/FCImpl_kernels.hpp
+++ b/include/aidge/backend/cpu/operator/FCImpl_kernels.hpp
@@ -96,21 +96,16 @@ void FCImpl_cpu_forward_kernel(const DimSize_t batchSize,
const B* biases = static_cast<const B*>(biases_);
O* output = static_cast<O*>(output_);
- if (biases == nullptr) {
- std::fill(output, output+(batchSize*outputFeatureSize), B(0));
- }
- else {
- for (std::size_t batch = 0; batch < batchSize; ++batch) {
- std::copy(biases, biases+outputFeatureSize, output+(batch*outputFeatureSize));
- }
- }
-
- for (std::size_t batch = 0; batch < batchSize; ++batch) {
- for (std::size_t out = 0; out < outputFeatureSize; ++out) {
+#ifdef _OPENMP
+ #pragma omp parallel for collapse(2) if (batchSize * outputFeatureSize > 32)
+#endif
+ for (int batch = 0; batch < static_cast<int>(batchSize); ++batch) {
+ for (int out = 0; out < static_cast<int>(outputFeatureSize); ++out) {
+ const auto biasVal = (biases) ? biases[out] : B(0);
output[out + batch*outputFeatureSize] = std::inner_product(input + batch*inputFeatureSize,
input + (batch + 1)*inputFeatureSize,
weights + out*inputFeatureSize,
- output[out + batch*outputFeatureSize]);
+ biasVal);
}
}
}
--
GitLab