From 6553b3f0e3e55fc30a99c7d4b5798aa86d18a114 Mon Sep 17 00:00:00 2001
From: Joseph Redmon <pjreddie@gmail.com>
Date: Mon, 30 Mar 2015 02:31:47 +0000
Subject: [PATCH] no comment
---
src/convolutional_kernels.cu | 6 +++---
1 files changed, 3 insertions(+), 3 deletions(-)
diff --git a/src/convolutional_kernels.cu b/src/convolutional_kernels.cu
index c28731f..d260a95 100644
--- a/src/convolutional_kernels.cu
+++ b/src/convolutional_kernels.cu
@@ -11,15 +11,15 @@
__global__ void bias_output_kernel(float *output, float *biases, int n, int size)
{
int offset = blockIdx.x * blockDim.x + threadIdx.x;
- int filter = blockIdx.y % n;
- int batch = blockIdx.y / n;
+ int filter = blockIdx.y;
+ int batch = blockIdx.z;
if(offset < size) output[(batch*n+filter)*size + offset] = biases[filter];
}
void bias_output_gpu(float *output, float *biases, int batch, int n, int size)
{
- dim3 dimGrid((size-1)/BLOCK + 1, n*batch, 1);
+ dim3 dimGrid((size-1)/BLOCK + 1, n, batch);
dim3 dimBlock(BLOCK, 1, 1);
bias_output_kernel<<<dimGrid, dimBlock>>>(output, biases, n, size);
--
Gitblit v1.10.0