From 809f924db2823b9e1eaf3efb9370380edc1f76ed Mon Sep 17 00:00:00 2001
From: Joseph Redmon <pjreddie@gmail.com>
Date: Fri, 23 Jan 2015 00:38:24 +0000
Subject: [PATCH] CUDA so fast
---
src/maxpool_layer.c | 78 ++-------------------------------------
1 files changed, 4 insertions(+), 74 deletions(-)
diff --git a/src/maxpool_layer.c b/src/maxpool_layer.c
index c05e939..834ebdb 100644
--- a/src/maxpool_layer.c
+++ b/src/maxpool_layer.c
@@ -1,4 +1,5 @@
#include "maxpool_layer.h"
+#include "cuda.h"
#include <stdio.h>
image get_maxpool_image(maxpool_layer layer)
@@ -32,9 +33,9 @@
layer->output = calloc(output_size, sizeof(float));
layer->delta = calloc(output_size, sizeof(float));
#ifdef GPU
- layer->indexes_cl = cl_make_int_array(layer->indexes, output_size);
- layer->output_cl = cl_make_array(layer->output, output_size);
- layer->delta_cl = cl_make_array(layer->delta, output_size);
+ layer->indexes_gpu = cuda_make_int_array(output_size);
+ layer->output_gpu = cuda_make_array(layer->output, output_size);
+ layer->delta_gpu = cuda_make_array(layer->delta, output_size);
#endif
return layer;
}
@@ -98,74 +99,3 @@
}
}
-#ifdef GPU
-cl_kernel get_forward_kernel()
-{
- static int init = 0;
- static cl_kernel kernel;
- if(!init){
- kernel = get_kernel("src/maxpool_layer.cl", "forward", 0);
- init = 1;
- }
- return kernel;
-}
-
-void forward_maxpool_layer_gpu(maxpool_layer layer, cl_mem input)
-{
- int h = (layer.h-1)/layer.stride + 1;
- int w = (layer.w-1)/layer.stride + 1;
- int c = layer.c;
- cl_kernel kernel = get_forward_kernel();
- cl_command_queue queue = cl.queue;
-
- cl_uint i = 0;
- cl.error = clSetKernelArg(kernel, i++, sizeof(layer.h), (void*) &layer.h);
- cl.error = clSetKernelArg(kernel, i++, sizeof(layer.w), (void*) &layer.w);
- cl.error = clSetKernelArg(kernel, i++, sizeof(layer.c), (void*) &layer.c);
- cl.error = clSetKernelArg(kernel, i++, sizeof(layer.stride), (void*) &layer.stride);
- cl.error = clSetKernelArg(kernel, i++, sizeof(layer.size), (void*) &layer.size);
- cl.error = clSetKernelArg(kernel, i++, sizeof(input), (void*) &input);
- cl.error = clSetKernelArg(kernel, i++, sizeof(layer.output_cl), (void*) &layer.output_cl);
- cl.error = clSetKernelArg(kernel, i++, sizeof(layer.indexes_cl), (void*) &layer.indexes_cl);
- check_error(cl);
-
- const size_t global_size[] = {h*w*c*layer.batch};
-
- cl.error = clEnqueueNDRangeKernel(queue, kernel, 1, 0, global_size, 0, 0, 0, 0);
- check_error(cl);
-}
-
-cl_kernel get_backward_kernel()
-{
- static int init = 0;
- static cl_kernel kernel;
- if(!init){
- kernel = get_kernel("src/maxpool_layer.cl", "backward", 0);
- init = 1;
- }
- return kernel;
-}
-
-void backward_maxpool_layer_gpu(maxpool_layer layer, cl_mem delta)
-{
- cl_kernel kernel = get_backward_kernel();
- cl_command_queue queue = cl.queue;
-
- cl_uint i = 0;
- cl.error = clSetKernelArg(kernel, i++, sizeof(layer.h), (void*) &layer.h);
- cl.error = clSetKernelArg(kernel, i++, sizeof(layer.w), (void*) &layer.w);
- cl.error = clSetKernelArg(kernel, i++, sizeof(layer.c), (void*) &layer.c);
- cl.error = clSetKernelArg(kernel, i++, sizeof(layer.stride), (void*) &layer.stride);
- cl.error = clSetKernelArg(kernel, i++, sizeof(layer.size), (void*) &layer.size);
- cl.error = clSetKernelArg(kernel, i++, sizeof(layer.delta_cl), (void*) &layer.delta_cl);
- cl.error = clSetKernelArg(kernel, i++, sizeof(delta), (void*) &delta);
- cl.error = clSetKernelArg(kernel, i++, sizeof(layer.indexes_cl), (void*) &layer.indexes_cl);
- check_error(cl);
-
- const size_t global_size[] = {layer.h*layer.w*layer.c*layer.batch};
-
- cl.error = clEnqueueNDRangeKernel(queue, kernel, 1, 0, global_size, 0, 0, 0, 0);
- check_error(cl);
-}
-
-#endif
--
Gitblit v1.10.0