From e36182cd8c5dd5c6d0aa1f77cf5cdca87e8bb1f0 Mon Sep 17 00:00:00 2001
From: Joseph Redmon <pjreddie@gmail.com>
Date: Fri, 21 Nov 2014 23:35:19 +0000
Subject: [PATCH] cleaned up data parsing a lot. probably nothing broken?
---
src/maxpool_layer.c | 89 ++++++++++++++++++++++++++++++++++++++++++--
1 files changed, 84 insertions(+), 5 deletions(-)
diff --git a/src/maxpool_layer.c b/src/maxpool_layer.c
index 01eed45..df19040 100644
--- a/src/maxpool_layer.c
+++ b/src/maxpool_layer.c
@@ -27,9 +27,15 @@
layer->c = c;
layer->size = size;
layer->stride = stride;
- layer->indexes = calloc(((h-1)/stride+1) * ((w-1)/stride+1) * c*batch, sizeof(int));
- layer->output = calloc(((h-1)/stride+1) * ((w-1)/stride+1) * c*batch, sizeof(float));
- layer->delta = calloc(((h-1)/stride+1) * ((w-1)/stride+1) * c*batch, sizeof(float));
+ int output_size = ((h-1)/stride+1) * ((w-1)/stride+1) * c * batch;
+ layer->indexes = calloc(output_size, sizeof(int));
+ 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);
+ #endif
return layer;
}
@@ -66,7 +72,7 @@
int index = cur_w + layer.w*(cur_h + layer.h*(k + b*layer.c));
int valid = (cur_h >= 0 && cur_h < layer.h &&
cur_w >= 0 && cur_w < layer.w);
- float val = (valid != 0) ? input[index] : -INFINITY;
+ float val = (valid != 0) ? input[index] : -FLT_MAX;
max_i = (val > max) ? index : max_i;
max = (val > max) ? val : max;
}
@@ -79,7 +85,7 @@
}
}
-void backward_maxpool_layer(const maxpool_layer layer, float *input, float *delta)
+void backward_maxpool_layer(const maxpool_layer layer, float *delta)
{
int i;
int h = (layer.h-1)/layer.stride + 1;
@@ -92,3 +98,76 @@
}
}
+#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_setup();
+ 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_setup();
+ 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