From 08b757a0bf76efe8c76b453063a1bb19315bcaa6 Mon Sep 17 00:00:00 2001
From: Joseph Redmon <pjreddie@gmail.com>
Date: Wed, 14 Jan 2015 20:18:57 +0000
Subject: [PATCH] Stable, needs to be way faster
---
src/axpy.c | 30 ++++++++++++++++++++----------
1 files changed, 20 insertions(+), 10 deletions(-)
diff --git a/src/axpy.c b/src/axpy.c
index c4ec1eb..857579c 100644
--- a/src/axpy.c
+++ b/src/axpy.c
@@ -1,24 +1,24 @@
#include "mini_blas.h"
-inline void axpy_cpu(int N, float ALPHA, float *X, int INCX, float *Y, int INCY)
+void axpy_cpu(int N, float ALPHA, float *X, int INCX, float *Y, int INCY)
{
int i;
for(i = 0; i < N; ++i) Y[i*INCY] += ALPHA*X[i*INCX];
}
-inline void scal_cpu(int N, float ALPHA, float *X, int INCX)
+void scal_cpu(int N, float ALPHA, float *X, int INCX)
{
int i;
for(i = 0; i < N; ++i) X[i*INCX] *= ALPHA;
}
-inline void copy_cpu(int N, float *X, int INCX, float *Y, int INCY)
+void copy_cpu(int N, float *X, int INCX, float *Y, int INCY)
{
int i;
for(i = 0; i < N; ++i) Y[i*INCY] = X[i*INCX];
}
-inline float dot_cpu(int N, float *X, int INCX, float *Y, int INCY)
+float dot_cpu(int N, float *X, int INCX, float *Y, int INCY)
{
int i;
float dot = 0;
@@ -65,7 +65,11 @@
void axpy_ongpu(int N, float ALPHA, cl_mem X, int INCX, cl_mem Y, int INCY)
{
- cl_setup();
+ axpy_ongpu_offset(N,ALPHA,X,0,INCX,Y,0,INCY);
+}
+
+void axpy_ongpu_offset(int N, float ALPHA, cl_mem X, int OFFX, int INCX, cl_mem Y, int OFFY, int INCY)
+{
cl_kernel kernel = get_axpy_kernel();
cl_command_queue queue = cl.queue;
@@ -73,39 +77,45 @@
cl.error = clSetKernelArg(kernel, i++, sizeof(N), (void*) &N);
cl.error = clSetKernelArg(kernel, i++, sizeof(ALPHA), (void*) &ALPHA);
cl.error = clSetKernelArg(kernel, i++, sizeof(X), (void*) &X);
+ cl.error = clSetKernelArg(kernel, i++, sizeof(OFFX), (void*) &OFFX);
cl.error = clSetKernelArg(kernel, i++, sizeof(INCX), (void*) &INCX);
cl.error = clSetKernelArg(kernel, i++, sizeof(Y), (void*) &Y);
+ cl.error = clSetKernelArg(kernel, i++, sizeof(OFFY), (void*) &OFFY);
cl.error = clSetKernelArg(kernel, i++, sizeof(INCY), (void*) &INCY);
check_error(cl);
const size_t global_size[] = {N};
- clEnqueueNDRangeKernel(queue, kernel, 1, 0, global_size, 0, 0, 0, 0);
+ cl.error = clEnqueueNDRangeKernel(queue, kernel, 1, 0, global_size, 0, 0, 0, 0);
check_error(cl);
}
void copy_ongpu(int N, cl_mem X, int INCX, cl_mem Y, int INCY)
{
- cl_setup();
+ copy_ongpu_offset(N,X,0,INCX,Y,0,INCY);
+}
+void copy_ongpu_offset(int N, cl_mem X, int OFFX, int INCX, cl_mem Y, int OFFY, int INCY)
+{
cl_kernel kernel = get_copy_kernel();
cl_command_queue queue = cl.queue;
cl_uint i = 0;
cl.error = clSetKernelArg(kernel, i++, sizeof(N), (void*) &N);
cl.error = clSetKernelArg(kernel, i++, sizeof(X), (void*) &X);
+ cl.error = clSetKernelArg(kernel, i++, sizeof(OFFX), (void*) &OFFX);
cl.error = clSetKernelArg(kernel, i++, sizeof(INCX), (void*) &INCX);
cl.error = clSetKernelArg(kernel, i++, sizeof(Y), (void*) &Y);
+ cl.error = clSetKernelArg(kernel, i++, sizeof(OFFY), (void*) &OFFY);
cl.error = clSetKernelArg(kernel, i++, sizeof(INCY), (void*) &INCY);
check_error(cl);
const size_t global_size[] = {N};
- clEnqueueNDRangeKernel(queue, kernel, 1, 0, global_size, 0, 0, 0, 0);
+ cl.error = clEnqueueNDRangeKernel(queue, kernel, 1, 0, global_size, 0, 0, 0, 0);
check_error(cl);
}
void scal_ongpu(int N, float ALPHA, cl_mem X, int INCX)
{
- cl_setup();
cl_kernel kernel = get_scal_kernel();
cl_command_queue queue = cl.queue;
@@ -118,7 +128,7 @@
const size_t global_size[] = {N};
- clEnqueueNDRangeKernel(queue, kernel, 1, 0, global_size, 0, 0, 0, 0);
+ cl.error = clEnqueueNDRangeKernel(queue, kernel, 1, 0, global_size, 0, 0, 0, 0);
check_error(cl);
}
#endif
--
Gitblit v1.10.0