From 1c0fd9bb4726f28b5ccf4491b8d108b00c884ec3 Mon Sep 17 00:00:00 2001
From: Joseph Redmon <pjreddie@gmail.com>
Date: Thu, 30 Oct 2014 06:26:41 +0000
Subject: [PATCH] im2col slightly faster
---
src/axpy.c | 13 +++++++++++++
1 files changed, 13 insertions(+), 0 deletions(-)
diff --git a/src/axpy.c b/src/axpy.c
index 10ffca4..eddfdc6 100644
--- a/src/axpy.c
+++ b/src/axpy.c
@@ -65,6 +65,11 @@
void axpy_ongpu(int N, float ALPHA, cl_mem X, int INCX, cl_mem Y, int INCY)
{
+ 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_setup();
cl_kernel kernel = get_axpy_kernel();
cl_command_queue queue = cl.queue;
@@ -73,8 +78,10 @@
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);
@@ -86,6 +93,10 @@
}
void copy_ongpu(int N, cl_mem X, int INCX, cl_mem Y, int INCY)
{
+ 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_setup();
cl_kernel kernel = get_copy_kernel();
cl_command_queue queue = cl.queue;
@@ -93,8 +104,10 @@
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);
--
Gitblit v1.10.0