beginnings of some cuda linear algebra operations and a kernel to test them.

1 files changed, 99 insertions, 0 deletions

diff --git a/linalg_test.cu b/linalg_test.cu
new file mode 100644
index 0000000..c4647a7
--- /dev/null
+++ b/linalg_test.cu
@@ -0,0 +1,99 @@
+//-*-c-*-
+
+#include "linalg.h"
+
+extern "C"
+{
+
+__global__ void add(int n, float3 *a, float3 *b, float3 *out)
+{
+	int i;
+	for (i=0; i < n; i++)
+		out[i] = a[i] + b[i];
+}
+
+__global__ void addequal(int n, float3 *a, float3 *b)
+{
+	int i;
+	for (i=0; i < n; i++)
+		a[i] += b[i];
+}
+
+__global__ void sub(int n, float3 *a, float3 *b, float3 *out)
+{
+	int i;
+	for (i=0; i < n; i++)
+		out[i] = a[i] - b[i];
+}
+
+__global__ void subequal(int n, float3 *a, float3 *b)
+{
+	int i;
+	for (i=0; i < n; i++)
+		a[i] -= b[i];
+}
+
+__gloabl__ void addfloat(int n, float3 *a, float c, float3 *out)
+{
+	int i;
+	for (i=0; i < n; i++)
+		out[i] = a[i] + c;
+}
+
+__global__ void addfloatequal(int n, float3 *a, float c)
+{
+	int i;
+	for (i=0; i < n; i++)
+		a += c;
+}
+
+__global__ void subfloat(int n, float3 *a, float c, float3 *out)
+{
+	int i;
+	for (i=0; i < n; i++)
+		out[i] = a[i] - c;
+}
+
+__global__ void subfloatequal(int n, float3 *a, float c)
+{
+	int i;
+	for (i=0; i < n; i++)
+		a[i] -= c;
+}
+
+__global__ void mulfloat(int n, float3 *a, float c, float3 *out)
+{
+	int i;
+	for (i=0; i < n; i++)
+		out[i] = a[i]*c;
+}
+
+__global__ void mulfloatequal(int n, float3 *a, float c)
+{
+	int i;
+	for (i=0; i < n; i++)
+		a *= c;
+}
+
+__global__ void divfloat(int n, float3 *a, float c, float3 *out)
+{
+	int i;
+	for (i=0; i < n; i++)
+		out[i] = a[i]/c;
+}
+
+__global__ void divfloatequal(int n, float3 *a, float c)
+{
+	int i;
+	for (i=0; i < n; i++)
+		a /= c;
+}
+
+__global__ void dot(int n, float3 *a, float3 *b, float* out)
+{
+	int i;
+	for (i=0; i < n; i++)
+		out[i] = dot(a,b);
+}
+
+} // extern "c"