skein: merge the double implementations in one

based on alexis skein kernels, tested ok on SM 2.1 and 3.0

code is a bit hard to read but... well... users dont care :p

cuda_vectors.h

@@ -0,0 +1,34 @@
+#include "cuda_helper.h"
+
+/* Macros for uint2 operations (used by skein) */
+
+__device__ __forceinline__
+uint2 ROR8(const uint2 a) {
+	uint2 result;
+	result.x = __byte_perm(a.x, a.y, 0x4321);
+	result.y = __byte_perm(a.y, a.x, 0x4321);
+	return result;
+}
+
+__device__ __forceinline__
+uint2 ROL24(const uint2 a) {
+	uint2 result;
+	result.x = __byte_perm(a.x, a.y, 0x0765);
+	result.y = __byte_perm(a.y, a.x, 0x0765);
+	return result;
+}
+
+static __device__ __forceinline__ uint2 operator+ (const uint2 a, const uint32_t b)
+{
+#if 0 && defined(__CUDA_ARCH__) && CUDA_VERSION < 7000
+	uint2 result;
+	asm(
+		"add.cc.u32 %0,%2,%4; \n\t"
+		"addc.u32 %1,%3,%5;   \n\t"
+	: "=r"(result.x), "=r"(result.y) : "r"(a.x), "r"(a.y), "r"(b), "r"(0));
+	return result;
+#else
+	return vectorize(devectorize(a) + b);
+#endif
+}
+