diff --git a/libethash-cl/ethash_cl_miner.cpp b/libethash-cl/ethash_cl_miner.cpp
index 226d019ec..44ae5efe7 100644
--- a/libethash-cl/ethash_cl_miner.cpp
+++ b/libethash-cl/ethash_cl_miner.cpp
@@ -327,8 +327,19 @@ bool ethash_cl_miner::init(
 
 		// use selected platform
 		_platformId = min<unsigned>(_platformId, platforms.size() - 1);
-		ETHCL_LOG("Using platform: " << platforms[_platformId].getInfo<CL_PLATFORM_NAME>().c_str());
 
+		string platformName = platforms[_platformId].getInfo<CL_PLATFORM_NAME>();
+		ETHCL_LOG("Using platform: " << platformName.c_str());
+
+		int platformId = 0;
+		if (platformName == "NVIDIA CUDA")
+		{
+			platformId = 1;
+		}
+		else if (platformName == "AMD Accelerated Parallel Processing")
+		{
+			platformId = 2;
+		}
 		// get GPU device of the default platform
 		vector<cl::Device> devices = getDevices(platforms, _platformId);
 		if (devices.empty())
@@ -367,6 +378,8 @@ bool ethash_cl_miner::init(
 		addDefinition(code, "DAG_SIZE", (unsigned)(_dagSize / ETHASH_MIX_BYTES));
 		addDefinition(code, "ACCESSES", ETHASH_ACCESSES);
 		addDefinition(code, "MAX_OUTPUTS", c_maxSearchResults);
+		addDefinition(code, "PLATFORM", platformId);
+
 		//debugf("%s", code.c_str());
 
 		// create miner OpenCL program
diff --git a/libethash-cl/ethash_cl_miner_kernel.cl b/libethash-cl/ethash_cl_miner_kernel.cl
index b513d0c5a..c1eed75c6 100644
--- a/libethash-cl/ethash_cl_miner_kernel.cl
+++ b/libethash-cl/ethash_cl_miner_kernel.cl
@@ -36,6 +36,33 @@ __constant uint2 const Keccak_f1600_RC[24] = {
 	(uint2)(0x80008008, 0x80000000),
 };
 
+#if PLATFORM == 1 // CUDA
+static uint2 ROL2(const uint2 a, const int offset) {
+	uint2 result;
+	if (offset >= 32) {
+		asm("shf.l.wrap.b32 %0, %1, %2, %3;" : "=r"(result.x) : "r"(a.x), "r"(a.y), "r"(offset));
+		asm("shf.l.wrap.b32 %0, %1, %2, %3;" : "=r"(result.y) : "r"(a.y), "r"(a.x), "r"(offset));
+	}
+	else {
+		asm("shf.l.wrap.b32 %0, %1, %2, %3;" : "=r"(result.x) : "r"(a.y), "r"(a.x), "r"(offset));
+		asm("shf.l.wrap.b32 %0, %1, %2, %3;" : "=r"(result.y) : "r"(a.x), "r"(a.y), "r"(offset));
+	}
+	return result;
+}
+#elif PLATFORM == 2 // APP
+#pragma OPENCL EXTENSION cl_amd_media_ops : enable
+static uint2 ROL2(const uint2 vv, const int r)
+{
+	if (r <= 32)
+	{
+		return amd_bitalign((vv).xy, (vv).yx, 32 - r);
+	}
+	else
+	{
+		return amd_bitalign((vv).yx, (vv).xy, 64 - r);
+	}
+}
+#else
 static uint2 ROL2(const uint2 v, const int n)
 {
 	uint2 result;
@@ -51,6 +78,7 @@ static uint2 ROL2(const uint2 v, const int n)
 	}
 	return result;
 }
+#endif
 
 static void keccak_f1600_round(uint2* a, uint r, uint out_size)
 {