diff --git a/tests/main.cpp b/tests/main.cpp
index 25d458f..3fda820 100644
--- a/tests/main.cpp
+++ b/tests/main.cpp
@@ -40,68 +40,12 @@
 #include <ttmath/ttmath.h>
 #include "uinttest.h"
 
-//#include <windows.h>
+
 
 const char uint_tests_file[] = "tests.uint32";
 
 
 
-
-bool test_lahf()
-{
-#ifdef TTMATH_PLATFORM64
-
-
-	#ifndef __GNUC__
-			#error "another compiler than GCC is currently not supported in 64bit mode"
-	#endif
-
-
-	std::cout << "Processor 64bit: checking LAHF and SAHF instructions....";
-
-	unsigned long int is_lahf;
-
-
-	__asm__ __volatile__ (
-		
-		"push %%rax               \n"
-		"push %%rcx               \n"
-		"push %%rdx               \n"
-
-		"mov $0x80000001, %%rax   \n"
-		"CPUID                    \n"
-		"test $1, %%rcx           \n"
-		"setnz %%al               \n"
-		"movzx %%al, %%rbx        \n"
-
-		"pop %%rdx                \n"
-		"pop %%rcx                \n"
-		"pop %%rax                \n"
-
-		: "=b" (is_lahf)
-		:
-	    : "cc"
-		);
-
-	if( is_lahf )
-	{
-		std::cout << "ok" << std::endl;
-		return true;
-	}
-
-	std::cout << "fail" << std::endl;
-
-
-	return false;
-
-#endif
-
-
-	// 32bit platform
-return true;
-}
-
-
 void test_uint()
 {
 	UIntTest test;
@@ -116,8 +60,6 @@ int main()
 {
 using namespace ttmath;
 	
-	if( !test_lahf() )
-		return 1;
 
 	test_uint();
 
diff --git a/ttmath/ttmathuint.h b/ttmath/ttmathuint.h
index 5a4e363..8848f53 100644
--- a/ttmath/ttmathuint.h
+++ b/ttmath/ttmathuint.h
@@ -245,6 +245,8 @@ public:
 	register uint * p1 = table;
 	register uint * p2 = const_cast<uint*>(ss2.table);
 
+		// we don't have to use TTMATH_REFERENCE_ASSERT here
+		// this algorithm doesn't require it
 
 		#ifndef __GNUC__
 			
@@ -354,6 +356,8 @@ public:
 	register uint * p1 = table;
 	register uint c;
 
+		TTMATH_ASSERT( index < value_size )
+
 		#ifndef __GNUC__
 
 			__asm
@@ -466,6 +470,8 @@ public:
 	register uint * p1 = table;
 	register uint c;
 
+		TTMATH_ASSERT( index < value_size - 1 )
+
 		#ifndef __GNUC__
 			__asm
 			{
@@ -566,6 +572,9 @@ public:
 	register uint * p1 = table;
 	register uint * p2 = const_cast<uint*>(ss2.table);
 
+		// we don't have to use TTMATH_REFERENCE_ASSERT here
+		// this algorithm doesn't require it
+
 		#ifndef __GNUC__
 
 			__asm
@@ -667,6 +676,8 @@ public:
 	register uint * p1 = table;
 	register uint c;
 
+		TTMATH_ASSERT( index < value_size )
+
 		#ifndef __GNUC__
 			__asm
 			{
@@ -1424,15 +1435,12 @@ public:
 				push eax
 				push edx
 
-				and edx,-1
-				bsr eax, x 
+				mov edx,-1
+				bsr eax,x 
 				cmovz eax,edx
 				mov result, eax
 
-				//
 				pop edx
-				//
-
 				pop eax
 			}
 		#endif
@@ -1441,11 +1449,10 @@ public:
 		#ifdef __GNUC__
 			__asm__  __volatile__(
 
-			"push %%edx			\n"
-			"andl $-1,%%edx		\n"
 			"bsrl %1, %0		\n"
-			"cmovz %%edx,%0		\n"
-			"pop %%edx			\n"
+			"jnz 1f				\n"
+			"movl $-1, %0		\n"
+			"1:					\n"
 
 			: "=R" (result)
 			: "R" (x)
@@ -1507,6 +1514,8 @@ public:
 	*/
 	static uint SetBitInWord(uint value, uint bit)
 	{
+		TTMATH_ASSERT( bit < TTMATH_BITS_PER_UINT )
+
 		#ifndef __GNUC__
 			__asm
 			{
diff --git a/ttmath/ttmathuint64.h b/ttmath/ttmathuint64.h
index 6643bd9..a2daab2 100644
--- a/ttmath/ttmathuint64.h
+++ b/ttmath/ttmathuint64.h
@@ -825,14 +825,13 @@ namespace ttmath
 		#ifdef __GNUC__
 			__asm__  __volatile__(
 
-			"push %%rdx			\n"
-			"andq $-1,%%rdx		\n"
-			"bsrq %%rbx, %%rax	\n"
-			"cmovz %%rdx,%%rax	\n"
-			"pop %%rdx			\n"
+			"bsrq %1, %0		\n"
+			"jnz 1f				\n"
+			"movq $-1, %0		\n"
+			"1:					\n"
 
-			: "=a" (result)
-			: "b" (x)
+			: "=R" (result)
+			: "R" (x)
 			: "cc" );
 
 		#endif