diff --git a/base_c/include/embb/base/c/internal/atomic/and_assign.h b/base_c/include/embb/base/c/internal/atomic/and_assign.h
index bf3ed84..b156b1f 100644
--- a/base_c/include/embb/base/c/internal/atomic/and_assign.h
+++ b/base_c/include/embb/base/c/internal/atomic/and_assign.h
@@ -131,14 +131,21 @@ EMBB_DEFINE_AND_ASSIGN(4, "")
 #define EMBB_DEFINE_AND_ASSIGN(EMBB_PARAMETER_SIZE_BYTE, EMBB_ATOMIC_TC_SIZE_SUFFIX)\
   EMBB_PLATFORM_INLINE void EMBB_CAT2(embb_internal__atomic_and_assign_, \
   EMBB_PARAMETER_SIZE_BYTE)(EMBB_CAT2(EMBB_BASE_BASIC_TYPE_SIZE_, EMBB_PARAMETER_SIZE_BYTE) volatile* pointer_to_value, EMBB_CAT2(EMBB_BASE_BASIC_TYPE_SIZE_, EMBB_PARAMETER_SIZE_BYTE) value) { \
-  EMBB_CAT2(EMBB_BASE_BASIC_TYPE_SIZE_, EMBB_PARAMETER_SIZE_BYTE) result=0; \
-  __asm__ __volatile__("dsync\n\t" \
-  "ld" EMBB_ATOMIC_TC_SIZE_SUFFIX " %[result], %[pointer_to_value]\n\t" \
-  "and %[result], %[value]\n\t" \
-  "st" EMBB_ATOMIC_TC_SIZE_SUFFIX " %[pointer_to_value], %[result]\n\t" \
-  : [pointer_to_value]"+m" (*pointer_to_value), [result]"+d" (result) \
-  : [value]"d" (value)\
-  : "memory"); \
+  	register uint64_t e_reg = sizeof(value) == sizeof(uint32_t) ? (uint64_t)value : 0xFFFFFFFFull; \
+	register uint32_t add = ((uint32_t)pointer_to_value) & 0xFFFFFFFC; \
+	uint32_t pos = ((uint32_t)pointer_to_value & 0x3) * 8; \
+	uint32_t width = sizeof(value) == sizeof(uint32_t) ? 31 : sizeof(value) * 8; \
+	register uint64_t e_ins = ((uint64_t)width << 32) | pos; \
+	  __asm__ __volatile__ ("dsync\n\t" \
+			  "mov.a %[pointer_to_value], %[add]\n\t" \
+			  "loop_:\tld.w %H0, [%2]0\n\t" \
+			  "insert %L0, %H0, %[value], %A1\n\t" \
+			  "and %L0, %H0\n\t" \
+			  "cmpswap.w [%2]0, %A0\n\t" \
+			  "jne %H0, %L0, loop_" \
+			  : "+d"(e_reg), [e_ins]"+d"(e_ins) \
+			  : [pointer_to_value]"a"(pointer_to_value), [value] "r" (value), [add] "d" (add) \
+			  : "memory"); \
   }
 #else
 #error "No atomic fetch and store implementation found"
@@ -146,7 +153,7 @@ EMBB_DEFINE_AND_ASSIGN(4, "")
 
 EMBB_DEFINE_AND_ASSIGN(1, ".b")
 EMBB_DEFINE_AND_ASSIGN(2, ".h")
-EMBB_DEFINE_AND_ASSIGN(4, ".w")
+EMBB_DEFINE_AND_ASSIGN(4, "")
 
 #else
 #error "Unknown architecture"
diff --git a/base_c/include/embb/base/c/internal/atomic/compare_and_swap.h b/base_c/include/embb/base/c/internal/atomic/compare_and_swap.h
index 1da58ba..447e50d 100644
--- a/base_c/include/embb/base/c/internal/atomic/compare_and_swap.h
+++ b/base_c/include/embb/base/c/internal/atomic/compare_and_swap.h
@@ -132,25 +132,25 @@ EMBB_DEFINE_COMPARE_AND_SWAP(4, "")
   EMBB_PLATFORM_INLINE int EMBB_CAT2(embb_internal__atomic_compare_and_swap_, \
   EMBB_PARAMETER_SIZE_BYTE)(EMBB_CAT2(EMBB_BASE_BASIC_TYPE_SIZE_, EMBB_PARAMETER_SIZE_BYTE) volatile* pointer_to_value, EMBB_CAT2(EMBB_BASE_BASIC_TYPE_SIZE_, EMBB_PARAMETER_SIZE_BYTE) volatile* expected, \
   EMBB_CAT2(EMBB_BASE_BASIC_TYPE_SIZE_, EMBB_PARAMETER_SIZE_BYTE) desired) { \
-  register uint64_t e_reg = ((uint64_t)*expected << 32U) | desired; \
-  __asm__ __volatile__ ("cmpswap" EMBB_ATOMIC_TC_SIZE_SUFFIX " [%1]0, %A0\n\t" \
-  : "+d"(e_reg) \
-  : "a"(pointer_to_value) \
+  EMBB_CAT2(EMBB_BASE_BASIC_TYPE_SIZE_, EMBB_PARAMETER_SIZE_BYTE) result; \
+  register uint64_t e_reg = desired; \
+  __asm__ __volatile__ ("dsync\n\t" \
+	  "ld.w %H0, [%3]0\n\t" \
+	  "cmpswap.w [%2]0, %A0\n\t" \
+	  "eq %[result], %H0, %L0" \
+  : "+d"(e_reg), [result]"=&d" (result) \
+  : "a"(pointer_to_value), "a"(expected) \
   : "memory"); \
-  __asm__ __volatile__ ("dsync" : : : "memory"); \
-  if(*pointer_to_value == desired) { \
-	  return 1; \
-  } else { \
-	  return 0; \
-  } \
+  return result; \
   }
 #else
 #error "No atomic fetch and store implementation found"
 #endif
 
+//TODO: CAS-byte and CAS-short not needed yet, implement if needed
 EMBB_DEFINE_COMPARE_AND_SWAP(1, ".b")
 EMBB_DEFINE_COMPARE_AND_SWAP(2, ".h")
-EMBB_DEFINE_COMPARE_AND_SWAP(4, ".w")
+EMBB_DEFINE_COMPARE_AND_SWAP(4, "")
 
 #else
 #error "Unknown architecture"
diff --git a/base_c/include/embb/base/c/internal/atomic/fetch_and_add.h b/base_c/include/embb/base/c/internal/atomic/fetch_and_add.h
index 5bab66e..56b75e7 100644
--- a/base_c/include/embb/base/c/internal/atomic/fetch_and_add.h
+++ b/base_c/include/embb/base/c/internal/atomic/fetch_and_add.h
@@ -123,16 +123,24 @@ EMBB_DEFINE_FETCH_AND_ADD(4, "")
 #define EMBB_DEFINE_FETCH_AND_ADD(EMBB_PARAMETER_SIZE_BYTE, EMBB_ATOMIC_TC_SIZE_SUFFIX) \
   EMBB_PLATFORM_INLINE EMBB_CAT2(EMBB_BASE_BASIC_TYPE_SIZE_, EMBB_PARAMETER_SIZE_BYTE) EMBB_CAT2(embb_internal__atomic_fetch_and_add_, EMBB_PARAMETER_SIZE_BYTE) \
   (EMBB_CAT2(EMBB_BASE_BASIC_TYPE_SIZE_, EMBB_PARAMETER_SIZE_BYTE) volatile* pointer_to_value, EMBB_CAT2(EMBB_BASE_BASIC_TYPE_SIZE_, EMBB_PARAMETER_SIZE_BYTE) new_value) { \
-  EMBB_CAT2(EMBB_BASE_BASIC_TYPE_SIZE_, EMBB_PARAMETER_SIZE_BYTE) \
-  result=0,tmp1=0; \
-  __asm__ __volatile__ ("dsync\n\t" \
-  "ld" EMBB_ATOMIC_TC_SIZE_SUFFIX " %[result], %[pointer_to_value]\n\t" \
-  "add %[tmp1],%[new_value], %[result]\n\t" \
-  "st" EMBB_ATOMIC_TC_SIZE_SUFFIX " %[pointer_to_value], %[tmp1]\n\t" \
-  : [result] "+d" (result), [tmp1]"+d" (tmp1), [pointer_to_value]"+m" (*pointer_to_value) \
-  : [new_value]"d" (new_value) \
-  : "memory", "cc" ); \
-  return result; \
+	EMBB_CAT2(EMBB_BASE_BASIC_TYPE_SIZE_, EMBB_PARAMETER_SIZE_BYTE) result; \
+  	register uint64_t e_reg = sizeof(value) == sizeof(uint32_t) ? (uint64_t)value : 0ull; \
+	register uint32_t add = (uint32_t)pointer_to_value & 0xFFFFFFFC; \
+	uint32_t pos = ((uint32_t)pointer_to_value & 0x3) * 8; \
+	uint32_t width = sizeof(value) == sizeof(uint32_t) ? 31 : sizeof(value) * 8; \
+	register uint64_t e_mask = ((uint64_t)width << 32) | pos; \
+	__asm__ __volatile__ ("dsync\n\t" \
+		"mov.a %[pointer_to_value], %[add]\n\t" \
+		"loop_:\tld.w %H0, [%3]0\n\t" \
+		"insert %L0, %L0, %[value], %A1\n\t" \
+		"add" EMBB_ATOMIC_TC_SIZE_SUFFIX " %L0, %L0, %H0\n\t" \
+		"cmpswap.w [%3]0, %A0\n\t" \
+		"jne %H0, %L0, loop_\n\t" \
+		"extr.u %[result], %H0, %A1" \
+		: "+d"(e_reg), [e_mask]"+d"(e_mask), [result] "=&d" (result) \
+		: [pointer_to_value]"a"(pointer_to_value), [value] "r" (value), [add] "d" (add) \
+		: "memory"); \
+	return result; \
   }
 #else
 #error "No atomic fetch and store implementation found"
@@ -140,7 +148,7 @@ EMBB_DEFINE_FETCH_AND_ADD(4, "")
 
 EMBB_DEFINE_FETCH_AND_ADD(1, ".b")
 EMBB_DEFINE_FETCH_AND_ADD(2, ".h")
-EMBB_DEFINE_FETCH_AND_ADD(4, ".w")
+EMBB_DEFINE_FETCH_AND_ADD(4, "")
 
 #else
 #error "Unknown architecture"
diff --git a/base_c/include/embb/base/c/internal/atomic/or_assign.h b/base_c/include/embb/base/c/internal/atomic/or_assign.h
index 695452a..8b1d1e1 100644
--- a/base_c/include/embb/base/c/internal/atomic/or_assign.h
+++ b/base_c/include/embb/base/c/internal/atomic/or_assign.h
@@ -116,15 +116,21 @@ EMBB_DEFINE_OR_ASSIGN(4, "")
 #define EMBB_DEFINE_OR_ASSIGN(EMBB_PARAMETER_SIZE_BYTE, EMBB_ATOMIC_TC_SIZE_SUFFIX) \
   EMBB_PLATFORM_INLINE void EMBB_CAT2(embb_internal__atomic_or_assign_, EMBB_PARAMETER_SIZE_BYTE)(\
   EMBB_CAT2(EMBB_BASE_BASIC_TYPE_SIZE_, EMBB_PARAMETER_SIZE_BYTE) volatile* pointer_to_value, EMBB_CAT2(EMBB_BASE_BASIC_TYPE_SIZE_, EMBB_PARAMETER_SIZE_BYTE) value) { \
-  EMBB_CAT2(EMBB_BASE_BASIC_TYPE_SIZE_, EMBB_PARAMETER_SIZE_BYTE) result=0; \
-  __asm__ __volatile__( \
-  "dsync\n\t" \
-  "ld" EMBB_ATOMIC_TC_SIZE_SUFFIX " %[result], %[pointer_to_value]\n\t"\
-  "or %[result], %[value]\n\t" \
-  "st" EMBB_ATOMIC_TC_SIZE_SUFFIX " %[pointer_to_value], %[result]\n\t" \
-  : [result]"+d" (result), [pointer_to_value]"+m" (*pointer_to_value) \
-  : [value]"d" (value) \
-  : "memory"); \
+  	register uint64_t e_reg = sizeof(value) == sizeof(uint32_t) ? (uint64_t)value : 0ull; \
+	register uint32_t add = (uint32_t)pointer_to_value & 0xFFFFFFFC; \
+	uint32_t pos = ((uint32_t)pointer_to_value & 0x3) * 8; \
+	uint32_t width = sizeof(value) == sizeof(uint32_t) ? 31 : sizeof(value) * 8; \
+	register uint64_t e_ins = ((uint64_t)width << 32) | pos; \
+	  __asm__ __volatile__ ("dsync\n\t" \
+			  "mov.a %[pointer_to_value], %[add]\n\t" \
+			  "loop_:\tld.w %H0, [%2]0\n\t" \
+			  "insert %L0, %L0, %[value], %A1\n\t" \
+			  "or %L0, %H0\n\t" \
+			  "cmpswap.w [%2]0, %A0\n\t" \
+			  "jne %H0, %L0, loop_" \
+			  : "+d"(e_reg), [e_ins]"+d"(e_ins) \
+			  : [pointer_to_value]"a"(pointer_to_value), [value] "r" (value), [add] "d" (add) \
+			  : "memory"); \
   }
 
 #else
@@ -133,7 +139,7 @@ EMBB_DEFINE_OR_ASSIGN(4, "")
 
 EMBB_DEFINE_OR_ASSIGN(1, ".b")
 EMBB_DEFINE_OR_ASSIGN(2, ".h")
-EMBB_DEFINE_OR_ASSIGN(4, ".w")
+EMBB_DEFINE_OR_ASSIGN(4, "")
 
 #else
 #error "Unknown architecture"
diff --git a/base_c/include/embb/base/c/internal/atomic/swap.h b/base_c/include/embb/base/c/internal/atomic/swap.h
index 026b182..2f0ae68 100644
--- a/base_c/include/embb/base/c/internal/atomic/swap.h
+++ b/base_c/include/embb/base/c/internal/atomic/swap.h
@@ -118,17 +118,12 @@ EMBB_DEFINE_SWAP(4, "")
 #if defined(EMBB_PLATFORM_COMPILER_GNUC)
 #define EMBB_DEFINE_SWAP(EMBB_PARAMETER_SIZE_BYTE, EMBB_ATOMIC_TC_SIZE_SUFFIX) \
   EMBB_PLATFORM_INLINE EMBB_CAT2(EMBB_BASE_BASIC_TYPE_SIZE_, EMBB_PARAMETER_SIZE_BYTE) EMBB_CAT2(embb_internal__atomic_swap_, EMBB_PARAMETER_SIZE_BYTE)(\
-  EMBB_CAT2(EMBB_BASE_BASIC_TYPE_SIZE_, EMBB_PARAMETER_SIZE_BYTE) volatile* pointer_to_value, EMBB_CAT2(EMBB_BASE_BASIC_TYPE_SIZE_, EMBB_PARAMETER_SIZE_BYTE) new_value)\
-  { \
-  EMBB_CAT2(EMBB_BASE_BASIC_TYPE_SIZE_, EMBB_PARAMETER_SIZE_BYTE) \
-  result=0; \
-  __asm__ __volatile__("dsync\n\t" \
-  "ld" EMBB_ATOMIC_TC_SIZE_SUFFIX "%[result], %[pointer_to_value]" \
-  "st" EMBB_ATOMIC_TC_SIZE_SUFFIX "%[pointer_to_value], %[new_value]" \
-  : [result]"+d"(result), [pointer_to_value]"+m" (*pointer_to_value) \
-  : [new_value]"d" (new_value) \
+  EMBB_CAT2(EMBB_BASE_BASIC_TYPE_SIZE_, EMBB_PARAMETER_SIZE_BYTE) volatile* pointer_to_value, EMBB_CAT2(EMBB_BASE_BASIC_TYPE_SIZE_, EMBB_PARAMETER_SIZE_BYTE) new_value) { \
+  __asm__ __volatile__("swap" EMBB_ATOMIC_TC_SIZE_SUFFIX " [%0]0, %[new_value]"  \
+  : "+a" (pointer_to_value), [new_value]"+d" (new_value) \
+  :  \
   : "memory"); \
-  return result; \
+  return new_value; \
   }
 
 #else
diff --git a/base_c/include/embb/base/c/internal/atomic/xor_assign.h b/base_c/include/embb/base/c/internal/atomic/xor_assign.h
index 04c14d8..0859854 100644
--- a/base_c/include/embb/base/c/internal/atomic/xor_assign.h
+++ b/base_c/include/embb/base/c/internal/atomic/xor_assign.h
@@ -119,15 +119,21 @@ EMBB_DEFINE_XOR_ASSIGN(4, "")
   EMBB_PLATFORM_INLINE void EMBB_CAT2(embb_internal__atomic_xor_assign_, EMBB_PARAMETER_SIZE_BYTE)(\
   EMBB_CAT2(EMBB_BASE_BASIC_TYPE_SIZE_, EMBB_PARAMETER_SIZE_BYTE) volatile* pointer_to_value, \
   EMBB_CAT2(EMBB_BASE_BASIC_TYPE_SIZE_, EMBB_PARAMETER_SIZE_BYTE) value) { \
-  EMBB_CAT2(EMBB_BASE_BASIC_TYPE_SIZE_, EMBB_PARAMETER_SIZE_BYTE) result=0; \
-  __asm__ __volatile__( \
-  "dsync\n\t" \
-  "ld" EMBB_ATOMIC_TC_SIZE_SUFFIX " %[result], %[pointer_to_value]\n\t" \
-  "xor %[result], %[value]\n\t" \
-  "st" EMBB_ATOMIC_TC_SIZE_SUFFIX " %[pointer_to_value], %[result]\n\t" \
-  : [value]"+d" (value), [result]"+d" (result) \
-  : [pointer_to_value]"m" (*pointer_to_value) \
-  : "memory"); \
+  	register uint64_t e_reg = sizeof(value) == sizeof(uint32_t) ? (uint64_t)value : 0ull; \
+	register uint32_t add = (uint32_t)pointer_to_value & 0xFFFFFFFC; \
+	uint32_t pos = ((uint32_t)pointer_to_value & 0x3) * 8; \
+	uint32_t width = sizeof(value) == sizeof(uint32_t) ? 31 : sizeof(value) * 8; \
+	register uint64_t e_ins = ((uint64_t)width << 32) | pos; \
+	  __asm__ __volatile__ ("dsync\n\t" \
+			  "mov.a %[pointer_to_value], %[add]\n\t" \
+			  "loop_:\tld.w %H0, [%2]0\n\t" \
+			  "insert %L0, %L0, %[value], %A1\n\t" \
+			  "xor %L0, %H0\n\t" \
+			  "cmpswap.w [%2]0, %A0\n\t" \
+			  "jne %H0, %L0, loop_" \
+			  : "+d"(e_reg), [e_ins]"+d"(e_ins) \
+			  : [pointer_to_value]"a"(pointer_to_value), [value] "r" (value), [add] "d" (add) \
+			  : "memory"); \
   }
 
 #else
@@ -136,7 +142,7 @@ EMBB_DEFINE_XOR_ASSIGN(4, "")
 
 EMBB_DEFINE_XOR_ASSIGN(1, ".b")
 EMBB_DEFINE_XOR_ASSIGN(2, ".h")
-EMBB_DEFINE_XOR_ASSIGN(4, ".w")
+EMBB_DEFINE_XOR_ASSIGN(4, "")
 
 #else
 #error "Unknown architecture"