diff --git a/OpenCL/inc_common.cl b/OpenCL/inc_common.cl
index 09946e52f..64300e80a 100644
--- a/OpenCL/inc_common.cl
+++ b/OpenCL/inc_common.cl
@@ -854,17 +854,11 @@ DECLSPEC u32x hc_byte_perm (const u32x a, const u32x b, const int c)
   #endif
 
   #if VECT_SIZE >= 4
-  __asm__ __volatile__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.s0) : "v"(b.s0), "v"(a.s0), "v"(c));
-  __asm__ __volatile__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.s1) : "v"(b.s1), "v"(a.s1), "v"(c));
   __asm__ __volatile__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.s2) : "v"(b.s2), "v"(a.s2), "v"(c));
   __asm__ __volatile__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.s3) : "v"(b.s3), "v"(a.s3), "v"(c));
   #endif
 
   #if VECT_SIZE >= 8
-  __asm__ __volatile__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.s0) : "v"(b.s0), "v"(a.s0), "v"(c));
-  __asm__ __volatile__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.s1) : "v"(b.s1), "v"(a.s1), "v"(c));
-  __asm__ __volatile__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.s2) : "v"(b.s2), "v"(a.s2), "v"(c));
-  __asm__ __volatile__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.s3) : "v"(b.s3), "v"(a.s3), "v"(c));
   __asm__ __volatile__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.s4) : "v"(b.s4), "v"(a.s4), "v"(c));
   __asm__ __volatile__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.s5) : "v"(b.s5), "v"(a.s5), "v"(c));
   __asm__ __volatile__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.s6) : "v"(b.s6), "v"(a.s6), "v"(c));
@@ -872,14 +866,6 @@ DECLSPEC u32x hc_byte_perm (const u32x a, const u32x b, const int c)
   #endif
 
   #if VECT_SIZE >= 16
-  __asm__ __volatile__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.s0) : "v"(b.s0), "v"(a.s0), "v"(c));
-  __asm__ __volatile__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.s1) : "v"(b.s1), "v"(a.s1), "v"(c));
-  __asm__ __volatile__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.s2) : "v"(b.s2), "v"(a.s2), "v"(c));
-  __asm__ __volatile__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.s3) : "v"(b.s3), "v"(a.s3), "v"(c));
-  __asm__ __volatile__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.s4) : "v"(b.s4), "v"(a.s4), "v"(c));
-  __asm__ __volatile__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.s5) : "v"(b.s5), "v"(a.s5), "v"(c));
-  __asm__ __volatile__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.s6) : "v"(b.s6), "v"(a.s6), "v"(c));
-  __asm__ __volatile__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.s7) : "v"(b.s7), "v"(a.s7), "v"(c));
   __asm__ __volatile__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.s8) : "v"(b.s8), "v"(a.s8), "v"(c));
   __asm__ __volatile__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.s9) : "v"(b.s9), "v"(a.s9), "v"(c));
   __asm__ __volatile__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.sa) : "v"(b.sa), "v"(a.sa), "v"(c));
@@ -919,17 +905,11 @@ DECLSPEC u32x hc_add3 (const u32x a, const u32x b, const u32x c)
   #endif
 
   #if VECT_SIZE >= 4
-  __asm__ __volatile__ ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.s0) : "v"(b.s0), "v"(a.s0), "v"(c.s0));
-  __asm__ __volatile__ ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.s1) : "v"(b.s1), "v"(a.s1), "v"(c.s1));
   __asm__ __volatile__ ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.s2) : "v"(b.s2), "v"(a.s2), "v"(c.s2));
   __asm__ __volatile__ ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.s3) : "v"(b.s3), "v"(a.s3), "v"(c.s3));
   #endif
 
   #if VECT_SIZE >= 8
-  __asm__ __volatile__ ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.s0) : "v"(b.s0), "v"(a.s0), "v"(c.s0));
-  __asm__ __volatile__ ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.s1) : "v"(b.s1), "v"(a.s1), "v"(c.s1));
-  __asm__ __volatile__ ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.s2) : "v"(b.s2), "v"(a.s2), "v"(c.s2));
-  __asm__ __volatile__ ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.s3) : "v"(b.s3), "v"(a.s3), "v"(c.s3));
   __asm__ __volatile__ ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.s4) : "v"(b.s4), "v"(a.s4), "v"(c.s4));
   __asm__ __volatile__ ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.s5) : "v"(b.s5), "v"(a.s5), "v"(c.s5));
   __asm__ __volatile__ ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.s6) : "v"(b.s6), "v"(a.s6), "v"(c.s6));
@@ -937,14 +917,6 @@ DECLSPEC u32x hc_add3 (const u32x a, const u32x b, const u32x c)
   #endif
 
   #if VECT_SIZE >= 16
-  __asm__ __volatile__ ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.s0) : "v"(b.s0), "v"(a.s0), "v"(c.s0));
-  __asm__ __volatile__ ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.s1) : "v"(b.s1), "v"(a.s1), "v"(c.s1));
-  __asm__ __volatile__ ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.s2) : "v"(b.s2), "v"(a.s2), "v"(c.s2));
-  __asm__ __volatile__ ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.s3) : "v"(b.s3), "v"(a.s3), "v"(c.s3));
-  __asm__ __volatile__ ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.s4) : "v"(b.s4), "v"(a.s4), "v"(c.s4));
-  __asm__ __volatile__ ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.s5) : "v"(b.s5), "v"(a.s5), "v"(c.s5));
-  __asm__ __volatile__ ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.s6) : "v"(b.s6), "v"(a.s6), "v"(c.s6));
-  __asm__ __volatile__ ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.s7) : "v"(b.s7), "v"(a.s7), "v"(c.s7));
   __asm__ __volatile__ ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.s8) : "v"(b.s8), "v"(a.s8), "v"(c.s8));
   __asm__ __volatile__ ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.s9) : "v"(b.s9), "v"(a.s9), "v"(c.s9));
   __asm__ __volatile__ ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.sa) : "v"(b.sa), "v"(a.sa), "v"(c.sa));