Changeset 1661

Timestamp:

Nov 5, 2011, 12:32:19 PM (8 years ago)

Author:

ksherdy

Message:

Added mvmd srli slli builtins and strategies.

Location:

trunk

Files:

• lib/idisa_cpp/idisa_avx.cpp (modified) (7 diffs)
• lib/idisa_cpp/idisa_sse2.cpp (modified) (14 diffs)
• lib/idisa_cpp/idisa_sse3.cpp (modified) (8 diffs)
• lib/idisa_cpp/idisa_sse4_1.cpp (modified) (8 diffs)
• lib/idisa_cpp/idisa_sse4_2.cpp (modified) (8 diffs)
• lib/idisa_cpp/idisa_ssse3.cpp (modified) (8 diffs)
• libgen/Library_Generator/SSE2Instructions.py (modified) (1 diff)
• libgen/Library_Generator/StrategyPool.py (modified) (1 diff)

trunk/lib/idisa_cpp/idisa_avx.cpp

@@ -101 +101 @@
 IDISA_ALWAYS_INLINE bitblock256_t simd_andc(bitblock256_t arg1, bitblock256_t arg2);
 IDISA_ALWAYS_INLINE bitblock256_t simd_or(bitblock256_t arg1, bitblock256_t arg2);
+IDISA_ALWAYS_INLINE bitblock256_t simd_xor(bitblock256_t arg1, bitblock256_t arg2);
 IDISA_ALWAYS_INLINE bitblock256_t simd_and(bitblock256_t arg1, bitblock256_t arg2);
-IDISA_ALWAYS_INLINE bitblock256_t simd_xor(bitblock256_t arg1, bitblock256_t arg2);
 template <> IDISA_ALWAYS_INLINE bitblock256_t simd256<1>::max(bitblock256_t arg1, bitblock256_t arg2);
 template <> IDISA_ALWAYS_INLINE bitblock256_t simd256<2>::max(bitblock256_t arg1, bitblock256_t arg2);
@@ -562 +562 @@
 
 //The total number of operations is 1
+IDISA_ALWAYS_INLINE bitblock256_t simd_xor(bitblock256_t arg1, bitblock256_t arg2)
+{
+        return _mm256_xor_ps(arg1, arg2);
+}
+
+//The total number of operations is 1
 IDISA_ALWAYS_INLINE bitblock256_t simd_and(bitblock256_t arg1, bitblock256_t arg2)
 {
         return _mm256_and_ps(arg1, arg2);
-}
-
-//The total number of operations is 1
-IDISA_ALWAYS_INLINE bitblock256_t simd_xor(bitblock256_t arg1, bitblock256_t arg2)
-{
-        return _mm256_xor_ps(arg1, arg2);
 }
 
@@ -3097 +3097 @@
 template <> template <uint64_t sh> IDISA_ALWAYS_INLINE bitblock256_t mvmd256<4>::srli(bitblock256_t arg1)
 {
-        return simd256<256>::srli<(sh*4)>(arg1);
+        return mvmd256<(2)>::srli<(sh*2)>(arg1);
 }
 
@@ -3109 +3109 @@
 template <> template <uint64_t sh> IDISA_ALWAYS_INLINE bitblock256_t mvmd256<16>::srli(bitblock256_t arg1)
 {
-        return simd256<256>::srli<(sh*16)>(arg1);
+        return mvmd256<(8)>::srli<(sh*2)>(arg1);
 }
 
@@ -3127 +3127 @@
 template <> template <uint64_t sh> IDISA_ALWAYS_INLINE bitblock256_t mvmd256<128>::srli(bitblock256_t arg1)
 {
-        return simd256<256>::srli<(sh*128)>(arg1);
+        return mvmd256<(64)>::srli<(sh*2)>(arg1);
 }
 
@@ -3241 +3241 @@
 template <> template <uint64_t sh> IDISA_ALWAYS_INLINE bitblock256_t mvmd256<16>::slli(bitblock256_t arg1)
 {
-        return simd256<256>::slli<(sh*16)>(arg1);
+        return mvmd256<(8)>::slli<(sh*2)>(arg1);
 }
 
@@ -3259 +3259 @@
 template <> template <uint64_t sh> IDISA_ALWAYS_INLINE bitblock256_t mvmd256<128>::slli(bitblock256_t arg1)
 {
-        return simd256<256>::slli<(sh*128)>(arg1);
+        return mvmd256<(64)>::slli<(sh*2)>(arg1);
 }
 

trunk/lib/idisa_cpp/idisa_sse2.cpp

@@ -106 +106 @@
 IDISA_ALWAYS_INLINE bitblock128_t simd_andc(bitblock128_t arg1, bitblock128_t arg2);
 IDISA_ALWAYS_INLINE bitblock128_t simd_or(bitblock128_t arg1, bitblock128_t arg2);
+IDISA_ALWAYS_INLINE bitblock128_t simd_xor(bitblock128_t arg1, bitblock128_t arg2);
 IDISA_ALWAYS_INLINE bitblock128_t simd_and(bitblock128_t arg1, bitblock128_t arg2);
-IDISA_ALWAYS_INLINE bitblock128_t simd_xor(bitblock128_t arg1, bitblock128_t arg2);
 template <> IDISA_ALWAYS_INLINE bitblock128_t simd128<1>::max(bitblock128_t arg1, bitblock128_t arg2);
 template <> IDISA_ALWAYS_INLINE bitblock128_t simd128<2>::max(bitblock128_t arg1, bitblock128_t arg2);
@@ -519 +519 @@
 
 //The total number of operations is 1
+IDISA_ALWAYS_INLINE bitblock128_t simd_xor(bitblock128_t arg1, bitblock128_t arg2)
+{
+        return _mm_xor_si128(arg1, arg2);
+}
+
+//The total number of operations is 1
 IDISA_ALWAYS_INLINE bitblock128_t simd_and(bitblock128_t arg1, bitblock128_t arg2)
 {
         return _mm_and_si128(arg1, arg2);
-}
-
-//The total number of operations is 1
-IDISA_ALWAYS_INLINE bitblock128_t simd_xor(bitblock128_t arg1, bitblock128_t arg2)
-{
-        return _mm_xor_si128(arg1, arg2);
 }
 
@@ -2547 +2547 @@
 }
 
-//The total number of operations is 9
+//The total number of operations is 3
 template <> template <uint64_t sh> IDISA_ALWAYS_INLINE bitblock128_t mvmd128<8>::dsrli(bitblock128_t arg1, bitblock128_t arg2)
 {
@@ -2553 +2553 @@
 }
 
-//The total number of operations is 9
+//The total number of operations is 3
 template <> template <uint64_t sh> IDISA_ALWAYS_INLINE bitblock128_t mvmd128<16>::dsrli(bitblock128_t arg1, bitblock128_t arg2)
 {
@@ -2559 +2559 @@
 }
 
-//The total number of operations is 9
+//The total number of operations is 3
 template <> template <uint64_t sh> IDISA_ALWAYS_INLINE bitblock128_t mvmd128<32>::dsrli(bitblock128_t arg1, bitblock128_t arg2)
 {
@@ -2565 +2565 @@
 }
 
-//The total number of operations is 9
+//The total number of operations is 3
 template <> template <uint64_t sh> IDISA_ALWAYS_INLINE bitblock128_t mvmd128<64>::dsrli(bitblock128_t arg1, bitblock128_t arg2)
 {
@@ -2571 +2571 @@
 }
 
-//The total number of operations is 9
+//The total number of operations is 3
 template <> template <uint64_t sh> IDISA_ALWAYS_INLINE bitblock128_t mvmd128<128>::dsrli(bitblock128_t arg1, bitblock128_t arg2)
 {
@@ -2790 +2790 @@
 template <> template <uint64_t sh> IDISA_ALWAYS_INLINE bitblock128_t mvmd128<4>::srli(bitblock128_t arg1)
 {
-        return simd128<128>::srli<(sh*4)>(arg1);
-}
-
-//The total number of operations is 4
+        return mvmd128<(2)>::srli<(sh*2)>(arg1);
+}
+
+//The total number of operations is 1
 template <> template <uint64_t sh> IDISA_ALWAYS_INLINE bitblock128_t mvmd128<8>::srli(bitblock128_t arg1)
 {
-        return simd128<128>::srli<(sh*8)>(arg1);
-}
-
-//The total number of operations is 4
+        return _mm_srli_si128(arg1, (int32_t)(sh));
+}
+
+//The total number of operations is 1
 template <> template <uint64_t sh> IDISA_ALWAYS_INLINE bitblock128_t mvmd128<16>::srli(bitblock128_t arg1)
 {
-        return simd128<128>::srli<(sh*16)>(arg1);
-}
-
-//The total number of operations is 4
+        return mvmd128<(8)>::srli<(sh*2)>(arg1);
+}
+
+//The total number of operations is 1
 template <> template <uint64_t sh> IDISA_ALWAYS_INLINE bitblock128_t mvmd128<32>::srli(bitblock128_t arg1)
 {
-        return simd128<128>::srli<(sh*32)>(arg1);
-}
-
-//The total number of operations is 4
+        return mvmd128<(16)>::srli<(sh*2)>(arg1);
+}
+
+//The total number of operations is 1
 template <> template <uint64_t sh> IDISA_ALWAYS_INLINE bitblock128_t mvmd128<64>::srli(bitblock128_t arg1)
 {
-        return simd128<128>::srli<(sh*64)>(arg1);
-}
-
-//The total number of operations is 4
+        return mvmd128<(32)>::srli<(sh*2)>(arg1);
+}
+
+//The total number of operations is 1
 template <> template <uint64_t sh> IDISA_ALWAYS_INLINE bitblock128_t mvmd128<128>::srli(bitblock128_t arg1)
 {
-        return simd128<128>::srli<(sh*128)>(arg1);
+        return mvmd128<(64)>::srli<(sh*2)>(arg1);
 }
 
@@ -2877 +2877 @@
 }
 
-//The total number of operations is 9
+//The total number of operations is 3
 template <> template <uint64_t sh> IDISA_ALWAYS_INLINE bitblock128_t mvmd128<8>::dslli(bitblock128_t arg1, bitblock128_t arg2)
 {
@@ -2883 +2883 @@
 }
 
-//The total number of operations is 9
+//The total number of operations is 3
 template <> template <uint64_t sh> IDISA_ALWAYS_INLINE bitblock128_t mvmd128<16>::dslli(bitblock128_t arg1, bitblock128_t arg2)
 {
@@ -2889 +2889 @@
 }
 
-//The total number of operations is 9
+//The total number of operations is 3
 template <> template <uint64_t sh> IDISA_ALWAYS_INLINE bitblock128_t mvmd128<32>::dslli(bitblock128_t arg1, bitblock128_t arg2)
 {
@@ -2895 +2895 @@
 }
 
-//The total number of operations is 9
+//The total number of operations is 3
 template <> template <uint64_t sh> IDISA_ALWAYS_INLINE bitblock128_t mvmd128<64>::dslli(bitblock128_t arg1, bitblock128_t arg2)
 {
@@ -2901 +2901 @@
 }
 
-//The total number of operations is 9
+//The total number of operations is 3
 template <> template <uint64_t sh> IDISA_ALWAYS_INLINE bitblock128_t mvmd128<128>::dslli(bitblock128_t arg1, bitblock128_t arg2)
 {
@@ -2916 +2916 @@
 template <> template <uint64_t sh> IDISA_ALWAYS_INLINE bitblock128_t mvmd128<4>::slli(bitblock128_t arg1)
 {
-        return simd128<128>::slli<(sh*4)>(arg1);
-}
-
-//The total number of operations is 4
+        return mvmd128<(2)>::slli<(sh*2)>(arg1);
+}
+
+//The total number of operations is 1
 template <> template <uint64_t sh> IDISA_ALWAYS_INLINE bitblock128_t mvmd128<8>::slli(bitblock128_t arg1)
 {
-        return simd128<128>::slli<(sh*8)>(arg1);
-}
-
-//The total number of operations is 4
+        return _mm_slli_si128(arg1, (int32_t)(sh));
+}
+
+//The total number of operations is 1
 template <> template <uint64_t sh> IDISA_ALWAYS_INLINE bitblock128_t mvmd128<16>::slli(bitblock128_t arg1)
 {
-        return simd128<128>::slli<(sh*16)>(arg1);
-}
-
-//The total number of operations is 4
+        return mvmd128<(8)>::slli<(sh*2)>(arg1);
+}
+
+//The total number of operations is 1
 template <> template <uint64_t sh> IDISA_ALWAYS_INLINE bitblock128_t mvmd128<32>::slli(bitblock128_t arg1)
 {
-        return simd128<128>::slli<(sh*32)>(arg1);
-}
-
-//The total number of operations is 4
+        return mvmd128<(16)>::slli<(sh*2)>(arg1);
+}
+
+//The total number of operations is 1
 template <> template <uint64_t sh> IDISA_ALWAYS_INLINE bitblock128_t mvmd128<64>::slli(bitblock128_t arg1)
 {
-        return simd128<128>::slli<(sh*64)>(arg1);
-}
-
-//The total number of operations is 4
+        return mvmd128<(32)>::slli<(sh*2)>(arg1);
+}
+
+//The total number of operations is 1
 template <> template <uint64_t sh> IDISA_ALWAYS_INLINE bitblock128_t mvmd128<128>::slli(bitblock128_t arg1)
 {
-        return simd128<128>::slli<(sh*128)>(arg1);
+        return mvmd128<(64)>::slli<(sh*2)>(arg1);
 }
 

trunk/lib/idisa_cpp/idisa_sse3.cpp

@@ -106 +106 @@
 IDISA_ALWAYS_INLINE bitblock128_t simd_andc(bitblock128_t arg1, bitblock128_t arg2);
 IDISA_ALWAYS_INLINE bitblock128_t simd_or(bitblock128_t arg1, bitblock128_t arg2);
+IDISA_ALWAYS_INLINE bitblock128_t simd_xor(bitblock128_t arg1, bitblock128_t arg2);
 IDISA_ALWAYS_INLINE bitblock128_t simd_and(bitblock128_t arg1, bitblock128_t arg2);
-IDISA_ALWAYS_INLINE bitblock128_t simd_xor(bitblock128_t arg1, bitblock128_t arg2);
 template <> IDISA_ALWAYS_INLINE bitblock128_t simd128<1>::max(bitblock128_t arg1, bitblock128_t arg2);
 template <> IDISA_ALWAYS_INLINE bitblock128_t simd128<2>::max(bitblock128_t arg1, bitblock128_t arg2);
@@ -519 +519 @@
 
 //The total number of operations is 1
+IDISA_ALWAYS_INLINE bitblock128_t simd_xor(bitblock128_t arg1, bitblock128_t arg2)
+{
+        return _mm_xor_si128(arg1, arg2);
+}
+
+//The total number of operations is 1
 IDISA_ALWAYS_INLINE bitblock128_t simd_and(bitblock128_t arg1, bitblock128_t arg2)
 {
         return _mm_and_si128(arg1, arg2);
-}
-
-//The total number of operations is 1
-IDISA_ALWAYS_INLINE bitblock128_t simd_xor(bitblock128_t arg1, bitblock128_t arg2)
-{
-        return _mm_xor_si128(arg1, arg2);
 }
 
@@ -2790 +2790 @@
 template <> template <uint64_t sh> IDISA_ALWAYS_INLINE bitblock128_t mvmd128<4>::srli(bitblock128_t arg1)
 {
-        return simd128<128>::srli<(sh*4)>(arg1);
+        return mvmd128<(2)>::srli<(sh*2)>(arg1);
 }
 
@@ -2808 +2808 @@
 template <> template <uint64_t sh> IDISA_ALWAYS_INLINE bitblock128_t mvmd128<32>::srli(bitblock128_t arg1)
 {
-        return simd128<128>::srli<(sh*32)>(arg1);
+        return mvmd128<(16)>::srli<(sh*2)>(arg1);
 }
 
@@ -2820 +2820 @@
 template <> template <uint64_t sh> IDISA_ALWAYS_INLINE bitblock128_t mvmd128<128>::srli(bitblock128_t arg1)
 {
-        return simd128<128>::srli<(sh*128)>(arg1);
+        return mvmd128<(64)>::srli<(sh*2)>(arg1);
 }
 
@@ -2916 +2916 @@
 template <> template <uint64_t sh> IDISA_ALWAYS_INLINE bitblock128_t mvmd128<4>::slli(bitblock128_t arg1)
 {
-        return simd128<128>::slli<(sh*4)>(arg1);
+        return mvmd128<(2)>::slli<(sh*2)>(arg1);
 }
 
@@ -2928 +2928 @@
 template <> template <uint64_t sh> IDISA_ALWAYS_INLINE bitblock128_t mvmd128<16>::slli(bitblock128_t arg1)
 {
-        return simd128<128>::slli<(sh*16)>(arg1);
+        return mvmd128<(8)>::slli<(sh*2)>(arg1);
 }
 
@@ -2940 +2940 @@
 template <> template <uint64_t sh> IDISA_ALWAYS_INLINE bitblock128_t mvmd128<64>::slli(bitblock128_t arg1)
 {
-        return simd128<128>::slli<(sh*64)>(arg1);
+        return mvmd128<(32)>::slli<(sh*2)>(arg1);
 }
 

trunk/lib/idisa_cpp/idisa_sse4_1.cpp

@@ -107 +107 @@
 IDISA_ALWAYS_INLINE bitblock128_t simd_andc(bitblock128_t arg1, bitblock128_t arg2);
 IDISA_ALWAYS_INLINE bitblock128_t simd_or(bitblock128_t arg1, bitblock128_t arg2);
+IDISA_ALWAYS_INLINE bitblock128_t simd_xor(bitblock128_t arg1, bitblock128_t arg2);
 IDISA_ALWAYS_INLINE bitblock128_t simd_and(bitblock128_t arg1, bitblock128_t arg2);
-IDISA_ALWAYS_INLINE bitblock128_t simd_xor(bitblock128_t arg1, bitblock128_t arg2);
 template <> IDISA_ALWAYS_INLINE bitblock128_t simd128<1>::max(bitblock128_t arg1, bitblock128_t arg2);
 template <> IDISA_ALWAYS_INLINE bitblock128_t simd128<2>::max(bitblock128_t arg1, bitblock128_t arg2);
@@ -524 +524 @@
 
 //The total number of operations is 1
+IDISA_ALWAYS_INLINE bitblock128_t simd_xor(bitblock128_t arg1, bitblock128_t arg2)
+{
+        return _mm_xor_si128(arg1, arg2);
+}
+
+//The total number of operations is 1
 IDISA_ALWAYS_INLINE bitblock128_t simd_and(bitblock128_t arg1, bitblock128_t arg2)
 {
         return _mm_and_si128(arg1, arg2);
-}
-
-//The total number of operations is 1
-IDISA_ALWAYS_INLINE bitblock128_t simd_xor(bitblock128_t arg1, bitblock128_t arg2)
-{
-        return _mm_xor_si128(arg1, arg2);
 }
 
@@ -2822 +2822 @@
 template <> template <uint64_t sh> IDISA_ALWAYS_INLINE bitblock128_t mvmd128<4>::srli(bitblock128_t arg1)
 {
-        return simd128<128>::srli<(sh*4)>(arg1);
+        return mvmd128<(2)>::srli<(sh*2)>(arg1);
 }
 
@@ -2840 +2840 @@
 template <> template <uint64_t sh> IDISA_ALWAYS_INLINE bitblock128_t mvmd128<32>::srli(bitblock128_t arg1)
 {
-        return simd128<128>::srli<(sh*32)>(arg1);
+        return mvmd128<(16)>::srli<(sh*2)>(arg1);
 }
 
@@ -2852 +2852 @@
 template <> template <uint64_t sh> IDISA_ALWAYS_INLINE bitblock128_t mvmd128<128>::srli(bitblock128_t arg1)
 {
-        return simd128<128>::srli<(sh*128)>(arg1);
+        return mvmd128<(64)>::srli<(sh*2)>(arg1);
 }
 
@@ -2948 +2948 @@
 template <> template <uint64_t sh> IDISA_ALWAYS_INLINE bitblock128_t mvmd128<4>::slli(bitblock128_t arg1)
 {
-        return simd128<128>::slli<(sh*4)>(arg1);
+        return mvmd128<(2)>::slli<(sh*2)>(arg1);
 }
 
@@ -2960 +2960 @@
 template <> template <uint64_t sh> IDISA_ALWAYS_INLINE bitblock128_t mvmd128<16>::slli(bitblock128_t arg1)
 {
-        return simd128<128>::slli<(sh*16)>(arg1);
+        return mvmd128<(8)>::slli<(sh*2)>(arg1);
 }
 
@@ -2972 +2972 @@
 template <> template <uint64_t sh> IDISA_ALWAYS_INLINE bitblock128_t mvmd128<64>::slli(bitblock128_t arg1)
 {
-        return simd128<128>::slli<(sh*64)>(arg1);
+        return mvmd128<(32)>::slli<(sh*2)>(arg1);
 }
 

trunk/lib/idisa_cpp/idisa_sse4_2.cpp

@@ -107 +107 @@
 IDISA_ALWAYS_INLINE bitblock128_t simd_andc(bitblock128_t arg1, bitblock128_t arg2);
 IDISA_ALWAYS_INLINE bitblock128_t simd_or(bitblock128_t arg1, bitblock128_t arg2);
+IDISA_ALWAYS_INLINE bitblock128_t simd_xor(bitblock128_t arg1, bitblock128_t arg2);
 IDISA_ALWAYS_INLINE bitblock128_t simd_and(bitblock128_t arg1, bitblock128_t arg2);
-IDISA_ALWAYS_INLINE bitblock128_t simd_xor(bitblock128_t arg1, bitblock128_t arg2);
 template <> IDISA_ALWAYS_INLINE bitblock128_t simd128<1>::max(bitblock128_t arg1, bitblock128_t arg2);
 template <> IDISA_ALWAYS_INLINE bitblock128_t simd128<2>::max(bitblock128_t arg1, bitblock128_t arg2);
@@ -524 +524 @@
 
 //The total number of operations is 1
+IDISA_ALWAYS_INLINE bitblock128_t simd_xor(bitblock128_t arg1, bitblock128_t arg2)
+{
+        return _mm_xor_si128(arg1, arg2);
+}
+
+//The total number of operations is 1
 IDISA_ALWAYS_INLINE bitblock128_t simd_and(bitblock128_t arg1, bitblock128_t arg2)
 {
         return _mm_and_si128(arg1, arg2);
-}
-
-//The total number of operations is 1
-IDISA_ALWAYS_INLINE bitblock128_t simd_xor(bitblock128_t arg1, bitblock128_t arg2)
-{
-        return _mm_xor_si128(arg1, arg2);
 }
 
@@ -2804 +2804 @@
 template <> template <uint64_t sh> IDISA_ALWAYS_INLINE bitblock128_t mvmd128<4>::srli(bitblock128_t arg1)
 {
-        return simd128<128>::srli<(sh*4)>(arg1);
+        return mvmd128<(2)>::srli<(sh*2)>(arg1);
 }
 
@@ -2822 +2822 @@
 template <> template <uint64_t sh> IDISA_ALWAYS_INLINE bitblock128_t mvmd128<32>::srli(bitblock128_t arg1)
 {
-        return simd128<128>::srli<(sh*32)>(arg1);
+        return mvmd128<(16)>::srli<(sh*2)>(arg1);
 }
 
@@ -2834 +2834 @@
 template <> template <uint64_t sh> IDISA_ALWAYS_INLINE bitblock128_t mvmd128<128>::srli(bitblock128_t arg1)
 {
-        return simd128<128>::srli<(sh*128)>(arg1);
+        return mvmd128<(64)>::srli<(sh*2)>(arg1);
 }
 
@@ -2930 +2930 @@
 template <> template <uint64_t sh> IDISA_ALWAYS_INLINE bitblock128_t mvmd128<4>::slli(bitblock128_t arg1)
 {
-        return simd128<128>::slli<(sh*4)>(arg1);
+        return mvmd128<(2)>::slli<(sh*2)>(arg1);
 }
 
@@ -2942 +2942 @@
 template <> template <uint64_t sh> IDISA_ALWAYS_INLINE bitblock128_t mvmd128<16>::slli(bitblock128_t arg1)
 {
-        return simd128<128>::slli<(sh*16)>(arg1);
+        return mvmd128<(8)>::slli<(sh*2)>(arg1);
 }
 
@@ -2954 +2954 @@
 template <> template <uint64_t sh> IDISA_ALWAYS_INLINE bitblock128_t mvmd128<64>::slli(bitblock128_t arg1)
 {
-        return simd128<128>::slli<(sh*64)>(arg1);
+        return mvmd128<(32)>::slli<(sh*2)>(arg1);
 }
 

trunk/lib/idisa_cpp/idisa_ssse3.cpp

@@ -107 +107 @@
 IDISA_ALWAYS_INLINE bitblock128_t simd_andc(bitblock128_t arg1, bitblock128_t arg2);
 IDISA_ALWAYS_INLINE bitblock128_t simd_or(bitblock128_t arg1, bitblock128_t arg2);
+IDISA_ALWAYS_INLINE bitblock128_t simd_xor(bitblock128_t arg1, bitblock128_t arg2);
 IDISA_ALWAYS_INLINE bitblock128_t simd_and(bitblock128_t arg1, bitblock128_t arg2);
-IDISA_ALWAYS_INLINE bitblock128_t simd_xor(bitblock128_t arg1, bitblock128_t arg2);
 template <> IDISA_ALWAYS_INLINE bitblock128_t simd128<1>::max(bitblock128_t arg1, bitblock128_t arg2);
 template <> IDISA_ALWAYS_INLINE bitblock128_t simd128<2>::max(bitblock128_t arg1, bitblock128_t arg2);
@@ -524 +524 @@
 
 //The total number of operations is 1
+IDISA_ALWAYS_INLINE bitblock128_t simd_xor(bitblock128_t arg1, bitblock128_t arg2)
+{
+        return _mm_xor_si128(arg1, arg2);
+}
+
+//The total number of operations is 1
 IDISA_ALWAYS_INLINE bitblock128_t simd_and(bitblock128_t arg1, bitblock128_t arg2)
 {
         return _mm_and_si128(arg1, arg2);
-}
-
-//The total number of operations is 1
-IDISA_ALWAYS_INLINE bitblock128_t simd_xor(bitblock128_t arg1, bitblock128_t arg2)
-{
-        return _mm_xor_si128(arg1, arg2);
 }
 
@@ -2825 +2825 @@
 template <> template <uint64_t sh> IDISA_ALWAYS_INLINE bitblock128_t mvmd128<4>::srli(bitblock128_t arg1)
 {
-        return simd128<128>::srli<(sh*4)>(arg1);
+        return mvmd128<(2)>::srli<(sh*2)>(arg1);
 }
 
@@ -2843 +2843 @@
 template <> template <uint64_t sh> IDISA_ALWAYS_INLINE bitblock128_t mvmd128<32>::srli(bitblock128_t arg1)
 {
-        return simd128<128>::srli<(sh*32)>(arg1);
+        return mvmd128<(16)>::srli<(sh*2)>(arg1);
 }
 
@@ -2855 +2855 @@
 template <> template <uint64_t sh> IDISA_ALWAYS_INLINE bitblock128_t mvmd128<128>::srli(bitblock128_t arg1)
 {
-        return simd128<128>::srli<(sh*128)>(arg1);
+        return mvmd128<(64)>::srli<(sh*2)>(arg1);
 }
 
@@ -2951 +2951 @@
 template <> template <uint64_t sh> IDISA_ALWAYS_INLINE bitblock128_t mvmd128<4>::slli(bitblock128_t arg1)
 {
-        return simd128<128>::slli<(sh*4)>(arg1);
+        return mvmd128<(2)>::slli<(sh*2)>(arg1);
 }
 
@@ -2963 +2963 @@
 template <> template <uint64_t sh> IDISA_ALWAYS_INLINE bitblock128_t mvmd128<16>::slli(bitblock128_t arg1)
 {
-        return simd128<128>::slli<(sh*16)>(arg1);
+        return mvmd128<(8)>::slli<(sh*2)>(arg1);
 }
 
@@ -2975 +2975 @@
 template <> template <uint64_t sh> IDISA_ALWAYS_INLINE bitblock128_t mvmd128<64>::slli(bitblock128_t arg1)
 {
-        return simd128<128>::slli<(sh*64)>(arg1);
+        return mvmd128<(32)>::slli<(sh*2)>(arg1);
 }
 

trunk/libgen/Library_Generator/SSE2Instructions.py

@@ -164 +164 @@
                 "fws":[8, 16, 32, 64],
         },
+        "mvmd_srli":\
+        {
+                "signature":"SIMD_type _mm_srli_si128(SIMD_type arg1, int sh)",
+                "args_type":{"arg1":"SIMD_type", "sh":"signed_int(32)"},
+                "return_type":"SIMD_type",
+                "fws":[8],
+        },
+        "mvmd_slli":\
+        {
+                "signature":"SIMD_type _mm_slli_si128(SIMD_type arg1, int sh)",
+                "args_type":{"arg1":"SIMD_type", "sh":"signed_int(32)"},
+                "return_type":"SIMD_type",
+                "fws":[8],
+        },      
         "mvmd_fill":\
         {

trunk/libgen/Library_Generator/StrategyPool.py

@@ -1796 +1796 @@
                 "Platforms":[configure.ALL],
                 },
-                
+
+                "mvmd_srli_halving":\
+                {
+                "body":r'''
+return (mvmd_srli(fw/2, sh*2, arg1))''',
+                "Ops":["mvmd_srli"],
+                "Fws":[-1],
+                "Platforms":[configure.ALL],
+                },
+
+                "mvmd_slli_halving":\
+                {
+                "body":r'''
+return (mvmd_slli(fw/2, sh*2, arg1))''',
+                "Ops":["mvmd_slli"],
+                "Fws":[-1],
+                "Platforms":[configure.ALL],
+                }, 
+                        
                 "simd_add_2_logic":\
                 {