Do SIMD XOR, where possible!

include/galois.h

@@ -51,9 +51,8 @@ extern int galois_single_multiply(int a, int b, int w);
 extern int galois_single_divide(int a, int b, int w);
 extern int galois_inverse(int x, int w);
 
-void galois_region_xor(           char *r1,         /* Region 1 */
-                                  char *r2,         /* Region 2 */
-                                  char *r3,         /* Sum region (r3 = r1 ^ r2) -- can be r1 or r2 */
+void galois_region_xor(           char *src,         /* Source Region */
+                                  char *dest,        /* Dest Region (holds result) */
                                   int nbytes);      /* Number of bytes in region */
 
 /* These multiply regions in w=8, w=16 and w=32.  They are much faster

src/galois.c

@@ -324,29 +324,9 @@ void galois_w32_region_xor(void *src, void *dest, int nbytes)
   gfp_array[32]->multiply_region.w32(gfp_array[32], src, dest, 1, nbytes, 1);
 }
 
-void galois_region_xor(char *r1,         /* Region 1 */
-                            char *r2,         /* Region 2 */
-                            char *r3,         /* Sum region (r3 = r1 ^ r2) -- can be r1 or r2 */
-                            int nbytes)       /* Number of bytes in region */
+void galois_region_xor(char *src, char *dest, int nbytes)
 {
-  long *l1;
-  long *l2;
-  long *l3;
-  long *ltop;
-  char *ctop;
-  
-  ctop = r1 + nbytes;
-  ltop = (long *) ctop;
-  l1 = (long *) r1;
-  l2 = (long *) r2;
-  l3 = (long *) r3;
- 
-  while (l1 < ltop) {
-    *l3 = ((*l1)  ^ (*l2));
-    l1++;
-    l2++;
-    l3++;
-  }
+  galois_w32_region_xor(src, dest, nbytes);
 }
 
 int galois_inverse(int y, int w)

src/jerasure.c

@@ -341,7 +341,7 @@ void jerasure_bitmatrix_dotprod(int k, int w, int *bitmatrix_row,
               jerasure_total_memcpy_bytes += packetsize;
               pstarted = 1;
             } else {
-              galois_region_xor(pptr, dptr, pptr, packetsize);
+              galois_region_xor(dptr, pptr, packetsize);
               jerasure_total_xor_bytes += packetsize;
             }
           }
@@ -360,7 +360,7 @@ void jerasure_do_parity(int k, char **data_ptrs, char *parity_ptr, int size)
   jerasure_total_memcpy_bytes += size;
   
   for (i = 1; i < k; i++) {
-    galois_region_xor(data_ptrs[i], parity_ptr, parity_ptr, size);
+    galois_region_xor(data_ptrs[i], parity_ptr, size);
     jerasure_total_xor_bytes += size;
   }
 }
@@ -599,7 +599,7 @@ void jerasure_matrix_dotprod(int k, int w, int *matrix_row,
         jerasure_total_memcpy_bytes += size;
         init = 1;
       } else {
-        galois_region_xor(sptr, dptr, dptr, size);
+        galois_region_xor(sptr, dptr, size);
         jerasure_total_xor_bytes += size;
       }
     }
@@ -1173,7 +1173,7 @@ void jerasure_do_scheduled_operations(char **ptrs, int **operations, int packets
       operations[op][2], 
       operations[op][3]); 
       printf("xor(0x%x, 0x%x -> 0x%x, %d)\n", sptr, dptr, dptr, packetsize); */
-      galois_region_xor(sptr, dptr, dptr, packetsize);
+      galois_region_xor(sptr, dptr, packetsize);
       jerasure_total_xor_bytes += packetsize;
     } else {
 /*      printf("memcpy(0x%x <- 0x%x)\n", dptr, sptr); */

src/reed_sol.c

@@ -213,7 +213,7 @@ int reed_sol_r6_encode(int k, int w, char **data_ptrs, char **coding_ptrs, int s
 
   memcpy(coding_ptrs[0], data_ptrs[0], size);
 
-  for (i = 1; i < k; i++) galois_region_xor(coding_ptrs[0], data_ptrs[i], coding_ptrs[0], size);
+  for (i = 1; i < k; i++) galois_region_xor(data_ptrs[i], coding_ptrs[0], size);
 
   /* Next, put the sum of (2^j)*Dj into coding region 1 */
 
@@ -227,7 +227,7 @@ int reed_sol_r6_encode(int k, int w, char **data_ptrs, char **coding_ptrs, int s
       default: return 0;
     }
 
-    galois_region_xor(coding_ptrs[1], data_ptrs[i], coding_ptrs[1], size);
+    galois_region_xor(data_ptrs[i], coding_ptrs[1], size);
   }
   return 1;
 }