X13 Fix

The modified fugue.cl included in the x15 changes broke x13 kernels.

kernel/fuguecoin.cl

@@ -79,6 +79,19 @@ __kernel void search(__global unsigned char* input, volatile __global uint* outp
 {
     uint gid = get_global_id(0);
 
+    //mixtab
+    __local sph_u32 mixtab0[256], mixtab1[256], mixtab2[256], mixtab3[256];
+    int init = get_local_id(0);
+    int step = get_local_size(0);
+    for (int i = init; i < 256; i += step)
+    {
+      mixtab0[i] = mixtab0_c[i];
+      mixtab1[i] = mixtab1_c[i];
+      mixtab2[i] = mixtab2_c[i];
+      mixtab3[i] = mixtab3_c[i];
+    }
+    barrier(CLK_GLOBAL_MEM_FENCE); 
+
     sph_u32 S00 = 0, S01 = 0, S02 = 0, S03 = 0, S04 = 0, S05 = 0, S06 = 0, S07 = 0, S08 = 0, S09 = 0; \
     sph_u32 S10 = 0, S11 = 0, S12 = 0, S13 = 0, S14 = 0, S15 = 0, S16 = 0, S17 = 0, S18 = 0, S19 = 0; \
     sph_u32 S20 = 0, S21 = 0, S22 = IV256[0], S23 = IV256[1], S24 = IV256[2], S25 = IV256[3], S26 = IV256[4], S27 = IV256[5], S28 = IV256[6], S29 = IV256[7];

kernel/marucoin-mod.cl

@@ -909,7 +909,8 @@ __attribute__((reqd_work_group_size(WORKSIZE, 1, 1)))
 __kernel void search12(__global hash_t* hashes, __global uint* output, const ulong target)
 {
   uint gid = get_global_id(0);
-    __global hash_t *hash = &(hashes[gid-get_global_offset(0)]);
+  uint offset = get_global_offset(0);
+  __global hash_t *hash = &(hashes[gid-offset]);
 
   //mixtab
   __local sph_u32 mixtab0[256], mixtab1[256], mixtab2[256], mixtab3[256];
@@ -922,7 +923,8 @@ __kernel void search12(__global hash_t* hashes, __global uint* output, const ulo
     mixtab2[i] = mixtab2_c[i];
     mixtab3[i] = mixtab3_c[i];
   }
-    barrier(CLK_LOCAL_MEM_FENCE);
+  barrier(CLK_GLOBAL_MEM_FENCE); 
+    
   // fugue
   sph_u32 S00, S01, S02, S03, S04, S05, S06, S07, S08, S09;
   sph_u32 S10, S11, S12, S13, S14, S15, S16, S17, S18, S19;
@@ -947,12 +949,15 @@ __kernel void search12(__global hash_t* hashes, __global uint* output, const ulo
   // apply round shift if necessary
   int i;
   
-        for (i = 0; i < 32; i ++) {
+  for (i = 0; i < 32; i ++) 
+  {
     ROR3;
     CMIX36(S00, S01, S02, S04, S05, S06, S18, S19, S20);
     SMIX(S00, S01, S02, S03);
   }
-        for (i = 0; i < 13; i ++) {
+  
+  for (i = 0; i < 13; i ++) 
+  {
     S04 ^= S00;
     S09 ^= S00;
     S18 ^= S00;
@@ -1002,7 +1007,8 @@ __kernel void search12(__global hash_t* hashes, __global uint* output, const ulo
 
   bool result = (hash->h8[3] <= target);
   if (result)
-        output[output[0xFF]++] = SWAP4(gid);
+    output[atomic_inc(output+0xFF)] = SWAP4(gid);
+
   barrier(CLK_GLOBAL_MEM_FENCE); 
 }
 

kernel/marucoin-modold.cl

@@ -895,6 +895,19 @@ __kernel void search10(__global hash_t* hashes, __global uint* output, const ulo
 
     }
 
+  //mixtab
+  __local sph_u32 mixtab0[256], mixtab1[256], mixtab2[256], mixtab3[256];
+  init = get_local_id(0);
+  step = get_local_size(0);
+  for (int i = init; i < 256; i += step)
+  {
+    mixtab0[i] = mixtab0_c[i];
+    mixtab1[i] = mixtab1_c[i];
+    mixtab2[i] = mixtab2_c[i];
+    mixtab3[i] = mixtab3_c[i];
+  }
+  barrier(CLK_GLOBAL_MEM_FENCE); 
+
   // fugue
 
     {