But also +4% is something it just shows that optimizations can still be in cooperated

Ubuntu 13.10, Catalyst 14.4, Radeon R9 290 900/1250 (downclocked)
Total 29.1 sol/s [dev0 30.2] 4 shares
Total 41.1 sol/s [dev0 42.0] 2 shares +40%

Ubuntu 16.04, NVidia 367, GeForce GTX1070
Total 196 solutions in 6588.2 ms (29.8 Sol/s)
Total 196 solutions in 5334.1 ms (36.7 Sol/s)  +20%

Ubuntu 16.04, amdgpu-pro 16.30, Radeon RX480
Total 50.4 sol/s [dev0 51.0] 4 shares
Total 53.1 sol/s [dev0 53.2] 14 shares +4%

#if NR_SLOTS <= (1 << 8 )
                    // note: this assumes slots can be encoded in 8 bits
                    collisions[nr_coll++] =
         ((ushort)j << 8 ) | ((ushort)i & 0xff);
#else
#error "unsupported NR_SLOTS"
#endif

 v_cmp_ge_u32  vcc, 53, v19                            // 00000000009C: 7D8C26B5
  s_and_saveexec_b64  s[24:25], vcc                     // 0000000000A0: BE98246A
  v_or_b32      v10, v6, v8                             // 0000000000A4: 38141106
  v_lshlrev_b32  v11, 1, v19                            // 0000000000A8: 34162681
 buffer_store_short  v10, v11, s[16:19], s14 offen glc // 0000000000AC: E0685000 0E040A0B

diff --git a/input.cl b/input.cl
index 91b7021..60a3ffe 100644
--- a/input.cl
+++ b/input.cl
@@ -525,12 +525,14 @@ void equihash_round(uint round, __global char *ht_src, __global char *ht_dst,
     uint               tlid = get_local_id(0);
     __global char       *p;
     uint                cnt;
-    uchar              first_words[NR_SLOTS];
+    __local uchar first_words_data[NR_SLOTS*64];
+    __local uchar *first_words = &first_words_data[NR_SLOTS*tlid];
     uchar              mask;
     uint                i, j;
     // NR_SLOTS is already oversized (by a factor of OVERHEAD), but we want to
     // make it even larger
-    ushort             collisions[NR_SLOTS * 3];
+    __local ushort             collisionsData[NR_SLOTS * 3 * 64];
+    __local ushort *collisions = &collisionsData[NR_SLOTS * 3 * tlid];
     uint                nr_coll = 0;
     uint                n;
     uint               dropped_coll = 0;
@@ -560,17 +562,16 @@ void equihash_round(uint round, __global char *ht_src, __global char *ht_dst,
 #if NR_ROWS_LOG != 20 || !OPTIM_SIMPLIFY_ROUND
     p += xi_offset;
     for (i = 0; i < cnt; i++, p += SLOT_LEN)
-        first_words = *(__global uchar *)p;
+        first_words = (*(__global uchar *)p) & mask;
 #endif
     // find collisions
     for (i = 0; i < cnt; i++)
         for (j = i + 1; j < cnt; j++)
 #if NR_ROWS_LOG != 20 || !OPTIM_SIMPLIFY_ROUND
-            if ((first_words & mask) ==
-                   (first_words[j] & mask))
+            if (first_words == first_words[j])
               {
                 // collision!
-                if (nr_coll >= sizeof (collisions) / sizeof (*collisions))
+                if (nr_coll >= NR_SLOTS*3)
                     dropped_coll++;
                 else
 #if NR_SLOTS <= (1 <<

Thank you for the pool source eXternal. I guess I'd test it inside my environment and let you know if everything goes fine.

@extremal my balance does not seem to be updating (is still at 9, has been all night after 10 hours of mining on 6x1070 GPUs), although the network recognizes my hashrate.  is there a bug?

coinsforall.io/#!/coin/ZCASH/wallet/t1TrGc86inDPnPn9vhbGHDthcHJ4MZjdPgH

better to say thank you eXtremal, you got your zec when other doors closed, even with high fee - GOOD WORK eXtremal all we get profit using your solution !

extremal, give money for mining from 5 to 6 date!

not only that, but he destroyed zec price too with his miner, every granny is now mining zec

same here, seem like no more people mine now

Trying to get caught up. Not sure if it was mentioned before, but can the Claymore miner be used to mine on Coinsforall yet?

stratum for silentarmy only?