Merge pull request #140 from guacamoleo/develop

AutoGemm performance data; sgemm add unroll=8 for benchmarking; gemm …

Author: Timmy

doc/performance/clBLAS_2.6.0/S9150/cgemmNT_S9150_14.50.2_2.6.0_8.csv

@@ -112,7 +112,7 @@ if( NOT DEFINED clBLAS_VERSION_MINOR )
 endif( )
 
 if( NOT DEFINED clBLAS_VERSION_PATCH )
-    set( clBLAS_VERSION_PATCH 0 )
+    set( clBLAS_VERSION_PATCH 1 )
 endif( )
 
 set( clBLAS_VERSION "${clBLAS_VERSION_MAJOR}.${clBLAS_VERSION_MINOR}.${clBLAS_VERSION_PATCH}")

doc/performance/clBLAS_2.6.0/S9150/dgemmNT_S9150_14.50.2_2.6.0_8.csv

@@ -9,15 +9,15 @@
 # [ size, fallback tile, [ valid tiles ] ],
   "s":[
     [ 4000, [ 16, 16,  6,  6], [ [ 16, 16,  6,  6] ] ],
-    [ 2496, [ 16, 16,  4,  4], [ [ 16, 16,  6,  6], [ 16, 16,  4,  4] ] ], 
+    [ 2496, [ 16, 16,  4,  4], [ [ 16, 16,  6,  6], [ 16, 16,  4,  4] ] ],
     [ 2448, [ 16, 16,  6,  6], [ [ 16, 16,  6,  6] ] ],
     [ 1600, [ 16, 16,  6,  6], [ [ 16, 16,  6,  6], [ 16, 16,  4,  4], [ 16, 16,  5,  5] ] ],
     [ 1008, [ 16, 16,  6,  6], [ [ 16, 16,  6,  6], [ 16, 16,  4,  4], [ 16, 16,  5,  5], [ 16, 16,  3,  3] ] ],
-    [  960, [ 16, 16,  2,  2], [ [ 16, 16,  4,  4], [ 16, 16,  3,  3], [ 16, 16,  5,  5], [ 16, 16,  2,  2] ] ], 
-    [  896, [ 16, 16,  2,  2], [ [ 16, 16,  4,  4], [ 16, 16,  6,  6], [ 16, 16,  3,  3], [ 16, 16,  5,  5], [ 16, 16,  2,  2] ] ], 
-    [  864, [ 16, 16,  2,  2], [ [ 16, 16,  6,  6], [ 16, 16,  3,  3], [ 16, 16,  5,  5], [ 16, 16,  4,  4], [ 16, 16,  2,  2] ] ], 
-    [  784, [ 16, 16,  2,  2], [ [ 16, 16,  3,  3], [ 16, 16,  5,  5], [ 16, 16,  4,  4], [ 16, 16,  2,  2], [ 16, 16,  1,  1] ] ], 
-    [  768, [ 16, 16,  2,  2], [ [ 16, 16,  3,  3], [ 16, 16,  5,  5], [ 16, 16,  4,  4], [ 16, 16,  2,  2], [ 16, 16,  1,  1] ] ], 
+    [  960, [ 16, 16,  2,  2], [ [ 16, 16,  4,  4], [ 16, 16,  3,  3], [ 16, 16,  5,  5], [ 16, 16,  2,  2] ] ],
+    [  896, [ 16, 16,  2,  2], [ [ 16, 16,  4,  4], [ 16, 16,  6,  6], [ 16, 16,  3,  3], [ 16, 16,  5,  5], [ 16, 16,  2,  2] ] ],
+    [  864, [ 16, 16,  2,  2], [ [ 16, 16,  6,  6], [ 16, 16,  3,  3], [ 16, 16,  5,  5], [ 16, 16,  4,  4], [ 16, 16,  2,  2] ] ],
+    [  784, [ 16, 16,  2,  2], [ [ 16, 16,  3,  3], [ 16, 16,  5,  5], [ 16, 16,  4,  4], [ 16, 16,  2,  2], [ 16, 16,  1,  1] ] ],
+    [  768, [ 16, 16,  2,  2], [ [ 16, 16,  3,  3], [ 16, 16,  5,  5], [ 16, 16,  4,  4], [ 16, 16,  2,  2], [ 16, 16,  1,  1] ] ],
     [  720, [ 16, 16,  2,  2], [ [ 16, 16,  4,  4], [ 16, 16,  5,  5], [ 16, 16,  4,  4], [ 16, 16,  6,  6], [ 16, 16,  3,  3] ] ],
     [  464, [ 16, 16,  3,  3], [ [ 16, 16,  3,  3], [ 16, 16,  4,  4], [ 16, 16,  2,  2], [ 16, 16,  5,  5] ] ],
     [  304, [ 16, 16,  2,  2], [ [ 16, 16,  3,  3], [ 16, 16,  2,  2], [ 16, 16,  1,  1] ] ],
@@ -76,7 +76,7 @@
 transposes = { "s":["N", "T"], "d":["N", "T"], \
     "c":["N", "T", "C"], "z":["N", "T", "C"] }
 
-unrolls = { "s":[16, 1], "d":[8, 1], "c":[8, 1], "z":[8, 1] }
+unrolls = { "s":[16, 8, 1], "d":[8, 1], "c":[8, 1], "z":[8, 1] }
 
 betas = [ 0, 1 ]