Added XgemvFastRot and Xgemm 16-bit tuning results: just defaults which are now automatically taken from 32-bit if there are no entries at all

5 files changed, 43 insertions, 7 deletions

diff --git a/scripts/database/database/clblast.py b/scripts/database/database/clblast.py
index beed46d9..8190f225 100644
--- a/scripts/database/database/clblast.py
+++ b/scripts/database/database/clblast.py
@@ -82,7 +82,7 @@ def print_cpp_database(database, output_dir):
     """Outputs the database as C++ code"""
 
     # Iterates over the kernel families
-    kernel_families = [s["kernel_family"] for s in database["sections"]]
+    kernel_families = sorted(set([s["kernel_family"] for s in database["sections"]]))
     for family_name in kernel_families:
         family_database = [s for s in database["sections"] if s["kernel_family"] == family_name]
 
@@ -92,15 +92,27 @@ def print_cpp_database(database, output_dir):
             f.write(get_cpp_header(family_name))
 
             # Loops over the different precision (e.g. 16, 32, 3232, 64, 6464)
-            precisions = sorted(set([s["precision"] for s in family_database]))
+            precisions = sorted(set([s["precision"] for s in database["sections"]]))  # Based on full database
             for precision in precisions:
                 precision_database = [s for s in family_database if s["precision"] == precision]
                 f.write(get_cpp_precision(family_name, precision))
 
-                # Loops over a combination of device vendors and device types (e.g. AMD GPU)
+                # In case there is nothing found at all (e.g. 16-bit): continue as if this was a precision of 32 but
+                # with the defaults only
+                if len(precision_database) == 0:
+                    print("[database] No results found for %s:%s, retrieving defaults from %s:32" %
+                          (family_name, precision, family_name))
+                    precision_database = [s for s in family_database if s["precision"] == "32"
+                                          and s["device_vendor"] == VENDOR_DEFAULT
+                                          and s["device_type"] == DEVICE_TYPE_DEFAULT
+                                          and s["device"] == DEVICE_NAME_DEFAULT]
+
+                # Loops over device vendors (e.g. AMD)
                 device_vendors = sorted(set([s["device_vendor"] for s in precision_database]))
                 for vendor in device_vendors:
                     vendor_database = [s for s in precision_database if s["device_vendor"] == vendor]
+
+                    # Loops over device types (e.g. GPU)
                     device_types = sorted(set([s["device_type"] for s in vendor_database]))
                     for device_type in device_types:
                         type_database = [s for s in vendor_database if s["device_type"] == device_type]
diff --git a/src/database/database.cpp b/src/database/database.cpp
index 38974b95..34c44a29 100644
--- a/src/database/database.cpp
+++ b/src/database/database.cpp
@@ -35,9 +35,9 @@ const std::vector<Database::DatabaseEntry> Database::database = {
   XdotHalf, XdotSingle, XdotDouble, XdotComplexSingle, XdotComplexDouble,
   XgemvHalf, XgemvSingle, XgemvDouble, XgemvComplexSingle, XgemvComplexDouble,
   XgemvFastHalf, XgemvFastSingle, XgemvFastDouble, XgemvFastComplexSingle, XgemvFastComplexDouble,
-  /* XgemvFastRotHalf, */ XgemvFastRotSingle, XgemvFastRotDouble, XgemvFastRotComplexSingle, XgemvFastRotComplexDouble,
+  XgemvFastRotHalf, XgemvFastRotSingle, XgemvFastRotDouble, XgemvFastRotComplexSingle, XgemvFastRotComplexDouble,
   XgerHalf, XgerSingle, XgerDouble, XgerComplexSingle, XgerComplexDouble,
-  /* XgemmHalf, */ XgemmSingle, XgemmDouble, XgemmComplexSingle, XgemmComplexDouble,
+  XgemmHalf, XgemmSingle, XgemmDouble, XgemmComplexSingle, XgemmComplexDouble,
   CopyHalf, CopySingle, CopyDouble, CopyComplexSingle, CopyComplexDouble,
   PadHalf, PadSingle, PadDouble, PadComplexSingle, PadComplexDouble,
   TransposeHalf, TransposeSingle, TransposeDouble, TransposeComplexSingle, TransposeComplexDouble,
diff --git a/src/database/database.hpp b/src/database/database.hpp
index 8d6d3863..a6ab49c5 100644
--- a/src/database/database.hpp
+++ b/src/database/database.hpp
@@ -72,9 +72,9 @@ class Database {
   static const DatabaseEntry XdotHalf, XdotSingle, XdotDouble, XdotComplexSingle, XdotComplexDouble;
   static const DatabaseEntry XgemvHalf, XgemvSingle, XgemvDouble, XgemvComplexSingle, XgemvComplexDouble;
   static const DatabaseEntry XgemvFastHalf, XgemvFastSingle, XgemvFastDouble, XgemvFastComplexSingle, XgemvFastComplexDouble;
-  static const DatabaseEntry /* XgemvFastRotHalf, */ XgemvFastRotSingle, XgemvFastRotDouble, XgemvFastRotComplexSingle, XgemvFastRotComplexDouble;
+  static const DatabaseEntry XgemvFastRotHalf, XgemvFastRotSingle, XgemvFastRotDouble, XgemvFastRotComplexSingle, XgemvFastRotComplexDouble;
   static const DatabaseEntry XgerHalf, XgerSingle, XgerDouble, XgerComplexSingle, XgerComplexDouble;
-  static const DatabaseEntry /* XgemmHalf, */ XgemmSingle, XgemmDouble, XgemmComplexSingle, XgemmComplexDouble;
+  static const DatabaseEntry XgemmHalf, XgemmSingle, XgemmDouble, XgemmComplexSingle, XgemmComplexDouble;
   static const DatabaseEntry CopyHalf, CopySingle, CopyDouble, CopyComplexSingle, CopyComplexDouble;
   static const DatabaseEntry PadHalf, PadSingle, PadDouble, PadComplexSingle, PadComplexDouble;
   static const DatabaseEntry TransposeHalf, TransposeSingle, TransposeDouble, TransposeComplexSingle, TransposeComplexDouble;
diff --git a/src/database/kernels/xgemm.hpp b/src/database/kernels/xgemm.hpp
index 7e793076..d19c55b5 100644
--- a/src/database/kernels/xgemm.hpp
+++ b/src/database/kernels/xgemm.hpp
@@ -14,6 +14,18 @@
 namespace clblast {
 // =================================================================================================
 
+const Database::DatabaseEntry Database::XgemmHalf = {
+  "Xgemm", Precision::kHalf, {
+    { // Default
+      kDeviceTypeAll, "default", {
+        { "default",                                         { {"KWG",16}, {"KWI",2}, {"MDIMA",8}, {"MDIMC",8}, {"MWG",32}, {"NDIMB",8}, {"NDIMC",8}, {"NWG",16}, {"SA",0}, {"SB",0}, {"STRM",0}, {"STRN",0}, {"VWM",1}, {"VWN",1} } },
+      }
+    },
+  }
+};
+
+// =================================================================================================
+
 const Database::DatabaseEntry Database::XgemmSingle = {
   "Xgemm", Precision::kSingle, {
     { // AMD GPUs
diff --git a/src/database/kernels/xgemv_fast_rot.hpp b/src/database/kernels/xgemv_fast_rot.hpp
index 3d2e0d3a..328094e1 100644
--- a/src/database/kernels/xgemv_fast_rot.hpp
+++ b/src/database/kernels/xgemv_fast_rot.hpp
@@ -14,6 +14,18 @@
 namespace clblast {
 // =================================================================================================
 
+const Database::DatabaseEntry Database::XgemvFastRotHalf = {
+  "XgemvFastRot", Precision::kHalf, {
+    { // Default
+      kDeviceTypeAll, "default", {
+        { "default",                                         { {"VW3",8}, {"WGS3",32}, {"WPT3",32} } },
+      }
+    },
+  }
+};
+
+// =================================================================================================
+
 const Database::DatabaseEntry Database::XgemvFastRotSingle = {
   "XgemvFastRot", Precision::kSingle, {
     { // AMD GPUs