HMMM. This is where the float/double co-existence stuff starts to bec…

…ome NOT NICE: code repetition at another level.

TODO: Better idea? --> Maybe namespaces and double kernel projects or compile via #define+#include-all-source-files hack collective source code pages? (Latter approach may become a problem when debugging, or will the compiler suite cope well? Will know only once done & tested.)

At least this is about the point where the function template solution stops to be useful. The run-time switching desire between float and double is doable, but not by using #ifdef/#else throughout, nor templating all the way up the TFloat usage calltree.

src/arch/intsimdmatrix.h

@@ -97,9 +97,12 @@ struct TESS_API IntSimdMatrix {
   // RoundInputs above.
   // The input will be over-read to the extent of the padding. There are no
   // alignment requirements.
-  using MatrixDotVectorFunction = void (*)(int, int, const int8_t *, const TFloat *, const int8_t *,
-                                           TFloat *);
-  MatrixDotVectorFunction matrixDotVectorFunction;
+  using MatrixDotVectorFunctionFP32 = void (*)(int, int, const int8_t *, const float *, const int8_t *,
+                                           float *);
+  using MatrixDotVectorFunctionFP64 = void (*)(int, int, const int8_t *, const double *, const int8_t *,
+                                           double *);
+  MatrixDotVectorFunctionFP32 matrixDotVectorFunctionFP32;
+  MatrixDotVectorFunctionFP64 matrixDotVectorFunctionFP64;
 
   // Number of 32 bit outputs held in each register.
   int num_outputs_per_register_;

src/lstm/weightmatrix.cpp

@@ -351,15 +351,25 @@ void WeightMatrix::MatrixDotVector(const TFloat *u, TFloat *v) const {
   MatrixDotVectorInternal(wf_, true, false, u, v);
 }
 
-void WeightMatrix::MatrixDotVector(const int8_t *u, TFloat *v) const {
+void WeightMatrix::MatrixDotVector(const int8_t *u, float *v) const {
   assert(int_mode_);
   if (IntSimdMatrix::intSimdMatrix) {
-    IntSimdMatrix::intSimdMatrix->matrixDotVectorFunction(wi_.dim1(), wi_.dim2(), &shaped_w_[0],
+    IntSimdMatrix::intSimdMatrix->matrixDotVectorFunctionFP32(wi_.dim1(), wi_.dim2(), &shaped_w_[0],
                                                           &scales_[0], u, v);
   } else {
     IntSimdMatrix::MatrixDotVector(wi_, scales_, u, v);
   }
 }
+void WeightMatrix::MatrixDotVector(const int8_t *u, double *v) const {
+  assert(int_mode_);
+  if (IntSimdMatrix::intSimdMatrix) {
+    IntSimdMatrix::intSimdMatrix->matrixDotVectorFunctionFP64(wi_.dim1(), wi_.dim2(), &shaped_w_[0],
+                                                          &scales_[0], u, v);
+  } else {
+    IntSimdMatrix::MatrixDotVector(wi_, scales_, u, v);
+  }
+}
+
 
 // MatrixDotVector for peep weights, MultiplyAccumulate adds the
 // component-wise products of *this[0] and v to inout.