Merge pull request #2 from GerHobbelt/tfloat-patch-1

consistent TFloat prototypes
stweil · Jul 13, 2021 · 3b8f329 · 3b8f329
2 parents b3adfdd + efc7601
commit 3b8f329
Show file tree

Hide file tree

Showing 2 changed files with 6 additions and 6 deletions.
diff --git a/src/arch/dotproductsse.cpp b/src/arch/dotproductsse.cpp
@@ -39,7 +39,7 @@ TFloat DotProductSSE(const TFloat *u, const TFloat *v, int n) {
   return total;
 }
 #else
-double DotProductSSE(const double *u, const double *v, int n) {
+TFloat DotProductSSE(const TFloat *u, const TFloat *v, int n) {
   int max_offset = n - 2;
   int offset = 0;
   // Accumulate a set of 2 sums in sum, by loading pairs of 2 values from u and

diff --git a/src/arch/intsimdmatrixsse.cpp b/src/arch/intsimdmatrixsse.cpp
@@ -73,15 +73,15 @@ static int32_t IntDotProductSSE(const int8_t *u, const int8_t *v, int n) {
 }
 
 // Computes part of matrix.vector v = Wu. Computes 1 result.
-static void PartialMatrixDotVector1(const int8_t *wi, const double *scales, const int8_t *u,
-                                    int num_in, double *v) {
-  double total = IntDotProductSSE(u, wi, num_in);
+static void PartialMatrixDotVector1(const int8_t *wi, const TFloat *scales, const int8_t *u,
+                                    int num_in, TFloat *v) {
+  TFloat total = IntDotProductSSE(u, wi, num_in);
   // Add in the bias and correct for integer values.
   *v = (total + wi[num_in] * INT8_MAX) * *scales;
 }
 
-static void matrixDotVector(int dim1, int dim2, const int8_t *wi, const double *scales,
-                            const int8_t *u, double *v) {
+static void matrixDotVector(int dim1, int dim2, const int8_t *wi, const TFloat *scales,
+                            const int8_t *u, TFloat *v) {
   const int num_out = dim1;
   const int num_in = dim2 - 1;
   int output = 0;