[fix] explicitly declare template specialization for readFromNpyFile
Not all compiler/linker can link the typed call side to the desired
type of specialized template function without having explicit
declaration in the header file. Dtype mismatch will happen during
reading numpy file when wrong caller-callee pair linked.
Change-Id: If97c2a12d9cb602e08c8c31e828f85577cf002e6
Signed-off-by: TatWai Chong <tatwai.chong@arm.com>
diff --git a/include/numpy_utils.h b/include/numpy_utils.h
index 83fbd5c..a788f81 100644
--- a/include/numpy_utils.h
+++ b/include/numpy_utils.h
@@ -139,4 +139,7 @@
static NPError writeNpyHeader(FILE* outfile, const std::vector<int32_t>& shape, const char* dtype_str);
};
+template <>
+NumpyUtilities::NPError NumpyUtilities::readFromNpyFile(const char* filename, const uint32_t elems, int32_t* databuf);
+
#endif // _TOSA_NUMPY_UTILS_H