Fix few multiply_add calls with templated versions.

Author: lntue

libc/src/math/generic/expm1f.cpp

@@ -103,7 +103,7 @@ LLVM_LIBC_FUNCTION(float, expm1f, (float x)) {
         // intermediate results as it is more efficient than using an emulated
         // version of FMA.
 #if defined(LIBC_TARGET_CPU_HAS_FMA_FLOAT)
-      return fputil::multiply_add<float>(x, x, x);
+      return fputil::multiply_add(x, x, x);
 #else
       double xd = x;
       return static_cast<float>(fputil::multiply_add(xd, xd, xd));

libc/src/math/generic/sincosf16_utils.h

@@ -40,7 +40,7 @@ constexpr float SIN_K_PI_OVER_32[64] = {
 
 LIBC_INLINE int32_t range_reduction_sincospif16(float x, float &y) {
   float kf = fputil::nearest_integer(x * 32);
-  y = fputil::multiply_add<float>(x, 32.0, -kf);
+  y = fputil::multiply_add(x, 32.0f, -kf);
 
   return static_cast<int32_t>(kf);
 }

libc/src/math/generic/sincosf_utils.h

@@ -108,7 +108,7 @@ LIBC_INLINE void sincosf_eval(double xd, uint32_t x_abs, double &sin_k,
 //   => pi * x = (k + y) * pi / 32
 static LIBC_INLINE int64_t range_reduction_sincospi(double x, double &y) {
   double kd = fputil::nearest_integer(x * 32);
-  y = fputil::multiply_add<double>(x, 32.0, -kd);
+  y = fputil::multiply_add(x, 32.0, -kd);
 
   return static_cast<int64_t>(kd);
 }