added support for lgammaf and lgamma

1. Implementation inside HIP

Change-Id: I657263b7276a57c56081d3336fef816b5f204eff


[ROCm/clr commit: 52859a8a40]

projects/clr/hipamd/include/hip/hcc_detail/math_functions.h

@@ -66,7 +66,7 @@ __device__ float j0f(float x);
 __device__ float j1f(float x);
 __device__ float jnf(int n, float x);
 __device__ float ldexpf(float x, int exp);
-//__device__ float lgammaf(float x);
+__device__ float lgammaf(float x);
 __device__ long long int llrintf(float x);
 __device__ long long int llroundf(float x);
 __device__ float log10f(float x);

projects/clr/hipamd/src/math_functions.cpp

@@ -188,8 +188,12 @@ __device__ float ldexpf(float x, int exp)
 }
 __device__ float lgammaf(float x)
 {
-    int sign;
-    return hc::precise_math::lgammaf(x, &sign);
+  float val = 0.0f;
+  float y = x - 1;
+  while(y > 0){
+    val += logf(y--);
+  }
+  return val;
 }
 __device__ long long int llrintf(float x)
 {
@@ -570,8 +574,12 @@ __device__ double ldexp(double x, int exp)
 }
 __device__ double lgamma(double x)
 {
-    int sign;
-    return hc::precise_math::lgamma(x, &sign);
+  double val = 0.0;
+  double y = x - 1;
+  while(y > 0){
+    val += log(y--);
+  }
+  return val;
 }
 __device__ long long int llrint(double x)
 {