[libclc] Simplify unary_def_scalarize.inc's use in __clc_erf/erfc/tgamma (#150181)

Also delete unary_def_via_fp32.inc. There are small changes in
amdgcn--amdhsa.bc due to vector conversion is scalarized, e.g.
  %2 = fpext <4 x half> %0 to <4 x float>
  %3 = extractelement <4 x float> %2, i64 0
  %4 = tail call float @llvm.fabs.f32(float %3)
->
  %2 = extractelement <4 x half> %0, i64 0
  %3 = tail call half @llvm.fabs.f16(half %2)
  %4 = fpext half %3 to float

Author: wenju-he

libclc/clc/include/clc/math/unary_def_via_fp32.inc

@@ -6,7 +6,6 @@
 //
 //===----------------------------------------------------------------------===//
 
-#include <clc/clcmacro.h>
 #include <clc/internal/clc.h>
 #include <clc/math/clc_exp.h>
 #include <clc/math/clc_fabs.h>
@@ -211,12 +210,6 @@ _CLC_OVERLOAD _CLC_DEF float __clc_erf(float x) {
   return ret;
 }
 
-#define __FLOAT_ONLY
-#define FUNCTION __clc_erf
-#define __CLC_BODY <clc/shared/unary_def_scalarize.inc>
-#include <clc/math/gentype.inc>
-#undef FUNCTION
-
 #ifdef cl_khr_fp64
 
 #pragma OPENCL EXTENSION cl_khr_fp64 : enable
@@ -500,24 +493,19 @@ _CLC_OVERLOAD _CLC_DEF double __clc_erf(double y) {
   return y < 0.0 ? -ret : ret;
 }
 
-#define __DOUBLE_ONLY
-#define FUNCTION __clc_erf
-#define __CLC_BODY <clc/shared/unary_def_scalarize.inc>
-#include <clc/math/gentype.inc>
-#undef FUNCTION
-
 #endif
 
 #ifdef cl_khr_fp16
 
-#include <clc/clc_convert.h>
-
 #pragma OPENCL EXTENSION cl_khr_fp16 : enable
 
 // Forward the half version of this builtin onto the float one
-#define __HALF_ONLY
-#define FUNCTION __clc_erf
-#define __CLC_BODY <clc/math/unary_def_via_fp32.inc>
-#include <clc/math/gentype.inc>
+_CLC_OVERLOAD _CLC_DEF half __clc_erf(half x) {
+  return (half)__clc_erf((float)x);
+}
 
 #endif
+
+#define FUNCTION __clc_erf
+#define __CLC_BODY <clc/shared/unary_def_scalarize.inc>
+#include <clc/math/gentype.inc>

libclc/clc/lib/generic/math/clc_erf.cl

@@ -6,7 +6,6 @@
 //
 //===----------------------------------------------------------------------===//
 
-#include <clc/clcmacro.h>
 #include <clc/internal/clc.h>
 #include <clc/math/clc_exp.h>
 #include <clc/math/clc_fabs.h>
@@ -211,12 +210,6 @@ _CLC_OVERLOAD _CLC_DEF float __clc_erfc(float x) {
   return ret;
 }
 
-#define __FLOAT_ONLY
-#define FUNCTION __clc_erfc
-#define __CLC_BODY <clc/shared/unary_def_scalarize.inc>
-#include <clc/math/gentype.inc>
-#undef FUNCTION
-
 #ifdef cl_khr_fp64
 
 #pragma OPENCL EXTENSION cl_khr_fp64 : enable
@@ -509,12 +502,6 @@ _CLC_OVERLOAD _CLC_DEF double __clc_erfc(double x) {
   return ret;
 }
 
-#define __DOUBLE_ONLY
-#define FUNCTION __clc_erfc
-#define __CLC_BODY <clc/shared/unary_def_scalarize.inc>
-#include <clc/math/gentype.inc>
-#undef FUNCTION
-
 #endif
 
 #ifdef cl_khr_fp16
@@ -524,9 +511,12 @@ _CLC_OVERLOAD _CLC_DEF double __clc_erfc(double x) {
 #pragma OPENCL EXTENSION cl_khr_fp16 : enable
 
 // Forward the half version of this builtin onto the float one
-#define __HALF_ONLY
-#define FUNCTION __clc_erfc
-#define __CLC_BODY <clc/math/unary_def_via_fp32.inc>
-#include <clc/math/gentype.inc>
+_CLC_OVERLOAD _CLC_DEF half __clc_erfc(half x) {
+  return (half)__clc_erfc((float)x);
+}
 
 #endif
+
+#define FUNCTION __clc_erfc
+#define __CLC_BODY <clc/shared/unary_def_scalarize.inc>
+#include <clc/math/gentype.inc>

libclc/clc/lib/generic/math/clc_erfc.cl

@@ -6,7 +6,6 @@
 //
 //===----------------------------------------------------------------------===//
 
-#include <clc/clcmacro.h>
 #include <clc/float/definitions.h>
 #include <clc/internal/clc.h>
 #include <clc/math/clc_exp.h>
@@ -32,12 +31,6 @@ _CLC_OVERLOAD _CLC_DEF float __clc_tgamma(float x) {
   return g;
 }
 
-#define __FLOAT_ONLY
-#define FUNCTION __clc_tgamma
-#define __CLC_BODY <clc/shared/unary_def_scalarize.inc>
-#include <clc/math/gentype.inc>
-#undef FUNCTION
-
 #ifdef cl_khr_fp64
 
 #pragma OPENCL EXTENSION cl_khr_fp64 : enable
@@ -59,24 +52,19 @@ _CLC_OVERLOAD _CLC_DEF double __clc_tgamma(double x) {
   return g;
 }
 
-#define __DOUBLE_ONLY
-#define FUNCTION __clc_tgamma
-#define __CLC_BODY <clc/shared/unary_def_scalarize.inc>
-#include <clc/math/gentype.inc>
-#undef FUNCTION
-
 #endif
 
 #ifdef cl_khr_fp16
 
-#include <clc/clc_convert.h>
-
 #pragma OPENCL EXTENSION cl_khr_fp16 : enable
 
 // Forward the half version of this builtin onto the float one
-#define __HALF_ONLY
-#define FUNCTION __clc_tgamma
-#define __CLC_BODY <clc/math/unary_def_via_fp32.inc>
-#include <clc/math/gentype.inc>
+_CLC_OVERLOAD _CLC_DEF half __clc_tgamma(half x) {
+  return (half)__clc_tgamma((float)x);
+}
 
 #endif
+
+#define FUNCTION __clc_tgamma
+#define __CLC_BODY <clc/shared/unary_def_scalarize.inc>
+#include <clc/math/gentype.inc>