From 1ded220cc039f081b09f86f41dfa52afca5303d1 Mon Sep 17 00:00:00 2001
From: Tom Stellard <thomas.stellard@amd.com>
Date: Tue, 31 Mar 2015 16:59:23 +0000
Subject: Implement fmax using __builtin_fmax

This ensures correct handling of NaNi.

This has been tested with piglit, OpenCV, and the ocl conformance tests.

llvm-svn: 233713
---
 libclc/generic/lib/math/fmax.cl  | 13 +++++++++----
 libclc/generic/lib/math/fmax.inc | 18 ++++++++++++++++++
 2 files changed, 27 insertions(+), 4 deletions(-)
 create mode 100644 libclc/generic/lib/math/fmax.inc

(limited to 'libclc/generic/lib')

diff --git a/libclc/generic/lib/math/fmax.cl b/libclc/generic/lib/math/fmax.cl
index 58583d6767a..239da3d3a61 100644
--- a/libclc/generic/lib/math/fmax.cl
+++ b/libclc/generic/lib/math/fmax.cl
@@ -1,11 +1,16 @@
 #include <clc/clc.h>
 
+#include "../clcmacro.h"
+
+_CLC_DEFINE_BINARY_BUILTIN(float, fmax, __builtin_fmaxf, float, float);
+
 #ifdef cl_khr_fp64
+
 #pragma OPENCL EXTENSION cl_khr_fp64 : enable
-#endif
 
-#define FUNCTION __clc_fmax
-#define FUNCTION_IMPL(x, y) ((x) < (y) ? (y) : (x))
+_CLC_DEFINE_BINARY_BUILTIN(double, fmax, __builtin_fmax, double, double);
+
+#endif
 
-#define __CLC_BODY <binary_impl.inc>
+#define __CLC_BODY <fmax.inc>
 #include <clc/math/gentype.inc>
diff --git a/libclc/generic/lib/math/fmax.inc b/libclc/generic/lib/math/fmax.inc
new file mode 100644
index 00000000000..8315c5fb6e0
--- /dev/null
+++ b/libclc/generic/lib/math/fmax.inc
@@ -0,0 +1,18 @@
+
+#if !defined(__CLC_SCALAR)
+
+_CLC_OVERLOAD _CLC_DEF __CLC_GENTYPE fmax(__CLC_GENTYPE x, float y) {
+  return fmax(x, (__CLC_GENTYPE)((__CLC_SCALAR_GENTYPE)y));
+}
+
+#ifdef cl_khr_fp64
+
+#pragma OPENCL EXTENSION cl_khr_fp64 : enable
+
+_CLC_OVERLOAD _CLC_DEF __CLC_GENTYPE fmax(__CLC_GENTYPE x, double y) {
+  return fmax(x, (__CLC_GENTYPE)((__CLC_SCALAR_GENTYPE)y));
+}
+
+#endif // ifdef cl_khr_fp64
+
+#endif // !defined(__CLC_SCALAR)
-- 
cgit v1.2.3