[Libclc-dev] [PATCH] Implementations for exp(float) and exp(double) v2

[Jeroen Ketema]

Use separate implementations instead of a macro
to ensure the constant multiplied with is of
higher precision.

v2: Use the correct formula, spotted by Dan Liew
---
generic/include/clc/math/exp.h       |  9 +++++++--
generic/include/clc/math/gentype.inc |  4 ++++
generic/lib/SOURCES                  |  1 +
generic/lib/math/exp.cl              |  8 ++++++++
generic/lib/math/exp.inc             | 10 ++++++++++
5 files changed, 30 insertions(+), 2 deletions(-)
create mode 100644 generic/lib/math/exp.cl
create mode 100644 generic/lib/math/exp.inc

diff --git a/generic/include/clc/math/exp.h b/generic/include/clc/math/exp.h
index dbc4b84..9866524 100644
--- a/generic/include/clc/math/exp.h
+++ b/generic/include/clc/math/exp.h
@@ -1,4 +1,9 @@
#undef exp

-// exp(x) = exp2(x * log2(e)
-#define exp(val) (__clc_exp2((val) * 1.44269504f))
+#define __CLC_BODY <clc/math/unary_decl.inc>
+#define __CLC_FUNCTION exp
+
+#include <clc/math/gentype.inc>
+
+#undef __CLC_BODY
+#undef __CLC_FUNCTION
diff --git a/generic/include/clc/math/gentype.inc b/generic/include/clc/math/gentype.inc
index bff4f56..9f79f6e 100644
--- a/generic/include/clc/math/gentype.inc
+++ b/generic/include/clc/math/gentype.inc
@@ -1,4 +1,5 @@
#define __CLC_SCALAR_GENTYPE float
+#define __CLC_FPSIZE 32

#define __CLC_GENTYPE float
#define __CLC_SCALAR
@@ -26,10 +27,12 @@
#include __CLC_BODY
#undef __CLC_GENTYPE

+#undef __CLC_FPSIZE
#undef __CLC_SCALAR_GENTYPE

#ifdef cl_khr_fp64
#define __CLC_SCALAR_GENTYPE double
+#define __CLC_FPSIZE 64

#define __CLC_SCALAR
#define __CLC_GENTYPE double
@@ -57,6 +60,7 @@
#include __CLC_BODY
#undef __CLC_GENTYPE

+#undef __CLC_FPSIZE
#undef __CLC_SCALAR_GENTYPE
#endif

diff --git a/generic/lib/SOURCES b/generic/lib/SOURCES
index 6ccdf48..a35542c 100644
--- a/generic/lib/SOURCES
+++ b/generic/lib/SOURCES
@@ -27,6 +27,7 @@ integer/sub_sat.cl
integer/sub_sat_if.ll
integer/sub_sat_impl.ll
integer/upsample.cl
+math/exp.cl
math/fmax.cl
math/fmin.cl
math/hypot.cl
diff --git a/generic/lib/math/exp.cl b/generic/lib/math/exp.cl
new file mode 100644
index 0000000..dbf4a93
--- /dev/null
+++ b/generic/lib/math/exp.cl
@@ -0,0 +1,8 @@
+#include <clc/clc.h>
+
+#ifdef cl_khr_fp64
+#pragma OPENCL EXTENSION cl_khr_fp64 : enable
+#endif
+
+#define __CLC_BODY <exp.inc>
+#include <clc/math/gentype.inc>
diff --git a/generic/lib/math/exp.inc b/generic/lib/math/exp.inc
new file mode 100644
index 0000000..525fb59
--- /dev/null
+++ b/generic/lib/math/exp.inc
@@ -0,0 +1,10 @@
+_CLC_OVERLOAD _CLC_DEF __CLC_GENTYPE exp(__CLC_GENTYPE val) {
+  // exp(x) = exp2(x * log2(e))
+#if __CLC_FPSIZE == 32
+  return exp2(val * M_LOG2E_F);
+#elif __CLC_FPSIZE == 64
+  return exp2(val * M_LOG2E);
+#else
+#error unknown _CLC_FPSIZE
+#endif
+}
-- 
1.8.5.2 (Apple Git-48)