[libclc] r217509 - math: Add acos implementation

[Aaron Watry]

Author: awatry
Date: Wed Sep 10 10:43:29 2014
New Revision: 217509

URL: http://llvm.org/viewvc/llvm-project?rev=217509&view=rev
Log:
math: Add acos implementation

Passes the tests that were submitted to the piglit list

Tested on R600 (Pitcairn)

Signed-off-by: Aaron Watry <awatry at gmail.com>
Reviewed-by: Jan Vesely <jan.vesely at rutgers.edu>

Added:
    libclc/trunk/generic/include/clc/math/acos.h
    libclc/trunk/generic/include/clc/math/acos.inc
    libclc/trunk/generic/lib/math/acos.cl
    libclc/trunk/generic/lib/math/acos.inc
Modified:
    libclc/trunk/generic/include/clc/clc.h
    libclc/trunk/generic/lib/SOURCES

Modified: libclc/trunk/generic/include/clc/clc.h
URL: http://llvm.org/viewvc/llvm-project/libclc/trunk/generic/include/clc/clc.h?rev=217509&r1=217508&r2=217509&view=diff
==============================================================================
--- libclc/trunk/generic/include/clc/clc.h (original)
+++ libclc/trunk/generic/include/clc/clc.h Wed Sep 10 10:43:29 2014
@@ -32,6 +32,7 @@
 #include <clc/workitem/get_group_id.h>
 
 /* 6.11.2 Math Functions */
+#include <clc/math/acos.h>
 #include <clc/math/atan.h>
 #include <clc/math/atan2.h>
 #include <clc/math/copysign.h>

Added: libclc/trunk/generic/include/clc/math/acos.h
URL: http://llvm.org/viewvc/llvm-project/libclc/trunk/generic/include/clc/math/acos.h?rev=217509&view=auto
==============================================================================
--- libclc/trunk/generic/include/clc/math/acos.h (added)
+++ libclc/trunk/generic/include/clc/math/acos.h Wed Sep 10 10:43:29 2014
@@ -0,0 +1,2 @@
+#define __CLC_BODY <clc/math/acos.inc>
+#include <clc/math/gentype.inc>

Added: libclc/trunk/generic/include/clc/math/acos.inc
URL: http://llvm.org/viewvc/llvm-project/libclc/trunk/generic/include/clc/math/acos.inc?rev=217509&view=auto
==============================================================================
--- libclc/trunk/generic/include/clc/math/acos.inc (added)
+++ libclc/trunk/generic/include/clc/math/acos.inc Wed Sep 10 10:43:29 2014
@@ -0,0 +1 @@
+_CLC_OVERLOAD _CLC_DECL __CLC_GENTYPE acos(__CLC_GENTYPE x);

Modified: libclc/trunk/generic/lib/SOURCES
URL: http://llvm.org/viewvc/llvm-project/libclc/trunk/generic/lib/SOURCES?rev=217509&r1=217508&r2=217509&view=diff
==============================================================================
--- libclc/trunk/generic/lib/SOURCES (original)
+++ libclc/trunk/generic/lib/SOURCES Wed Sep 10 10:43:29 2014
@@ -29,6 +29,7 @@ integer/sub_sat.cl
 integer/sub_sat_if.ll
 integer/sub_sat_impl.ll
 integer/upsample.cl
+math/acos.cl
 math/atan.cl
 math/atan2.cl
 math/copysign.cl

Added: libclc/trunk/generic/lib/math/acos.cl
URL: http://llvm.org/viewvc/llvm-project/libclc/trunk/generic/lib/math/acos.cl?rev=217509&view=auto
==============================================================================
--- libclc/trunk/generic/lib/math/acos.cl (added)
+++ libclc/trunk/generic/lib/math/acos.cl Wed Sep 10 10:43:29 2014
@@ -0,0 +1,8 @@
+#include <clc/clc.h>
+
+#ifdef cl_khr_fp64
+#pragma OPENCL EXTENSION cl_khr_fp64 : enable
+#endif
+
+#define __CLC_BODY <acos.inc>
+#include <clc/math/gentype.inc>

Added: libclc/trunk/generic/lib/math/acos.inc
URL: http://llvm.org/viewvc/llvm-project/libclc/trunk/generic/lib/math/acos.inc?rev=217509&view=auto
==============================================================================
--- libclc/trunk/generic/lib/math/acos.inc (added)
+++ libclc/trunk/generic/lib/math/acos.inc Wed Sep 10 10:43:29 2014
@@ -0,0 +1,21 @@
+/*
+ * There are multiple formulas for calculating arccosine of x:
+ * 1) acos(x) = (1/2*pi) + i * ln(i*x + sqrt(1-x^2)) (notice the 'i'...)
+ * 2) acos(x) = pi/2 + asin(-x) (asin isn't implemented yet)
+ * 3) acos(x) = pi/2 - asin(x) (ditto)
+ * 4) acos(x) = 2*atan2(sqrt(1-x), sqrt(1+x))
+ * 5) acos(x) = pi/2 - atan2(x, ( sqrt(1-x^2) ) )
+ *
+ * Options 1-3 are not currently usable, #5 generates more concise radeonsi
+ * bitcode and assembly than #4 (134 vs 132 instructions on radeonsi), but
+ * precision of #4 may be better.
+ */
+
+_CLC_OVERLOAD _CLC_DEF __CLC_GENTYPE acos(__CLC_GENTYPE x) {
+  return (
+    (__CLC_GENTYPE) 2.0 * atan2(
+      sqrt((__CLC_GENTYPE) 1.0 - x),
+      sqrt((__CLC_GENTYPE) 1.0 + x)
+    )
+  );
+}