[Beignet] [PATCH] Improve the accuracy of built-in function asin.

[Xing, Homer]

Good idea. This patch looks good to me.

Homer

-----Original Message-----
From: beignet-bounces+homer.xing=intel.com at lists.freedesktop.org [mailto:beignet-bounces+homer.xing=intel.com at lists.freedesktop.org] On Behalf Of Yi Sun
Sent: Tuesday, August 6, 2013 10:03 PM
To: beignet at lists.freedesktop.org
Cc: Sun, Yi; Jin, Gordon
Subject: [Beignet] [PATCH] Improve the accuracy of built-in function asin.

Method: asin(x) = x + (1 * x^3)/(2 * 3) + (1 * 3 * x^5)/(2*4 * 5) + \
	(1 * 3 * 5 * x^7)/(2*4*6 * 7) + ...
Iterate this for 30 times.

Signed-off-by: Yi Sun <yi.sun at intel.com>
---
 backend/src/ocl_stdlib.tmpl.h |   11 ++++++++++-
 1 files changed, 10 insertions(+), 1 deletions(-)

diff --git a/backend/src/ocl_stdlib.tmpl.h b/backend/src/ocl_stdlib.tmpl.h index c972a3e..8002a85 100644
--- a/backend/src/ocl_stdlib.tmpl.h
+++ b/backend/src/ocl_stdlib.tmpl.h
@@ -562,7 +562,16 @@ INLINE_OVERLOADABLE float __gen_ocl_internal_tanh(float x) {
   return (1 - y) / (1 + y);
 }
 INLINE_OVERLOADABLE float __gen_ocl_internal_asin(float x) {
-  return x + __gen_ocl_pow(x, 3) / 6 + __gen_ocl_pow(x, 5) * 3 / 40 + __gen_ocl_pow(x, 7) * 5 / 112;
+  float sum = x, c = x, m = 1.0;
+  int n = 1;
+  do
+  {
+    c *= (2 * n - 1) * x * x;
+    m *= (2 * n);
+    sum += ( c / m / (2 * n + 1));
+    n++;
+  }while( n < 30);
+  return sum;
 }
 INLINE_OVERLOADABLE float __gen_ocl_internal_asinpi(float x) {
   return __gen_ocl_internal_asin(x) / M_PI_F;
--
1.7.6.4

_______________________________________________
Beignet mailing list
Beignet at lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/beignet