summaryrefslogtreecommitdiff
path: root/backend
diff options
context:
space:
mode:
authorrander.wang <rander.wang@intel.com>2017-05-15 16:12:24 +0800
committerYang Rong <rong.r.yang@intel.com>2017-05-17 18:11:00 +0800
commit4e1b94f00ed0a09f62b94b5f6889a96c92355c46 (patch)
tree5ed32f1dd1d6fd482652451602e7a17c9a5b1774 /backend
parentcbe117343fa14351f4886cb8c9102f0ebbf8b340 (diff)
downloadbeignet-4e1b94f00ed0a09f62b94b5f6889a96c92355c46.tar.gz
backend: refine acos
refine algorithm to remove branch Signed-off-by: rander.wang <rander.wang@intel.com> Tested-by: Yang Rong <rong.r.yang@intel.com>
Diffstat (limited to 'backend')
-rw-r--r--backend/src/libocl/tmpl/ocl_math_common.tmpl.cl13
1 files changed, 9 insertions, 4 deletions
diff --git a/backend/src/libocl/tmpl/ocl_math_common.tmpl.cl b/backend/src/libocl/tmpl/ocl_math_common.tmpl.cl
index 576033e9..fb6e412c 100644
--- a/backend/src/libocl/tmpl/ocl_math_common.tmpl.cl
+++ b/backend/src/libocl/tmpl/ocl_math_common.tmpl.cl
@@ -1187,10 +1187,15 @@ OVERLOADABLE float __gen_ocl_internal_asinpi(float x) {
return __gen_ocl_internal_asin(x) / M_PI_F;
}
OVERLOADABLE float __gen_ocl_internal_acos(float x) {
- if(x > 0.5)
- return 2 * __gen_ocl_asin_util(native_sqrt((1-x)/2));
- else
- return M_PI_2_F - __gen_ocl_internal_asin(x);
+ float absX = fabs(x);
+ float asinX2 =__gen_ocl_asin_util(x);
+ float tmp = __gen_ocl_asin_util(native_sqrt(mad(-0.5f, absX, 0.5f)));
+ float asinX1 = mad(2.0f ,tmp, -M_PI_2_F);
+
+ float retVal = (x < 0.0f)?asinX1:-asinX1;
+ retVal = (absX > 0.5f)?retVal:asinX2;
+ retVal = (x <= 0.5f) ? M_PI_2_F - retVal:2.0f*tmp;
+ return retVal;
}
OVERLOADABLE float __gen_ocl_internal_acospi(float x) {
return __gen_ocl_internal_acos(x) / M_PI_F;