[clfft] 92/128: Double Precision pragma enable in Square Transpose generator

[Ghislain Vaillant]

This is an automated email from the git hooks/post-receive script.

ghisvail-guest pushed a commit to branch master
in repository clfft.

commit da3965cc928bc8b962b0294d8f60b5062b441f83
Author: Pradeep <pradeep.rao at amd.com>
Date:   Thu Oct 1 12:05:51 2015 +0530

    Double Precision pragma enable in Square Transpose generator
---
 src/library/generator.transpose.square.cpp | 8 ++++++++
 1 file changed, 8 insertions(+)

diff --git a/src/library/generator.transpose.square.cpp b/src/library/generator.transpose.square.cpp
index 357dfad..230c32d 100644
--- a/src/library/generator.transpose.square.cpp
+++ b/src/library/generator.transpose.square.cpp
@@ -278,6 +278,14 @@ static clfftStatus genTransposeKernel( const FFTGeneratedTransposeSquareAction::
     case CLFFT_DOUBLE_FAST:
         dtPlanar = "double";
         dtComplex = "double2";
+
+		// Emit code that enables double precision in the kernel
+        clKernWrite( transKernel, 0 ) << "#ifdef cl_khr_fp64" << std::endl;
+        clKernWrite( transKernel, 3 ) << "#pragma OPENCL EXTENSION cl_khr_fp64 : enable" << std::endl;
+        clKernWrite( transKernel, 0 ) << "#else" << std::endl;
+        clKernWrite( transKernel, 3 ) <<  "#pragma OPENCL EXTENSION cl_amd_fp64 : enable" << std::endl;
+        clKernWrite( transKernel, 0 ) << "#endif\n" << std::endl;
+
         break;
     default:
         return CLFFT_TRANSPOSED_NOTIMPLEMENTED;

-- 
Alioth's /usr/local/bin/git-commit-notice on /srv/git.debian.org/git/debian-science/packages/clfft.git