[clfft] 08/128: Precallback - skip redundant pointer declaration in transpose kernel if precallback set

Ghislain Vaillant ghisvail-guest at moszumanska.debian.org
Thu Oct 22 14:54:32 UTC 2015


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

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

commit 377e665ef79d1deb2293d2974f7325f199cc6c67
Author: Pradeep <pradeep.rao at amd.com>
Date:   Tue Jul 28 09:17:23 2015 +0530

    Precallback - skip redundant pointer declaration in transpose kernel if precallback set
---
 src/library/generator.transpose.gcn.cpp | 14 +++++++++++---
 1 file changed, 11 insertions(+), 3 deletions(-)

diff --git a/src/library/generator.transpose.gcn.cpp b/src/library/generator.transpose.gcn.cpp
index 40736b2..aebdfdc 100644
--- a/src/library/generator.transpose.gcn.cpp
+++ b/src/library/generator.transpose.gcn.cpp
@@ -495,11 +495,19 @@ static clfftStatus genTransposeKernel( const FFTGeneratedTransposeGCNAction::Sig
 		switch( params.fft_inputLayout )
 		{
 		case CLFFT_COMPLEX_INTERLEAVED:
-			clKernWrite( transKernel, 3 ) << "global " << dtInput << "* tileIn = " << pmComplexIn << " + iOffset;" << std::endl;
+			//No need of tileIn declaration when precallback is set as the global buffer is used directly
+			if (!params.fft_hasPreCallback)
+			{
+				clKernWrite( transKernel, 3 ) << "global " << dtInput << "* tileIn = " << pmComplexIn << " + iOffset;" << std::endl;
+			}
 			break;
 		case CLFFT_COMPLEX_PLANAR:
-			clKernWrite( transKernel, 3 ) << "global " << dtInput << "* realTileIn = " << pmRealIn << " + iOffset;" << std::endl;
-			clKernWrite( transKernel, 3 ) << "global " << dtInput << "* imagTileIn = " << pmImagIn << " + iOffset;" << std::endl;
+			//No need of tileIn declaration when precallback is set as the global buffer is used directly
+			if (!params.fft_hasPreCallback)
+			{
+				clKernWrite( transKernel, 3 ) << "global " << dtInput << "* realTileIn = " << pmRealIn << " + iOffset;" << std::endl;
+				clKernWrite( transKernel, 3 ) << "global " << dtInput << "* imagTileIn = " << pmImagIn << " + iOffset;" << std::endl;
+			}
 			break;
 		case CLFFT_HERMITIAN_INTERLEAVED:
 		case CLFFT_HERMITIAN_PLANAR:

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



More information about the debian-science-commits mailing list