From 75fe8235f78520fbbfff7c9c035ecd9f1aa3e6f6 Mon Sep 17 00:00:00 2001 From: Cedric Nugteren Date: Sat, 23 Jul 2016 10:20:11 +0200 Subject: Improved the XgemvFastRot kernel by tiled loading of the input matrix A, enabling better memory performance --- src/database/kernels/xgemv.hpp | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src/database/kernels') diff --git a/src/database/kernels/xgemv.hpp b/src/database/kernels/xgemv.hpp index 65f4b5c8..6fb68858 100644 --- a/src/database/kernels/xgemv.hpp +++ b/src/database/kernels/xgemv.hpp @@ -36,7 +36,7 @@ const Database::DatabaseEntry Database::XgemvSingle = { "Xgemv", Precision::kSingle, { { // AMD GPUs kDeviceTypeGPU, "AMD", { - { "AMD Radeon R9 M370X Compute Engine", { {"WGS1",128}, {"WPT1",1}, {"VW2",1}, {"WGS2",128}, {"WPT2",1}, {"VW3",1}, {"WGS3",64}, {"WPT3",1} } }, + { "AMD Radeon R9 M370X Compute Engine", { {"WGS1",128}, {"WPT1",1}, {"VW2",1}, {"WGS2",128}, {"WPT2",1}, {"VW3",4}, {"WGS3",32}, {"WPT3",32} } }, { "Hawaii", { {"WGS1",128}, {"WPT1",1}, {"VW2",1}, {"WGS2",64}, {"WPT2",1}, {"VW3",1}, {"WGS3",64}, {"WPT3",1} } }, { "Oland", { {"WGS1",128}, {"WPT1",1}, {"VW2",1}, {"WGS2",64}, {"WPT2",1}, {"VW3",4}, {"WGS3",256}, {"WPT3",4} } }, { "Pitcairn", { {"WGS1",256}, {"WPT1",1}, {"VW2",1}, {"WGS2",64}, {"WPT2",1}, {"VW3",1}, {"WGS3",64}, {"WPT3",1} } }, -- cgit v1.2.3