From 6925003e45e5c681aaeb26c95ffa29275cebcaac Mon Sep 17 00:00:00 2001 From: Cedric Nugteren Date: Wed, 8 Jun 2016 10:13:37 +0200 Subject: Added global memory synchronisation for better cache performance on ARM Mali GPUs --- src/kernels/level3/xgemm_part2.opencl | 3 +++ 1 file changed, 3 insertions(+) (limited to 'src/kernels/level3/xgemm_part2.opencl') diff --git a/src/kernels/level3/xgemm_part2.opencl b/src/kernels/level3/xgemm_part2.opencl index 56ccdb96..42c1127c 100644 --- a/src/kernels/level3/xgemm_part2.opencl +++ b/src/kernels/level3/xgemm_part2.opencl @@ -258,6 +258,9 @@ inline void XgemmBody(const int kSizeM, const int kSizeN, const int kSizeK, barrier(CLK_LOCAL_MEM_FENCE); #endif } + #if GLOBAL_MEM_FENCE == 1 + barrier(CLK_GLOBAL_MEM_FENCE); + #endif } // ================================================================================================= -- cgit v1.2.3