Commit 952757a8 authored by Peter Eastman's avatar Peter Eastman
Browse files

Reduced number of force buffers on AMD

parent 95b02c8e
...@@ -55,7 +55,7 @@ OpenCLNonbondedUtilities::OpenCLNonbondedUtilities(OpenCLContext& context) : con ...@@ -55,7 +55,7 @@ OpenCLNonbondedUtilities::OpenCLNonbondedUtilities(OpenCLContext& context) : con
else { else {
numForceThreadBlocks = context.getNumThreadBlocks(); numForceThreadBlocks = context.getNumThreadBlocks();
forceThreadBlockSize = OpenCLContext::ThreadBlockSize; forceThreadBlockSize = OpenCLContext::ThreadBlockSize;
numForceBuffers = numForceThreadBlocks*forceThreadBlockSize/OpenCLContext::TileSize; numForceBuffers = numForceThreadBlocks;
if (numForceBuffers >= context.getNumAtomBlocks()) { if (numForceBuffers >= context.getNumAtomBlocks()) {
// For small systems, it is more efficient to have one force buffer per block of 32 atoms instead of one per warp. // For small systems, it is more efficient to have one force buffer per block of 32 atoms instead of one per warp.
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment