Daniel/master-thesis
1
0
Fork 0
Code Issues 3 Pull Requests Actions 1 Packages Projects 1 Releases Wiki Activity

benchmarking: updated blocksize to 256 with moderate improvements
Some checks failed
CI / Julia ${{ matrix.version }} - ${{ matrix.os }} - ${{ matrix.arch }} - ${{ github.event_name }} (x64, ubuntu-latest, 1.10) (push) Has been cancelled
Details
CI / Julia ${{ matrix.version }} - ${{ matrix.os }} - ${{ matrix.arch }} - ${{ github.event_name }} (x64, ubuntu-latest, 1.6) (push) Has been cancelled
Details
CI / Julia ${{ matrix.version }} - ${{ matrix.os }} - ${{ matrix.arch }} - ${{ github.event_name }} (x64, ubuntu-latest, pre) (push) Has been cancelled
Details

Browse Source
This commit is contained in:
Daniel
2025-03-30 13:56:25 +02:00
parent 1dc0c1898d
commit d9c83caad9
5 changed files with 129 additions and 100 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
package/src/Transpiler.jl
View File

@ -55,8 +55,8 @@ function evaluate(expressions::Vector{Expr}, variables::Matrix{Float32}, paramet
# execute each kernel (also try doing this with Threads.@threads. Since we can have multiple grids, this might improve performance)
for i in eachindex(kernels)
config = launch_configuration(kernels[i])
threads = min(variableCols, config.threads)
# config = launch_configuration(kernels[i])
threads = min(variableCols, 256)
blocks = cld(variableCols, threads)
cudacall(kernels[i], (CuPtr{Float32},CuPtr{Float32},CuPtr{Float32}), cudaVars, cudaParams, cudaResults; threads=threads, blocks=blocks)