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

benchmarking: moved compilation of kernel to evaluate function, as it required too much memory
Some checks are pending
CI / Julia ${{ matrix.version }} - ${{ matrix.os }} - ${{ matrix.arch }} - ${{ github.event_name }} (x64, ubuntu-latest, 1.10) (push) Waiting to run
Details
CI / Julia ${{ matrix.version }} - ${{ matrix.os }} - ${{ matrix.arch }} - ${{ github.event_name }} (x64, ubuntu-latest, 1.6) (push) Waiting to run
Details
CI / Julia ${{ matrix.version }} - ${{ matrix.os }} - ${{ matrix.arch }} - ${{ github.event_name }} (x64, ubuntu-latest, pre) (push) Waiting to run
Details

Browse Source
This commit is contained in:
Daniel
2025-05-19 11:58:24 +02:00
parent f33551e25f
commit e29199d865
2 changed files with 7 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
package/src/ExpressionExecutorCuda.jl
View File

@ -56,19 +56,19 @@ function evaluate_gpu(expressions::Vector{Expr}, X::Matrix{Float32}, p::Vector{V
largestParameterSetSize = Utils.get_max_inner_length(p) # parameters get transformed into matrix. Will be nr. of rows in parameter matrix
compiledKernels = Vector{CuFunction}(undef, length(expressions))
ptxKernels = Vector{String}(undef, length(expressions))
kernelName = "evaluate_gpu"
@inbounds Threads.@threads for i in eachindex(expressions)
ex = ExpressionProcessing.expr_to_postfix(expressions[i])
ptxKernel = Transpiler.transpile(ex, variableSetSize, largestParameterSetSize, numVariableSets, i-1, kernelName) # i-1 because julia is 1-based but PTX needs 0-based indexing
compiledKernels[i] = Transpiler.CompileKernel(ptxKernel, kernelName)
ptxKernels[i] = Transpiler.transpile(ex, variableSetSize, largestParameterSetSize, numVariableSets, i-1, kernelName) # i-1 because julia is 1-based but PTX needs 0-based indexing
# compiledKernels[i] = Transpiler.CompileKernel(ptxKernel, kernelName)
end
results = Matrix{Float32}(undef, numVariableSets, length(exprs))
for i in 1:repetitions # Simulate parameter tuning -> local search (X remains the same, p gets changed in small steps and must be performed sequentially, which it is with this impl)
# evaluate
# results = Transpiler.evaluate(exprs, variables, numVariableSets, variableSetSize, p)
results = Transpiler.evaluate(compiledKernels, variables, variableSetSize, p)
results = Transpiler.evaluate(ptxKernels, variables, variableSetSize, p, kernelName)
end
return results

5
package/src/Transpiler.jl
View File

@ -35,7 +35,7 @@ end
"
A simplified version of the evaluate function. It takes a list of already compiled kernels to be executed. This should yield better performance, where the same expressions should be evaluated multiple times i.e. for parameter optimisation.
"
function evaluate(kernels::Vector{CuFunction}, cudaVars::CuArray{Float32}, nrOfVariableSets::Integer, parameters::Vector{Vector{Float32}})::Matrix{Float32}
function evaluate(kernels::Vector{String}, cudaVars::CuArray{Float32}, nrOfVariableSets::Integer, parameters::Vector{Vector{Float32}}, kernelName::String)::Matrix{Float32}
cudaParams = Utils.create_cuda_array(parameters, NaN32) # maybe make constant (see PerformanceTests.jl for more info)
@ -46,7 +46,8 @@ function evaluate(kernels::Vector{CuFunction}, cudaVars::CuArray{Float32}, nrOfV
blocks = cld(nrOfVariableSets, threads)
@inbounds Threads.@threads for i in eachindex(kernels)
cudacall(kernels[i], (CuPtr{Float32},CuPtr{Float32},CuPtr{Float32}), cudaVars, cudaParams, cudaResults; threads=threads, blocks=blocks)
compiledKernel = CompileKernel(kernel[i], kernelName)
cudacall(compiledKernel, (CuPtr{Float32},CuPtr{Float32},CuPtr{Float32}), cudaVars, cudaParams, cudaResults; threads=threads, blocks=blocks)
end
return cudaResults