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

benchmarking: moved frontend calls and sending postfixExprs+vars outside to drastically reduce amount of calculations
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
2025-05-17 18:32:04 +02:00
parent 88ee8d20bd
commit a5518dd63e
8 changed files with 79 additions and 63 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
package/test/PerformanceTests.jl
View File

@ -10,6 +10,7 @@ using .ExpressionProcessing
include("parser.jl") # to parse expressions from a file
# ATTENTAION: Evaluation information at the very bottom
const BENCHMARKS_RESULTS_PATH = "./results-fh-new"
# Number of expressions can get really big (into millions)
@ -68,7 +69,7 @@ suite["GPUT"]["nikuradse_1"] = @benchmarkable evaluate_gpu(exprs, X_t, parameter
loadparams!(suite, BenchmarkTools.load("params.json")[1], :samples, :evals, :gctrial, :time_tolerance, :evals_set, :gcsample, :seconds, :overhead, :memory_tolerance)
results = run(suite, verbose=true, seconds=28800) # 8 hour timeout
results = run(suite, verbose=true, seconds=43200) # 12 hour timeout
resultsCPU = BenchmarkTools.load("$BENCHMARKS_RESULTS_PATH/cpu.json")[1]
if compareWithCPU
@ -139,3 +140,8 @@ else
println(oldVsGPUT_std)
end
# Initial implementation:
# - Interpreter: no cache; 256 blocksize; exprs pre-processed and sent to GPU on every call; vars sent on every call; frontend + dispatch are multithreaded
# - Transpiler: no cahce; 256 blocksize; exprs pre-processed and transpiled on every call; vars sent on every call; frontend + transpilation + dispatch are multithreaded