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

transpiler: invalid memory access error finally fixed
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
2025-03-27 22:32:24 +01:00
parent 561b37160b
commit 9df78ca72e
7 changed files with 133 additions and 30 deletions
Download Patch File Download Diff File Expand all files Collapse all files

22
package/test/CpuInterpreterTests.jl
View File

@ -1,4 +1,5 @@
using LinearAlgebra
using BenchmarkTools
function test_cpu_interpreter(nrows; parallel = false)
exprs = [
@ -18,16 +19,27 @@ function test_cpu_interpreter(nrows; parallel = false)
reps= 100
if parallel
t_sec = @elapsed fetch.([Threads.@spawn interpret_cpu(exprs, X, p; repetitions=expr_reps) for i in 1:reps])
println("~ $(round(30 * reps * expr_reps * nrows / 1e9 / t_sec, digits=2)) GFLOPS ($(Threads.nthreads()) threads) ($(round(peakflops(1000, eltype=Float32, ntrials=1) / 1e9, digits=2)) GFLOPS (peak, single-core))")
# t_sec = @elapsed fetch.([Threads.@spawn interpret_cpu(exprs, X, p; repetitions=expr_reps) for i in 1:reps])
@btime parallel(exprs, X, p, expr_reps, reps)
println("~ $(round(30 * reps * expr_reps * nrows / 1e9 / t_sec, digits=2)) GFLOPS ($(Threads.nthreads()) threads) ($(round(LinearAlgebra.peakflops(1000, eltype=Float32, ntrials=1) / 1e9, digits=2)) GFLOPS (peak, single-core))")
else
t_sec = @elapsed for i in 1:reps interpret_cpu(exprs, X, p; repetitions=expr_reps) end
println("~ $(round(30 * reps * expr_reps * nrows / 1e9 / t_sec, digits=2)) GFLOPS (single-core) ($(round(peakflops(1000, eltype=Float32, ntrials=1) / 1e9, digits=2)) GFLOPS (peak, single-core))")
# t_sec = @elapsed for i in 1:reps interpret_cpu(exprs, X, p; repetitions=expr_reps) end
@btime single(exprs, X, p, expr_reps, reps)
println("~ $(round(30 * reps * expr_reps * nrows / 1e9 / t_sec, digits=2)) GFLOPS (single-core) ($(round(LinearAlgebra.peakflops(1000, eltype=Float32, ntrials=1) / 1e9, digits=2)) GFLOPS (peak, single-core))")
end
true
end
LinearAlgebra.BLAS.set_num_threads(1) # only use a single thread for peakflops
function parallel(exprs, X, p, expr_reps, reps)
fetch.([Threads.@spawn interpret_cpu(exprs, X, p; repetitions=expr_reps) for i in 1:reps])
end
function single(exprs, X, p, expr_reps, reps)
for i in 1:reps interpret_cpu(exprs, X, p; repetitions=expr_reps) end
end
# LinearAlgebra.BLAS.set_num_threads(1) # only use a single thread for peakflops
@test test_cpu_interpreter(1000)
@test test_cpu_interpreter(1000, parallel=true) # start julia -t 6 for six threads