rubydragon/metagraphoptimization.jl
1
0
Fork 0
Code Issues 6 Pull Requests Actions Packages Projects Releases Wiki Activity

experiments #1

Merged
rubydragon merged 39 commits from experiments into main 2024-05-08 12:03:28 +02:00
Conversation 0 Commits 39 Files Changed 155 +135 -32
4 changed files with 135 additions and 32 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files
Showing only changes of commit 4584285126 - Show all commits

80
examples/qed_bench_reduction_steps.jl
View File

@ -47,29 +47,29 @@ function bench(func, inputs)
)
end
# preparation of machine
machine = Machine(
[
MetagraphOptimization.NumaNode(
0,
1,
MetagraphOptimization.default_strategy(MetagraphOptimization.NumaNode),
-1.0,
UUIDs.uuid1(),
),
],
[-1.0;;],
)
# bench and produce data
n_inputs = 10_000
n_inputs = 50_000
optimizer = ReductionOptimizer()
processes = [("ke->kke", 50), ("ke->ke", 1), ("ke->kke", 1), ("ke->kkke", 5), ("ke->kkkke", 5), ("ke->kkkkke", 10), ("ke->kkkkkke", 20)]
processes = [("ke->kke", 5), ("ke->ke", 1), ("ke->kke", 1), ("ke->kkke", 1), ("ke->kkkke", 1), ("ke->kkkkke", 1)]
for (process_str, STEPSIZE) in processes
n = 0
opt_time_cum = 0
# preparation of graph
machine = Machine(
[
MetagraphOptimization.NumaNode(
0,
1,
MetagraphOptimization.default_strategy(MetagraphOptimization.NumaNode),
-1.0,
UUIDs.uuid1(),
),
],
[-1.0;;],
)
process = parse_process(process_str, QEDModel())
graph = gen_graph(process)
inputs = [gen_process_input(process) for _ in 1:n_inputs]
@ -113,3 +113,51 @@ for (process_str, STEPSIZE) in processes
end
CSV.write(results_filename, df)
for (process_str, STEPSIZE) in [("AB->AB", 1), ("AB->ABBB", 1), ("AB->ABBBBB", 1)]
n = 0
opt_time_cum = 0
process = parse_process(process_str, ABCModel())
graph = parse_dag("input/$process_str.txt", ABCModel())
inputs = [gen_process_input(process) for _ in 1:n_inputs]
get_compute_function(graph, process, machine)
while true
func_gen_time = @elapsed func = get_compute_function(graph, process, machine)
res = bench(func, inputs)
graph_properties = get_properties(graph)
push!(
df,
(
Threads.nthreads(),
process_str,
n,
opt_time_cum,
graph_properties.noNodes,
graph_properties.noEdges,
graph_properties.computeEffort,
graph_properties.data,
graph_properties.computeIntensity,
func_gen_time,
res.cpu_compile_time,
res.cpu_single_thread_time,
res.cpu_multi_thread_time,
res.gpu_compile_time,
res.gpu_time,
),
)
CSV.write(results_filename, df)
if fixpoint_reached(optimizer, graph)
break
end
opt_time_cum += @elapsed optimize!(optimizer, graph, STEPSIZE)
n += STEPSIZE
end
end
CSV.write(results_filename, df)

85
examples/qed_bench_reduction_steps_gpu.jl
View File

@ -50,30 +50,35 @@ function bench(func, inputs)
)
end
log("CUDA devices: $(CUDA.devices())")
for dev in CUDA.devices()
log("CUDA device: $(dev)")
end
# preparation of machine
machine = Machine(
[
MetagraphOptimization.NumaNode(
0,
1,
MetagraphOptimization.default_strategy(MetagraphOptimization.NumaNode),
-1.0,
UUIDs.uuid1(),
),
],
[-1.0;;],
)
# bench and produce data
n_inputs = 10_000
n_inputs = 50_000
optimizer = ReductionOptimizer()
processes = [("ke->kke", 50), ("ke->ke", 1), ("ke->kke", 1), ("ke->kkke", 1), ("ke->kkkke", 5)]
processes = [("ke->kke", 50), ("ke->ke", 1), ("ke->kke", 1), ("ke->kkke", 1)]
for (process_str, STEPSIZE) in processes
n = 0
opt_time_cum = 0
# preparation of graph
machine = Machine(
[
MetagraphOptimization.NumaNode(
0,
1,
MetagraphOptimization.default_strategy(MetagraphOptimization.NumaNode),
-1.0,
UUIDs.uuid1(),
),
],
[-1.0;;],
)
process = parse_process(process_str, QEDModel())
graph = gen_graph(process)
inputs = CuVector([gen_process_input(process) for _ in 1:n_inputs])
@ -117,3 +122,51 @@ for (process_str, STEPSIZE) in processes
end
CSV.write(results_filename, df)
for (process_str, STEPSIZE) in [("AB->AB", 1), ("AB->ABBB", 1), ("AB->ABBBBB", 1)]
n = 0
opt_time_cum = 0
process = parse_process(process_str, ABCModel())
graph = parse_dag("input/$process_str.txt", ABCModel())
inputs = CuVector([gen_process_input(process) for _ in 1:n_inputs])
get_compute_function(graph, process, machine)
while true
func_gen_time = @elapsed func = get_compute_function(graph, process, machine)
res = bench(func, inputs)
graph_properties = get_properties(graph)
push!(
df,
(
Threads.nthreads(),
process_str,
n,
opt_time_cum,
graph_properties.noNodes,
graph_properties.noEdges,
graph_properties.computeEffort,
graph_properties.data,
graph_properties.computeIntensity,
func_gen_time,
res.cpu_compile_time,
res.cpu_single_thread_time,
res.cpu_multi_thread_time,
res.gpu_compile_time,
res.gpu_time,
),
)
CSV.write(results_filename, df)
if fixpoint_reached(optimizer, graph)
break
end
opt_time_cum += @elapsed optimize!(optimizer, graph, STEPSIZE)
n += STEPSIZE
end
end
CSV.write(results_filename, df)

1
experiments/reduce_bench_hemera_gpu.sh
View File

@ -1,6 +1,7 @@
#!/bin/bash
#SBATCH --job-name=qed_bench
#SBATCH --partition=casus_a100
#SBATCH --account=casus
#SBATCH --time=48:00:00
#SBATCH --nodes=1
#SBATCH --ntasks=1

1
experiments/run_reduce_bench_gpu.sh
View File

@ -18,6 +18,7 @@ lspci > results/pci_bench_reduce_gpu.txt
#echo "Initiating julia..."
#julia --threads=8 --project=./ -e 'using Pkg; Pkg.instantiate(); Pkg.add(url="https://github.com/QEDjl-project/QEDprocesses.jl/")' >> $LOG_FILE 2>&1 || exit 1 # need current dev version of QEDprocesses
#julia --threads=8 -e 'using Pkg; Pkg.add("CSV"); Pkg.add("DataFrames"); Pkg.add("LIKWID"); Pkg.add("CUDA"); Pkg.add("Random"); Pkg.add("BenchmarkTools"); Pkg.add("Dates")' >> $LOG_FILE 2>&1 || exit 1 # add requirements for the bench script
julia --project -e 'using CUDA; CUDA.set_runtime_version!(VersionNumber("12.1"))' >> $LOG_FILE 2>&1 || echo "Failed to set CUDA version number"
echo "Benchmarking Reduction 32 Threads, *GPU*"
julia --project --threads=32 examples/qed_bench_reduction_steps_gpu.jl >> $LOG_FILE 2>&1 || echo "-- Something went wrong, check logs --"