1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
|
#!/bin/sh
data=$(mktemp -d)
echo
echo SDK with transformation
echo
DFATOOL_ULS_MIN_BOUND=0 DFATOOL_PARAM_RELEVANCE_THRESHOLD=0.9 DFATOOL_ULS_FUNCTIONS=roofline,linear \
analyze-log.py \
--filter-param='n_elements_per_dpu=1048576' \
--ignore-param='n_dpus,n_nops,numa_node_cpu,numa_node_in,numa_node_out,numa_node_rank,n_elements' \
--filter-observation='NMC-transfer:throughput_dram_mram_Bps,NMC-transfer:throughput_mram_dram_Bps' \
--normalize-nfp='writeThroughputGBps=throughput_dram_mram_Bps=/1e9;readThroughputGBps=throughput_mram_dram_Bps=/1e9' \
--export-pgf-unparam ${data}/orig- \
--cross-validate=kfold:10 --progress \
--show-model=param --show-model-error --show-model-precision=6 \
log/tinos/upvec-2025.1.0-orig.txt
echo
echo SDK without transformation
echo
DFATOOL_ULS_MIN_BOUND=0 DFATOOL_PARAM_RELEVANCE_THRESHOLD=0.9 DFATOOL_ULS_FUNCTIONS=roofline,linear \
analyze-log.py \
--filter-param='n_elements_per_dpu=1048576' \
--ignore-param='n_dpus,n_nops,numa_node_cpu,numa_node_in,numa_node_out,numa_node_rank,n_elements' \
--filter-observation='NMC-transfer:throughput_dram_mram_Bps,NMC-transfer:throughput_mram_dram_Bps' \
--normalize-nfp='writeThroughputGBps=throughput_dram_mram_Bps=/1e9;readThroughputGBps=throughput_mram_dram_Bps=/1e9' \
--export-pgf-unparam ${data}/notransform- \
--cross-validate=kfold:10 --progress \
--show-model=param --show-model-error --show-model-precision=6 \
log/tinos/upvec-2025.1.0-notransform.txt
for op in read write; do
cp util/upvec-${op}.tex ${data}
lualatex -output-directory ${data} ${data}/upvec-${op}
cp ${data}/upvec-${op}.pdf util
done
rm -rf ${data}
|
