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
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
|
#!/bin/bash
# Copyright (c) 2025, Intel Corporation
# SPDX-License-Identifier: BSD-3-Clause
show_usage() {
cat <<EOF
Usage: $0 [BASE_BUILD] [PEAK_BUILD] [TEST_NAME] [FILTER]
Run and compare ISPC benchmark performance between two builds.
ARGUMENTS:
BASE_BUILD Directory name of base build (default: avx512icl-x32)
PEAK_BUILD Directory name of peak build (default: avx512icl-x32-new)
TEST_NAME Name of the benchmark test to run (default: 08_masked_load_store)
FILTER Optional benchmark filter pattern (default: none)
EXAMPLES:
$0 # Run default comparison
$0 avx2-base avx2-peak 01_simple # Compare specific builds and test
$0 base peak 08_masked_load_store ".*load" # Run with benchmark filter
The script will:
1. Run the benchmark test on BASE_BUILD with 3 repetitions
2. Run the benchmark test on PEAK_BUILD with 3 repetitions
3. Compare results using Google Benchmark's compare.py tool
4. Show median performance differences
Output files:
- {BASE_BUILD}-{TEST_NAME}.json
- {PEAK_BUILD}-{TEST_NAME}.json
Requirements:
- Both build directories must exist with compiled benchmarks
- Google Benchmark compare.py tool must be available
- Tests run with CPU affinity (taskset) and ASLR disabled (setarch -R)
EOF
}
if [[ "${1:-}" == "-h" || "${1:-}" == "--help" ]]; then
show_usage
exit 0
fi
BASE=${1:-avx512icl-x32}
PEAK=${2:-avx512icl-x32-new}
TEST=${3:-08_masked_load_store}
FILTER=${4:-}
SCRIPTS_DIR=$(cd "$(dirname "$0")" && pwd)
COMPARE_PY="$SCRIPTS_DIR/../benchmarks/vendor/google/benchmark/tools/compare.py"
# Build filter argument if provided
FILTER_ARG=""
if [ -n "$FILTER" ]; then
FILTER_ARG="--benchmark_filter=$FILTER"
fi
setarch -R x86_64 taskset --cpu-list 1 ./$BASE/bin/$TEST --benchmark_repetitions=3 --benchmark_out=$BASE-$TEST.json --benchmark_out_format=json --benchmark_report_aggregates_only=true $FILTER_ARG
setarch -R x86_64 taskset --cpu-list 1 ./$PEAK/bin/$TEST --benchmark_repetitions=3 --benchmark_out=$PEAK-$TEST.json --benchmark_out_format=json --benchmark_report_aggregates_only=true $FILTER_ARG
python3 $COMPARE_PY benchmarks ./$BASE-$TEST.json ./$PEAK-$TEST.json | grep median
|
