|
1 | 1 | # Parsec Benchmark
|
2 | 2 |
|
3 |
| -<http://parsec.cs.princeton.edu/> 3.0-beta-20150206 ported to Ubuntu 16.04 and with proper version control. |
| 3 | +<http://parsec.cs.princeton.edu/> 3.0-beta-20150206 ported to Ubuntu 16.04 and SPLASH2 ported to Buildroot 2017.08 cross compilation (ARM, MIPS, etc.). |
4 | 4 |
|
5 |
| -## Getting started |
| 5 | +## Getting started Buildroot cross compilation |
| 6 | + |
| 7 | +See the instructions at: <https://github.com/cirosantilli/linux-kernel-module-cheat#parsec-benchmark> The Buildroot package is in that repo. |
| 8 | + |
| 9 | +Only SPLASH2 was ported currently, not the other benchmarks. |
| 10 | + |
| 11 | +PARSEC's build was designed for multiple archs, this can be seen at [bin/parsecmgmt](bin/parsecmgmt), but not for cross compilation. Some of the changes we've had to make: |
| 12 | + |
| 13 | +- use `CC` everywhere instead of hardcoded `gcc` |
| 14 | +- use `HOST_CC` for `.c` utilities used during compilation |
| 15 | +- remove absolute paths, e.g. `-I /usr/include` |
| 16 | + |
| 17 | +## Getting started Ubuntu 16.04 |
6 | 18 |
|
7 | 19 | ./configure
|
8 | 20 |
|
9 |
| -Build all: |
| 21 | +Before doing anything else, you must get the `parecmgmt` command with: |
10 | 22 |
|
11 | 23 | . env.sh
|
12 |
| - parsecmgmt -a build -p all |
13 | 24 |
|
14 |
| -Build just one to save time: |
| 25 | +Build all: |
15 | 26 |
|
16 |
| - parsecmgmt -a build -p splash2 |
| 27 | + parsecmgmt -a build -p all |
17 | 28 |
|
18 |
| -Run all: |
| 29 | +Build all SPLASH2 benchmarks: |
19 | 30 |
|
20 |
| - parsecmgmt -a run -p all |
| 31 | + parsecmgmt -a build -p splash2x |
21 | 32 |
|
22 |
| -TODO some tests are broken. We will maintain a list. |
| 33 | +Build just one SPLASH2 benchmark: |
| 34 | + |
| 35 | + parsecmgmt -a build -p splash2x.barnes |
23 | 36 |
|
24 | 37 | Run one benchmark with one input size, listed in by increasing size:
|
25 | 38 |
|
26 |
| - parsecmgmt -a run -p splash2 -i simsmall |
27 |
| - parsecmgmt -a run -p splash2 -i simmedium |
28 |
| - parsecmgmt -a run -p splash2 -i simlarge |
29 |
| - parsecmgmt -a run -p splash2 -i native |
| 39 | + parsecmgmt -a run -p splash2x.barnes -i test |
| 40 | + parsecmgmt -a run -p splash2x.barnes -i simdev |
| 41 | + parsecmgmt -a run -p splash2x.barnes -i simsmall |
| 42 | + parsecmgmt -a run -p splash2x.barnes -i simmedium |
| 43 | + parsecmgmt -a run -p splash2x.barnes -i simlarge |
| 44 | + parsecmgmt -a run -p splash2x.barnes -i native |
| 45 | + |
| 46 | +For some reason, the `splash2` version (without the X) does not have any test data besides `-i test`, making it basically useless. So just use the X version instead. TODO why? Can we just remove it then? When running `splash2`, it says: |
| 47 | + |
| 48 | + NOTE: SPLASH-2 only supports "test" input sets. |
| 49 | + |
| 50 | +so likely not a bug. |
| 51 | + |
| 52 | +The tests are distributed separately as: |
| 53 | + |
| 54 | +* `test` tests come with the smallest possible distribution `core`, and are tiny sanity checks as the name suggests. We have however removed them from this repo, since they are still blobs, and blobs are evil. |
| 55 | +* `sim*` tests require `parsec-3.0-input-sim.tar.gz` which we install by default |
| 56 | +* `native` requires `parsec-3.0-input-native.tar.gz`, which we don't install by default because it is huge. These huge instances are intended for real silicon. |
| 57 | + |
| 58 | +Run all packages with the default `test` input size: |
| 59 | + |
| 60 | + parsecmgmt -a run -p all |
| 61 | + |
| 62 | +TODO some tests are broken. We will maintain a list. |
30 | 63 |
|
31 |
| -`sim*` tests require `parsec-3.0-input-sim.tar.gz` which we install by default, and `native` requires `parsec-3.0-input-native.tar.gz`, which we don't install by default. |
| 64 | +Not every benchmark has every input size, e.g. `splash2.barnes` only has `test` input inside of `core` and `input-sim` |
32 | 65 |
|
33 | 66 | TODO runs all sizes, or just one default size:
|
34 | 67 |
|
35 |
| - parsecmgmt -a run -p splash2 |
| 68 | + parsecmgmt -a run -p splash2x |
36 | 69 |
|
37 | 70 | TODO how to read run output?
|
38 | 71 |
|
|
0 commit comments