NVIDIA/nvbench
1
0
Fork 0
mirror of https://github.com/NVIDIA/nvbench.git synced 2026-04-20 14:58:54 +00:00
Code Issues Packages Projects Releases Wiki Activity
156 Commits 8 Branches 5 Tags
b1269b2f7f24c60c2ca19246271ed47b0cbe3d39
Commit Graph

156 Commits

This Branch
This Branch
All Branches
Author SHA1 Message Date
Allison Vacanti
b1269b2f7f Add json_printer. 
The new dependency emits some warnings from it's header. We'll need
to suppress / fix those.
 2021-03-05 10:50:41 -05:00
Allison Vacanti
acf076ec02 Add missing [[nodiscard]]. 2021-03-04 23:24:51 -05:00
Allison Vacanti
33fa0c773f Typo. 2021-03-04 23:24:37 -05:00
Allison Vacanti
65bc2c1e3f Documentation overhaul. 
Revamp README, split into multiple files. Add docs on CLI.

Add `--help` and `--help-axis`.
 2021-03-04 18:40:23 -05:00
Allison Vacanti
92fb4c7486 Link to NVBench Demo from README. 2021-03-04 16:26:01 -05:00
Allison Vacanti
b8876db858 Added CODE_OF_CONDUCT.md. 2021-03-04 16:17:53 -05:00
Allison Vacanti
430d5cedee Add .clang-format. 2021-03-04 16:03:11 -05:00
Allison Vacanti
078919b6b4 Add LICENSE file and boilerplate. 
for i in `find  ./ -regex ".*\.cuh?$"`
do
  sh -c "cat ~/boilerplate ${i} > tmp && mv tmp ${i}"
done
 2021-03-04 16:03:02 -05:00
Allison Vacanti
b1327b5f2b Expand summary documentation. 2021-03-04 07:19:37 -05:00
Allison Vacanti
ba6c8075c1 Update summary name to current conventions. 2021-03-04 07:19:18 -05:00
Allison Vacanti
b562934a46 Style fix for enum. 2021-03-04 07:00:14 -05:00
Allison Vacanti
a588af92fa Simplify file opening logic. 2021-03-03 21:37:09 -05:00
Allison Vacanti
b858b2c06d Update README example to use current macro names. 2021-03-03 21:01:05 -05:00
Allison Vacanti
47c69b83c9 More README cleanup. 2021-03-03 17:29:15 -05:00
Allison Vacanti
d7c34c835d More README cleanup. 2021-03-03 17:25:12 -05:00
Allison Vacanti
21e13f002d Fix error in README. 2021-03-03 17:22:35 -05:00
Allison Vacanti
544deaf539 Fix warning. 2021-03-03 16:57:17 -05:00
Allison Vacanti
75439d3ef8 Fix device global arg storage. 2021-03-03 16:49:43 -05:00
Allison Vacanti
9ff22cb12d Update README to use current macro names. 2021-03-03 16:30:45 -05:00
Allison Vacanti
a6b26ef7be Add initial README.md. 2021-03-03 16:00:11 -05:00
Allison Vacanti
cf71f6ee15 Update NVBench build system with initial standalone support. 2021-03-03 13:59:29 -05:00
Allison Vacanti
2ce12d2201 More printer cleanups. 
- Initialize color from env_var.
- Change ivar to m_global_benchmark_args to clarify usage.
 2021-03-02 17:36:19 -05:00
Allison Vacanti
3f3c648358 Update option_parser for recent refactorings. 2021-03-02 17:13:52 -05:00
Allison Vacanti
c83bf8cdb8 s/output_multiplex/printer_multiplex/g 2021-03-02 17:12:22 -05:00
Allison Vacanti
9d2194f2ab s/csv_format/csv_printer/g 2021-03-02 17:10:09 -05:00
Allison Vacanti
015b8d1fb1 s/markdown_format/markdown_printer/g 2021-03-02 17:08:54 -05:00
Allison Vacanti
780dc3b649 s/output_format/printer_base/g 2021-03-02 17:06:13 -05:00
Allison Vacanti
4a1e670f50 Use a more robust method to add the stdout printer, add --quiet. 2021-03-02 16:57:03 -05:00
Allison Vacanti
9cd0d10fe1 Route log messages through output formats. 2021-03-02 16:36:28 -05:00
Allison Vacanti
52fbbbcc7a Add --markdown / --csv options to option_parser. 2021-03-01 17:13:11 -05:00
Allison Vacanti
630aefda93 Make output_format explicitly move-only. 2021-03-01 17:10:33 -05:00
Allison Vacanti
0112993de2 Add output_multiplex::get_output_count. 2021-03-01 17:10:13 -05:00
Allison Vacanti
6c28a6a791 Refactor to keep stream includes out of headers. 2021-03-01 17:09:56 -05:00
Allison Vacanti
33a069af2b Add output_multiplex. 
This allows an arbitrary number of output_formats to be wrapped up
into a single object. This output format just forwards all calls to its
children.
 2021-03-01 15:16:43 -05:00
Allison Vacanti
14d41bb7e1 Add initial implementation of csv_format. 2021-02-25 16:10:28 -05:00
Allison Vacanti
17db5d31cc Split table_builder out from markdown_table. 2021-02-25 16:09:06 -05:00
Allison Vacanti
359db2c592 Initial pass at output_format.cuh, ported markdown_format. 2021-02-23 16:28:30 -05:00
Allison Vacanti
8d6d934dfe Add default axis names. 
Also cleaned up the annoying quirk where `set_type_axes_names` *had*
to be called on all benchmarks with type axes.

Default names are {"T", "U", "V", "W"} for up-to four type axes. For
five or more, {"T0", "T1", ...} is used instead.
 2021-02-19 12:37:05 -05:00
Allison Vacanti
324b0d107e Add "global args" to option parser. 
If a benchmark modifier is passed before `--benchmark`, the modifier
will apply to all benchmarks.
 2021-02-19 10:37:00 -05:00
Allison Vacanti
a747982415 Add nvbench::main CMake target. 
Linking to this instead of `nvbench::nvbench` will automatically include
the `NVBENCH_MAIN` macro.
 2021-02-19 09:34:02 -05:00
Allison Vacanti
2cc9bf41e3 Add demangle<T>() convenience overload. 2021-02-18 23:41:49 -05:00
Allison Vacanti
543488ef75 Make kernel wrapper into an lvalue. 2021-02-18 23:41:23 -05:00
Allison Vacanti
b5443e98c8 Use std::size_t for element counts, buffer size metadata. 2021-02-18 18:23:14 -05:00
Allison Vacanti
7dd46b0021 Update old benchmarks to use nvbench, remove old scaffolding. 
Remove the original attempt to adapt gbench to do CUDA stuff.

Update all benchmarks to use some conventions:

- Element count -> "Elements" [16:32]
- Throughput calcs
- Add input buffer column: "Size"
 2021-02-18 18:22:50 -05:00
Allison Vacanti
7657036f9c Add helper methods to configure throughput. 
Instead of:

```
state.set_element_count(size);
state.set_global_memory_bytes_accessed(
  size * (sizeof(InT) + sizeof(OutT)));
```

do:

```
state.add_element_count(size, "Elements");
state.add_global_memory_read<InT>(size, "InputSize");
state.add_global_memory_write<InT>(size, "OutputSize");
```

The string arguments are optional. If provided, a new column will
be added to the output with the indicated name and number
of bytes (or elements for `add_element_count`).
 2021-02-18 15:47:59 -05:00
Allison Vacanti
dcd5d1ffa6 Update markdown output format. 2021-02-18 14:44:17 -05:00
Allison Vacanti
ef3e1594eb Implement manual timers. 
See the new thrust/sort/basic.cu benchmark for usage.

Other notable changes:

- Updated summary column names:
  - Cold GPU -> GPU Time
  - Cold CPU -> CPU Time
  - Hot GPU  -> Batch GPU
- Removed CPU timings from measure_hot
  - They'd been hidden for a while, and aren't really useful.
- Moved the throughput calcs to measure_cold
  - `timer` will disable `hot` timings, still want throughput
  - `cold` timings make more sense for throughput, global BW numbers
    are meaningless if the data is sitting in L2.
 2021-02-17 18:48:26 -05:00
Allison Vacanti
385d4f77ba Teach markdown_format about sample_sizes. 2021-02-17 18:34:35 -05:00
Allison Vacanti
8a1f017a4e Inline some methods used in benchmark loops. 2021-02-17 18:34:09 -05:00
Allison Vacanti
f61be70a93 Add initial implementation of exec_tag dispatching. 
nvbench::exec_tags are used to request measurement types and share
information about the kernel. They are used to ensure that templated
measurement code is not instantiated unless actually used.

Replaces the nvbench::exec(state, launcher, tags) pattern with:

state.exec(tags, launcher);
state.exec(launcher); // defaults to hot/cold cuda measurements
 2021-02-16 23:47:36 -05:00