Skip to content

Improve benchmark anchors #228

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
mborland merged 1 commit into from
Jul 6, 2026
+59 −1
Merged

Improve benchmark anchors #228

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
33 changes: 33 additions & 0 deletions doc/modules/ROOT/nav.adoc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -54,4 +54,37 @@
** xref:functional.adoc[]
* xref:comparisons.adoc[]
* xref:benchmarks.adoc[]
** xref:benchmarks.adoc#run_benchmarks_[How to run the Benchmarks]
** xref:benchmarks.adoc#benchmarks_methodology[Methodology]
** Linux
*** xref:benchmarks.adoc#linux_x64_benchmarks[x64]
**** xref:benchmarks.adoc#linux_x64_floating_point[Floating-Point]
**** xref:benchmarks.adoc#linux_x64_signed_integers[Signed Integers]
**** xref:benchmarks.adoc#linux_x64_unsigned_integers[Unsigned Integers]
*** xref:benchmarks.adoc#linux_x86_32_benchmarks[x86 (32-bit)]
**** xref:benchmarks.adoc#linux_x86_32_floating_point[Floating-Point]
**** xref:benchmarks.adoc#linux_x86_32_signed_integers[Signed Integers]
**** xref:benchmarks.adoc#linux_x86_32_unsigned_integers[Unsigned Integers]
*** xref:benchmarks.adoc#linux_arm64_benchmarks[ARM64]
**** xref:benchmarks.adoc#linux_arm64_floating_point[Floating-Point]
**** xref:benchmarks.adoc#linux_arm64_signed_integers[Signed Integers]
**** xref:benchmarks.adoc#linux_arm64_unsigned_integers[Unsigned Integers]
** macOS
*** xref:benchmarks.adoc#macos_arm64_benchmarks[ARM64]
**** xref:benchmarks.adoc#macos_arm64_floating_point[Floating-Point]
**** xref:benchmarks.adoc#macos_arm64_signed_integers[Signed Integers]
**** xref:benchmarks.adoc#macos_arm64_unsigned_integers[Unsigned Integers]
** Windows
*** xref:benchmarks.adoc#windows_x64_benchmarks[x64]
**** xref:benchmarks.adoc#windows_x64_floating_point[Floating-Point]
**** xref:benchmarks.adoc#windows_x64_signed_integers[Signed Integers]
**** xref:benchmarks.adoc#windows_x64_unsigned_integers[Unsigned Integers]
*** xref:benchmarks.adoc#windows_x86_32_benchmarks[x86 (32-bit)]
**** xref:benchmarks.adoc#windows_x86_32_floating_point[Floating-Point]
**** xref:benchmarks.adoc#windows_x86_32_signed_integers[Signed Integers]
**** xref:benchmarks.adoc#windows_x86_32_unsigned_integers[Unsigned Integers]
*** xref:benchmarks.adoc#windows_arm64_benchmarks[ARM64]
**** xref:benchmarks.adoc#windows_arm64_floating_point[Floating-Point]
**** xref:benchmarks.adoc#windows_arm64_signed_integers[Signed Integers]
**** xref:benchmarks.adoc#windows_arm64_unsigned_integers[Unsigned Integers]
* xref:reference.adoc[]
24 changes: 23 additions & 1 deletion doc/modules/ROOT/pages/benchmarks.adoc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -22,8 +22,8 @@ Because the baseline differs by width, every table states the type it is measure

NOTE: Boost.SafeNumerics has no floating-point type and no 128-bit type, so the floating-point and 128-bit tables list only the built-in type and Boost.SafeNumbers. See xref:comparisons.adoc[] for the details of that limitation.

== How to run the Benchmarks
[#run_benchmarks_]
== How to run the Benchmarks

The benchmarks live in `test/benchmarks/` and are built with https://github.com/google/benchmark[Google Benchmark].
They are not part of the default build: configure with `-DBOOST_SAFE_NUMBERS_BUILD_BENCHMARKS=ON`, which pulls Google Benchmark in with CMake `FetchContent` so no system install is required.
Expand All @@ -42,6 +42,7 @@ cmake --build __build__ --config Release \

The `benchmarks.yml` GitHub Actions workflow runs the same build across a spread of native runners, emits each suite's JSON, and renders the per-platform sections below with `test/benchmarks/render_results.py`.

[#benchmarks_methodology]
== Methodology

Each benchmark fills a vector of 1,000,000 random values once, outside the timed region, and then for every operation evaluates `op(vec[i], vec[i + 1])` across the whole vector.
Expand All @@ -65,6 +66,7 @@ NOTE: The numbers come from shared CI runners and reflect a single run each, so

Run on the GitHub Actions `ubuntu-latest` runner using GCC 14 in release mode (`-O2`, pass:[C++]20).

[#linux_x64_floating_point]
=== Floating-Point

.32-bit floating point (ratio relative to `float`)
Expand All @@ -87,6 +89,7 @@ Run on the GitHub Actions `ubuntu-latest` runner using GCC 14 in release mode (`
| Division | 1.41 | 1.42 | 1.00
|===

[#linux_x64_signed_integers]
=== Signed Integers

.8-bit signed (ratio relative to `std::int8_t`)
Expand Down Expand Up @@ -144,6 +147,7 @@ Run on the GitHub Actions `ubuntu-latest` runner using GCC 14 in release mode (`
| Modulo | 6.56 | 7.19 | 1.10
|===

[#linux_x64_unsigned_integers]
=== Unsigned Integers

.8-bit unsigned (ratio relative to `std::uint8_t`)
Expand Down Expand Up @@ -206,6 +210,7 @@ Run on the GitHub Actions `ubuntu-latest` runner using GCC 14 in release mode (`

Run on the GitHub Actions `ubuntu-latest` runner using GCC 14 targeting 32-bit x86 (`-m32`) in release mode (`-O2`, pass:[C++]20).

[#linux_x86_32_floating_point]
=== Floating-Point

.32-bit floating point (ratio relative to `float`)
Expand All @@ -228,6 +233,7 @@ Run on the GitHub Actions `ubuntu-latest` runner using GCC 14 targeting 32-bit x
| Division | 6.32 | 6.21 | 0.98
|===

[#linux_x86_32_signed_integers]
=== Signed Integers

.8-bit signed (ratio relative to `std::int8_t`)
Expand Down Expand Up @@ -285,6 +291,7 @@ Run on the GitHub Actions `ubuntu-latest` runner using GCC 14 targeting 32-bit x
| Modulo | 13.47 | 16.47 | 1.22
|===

[#linux_x86_32_unsigned_integers]
=== Unsigned Integers

.8-bit unsigned (ratio relative to `std::uint8_t`)
Expand Down Expand Up @@ -347,6 +354,7 @@ Run on the GitHub Actions `ubuntu-latest` runner using GCC 14 targeting 32-bit x

Run on the GitHub Actions `ubuntu-24.04-arm` runner using GCC 14 in release mode (`-O2`, pass:[C++]20).

[#linux_arm64_floating_point]
=== Floating-Point

.32-bit floating point (ratio relative to `float`)
Expand All @@ -369,6 +377,7 @@ Run on the GitHub Actions `ubuntu-24.04-arm` runner using GCC 14 in release mode
| Division | 1.48 | 1.75 | 1.19
|===

[#linux_arm64_signed_integers]
=== Signed Integers

.8-bit signed (ratio relative to `std::int8_t`)
Expand Down Expand Up @@ -426,6 +435,7 @@ Run on the GitHub Actions `ubuntu-24.04-arm` runner using GCC 14 in release mode
| Modulo | 5.04 | 5.18 | 1.03
|===

[#linux_arm64_unsigned_integers]
=== Unsigned Integers

.8-bit unsigned (ratio relative to `std::uint8_t`)
Expand Down Expand Up @@ -488,6 +498,7 @@ Run on the GitHub Actions `ubuntu-24.04-arm` runner using GCC 14 in release mode

Run on the GitHub Actions `macos-latest` runner (Apple Silicon) using Apple Clang in release mode (`-O2`, pass:[C++]20).

[#macos_arm64_floating_point]
=== Floating-Point

.32-bit floating point (ratio relative to `float`)
Expand All @@ -510,6 +521,7 @@ Run on the GitHub Actions `macos-latest` runner (Apple Silicon) using Apple Clan
| Division | 3.64 | 3.69 | 1.01
|===

[#macos_arm64_signed_integers]
=== Signed Integers

.8-bit signed (ratio relative to `std::int8_t`)
Expand Down Expand Up @@ -567,6 +579,7 @@ Run on the GitHub Actions `macos-latest` runner (Apple Silicon) using Apple Clan
| Modulo | 7.20 | 7.92 | 1.10
|===

[#macos_arm64_unsigned_integers]
=== Unsigned Integers

.8-bit unsigned (ratio relative to `std::uint8_t`)
Expand Down Expand Up @@ -629,6 +642,7 @@ Run on the GitHub Actions `macos-latest` runner (Apple Silicon) using Apple Clan

Run on the GitHub Actions `windows-latest` runner using MSVC in release mode (`/O2`, pass:[C++]20).

[#windows_x64_floating_point]
=== Floating-Point

.32-bit floating point (ratio relative to `float`)
Expand All @@ -651,6 +665,7 @@ Run on the GitHub Actions `windows-latest` runner using MSVC in release mode (`/
| Division | 3.82 | 3.91 | 1.02
|===

[#windows_x64_signed_integers]
=== Signed Integers

.8-bit signed (ratio relative to `std::int8_t`)
Expand Down Expand Up @@ -708,6 +723,7 @@ Run on the GitHub Actions `windows-latest` runner using MSVC in release mode (`/
| Modulo | 15.63 | 17.36 | 1.11
|===

[#windows_x64_unsigned_integers]
=== Unsigned Integers

.8-bit unsigned (ratio relative to `std::uint8_t`)
Expand Down Expand Up @@ -770,6 +786,7 @@ Run on the GitHub Actions `windows-latest` runner using MSVC in release mode (`/

Run on the GitHub Actions `windows-latest` runner using MSVC targeting 32-bit x86 in release mode (`/O2`, pass:[C++]20).

[#windows_x86_32_floating_point]
=== Floating-Point

.32-bit floating point (ratio relative to `float`)
Expand All @@ -792,6 +809,7 @@ Run on the GitHub Actions `windows-latest` runner using MSVC targeting 32-bit x8
| Division | 3.82 | 17.36 | 4.55
|===

[#windows_x86_32_signed_integers]
=== Signed Integers

.8-bit signed (ratio relative to `std::int8_t`)
Expand Down Expand Up @@ -849,6 +867,7 @@ Run on the GitHub Actions `windows-latest` runner using MSVC targeting 32-bit x8
| Modulo | 26.04 | 31.25 | 1.20
|===

[#windows_x86_32_unsigned_integers]
=== Unsigned Integers

.8-bit unsigned (ratio relative to `std::uint8_t`)
Expand Down Expand Up @@ -911,6 +930,7 @@ Run on the GitHub Actions `windows-latest` runner using MSVC targeting 32-bit x8

Run on the GitHub Actions `windows-11-arm` runner using MSVC in release mode (`/O2`, pass:[C++]20).

[#windows_arm64_floating_point]
=== Floating-Point

.32-bit floating point (ratio relative to `float`)
Expand All @@ -933,6 +953,7 @@ Run on the GitHub Actions `windows-11-arm` runner using MSVC in release mode (`/
| Division | 2.08 | 3.65 | 1.75
|===

[#windows_arm64_signed_integers]
=== Signed Integers

.8-bit signed (ratio relative to `std::int8_t`)
Expand Down Expand Up @@ -990,6 +1011,7 @@ Run on the GitHub Actions `windows-11-arm` runner using MSVC in release mode (`/
| Modulo | 7.81 | 8.33 | 1.07
|===

[#windows_arm64_unsigned_integers]
=== Unsigned Integers

.8-bit unsigned (ratio relative to `std::uint8_t`)
Expand Down
3 changes: 3 additions & 0 deletions test/benchmarks/render_results.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -205,20 +205,23 @@ def render_section(title, anchor, desc, unsigned_json, signed_json, float_json):

if float_json:
data = load_ns_per_op(float_json)
out.append(f"[#{anchor}_floating_point]")
out.append("=== Floating-Point")
out.append("")
for bits in FLOAT_BITS:
emit_float_table(out, data, bits)

if signed_json:
data = load_ns_per_op(signed_json)
out.append(f"[#{anchor}_signed_integers]")
out.append("=== Signed Integers")
out.append("")
for bits in INT_BITS:
emit_int_table(out, data, "signed", bits)

if unsigned_json:
data = load_ns_per_op(unsigned_json)
out.append(f"[#{anchor}_unsigned_integers]")
out.append("=== Unsigned Integers")
out.append("")
for bits in INT_BITS:
Expand Down
Loading