coffee/fast_float
1
0
Fork 0
mirror of https://github.com/fastfloat/fast_float.git synced 2026-10-28 18:07:16 +08:00
Code Issues Packages Projects Releases Wiki Activity

Readme update.

Browse Source
This commit is contained in:
HedgehogInTheCPP 2026-09-23 03:54:18 +03:00
parent 2f67e94234
commit b5a2f54e30
2 changed files with 8 additions and 7 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

13
README.md
View File

@ -12,17 +12,16 @@
[![vs18 clang](https://github.com/irainman/fast_float/actions/workflows/vs18-clang-ci.yml/badge.svg)](https://github.com/irainman/fast_float/actions/workflows/vs18-clang-ci.yml)
[![CodeFactor](https://www.codefactor.io/repository/github/irainman/fast_float/badge)](https://www.codefactor.io/repository/github/irainman/fast_float)
## This is a fork of [fast_float](https://github.com/fastfloat/fast_float) made by HedgehogInTheCPP with some additional options, code refactoring and big cleanup to maximize performance, reduce size and improve mainteinability, please give a star to this repo ^^
## This is a fork of [fast_float](https://github.com/fastfloat/fast_float) made by HedgehogInTheCPP with some additional options, code refactoring and big cleanup to maximize performance, reduce size and improve mainteinability, please give a star to this repo ^^
### 🚀 Performance and Code Size Improvements
* Optimized parsing logic for use as an **internal lightweight parser** (for example, inside other libraries).
* Optimized parsing logic for use as an **internal lightweight parser** (for example, inside other libraries or in compilers).
* Reduced binary size by conditionally excluding unneeded parsing features.
* The `from_chars_result_t` structure is reduced to **4 bytes** for better memory efficiency.
* Improved the `parsed_number_string_t` layout and increased `constexpr`/`consteval` propagation to enable compile-time optimizations.
* Reduced register and cache pressure and branching in parsing hot paths.
* Reduced register and cache pressure and reduce branching especially in parsing hot paths.
* Improved performance in both 64-bit and 32-bit builds for all supported types.
* Added more optimized x86 specific code that uses up to SSE4.2 instructions in algorithms and significantly improve speed parsing especially for big numbers.
* Added more optimized x86 specific code that uses up to SSE4.2 instructions in algorithms and significantly improve speed parsing especially for numbers with many digits.
---
@ -31,10 +30,10 @@
Introduced new optional macros to minimize overhead when certain parsing features are not required:
* **`FASTFLOAT_ONLY_POSITIVE_C_NUMBER_WO_INF_NAN`**
Restricts parsing to **positive C-style numbers only** — no sign characters, no `INF`, `INFINITY`, or `NaN` and no additional options like skip white spaces or support for Fortran or JSON.
Restricts parsing to **positive C-style numbers only** — no sign characters, no `INF`, `INFINITY`, or `NaN` and no additional options like skip white spaces or support for Fortran/JSON/JS.
* **`FASTFLOAT_ONLY_ROUNDS_TO_NEAREST_SUPPORTED`**
Assumes that only the **IEEE 754 “round-to-nearest”** rounding mode is needed, removing crutches support code for other modes.
Assumes that only the **IEEE 754 “round-to-nearest”** rounding mode is needed, removing crutches and support code for other modes.
* **`FASTFLOAT_ISNOT_CHECKED_BOUNDS`**
Disables bounds checking for input ranges that are assumed to be valid.

2
tests/exhaustive16_midpoint.cpp
View File

@ -163,6 +163,8 @@ int main() {
if (f16.count != 0 || bf16.count != 0) {
return EXIT_FAILURE;
}
#else
std::printf("float16x type unsupported, so this test isn't applied\n");
#endif
std::printf("all ok\n");
return EXIT_SUCCESS;