Update to non-avx builds

This commit is contained in:
Pro7ech
2025-11-21 15:39:04 +01:00
parent 0fb88c9bd3
commit 3c818d292b
24 changed files with 356 additions and 128 deletions

View File

@@ -1,8 +1,51 @@
# 🐙 Poulpy-CPU-REF
# 🐙 Poulpy-CPU-AVX
**Poulpy-Backend-CPU-AVX** is a Rust crate that provides an AVX accelerated CPU implementation of **`poulpy-hal`**. This crate is used to instantiate projects implemented with **`poulpy-hal`**, **`poulpy-core`** and/or **`poulpy-schemes`**.
**Poulpy-CPU-AVX** is a Rust crate that provides an **AVX2 + FMA accelerated CPU backend for Poulpy**.
## Example
This backend implements the Poulpy HAL extension traits and can be used by:
- [`poulpy-hal`](https://github.com/phantomzone-org/poulpy/tree/main/poulpy-hal)
- [`poulpy-core`](https://github.com/phantomzone-org/poulpy/tree/main/poulpy-core)
- [`poulpy-schemes`](https://github.com/phantomzone-org/poulpy/tree/main/poulpy-schemes)
## 🚩 Safety and Requirements
To avoid illegal hardware instructions (SIGILL) on unsupported CPUs, this backend is **opt-in** and **only builds when explicitly requested**.
| Requirement | Status |
|------------|--------|
| Cargo feature flag | `--features enable-avx` **must be enabled** |
| CPU architecture | `x86_64` |
| CPU target features | `AVX2` + `FMA` |
If `enable-avx` is enabled but the target does not provide these capabilities, the build **fails immediately with a clear error message**, rather than generating invalid binaries.
When `enable-avx` is **not** enabled, this crate is simply skipped and Poulpy automatically falls back to the portable `poulpy-cpu-ref` backend. This ensure that Poulpy's workspace remains portable (e.g. for macOS ARM).
## ⚙️ Building with the AVX backend enabled
Because the compiler must generate AVX2 + FMA instructions, both the Cargo feature and CPU target flags must be specified:
```bash
RUSTFLAGS="-C target-feature=+avx2,+fma" \
cargo build --features enable-avx
````
### Running an example
```bash
RUSTFLAGS="-C target-feature=+avx2,+fma" \
cargo run --example <name> --features enable-avx
```
### Running benchmarks
```bash
RUSTFLAGS="-C target-feature=+avx2,+fma" \
cargo bench --features enable-avx
```
## Basic Usage
```rust
use poulpy_backend_cpu_avx::FFT64Avx;
@@ -12,7 +55,24 @@ let log_n: usize = 10;
let module = Module<FFT64Avx> = Module<FFT64Avx>::new(1<<log_n);
```
## Contributors
Once compiled with `enable-avx`, the backend is usable transparently anywhere Poulpy expects a backend type (`poulpy-hal`, `poulpy-core`, `poulpy-schemes`).
To add your own backend, implement the open extension traits from **`poulpy-hal/oep`** for a struct that implements the `Backend` trait.
This will automatically make your backend compatible with the API of **`poulpy-hal`**, **`poulpy-core`** and **`poulpy-schemes`**.
## 🤝 Contributors
To implement your own Poulpy backend (SIMD or accelerator):
1. Define a backend struct
2. Implement the open extension traits from `poulpy-hal/oep`
3. Implement the `Backend` trait
Your backend will automatically integrate with:
* `poulpy-hal`
* `poulpy-core`
* `poulpy-schemes`
No modifications to those crates are required — the HAL provides the extension points.
---
For questions or guidance, feel free to open an issue or discussion in the repository.