rust-math / fftw Goto Github PK

View Code? Open in Web Editor NEW

54.0 4.0 23.0 4.68 MB

FFTW3 binding for Rust

License: Other

Rust 97.71% Shell 2.22% C 0.06%

fftw rust fftw3-binding science

fftw's People

Contributors

Stargazers

Watchers

Forkers

gdkar adamnemecek amilajack ajsyp maxk11 hombit tginick astrojhgu derekdreery songww tuxzz icetdrinker apere3 sinhasantos atouchet samcrow miili iq-scm skailasa anghd

fftw's Issues

fftw-src doesn't compile

Probably the problem is with fs_extra 1.2:

error[E0063]: missing field `content_only` in initializer of `fs_extra::dir::CopyOptions`
  --> fftw-src/build.rs:52:10
   |
52 |         &fs_extra::dir::CopyOptions {
   |          ^^^^^^^^^^^^^^^^^^^^^^^^^^ missing `content_only`

error: aborting due to previous error

For more information about this error, try `rustc --explain E0063`.
error: could not compile `fftw-src`.

Support system mkl

Please, can you add support of system mkl in the same way it made for FFTW? Probably it is already supported by intel-mkl-src 0.5.0

Support features=system on macOS

I don't know how to mange system libraries on macOS 😢

Support DHT, DCT, and DST in R2R

http://www.fftw.org/fftw3_doc/Real_002dto_002dReal-Transform-Kinds.html

Add static linking

Please, can you add a feature to allow static linking with FFTW/MKL?

helpers for half-complex

Would you accept a PR for helpers to handle half-complex arrays. I'm thinking for half-complex arrays x, y,

x * y
mod(x) or abs(x)
arg(x) (or phase(x) if preferred)

the signature could be e.g.

/// contents of y are replaced with the result
fn half_complex_mult<X>(x: &[X], y: &mut [X]) 
where X: f32 or f64 // (either define a trait or 2 functions)
{
    assert_eq!(x.len(), y.len());
    let n = x.len();

    y[0] = y[0] * x[0];
    for i in 1..=(n - 1) / 2 {
        let xre = x[i];
        let yre = y[i];
        let xim = x[n - i];
        let yim = y[n - i];
        y[i] = xre * yre - xim * yim;
        y[n - i] = xre * yim + xre * yim;
    }
    if n % 2 == 0 {
        let n2 = n / 2;
        y[n2] = y[n2] * x[n2];
    }
}

ArrayView interface for multi-dimensional FFT

Switch license of wrapper codes to MIT-license

Support for Windows builds with GNU ABI

This library is currently not compatible with toolchains like stable-x86_64-pc-windows-gnu. I wasn't able to modify the build script to support building with MSYS/MinGW, but it could be possible to offer pre-built binaries for these platforms.

Static linking on Windows

I never noticed this since I always use Linux myself, but while the Linux and macOS versions statically link to the built version of fftw by default, the Windows version always dynamically links to a downloaded binary. This makes distributing binaries built against this crate a lot more difficult on Windows. Is there an easy way to get it to statically link instead?

Benchmark comparing FFTW, Intel MKL, and RustFFT

Buffer: in-place transformation

Plan-based implementation in FFTW is unconformable with the borrow checker in Rust since fftw_execute(plan) modify in/out arrays.
The transformation in FFTW has two types: in-place and out-place. Current implementation using Pair corresponding to out-place conversion.
I will implement in-place by Buffer struct.

Upgrade intel-mkl-src to 0.6.0+mkl2020.1

https://github.com/rust-math/intel-mkl-src/releases/tag/0.6.0%2Bmkl2020.1

New-array Execute Functions

Current structure has a problem that the transform cannot be executed safely since the Plan change the array without calling with them. We will rewrite them using new-array execute functions
http://www.fftw.org/fftw3_doc/New_002darray-Execute-Functions.html#New_002darray-Execute-Functions
which get input/ouput array as arguments.

Typing mismatch in mod.rs and timeval.rs for macOS 2.2.0 installation

I'm working with the Pulse Audio binding to do a simple FFT visualizer. I'm getting the following traceback when trying to compile basic usage of the standard mainloop and context and stream stuff:

   Compiling libpulse-binding v2.2.0
error[E0308]: mismatched types
  --> /Users/noahbkim/.cargo/registry/src/github.com-1ecc6299db9ec823/libpulse-binding-2.2.0/src/time/mod.rs:55:35
   |
55 |         Timeval::new(secs as i64, usecs as i64)
   |                                   ^^^^^^^^^^^^ expected i32, found i64

error[E0308]: mismatched types
  --> /Users/noahbkim/.cargo/registry/src/github.com-1ecc6299db9ec823/libpulse-binding-2.2.0/src/time/mod.rs:74:42
   |
74 |         Timeval::new(t.as_secs() as i64, (t.subsec_nanos() / NANOS_PER_MILLI) as i64)
   |                                          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ expected i32, found i64

error[E0308]: mismatched types
   --> /Users/noahbkim/.cargo/registry/src/github.com-1ecc6299db9ec823/libpulse-binding-2.2.0/src/time/timeval.rs:100:41
    |
100 |             true => { self.0.tv_usec |= PA_TIMEVAL_RTCLOCK; },
    |                                         ^^^^^^^^^^^^^^^^^^ expected i32, found i64

error[E0277]: no implementation for `i32 |= i64`
   --> /Users/noahbkim/.cargo/registry/src/github.com-1ecc6299db9ec823/libpulse-binding-2.2.0/src/time/timeval.rs:100:38
    |
100 |             true => { self.0.tv_usec |= PA_TIMEVAL_RTCLOCK; },
    |                                      ^^ no implementation for `i32 |= i64`
    |
    = help: the trait `std::ops::BitOrAssign<i64>` is not implemented for `i32`

I'm working on the current version of macOS with a vanilla installation of Rust. Let me know if you need any other information.

travis not configured correctly

It seems that only WIP tests are running (not Travsi tests)

how to enable openmp when use fftw3?

thread 'main' panicked at 'Failed to check md5 sum

I'm new to Rust , so please over look my ignorance of some its functions. I have iMac osx 10.13.6
My issue is the following:error:
failed to run custom build command for fftw-src v0.2.2 (file:///Users/jim/Downloads/rust-fftw3-master-2/fftw-src)
process didn't exit successfully: /Users/jim/Downloads/rust-fftw3-master-2/target/debug/build/fftw-src-a110d0ce96cf96b8/build-script-build (exit code: 101)
--- stderr
thread 'main' panicked at 'Failed to check md5 sum: Os { code: 2, kind: NotFound, message: "No such file or directory" }', libcore/result.rs:945:5

Any help will be greatly appreciated.

ocipkg-based intel-mkl-src does not work on Windows

intel-mkl-src 0.8.0 (upgraded at #121) switched download source from AWS S3 to GitHub Container registry using ocipkg. It only works with static library to avoid runtime search path issue.

However, it causes runtime error:

  process didn't exit successfully: `D:\a\fftw\fftw\target\debug\deps\c2c-8b0385d3c022d1c2.exe` (exit code: 0xc0000135, STATUS_DLL_NOT_FOUND)

https://github.com/rust-math/fftw/actions/runs/4837801529/jobs/8621864073

This does not occur if we use NuGet to download MKL from NuGet registry, and add its path to runtime search path.

Some attempt has been done in #122, but not resolved.

Maybe related to #118

cargo test --doc cannot find DLL on Windows

https://github.com/rust-math/fftw/runs/1044484751

running 3 tests
test src\array.rs - array::alignment_of (line 109) ... FAILED
test src\lib.rs -  (line 10) ... FAILED
test src\lib.rs -  (line 29) ... FAILED

failures:

---- src\array.rs - array::alignment_of (line 109) stdout ----
Test executable failed (exit code -1073741515).

---- src\lib.rs -  (line 10) stdout ----
Test executable failed (exit code -1073741515).

---- src\lib.rs -  (line 29) stdout ----
Test executable failed (exit code -1073741515).

Test Coverage

Add codecov to CI to manage line coverage of tests

Thread safety

In the documentation of FFTW3 it is stated that the creation of a plan is not thread save.
http://www.fftw.org/fftw3_doc/Thread-safety.html
Is this binding thread save? I cant find any mechanism which would prevent two threads to create 2 Plans at the same time.

Document need for wget

Otherwise, calling cargo build will fail with an error that took me a few minutes to chase down:

error: failed to run custom build command for `fftw-src v0.2.1`
process didn't exit successfully: `/Users/duane/Projects/audio/spectrogram/target/debug/build/fftw-src-1e03addce5c52064/build-script-build` (exit code: 101)
--- stdout
Running: "wget" "http://www.fftw.org/fftw-3.3.6-pl1.tar.gz"

--- stderr
thread 'main' panicked at 'failed to execute `"wget" "http://www.fftw.org/fftw-3.3.6-pl1.tar.gz"`: No such file or directory (os error 2)', /Users/duane/.cargo/registry/src/github.com-1ecc6299db9ec823/fftw-src-0.2.1/build.rs:61
note: Run with `RUST_BACKTRACE=1` for a backtrace.

Note at first I thought the target file was missing, not wget.

Alternatively, you could just use curl on darwin.

Still Maintained?

This repo appears vacant. Is it still maintained? @termoshtt

Drop reqwest dependency

From #83

For Linux and macOS, #88 make possible to drop download phase
For Windows, however, downloading is still needed to get compiled binary. We have some choice:
- Also compile with GNU toolchain, and drop msvc support
- Use reqwest or more light downloader, e.g. curl crate which is used in intel-mkl-src if download feature is enabled, only for Windows
- or combine them (build for gnu, and download for msvc)

Bundle FFTW source code with fftw-src

Downloading the source code of FFTW on every compilation could be inefficient or even impossible in some environments. For example, network access is blocked on docs.rs, that caused fftw-src 0.3.1 docs build fail, and can cause similar problem for dependent crates. This is why I propose to include FFTW source code into fftw-src crate.

Can plan executions take &self instead of &mut self?

I am asking this, because I am having a hard time trying to check whether this is possible or not. According to this documentation, execution of plans should take a const pointer to the plan (which AFAIK is nearly meaningless in the C world, because you are totally safe (ahem) to change nested values...).

Maybe it could be possible, but I am completely unable to hack through the code of FFTW3 to find the answer by myself. I could not find a similar question, therefore, even if c2c/r2c/c2r really need to use &mut self, at least there is an issue explicitly discussing this.

num-complex v0.2.0 support seems to be broken

It looks like the fftw-sys dependency is stuck on v0.2.4 which uses num-complex v0.1.43.

Creating a minimal project with the following dependencies

[dependencies]
fftw = "0.5.0"
num-complex = "0.2.0"

causes cargo to pull down both num-complex v0.2.0 and num-complex v0.1.43. Cargo reports that it's using v0.3.0 of fftw-sys, but the actual Cargo.toml in the registry (~/.cargo/registry/src/github.com-1ecc6299db9ec823/fftw-sys-0.3.0) shows version 0.2.4 with the old num-complex dependency.

use num_complex::Complex;
use fftw::array::AlignedVec;

fn main() {
    let v: AlignedVec<Complex<f64>> = AlignedVec::new(1);
}

errors with "the trait fftw::array::AlignedAllocable is not implemented for num_complex::Complex<f64>"

Changing the num-complex dependency version to 0.1.43 allows the above to compile.

Support 2d, 3d transformation

Basic structure of this crate has been clarified. We have to add 2d, 3d, and higher dimensional interface.

The source and intel-mkl feature flags are mutually exclusive

I've noticed that if the source and intel-mkl feature flags are both set, it leads to a build failure:

   Compiling fftw-sys v0.6.0
error[E0259]: the name `ffi` is defined multiple times
 --> /home/bowlofeggs/.cargo/registry/src/github.com-1ecc6299db9ec823/fftw-sys-0.6.0/src/lib.rs:7:1
  |
4 | extern crate fftw_src as ffi;
  | ----------------------------- previous import of the extern crate `ffi` here
...
7 | extern crate intel_mkl_src as ffi;
  | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ `ffi` reimported here
  |
  = note: `ffi` must be defined only once in the type namespace of this module
help: you can use `as` to change the binding name of the import
  |
7 | extern crate intel_mkl_src as other_ffi;
  |

error: aborting due to previous error

For more information about this error, try `rustc --explain E0259`.
error: could not compile `fftw-sys`

The Cargo Book suggests avoiding this scenario if possible:

There are rare cases where features may be mutually incompatible with one another. This should be avoided if at all possible, because it requires coordinating all uses of the package in the dependency graph to cooperate to avoid enabling them together.

Since source is the default feature, would it make sense to use intel-mkl if that feature is set, and fall back to source if it is not set?

Context

My particular problem is that my application also has an intel-mkl feature flag. If that flag is set, I have my application configured to set the same flag on fftw. However, if that flag is not set I would like to use the source feature. I am unfamiliar with a way to express that I want the source flag set if my feature is not set, but set if it is. If fftw made these flags additive rather than mutually exclusive, I could simply use the default feature set in fftw (giving me source), and then I can add the intel-mkl flag to fftw if my corresponding feature is enabled without hitting this compilation issue.

Support "system" feature for Windows using vcpkg

Like blas-lapack-rs/openblas-src#35

vcpkg has fftw port https://github.com/microsoft/vcpkg/tree/master/ports/fftw3

Failed to compile on Windows

x fftw-3.3.6-pl1/doc/fftw3.info-2
x fftw-3.3.6-pl1/tools/
x fftw-3.3.6-pl1/tools/Makefile.am
x fftw-3.3.6-pl1/tools/fftw-wisdom-to-conf.1
x fftw-3.3.6-pl1/tools/fftwf-wisdom.1
x fftw-3.3.6-pl1/tools/Makefile.in
x fftw-3.3.6-pl1/tools/fftw-wisdom-to-conf.in
x fftw-3.3.6-pl1/tools/fftw_wisdom.1.in
x fftw-3.3.6-pl1/tools/fftw-wisdom.c
thread 'main' panicked at 'failed to execute `"./configure" "--enable-static" "--with-pic" "--enable-single"`: The system cannot find the file specified. (os error 2)', C:\Users\username\.cargo\registry\src\github.com-1ecc6299db9ec823\fftw-src-0.2.3\build.rs:101:13
note: Run with `RUST_BACKTRACE=1` environment variable to display a backtrace.

System: Windows 10 Pro 17134 x64
Rust Toolchain: nightly-2019-01-26-x86_64-pc-windows-gnu
System Compiler: MSVC 2017

FFTW offically supports Windows. How can I build this crate on windows?

Build failed on docs.rs

due to wget does not exists in docs.rs builder

https://docs.rs/crate/fftw/0.2.2/builds

Auto-generate document for master branch

docs.rs does not work.

Pair for NAE interface

Multi dimension fft

First of all great effort to port a amazing fft library to an amazing programming language. I am relatively new to rust and come from a more c++ heavy background. I am struggling a bit in understanding how to use the multi dimensional fft. I saw that there are some commits related to this feature. When trying to figure out how to actually create the plan for this and pass the data I fail though. Would be amazing if someone could point me into the right direction.
Best wishes,
Malte

License question

Hey there:

We are very interested in using these RUST bindings for FFTW.

We are wondering if we can buy a license for the RUST bindings to be able to try it out in production.

We currently have an FFTW 3.3.x license, but would really like to extend that to RUST. Unfortunately, we can't use GPL :(

Let us know what you think :)

Feature request: cuFFT support

Nvidia Cuda FFT library cuFFT has an FFTW interface. It would be nice to have it as a feature along with FFTW and MKL

Recommend Projects

React

A declarative, efficient, and flexible JavaScript library for building user interfaces.
Vue.js

🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
Typescript

TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
TensorFlow

An Open Source Machine Learning Framework for Everyone
Django

The Web framework for perfectionists with deadlines.
Laravel

A PHP framework for web artisans
D3

Bring data to life with SVG, Canvas and HTML. 📊📈🎉

Recommend Topics

javascript

JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
web

Some thing interesting about web. New door for the world.
server

A server is a program made to process requests and deliver data to clients.
Machine learning

Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Visualization

Some thing interesting about visualization, use data art
Game

Some thing interesting about game, make everyone happy.

Recommend Org

Facebook

We are working to build community through open source technology. NB: members must have two-factor auth.
Microsoft

Open source projects and samples from Microsoft.
Google

Google ❤️ Open Source for everyone.
Alibaba

Alibaba Open Source for everyone
D3

Data-Driven Documents codes.
Tencent

China tencent open source team.