feat: make it #[no_std] compatible (#280)

* make it #[no_std] comptabile

* chore: bump rust to 1.78, msrv to 1.73 and use cargo clippy

* fix: use std when its possible for floats

* fix: CI

* refactor: remove unnecessary qualification

* ci: set features correctly on opencl build

* chore: add Unicode 3.0 to allow licenses

.github/workflows/rust.yml

@@ -112,7 +112,7 @@ jobs:
       - run: nvcc --version
       - name: CUDA tests
         run: |
-          cargo nextest run --release --no-default-features --features cuda,pasta,bls,arity2,arity4,arity8,arity11,arity16,arity24,arity36
+          cargo nextest run --release --no-default-features --features std,cuda,pasta,bls,arity2,arity4,arity8,arity11,arity16,arity24,arity36
 
     # Runs the test suite on a self-hosted GPU machine with CUDA and OpenCL enabled (that is using the OpenCL backend for NVIDIA GPUs)
   test-opencl:
@@ -152,4 +152,4 @@ jobs:
       - run: clinfo
       - name: OpenCL tests
         run: |
-          cargo nextest run --release --all-features
+          cargo nextest run --release --no-default-features --features std,strengthened,abomonation,opencl,pasta,bls,arity2,arity4,arity8,arity11,arity16,arity24,arity36

Cargo.toml

@@ -1,25 +1,25 @@
 [package]
 name = "neptune"
 description = "Poseidon hashing over BLS12-381 for Filecoin."
-version = "13.0.0"
+version = "14.0.0"
 authors = ["porcuquine <[email protected]>"]
 edition = "2021"
 license = "MIT OR Apache-2.0"
 repository = "https://github.com/argumentcomputer/neptune"
-rust-version = "1.71.0"
+rust-version = "1.73.0"
 
 [dependencies]
-bellpepper = { workspace = true }
-bellpepper-core = { workspace = true }
+bellpepper = { workspace = true, optional = true }
+bellpepper-core = { workspace = true, optional = true }
 blake2s_simd = { workspace = true }
 blstrs = { workspace = true, optional = true }
-byteorder = { workspace = true }
 ec-gpu = { workspace = true, optional = true }
 ec-gpu-gen = { workspace = true, optional = true }
-ff ={ workspace = true }
+libm = { workspace = true }
+ff = { workspace = true }
 generic-array = { workspace = true }
 pasta_curves = { workspace = true, features = ["serde"] }
-serde = { version = "1.0", features = ["derive"] }
+serde = { version = "1.0", features = ["derive"], optional = true }
 trait-set = "0.3.0"
 abomonation = { version = "0.7.3", optional = true }
 abomonation_derive = { version = "0.1.0", package = "abomonation_derive_ng", optional = true }
@@ -57,7 +57,7 @@ incremental = false
 codegen-units = 1
 
 [features]
-default = ["bls", "pasta"]
+default = ["std", "bls", "pasta"]
 cuda = ["ec-gpu-gen/cuda", "ec-gpu"]
 opencl = ["ec-gpu-gen/opencl", "ec-gpu"]
 # The supported arities for Poseidon running on the GPU are specified at compile-time.
@@ -74,6 +74,7 @@ strengthened = []
 bls = ["blstrs/gpu"]
 pasta = ["pasta_curves/gpu"]
 portable = ["blstrs/portable"]
+std = ["dep:serde", "dep:bellpepper", "dep:bellpepper-core"]
 # Unsafe Abomonation-based serialization
 abomonation = ["dep:abomonation", "dep:abomonation_derive"]
 
@@ -90,6 +91,7 @@ bellpepper = { version = "0.4.0", default-features = false }
 blake2s_simd = "1.0.1"
 blstrs = { version = "0.7.0" }
 ff = "0.13.0"
+libm = "0.2.8"
 generic-array = "1.0"
 pasta_curves = { version = "0.5" }
 ec-gpu = { version = "0.2.0" }

deny.toml

@@ -109,6 +109,7 @@ allow = [
     "CC0-1.0",
     "Apache-2.0",
     "Unicode-DFS-2016",
+    "Unicode-3.0",
 ]
 # List of explicitly disallowed licenses
 # See https://spdx.org/licenses/ for list of possible licenses

rust-toolchain.toml

@@ -1,4 +1,4 @@
 [toolchain]
 # The default profile includes rustc, rust-std, cargo, rust-docs, rustfmt and clippy.
 profile = "default"
-channel = "1.76"
+channel = "1.78"

src/column_tree_builder.rs

@@ -49,7 +49,7 @@ where
         let end = start + column_count;
 
         if end > self.leaf_count {
-            return Err(Error::Other("too many columns".to_string()));
+            return Err(Error::Other("too many columns"));
         }
 
         match self.column_batcher {

src/error.rs

@@ -1,4 +1,4 @@
-use std::{error, fmt};
+use core::fmt;
 
 #[derive(Debug, Clone)]
 #[cfg(any(feature = "cuda", feature = "opencl"))]
@@ -47,10 +47,11 @@ pub enum Error {
     FullBuffer,
     /// Attempt to reference an index element that is out of bounds
     IndexOutOfBounds,
+    #[cfg(any(feature = "cuda", feature = "opencl"))]
     GpuError(String),
     #[cfg(any(feature = "cuda", feature = "opencl"))]
     ClError(ClError),
-    Other(String),
+    Other(&'static str),
 }
 
 #[cfg(any(feature = "cuda", feature = "opencl"))]
@@ -67,7 +68,8 @@ impl From<ec_gpu_gen::EcError> for Error {
     }
 }
 
-impl error::Error for Error {}
+#[cfg(feature = "std")]
+impl std::error::Error for Error {}
 
 impl fmt::Display for Error {
     fn fmt(&self, f: &mut fmt::Formatter<'_>) -> Result<(), fmt::Error> {
@@ -77,6 +79,7 @@ impl fmt::Display for Error {
                 "The size of the buffer cannot be greater than the hash arity."
             ),
             Error::IndexOutOfBounds => write!(f, "The referenced index is outs of bounds."),
+            #[cfg(any(feature = "cuda", feature = "opencl"))]
             Error::GpuError(s) => write!(f, "GPU Error: {s}"),
             #[cfg(any(feature = "cuda", feature = "opencl"))]
             Error::ClError(e) => write!(f, "OpenCL Error: {e}"),

src/hash_type.rs

@@ -16,11 +16,16 @@ use abomonation::Abomonation;
 #[cfg(feature = "abomonation")]
 use abomonation_derive::Abomonation;
 use ff::PrimeField;
+#[cfg(feature = "std")]
 use serde::{Deserialize, Serialize};
 
-#[derive(Clone, Debug, PartialEq, Eq, Serialize, Deserialize)]
+#[derive(Clone, Debug, PartialEq, Eq)]
+#[cfg_attr(feature = "std", derive(Serialize, Deserialize))]
+#[cfg_attr(
+    feature = "std",
+    serde(bound(serialize = "F: Serialize", deserialize = "F: Deserialize<'de>"))
+)]
 #[cfg_attr(feature = "abomonation", derive(Abomonation))]
-#[serde(bound(serialize = "F: Serialize", deserialize = "F: Deserialize<'de>"))]
 #[cfg_attr(feature = "abomonation", abomonation_omit_bounds)]
 pub enum HashType<F: PrimeField, A: Arity<F>> {
     MerkleTree,
@@ -73,7 +78,8 @@ impl<F: PrimeField, A: Arity<F>> HashType<F, A> {
     }
 }
 
-#[derive(Clone, Debug, PartialEq, Eq, Serialize, Deserialize)]
+#[derive(Clone, Debug, PartialEq, Eq)]
+#[cfg_attr(feature = "std", derive(Serialize, Deserialize))]
 #[cfg_attr(feature = "abomonation", derive(Abomonation))]
 #[cfg_attr(feature = "abomonation", abomonation_omit_bounds)]
 pub enum CType<F: PrimeField, A: Arity<F>> {
@@ -82,7 +88,7 @@ pub enum CType<F: PrimeField, A: Arity<F>> {
     // This is a bit of a hack, but since `serde(skip)` tags the last variant arm,
     // the generated code ends up being correct. But, in the future, do not
     // carelessly add new variants to this enum.
-    #[serde(skip)]
+    #[cfg_attr(feature = "std", serde(skip))]
     #[cfg_attr(feature = "abomonation", abomonation_skip)]
     _Phantom((F, A)),
 }

src/lib.rs

@@ -1,6 +1,10 @@
+#![cfg_attr(not(feature = "std"), no_std)]
 #![allow(dead_code)]
 #![allow(unused_imports)]
 
+extern crate alloc;
+pub use alloc::vec::Vec;
+
 pub use crate::poseidon::{Arity, Poseidon};
 use crate::round_constants::generate_constants;
 use crate::round_numbers::{round_numbers_base, round_numbers_strengthened};
@@ -11,8 +15,8 @@ use blstrs::Scalar as Fr;
 pub use error::Error;
 use ff::PrimeField;
 use generic_array::GenericArray;
+#[cfg(feature = "std")]
 use serde::{Deserialize, Serialize};
-use std::fmt;
 use trait_set::trait_set;
 
 // See https://www.lurklurk.org/effective-rust/re-export.html
@@ -45,8 +49,11 @@ compile_error!("The `strengthened` feature needs the `cuda` and/or `opencl` feat
 compile_error!("The `cuda` and `opencl` features need the `bls` and/or `pasta` feature to be set");
 
 /// Poseidon circuit
+#[cfg(feature = "std")]
 pub mod circuit;
+#[cfg(feature = "std")]
 pub mod circuit2;
+#[cfg(feature = "std")]
 pub mod circuit2_witness;
 pub mod error;
 mod matrix;
@@ -60,6 +67,7 @@ mod round_constants;
 mod round_numbers;
 
 /// Sponge
+#[cfg(any(test, feature = "std"))]
 pub mod sponge;
 
 /// Hash types and domain separation tags.
@@ -97,21 +105,23 @@ trait_set! {
    pub trait NeptuneField = PrimeField + ec_gpu::GpuName;
 }
 
+#[cfg(feature = "std")]
 mod serde_impl;
 
 pub(crate) const TEST_SEED: [u8; 16] = [
     0x59, 0x62, 0xbe, 0x5d, 0x76, 0x3d, 0x31, 0x8d, 0x17, 0xdb, 0x37, 0x32, 0x54, 0x06, 0xbc, 0xe5,
 ];
 
-#[derive(Copy, Clone, Debug, PartialEq, Eq, Serialize, Deserialize)]
+#[derive(Copy, Clone, Debug, PartialEq, Eq)]
+#[cfg_attr(feature = "std", derive(Serialize, Deserialize))]
 #[cfg_attr(feature = "abomonation", derive(Abomonation))]
 pub enum Strength {
     Standard,
     Strengthened,
 }
 
-impl fmt::Display for Strength {
-    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
+impl core::fmt::Display for Strength {
+    fn fmt(&self, f: &mut core::fmt::Formatter<'_>) -> core::fmt::Result {
         match self {
             Self::Standard => write!(f, "standard"),
             Self::Strengthened => write!(f, "strengthened"),
@@ -183,7 +193,7 @@ fn round_constants<F: PrimeField>(arity: usize, strength: &Strength) -> Vec<F> {
 
     let fr_num_bits = F::NUM_BITS;
     let field_size = {
-        assert!(fr_num_bits <= u32::from(std::u16::MAX));
+        assert!(fr_num_bits <= u32::from(core::u16::MAX));
         // It's safe to convert to u16 for compatibility with other types.
         fr_num_bits as u16
     };

src/matrix.rs

@@ -1,6 +1,7 @@
 // Allow `&Matrix` in function signatures.
 #![allow(clippy::ptr_arg)]
 
+use crate::Vec;
 use ff::PrimeField;
 
 /// Matrix functions here are, at least for now, quick and dirty — intended only to support precomputation of poseidon optimization.
@@ -116,7 +117,7 @@ pub(crate) fn left_apply_matrix<F: PrimeField>(m: &Matrix<F>, v: &[F]) -> Vec<F>
         "Matrix can only be applied to vector of same size."
     );
 
-    let mut result = vec![F::ZERO; v.len()];
+    let mut result = alloc::vec![F::ZERO; v.len()];
 
     for (result, row) in result.iter_mut().zip(m.iter()) {
         for (mat_val, vec_val) in row.iter().zip(v) {
@@ -137,7 +138,7 @@ pub(crate) fn apply_matrix<F: PrimeField>(m: &Matrix<F>, v: &[F]) -> Vec<F> {
         "Matrix can only be applied to vector of same size."
     );
 
-    let mut result = vec![F::ZERO; v.len()];
+    let mut result = alloc::vec![F::ZERO; v.len()];
     for (j, val) in result.iter_mut().enumerate() {
         for (i, row) in m.iter().enumerate() {
             let mut tmp = row[j];
@@ -165,7 +166,7 @@ pub(crate) fn transpose<F: PrimeField>(matrix: &Matrix<F>) -> Matrix<F> {
 
 #[allow(clippy::needless_range_loop)]
 pub(crate) fn make_identity<F: PrimeField>(size: usize) -> Matrix<F> {
-    let mut result = vec![vec![F::ZERO; size]; size];
+    let mut result = alloc::vec![alloc::vec![F::ZERO; size]; size];
     for i in 0..size {
         result[i][i] = F::ONE;
     }

src/mds.rs

@@ -6,15 +6,18 @@ use abomonation::Abomonation;
 #[cfg(feature = "abomonation")]
 use abomonation_derive::Abomonation;
 use ff::PrimeField;
+#[cfg(feature = "std")]
 use serde::{Deserialize, Serialize};
 
 use crate::matrix;
 use crate::matrix::{
     apply_matrix, invert, is_identity, is_invertible, is_square, left_apply_matrix, mat_mul, minor,
     transpose, Matrix,
 };
+use crate::Vec;
 
-#[derive(Clone, Debug, PartialEq, Eq, Serialize, Deserialize)]
+#[derive(Clone, Debug, PartialEq, Eq)]
+#[cfg_attr(feature = "std", derive(Serialize, Deserialize))]
 #[cfg_attr(feature = "abomonation", derive(Abomonation))]
 #[cfg_attr(feature = "abomonation", abomonation_bounds(where F::Repr: Abomonation))]
 pub struct MdsMatrices<F: PrimeField> {
@@ -58,7 +61,8 @@ pub(crate) fn derive_mds_matrices<F: PrimeField>(m: Matrix<F>) -> MdsMatrices<F>
 /// This means its first row and column are each dense, and the interior matrix
 /// (minor to the element in both the row and column) is the identity.
 /// We will pluralize this compact structure `sparse_matrixes` to distinguish from `sparse_matrices` from which they are created.
-#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize)]
+#[derive(Debug, Clone, PartialEq, Eq)]
+#[cfg_attr(feature = "std", derive(Serialize, Deserialize))]
 #[cfg_attr(feature = "abomonation", derive(Abomonation))]
 #[cfg_attr(feature = "abomonation", abomonation_bounds(where F::Repr: Abomonation))]
 pub struct SparseMatrix<F: PrimeField> {
@@ -174,12 +178,12 @@ fn make_prime<F: PrimeField>(m: &Matrix<F>) -> Matrix<F> {
         .enumerate()
         .map(|(i, row)| match i {
             0 => {
-                let mut new_row = vec![F::ZERO; row.len()];
+                let mut new_row = alloc::vec![F::ZERO; row.len()];
                 new_row[0] = F::ONE;
                 new_row
             }
             _ => {
-                let mut new_row = vec![F::ZERO; row.len()];
+                let mut new_row = alloc::vec![F::ZERO; row.len()];
                 new_row[1..].copy_from_slice(&row[1..]);
                 new_row
             }
@@ -201,7 +205,7 @@ fn make_double_prime<F: PrimeField>(m: &Matrix<F>, m_hat_inv: &Matrix<F>) -> Mat
                 new_row
             }
             _ => {
-                let mut new_row = vec![F::ZERO; row.len()];
+                let mut new_row = alloc::vec![F::ZERO; row.len()];
                 new_row[0] = w_hat[i - 1];
                 new_row[i] = F::ONE;
                 new_row

src/poseidon.rs

@@ -6,16 +6,18 @@ use crate::mds::{
 };
 use crate::poseidon_alt::{hash_correct, hash_optimized_dynamic};
 use crate::preprocessing::compress_round_constants;
+use crate::Vec;
 use crate::{matrix, quintic_s_box, BatchHasher, Strength, DEFAULT_STRENGTH};
 use crate::{round_constants, round_numbers, Error};
 #[cfg(feature = "abomonation")]
 use abomonation::Abomonation;
 #[cfg(feature = "abomonation")]
 use abomonation_derive::Abomonation;
+use core::marker::PhantomData;
 use ff::PrimeField;
 use generic_array::{sequence::GenericSequence, typenum, ArrayLength, GenericArray};
+#[cfg(feature = "std")]
 use serde::{Deserialize, Serialize};
-use std::marker::PhantomData;
 use typenum::marker_traits::Unsigned;
 use typenum::*;
 
@@ -868,12 +870,12 @@ where
             }
         }
 
-        let _ = std::mem::replace(&mut self.elements, result);
+        let _ = core::mem::replace(&mut self.elements, result);
     }
 
     pub(crate) fn product_mds_with_matrix_left(&mut self, matrix: &Matrix<F>) {
         let result = left_apply_matrix(matrix, &self.elements);
-        let _ = std::mem::replace(
+        let _ = core::mem::replace(
             &mut self.elements,
             GenericArray::<F, A::ConstantsSize>::generate(|i| result[i]),
         );
@@ -900,9 +902,10 @@ where
             val.add_assign(&tmp);
         }
 
-        let _ = std::mem::replace(&mut self.elements, result);
+        let _ = core::mem::replace(&mut self.elements, result);
     }
 
+    #[cfg(feature = "std")]
     pub(crate) fn debug(&self, msg: &str) {
         dbg!(msg, &self.constants_offset, &self.elements);
     }

src/poseidon_alt.rs

@@ -2,6 +2,7 @@
 //! These are tested (in `poseidon::test`) to be equivalent to the 'static optimized' version
 //! used for actual hashing by the neptune library.
 use crate::poseidon::{Arity, Poseidon};
+use crate::Vec;
 use crate::{matrix, quintic_s_box};
 use ff::PrimeField;