Commit graph

27 commits

Author SHA1 Message Date
pinkforest
40cf5aff99
Workspace curve25519 under curve25519-dalek 2023-06-27 04:00:12 +00:00
pinkforest(she/her)
e429bde88d
Clean up backend features and vendor curve25519_dalek_derive (#531)
* Vendor import unsafe_target_features as curve25519-dalek-derive

Co-authored-by: Jan Bujak <jan@parity.io>

* Remove feature gates from avx2/ifma

* Add buildtime compile diagnostics about backend selection

* Add build script tests

* Documentation changes

* Disable simd related features unless simd was determined via build

* Add note and test about the override warning when unsuccesful

* Reduce complexity in build gating via compile_error

---------

Co-authored-by: Jan Bujak <jan@parity.io>
Co-authored-by: Michael Rosenberg <michael@mrosenberg.pub>
2023-06-22 01:46:27 -04:00
Jan Bujak
0db8783be8
Runtime backend autodetection 2023-04-11 20:13:20 +09:00
Koute
4583c472f5
Support SIMD on Rust stable (#520)
* Remove dependency on `packed_simd`

* Support SIMD on stable Rust

* Move `packed_simd.rs` to `vector` module

* Add comment header to `packed_simd.rs`

* Initialize SIMD registers using intrinsics instead of `transmute`

* Use a splat inside of `unpack_pair`

* Update README: the AVX2 backend now works on stable Rust

* Add a CI job to also build the AVX2 SIMD backend on Rust stable

* Added SIMD MSRV test
2023-03-30 02:16:18 -04:00
Tony Arcieri
6f237a0810
Use inherent constants for ZERO, ONE, and MINUS_ONE (#470)
For the field element types `FieldElement` and `Scalar`, use inherent
constants instead of (non-const) functions to return these constant
values.

It's likely the original functions predate support for inherent
constants, but now that they're available, they're a better fit for
these sort of constant values.
2022-12-12 01:04:42 -05:00
pinkforest(she/her)
e01bb1bdc6
Fix all clippy warnings replay (#441)
Also fixes CI not running on all branches

Co-authored-by: Anthony Ramine <nox@nox.paris>
2022-12-04 03:40:51 -05:00
Michael Rosenberg
8fa201639a
Fix AVX2 and AVX-512 builds (#419)
Build was broken by 5758b8c.
This adds a regression test to CI so it doesn't happen again
2022-10-28 13:10:44 -04:00
Isis Lovecruft
1491f0db36
Update copyright years. 2021-03-25 04:10:55 +00:00
Isis Lovecruft
7c5ba69491
Update copyright year. 2019-10-23 19:43:56 +00:00
Henry de Valence
cfa09d859f Use upstream IFMA intrinsics now that they exist. 2019-08-06 17:14:31 -07:00
Henry de Valence
5700add507 Clean up warnings. 2019-01-18 14:19:30 -08:00
Henry de Valence
fa42708257 Change naming scheme for multiplication accumulators. 2019-01-18 14:19:30 -08:00
Henry de Valence
a84f9eafe4 Allow non_snake_case in the IFMA field and edwards modules. 2019-01-18 14:19:30 -08:00
Henry de Valence
da62569355 Add lookup table support. 2019-01-18 14:19:30 -08:00
Henry de Valence
6cec313f16 Add impl of Sub and Neg for Edwards points. 2019-01-18 14:19:30 -08:00
Henry de Valence
8e38ff2859 Add Edwards doubling implementation. 2019-01-18 14:19:30 -08:00
Henry de Valence
c4617b28a9 Add squaring implementation. 2019-01-18 14:19:30 -08:00
Henry de Valence
3f21bb3b58 Add skeleton of parallel Edwards formulas. 2019-01-18 14:19:30 -08:00
Henry de Valence
5880fced2b Add skeleton of shuffle and blend functions. 2019-01-18 14:19:30 -08:00
Henry de Valence
d90bb779ca Delay all computations using x[0], y[0]
Since computation of the 0 term in reduction requires a multiplication with a
4-cycle latency, this ensures that the rest of the computation can start before
the 0 term is finished
2019-01-18 14:19:30 -08:00
Henry de Valence
c925c76342 Change the strategy for reducing the high half of the product. 2019-01-18 14:19:30 -08:00
Henry de Valence
607b839032 Change size of limbs in the reduction test. 2019-01-18 14:19:30 -08:00
Henry de Valence
14163131f9 Lift the reduction state of the field elements into the type system.
This splits the `FieldElement51x4` type into two types:

- `F51x4Reduced` (with reduced limbs)
- `F51x4Unreduced` (with unreduced limbs)

The reduction is implemented as a `From` impl to convert one type to the other.

The output of a multiplication is now a `F51x4Unreduced`.  The reason is that
the inputs to IFMA operations must be at most 52 bits, so it's not possible to
perform an addition of (51+epsilon)-bit values and still be small enough to be
used as an input to multiplication.  So, it doesn't make sense to perform a
reduction at the end of a multiplication, because the reduced values will be
fed into an addition or subtraction, which then needs to be re-reduced.
2019-01-18 14:19:30 -08:00
Henry de Valence
ff49673214 Add impl Mul<(u32, u32, u32, u32)> for FieldElement51x4. 2019-01-18 14:19:30 -08:00
Henry de Valence
70199d6094 Add impl Mul<FieldElement51x4> for FieldElement51x4. 2019-01-18 14:19:30 -08:00
Henry de Valence
aa73d7b1bc Add a FieldElement51x4 type. 2019-01-18 14:19:30 -08:00
Henry de Valence
9ed2128a10 Add stub code for IFMA intrinsics in Rust. 2019-01-18 01:52:17 -08:00