Add x86_64 asm codegen for PrimeField mul and square

README.md

@@ -64,9 +64,9 @@ cargo build --release
 This library comes with unit tests for each of the provided crates. Run the tests with:
 ```bash
 cargo test
-``` 
+```
 
-Lastly, this library comes with benchmarks for the following crates:
+This library comes with benchmarks for the following crates:
 
 - [`algebra`](algebra)
 - [`dpc`](dpc)
@@ -76,6 +76,18 @@ These benchmarks require the nightly Rust toolchain; to install this, run `rustu
 cargo +nightly bench
 ```
 
+Compiling with `adcxq`, `adoxq` and `mulxq` instructions can lead to a 30-70% speedup. These are available on most `x86_64` platforms (Broadwell onwards for Intel and Ryzen onwards for AMD). Run the following command:
+```bash
+RUSTFLAGS="-C target-feature=+bmi2,+adx" cargo +nightly test/build/bench --features asm
+```
+Tip: If optimising for performance, your mileage may vary with passing `--emit=asm` to `RUSTFLAGS`.
+
+To bench `algebra-benches` with greater accuracy, especially for functions with execution times on the order of nanoseconds, use the `n_fold` feature to run selected functions 1000x per iteration. To run with multiple features, make sure to double quote the features.
+```bash
+cargo +nightly bench --features "n_fold bls12_381"
+```
+
+
 ## License
 
 ZEXE is licensed under either of the following licenses, at your discretion.

algebra-benches/Cargo.toml

@@ -19,6 +19,7 @@ include = ["Cargo.toml", "src", "README.md", "LICENSE-APACHE", "LICENSE-MIT"]
 license = "MIT/Apache-2.0"
 edition = "2018"
 publish = false
+build = "build.rs"
 
 ################################# Dependencies ################################
 
@@ -27,3 +28,20 @@ algebra = { path = "../algebra", features = [ "full" ] }
 blake2 = "0.8.1"
 rand = "0.7"
 rand_xorshift = { version = "0.2" }
+paste = "0.1"
+
+[features]
+asm = [ "algebra/asm"]
+n_fold = []
+mnt4_298 = []
+mnt6_298 = []
+mnt4_753 = []
+mnt6_753 = []
+bls12_381 = []
+bls12_377 = []
+sw6 = []
+nightly = []
+stable = []
+
+[build-dependencies]
+rustc_version = "0.1.*"