use avx256 by default (#109)

* wip

* fix

* fix

* simplify

* system test

* fix

* Update m31.rs

.github/workflows/ci.yml

@@ -67,13 +67,13 @@ jobs:
             feature: avx2
             field: gf2ext128
           - os: 7950x3d
-            feature: avx512
+            feature: avx512f
             field: m31ext3
           - os: 7950x3d
-            feature: avx512
+            feature: avx512f
             field: fr
           - os: 7950x3d
-            feature: avx512
+            feature: avx512f
             field: gf2ext128
     steps:
       - uses: actions/checkout@v4

Cargo.toml

@@ -43,7 +43,6 @@ path = "src/utils.rs"
 default = []
 # default = [ "grinding" ]
 grinding = []
-avx256 = ["arith/avx256"]
 
 [workspace]
 members = ["arith", "bi-kzg"]

arith/Cargo.toml

@@ -27,5 +27,3 @@ harness = false
 name = "ext_field"
 harness = false
 
-[features]
-avx256 = []

arith/benches/ext_field.rs

@@ -1,5 +1,3 @@
-#[cfg(target_arch = "x86_64")]
-use arith::GF2_128x8_256;
 use arith::{ExtensionField, Field, GF2_128x8, M31Ext3, M31Ext3x16, GF2_128};
 use ark_std::test_rng;
 use criterion::{criterion_group, criterion_main, BatchSize, Criterion};
@@ -156,8 +154,6 @@ fn ext_by_base_benchmark(c: &mut Criterion) {
     bench_field::<M31Ext3x16>(c);
     bench_field::<GF2_128>(c);
     bench_field::<GF2_128x8>(c);
-    #[cfg(target_arch = "x86_64")]
-    bench_field::<GF2_128x8_256>(c);
 }
 
 criterion_group!(ext_by_base_benches, ext_by_base_benchmark);

arith/benches/field.rs

@@ -1,8 +1,6 @@
 // this module benchmarks the performance of different field operations
 
 use arith::{Field, GF2_128x8, GF2x8, M31Ext3, M31Ext3x16, M31x16, GF2, GF2_128, M31};
-#[cfg(target_arch = "x86_64")]
-use arith::{GF2_128x8_256, M31x16_256};
 use ark_std::test_rng;
 use criterion::{criterion_group, criterion_main, BatchSize, Criterion};
 use halo2curves::bn256::Fr;
@@ -176,17 +174,13 @@ pub(crate) fn bench_field<F: Field>(c: &mut Criterion) {
 fn criterion_benchmark(c: &mut Criterion) {
     bench_field::<M31>(c);
     bench_field::<M31x16>(c);
-    #[cfg(target_arch = "x86_64")]
-    bench_field::<M31x16_256>(c);
     bench_field::<M31Ext3>(c);
     bench_field::<M31Ext3x16>(c);
     bench_field::<Fr>(c);
     bench_field::<GF2>(c);
     bench_field::<GF2x8>(c);
     bench_field::<GF2_128>(c);
     bench_field::<GF2_128x8>(c);
-    #[cfg(target_arch = "x86_64")]
-    bench_field::<GF2_128x8_256>(c);
 }
 
 criterion_group!(benches, criterion_benchmark);

arith/src/extension_field.rs

@@ -1,16 +1,12 @@
 mod fr_ext;
-// mod gf2_127;
 mod gf2_128;
 mod gf2_128x8;
 mod m31_ext;
 mod m31_ext3x16;
 use crate::{Field, FieldSerde};
 
-// pub use gf2_127::*;
 pub use gf2_128::*;
 pub use gf2_128x8::GF2_128x8;
-#[cfg(target_arch = "x86_64")]
-pub use gf2_128x8::GF2_128x8_256;
 pub use m31_ext::M31Ext3;
 pub use m31_ext3x16::M31Ext3x16;