Removed nightly feature flag now that neon intrinsics are (sort of) s…

…table Note, neon still requires a nightly compiler to enable, this just means we don't need an explicit flag to detect this. And this code should "just work" when neon becomes stable. The state of things is a bit confusing. SIMD in Rust: rust-lang/rust#48556 NEON in Rust: rust-lang/rust#90972
geky · May 8, 2022 · 1e870c2 · 1e870c2
1 parent 83c44d3
commit 1e870c2
Show file tree

Hide file tree

Showing 13 changed files with 46 additions and 80 deletions.
diff --git a/Cargo.toml b/Cargo.toml
@@ -4,17 +4,10 @@ description = "A Rust library containing Galois-field types and utilities"
 authors = ["Christopher Haster <chaster@utexas.edu>"]
 repository = "https://github.com/geky/gf256"
 version = "0.2.0"
-edition = "2018"
+edition = "2021"
 license = "BSD-3-Clause"
 
 [features]
-# Enables features only found on a nightly compiler
-#
-# This is required to leverage pmull on aarch64, tracking issue:
-# https://github.com/rust-lang/rust/issues/48556
-#
-nightly = []
-
 # Disable carry-less multiplication instructions, forcing the use
 # of naive bitwise implementations
 #

diff --git a/Makefile b/Makefile
@@ -2,59 +2,58 @@
 override RUSTFLAGS += -Ctarget-cpu=native
 
 # this indirection is so commands with env are easily copied on the terminal
-override ENV += RUSTFLAGS="$(RUSTFLAGS)"
+CARGO ?= RUSTFLAGS="$(RUSTFLAGS)" cargo +nightly
 
 .PHONY: all build
 all build:
-	$(ENV) cargo +nightly build --features nightly,thread-rng,lfsr,crc,shamir,raid,rs
+	$(CARGO) build --features thread-rng,lfsr,crc,shamir,raid,rs
 
 .PHONY: test
 test:
-	$(ENV) cargo +nightly test --features nightly,thread-rng,lfsr,crc,shamir,raid,rs --lib
-	$(ENV) cargo +nightly test --features nightly,thread-rng,lfsr,crc,shamir,raid,rs --example find-p
-	$(ENV) cargo +nightly run --features nightly,thread-rng,lfsr,crc,shamir,raid,rs --example find-p -- -w9 -n4 -m1 -q
-	$(ENV) cargo +nightly run --features nightly,thread-rng,lfsr,crc,shamir,raid,rs --example lfsr
-	$(ENV) cargo +nightly run --features nightly,thread-rng,lfsr,crc,shamir,raid,rs --example crc
-	$(ENV) cargo +nightly run --features nightly,thread-rng,lfsr,crc,shamir,raid,rs --example shamir
-	$(ENV) cargo +nightly run --features nightly,thread-rng,lfsr,crc,shamir,raid,rs --example raid
-	$(ENV) cargo +nightly run --features nightly,thread-rng,lfsr,crc,shamir,raid,rs --example rs
+	$(CARGO) test --features thread-rng,lfsr,crc,shamir,raid,rs --lib
+	$(CARGO) test --features thread-rng,lfsr,crc,shamir,raid,rs --example find-p
+	$(CARGO) run --features thread-rng,lfsr,crc,shamir,raid,rs --example find-p -- -w9 -n4 -m1 -q
+	$(CARGO) run --features thread-rng,lfsr,crc,shamir,raid,rs --example lfsr
+	$(CARGO) run --features thread-rng,lfsr,crc,shamir,raid,rs --example crc
+	$(CARGO) run --features thread-rng,lfsr,crc,shamir,raid,rs --example shamir
+	$(CARGO) run --features thread-rng,lfsr,crc,shamir,raid,rs --example raid
+	$(CARGO) run --features thread-rng,lfsr,crc,shamir,raid,rs --example rs
 
 .PHONY: test-configs
 test-configs:
-	$(ENV) cargo test --lib
-	$(ENV) cargo test --features thread-rng,lfsr,crc,shamir,raid,rs --lib
-	$(ENV) cargo test --features no-xmul,thread-rng,lfsr,crc,shamir,raid,rs --lib
-	$(ENV) cargo test --features no-tables,thread-rng,lfsr,crc,shamir,raid,rs --lib
-	$(ENV) cargo test --features small-tables,thread-rng,lfsr,crc,shamir,raid,rs --lib
-	$(ENV) cargo +nightly test --features nightly,thread-rng,lfsr,crc,shamir,raid,rs --lib
+	$(CARGO) test --lib
+	$(CARGO) test --features thread-rng,lfsr,crc,shamir,raid,rs --lib
+	$(CARGO) test --features no-xmul,thread-rng,lfsr,crc,shamir,raid,rs --lib
+	$(CARGO) test --features no-tables,thread-rng,lfsr,crc,shamir,raid,rs --lib
+	$(CARGO) test --features small-tables,thread-rng,lfsr,crc,shamir,raid,rs --lib
 
 .PHONY: docs
 docs:
-	$(ENV) cargo +nightly doc --no-deps --features nightly,thread-rng,lfsr,crc,shamir,raid,rs
-	$(ENV) cargo +nightly test --features nightly,thread-rng,lfsr,crc,shamir,raid,rs --doc
+	$(CARGO) doc --no-deps --features thread-rng,lfsr,crc,shamir,raid,rs
+	$(CARGO) test --features thread-rng,lfsr,crc,shamir,raid,rs --doc
 
 .PHONY: bench
 bench:
-	$(ENV) cargo +nightly bench --features nightly,thread-rng,lfsr,crc,shamir,raid,rs --bench xmul   -- --noplot
-	$(ENV) cargo +nightly bench --features nightly,thread-rng,lfsr,crc,shamir,raid,rs --bench gf     -- --noplot
-	$(ENV) cargo +nightly bench --features nightly,thread-rng,lfsr,crc,shamir,raid,rs --bench find-p -- --noplot
-	$(ENV) cargo +nightly bench --features nightly,thread-rng,lfsr,crc,shamir,raid,rs --bench lfsr   -- --noplot
-	$(ENV) cargo +nightly bench --features nightly,thread-rng,lfsr,crc,shamir,raid,rs --bench crc    -- --noplot
-	$(ENV) cargo +nightly bench --features nightly,thread-rng,lfsr,crc,shamir,raid,rs --bench shamir -- --noplot
-	$(ENV) cargo +nightly bench --features nightly,thread-rng,lfsr,crc,shamir,raid,rs --bench raid   -- --noplot
-	$(ENV) cargo +nightly bench --features nightly,thread-rng,lfsr,crc,shamir,raid,rs --bench rs     -- --noplot
+	$(CARGO) bench --features thread-rng,lfsr,crc,shamir,raid,rs --bench xmul   -- --noplot
+	$(CARGO) bench --features thread-rng,lfsr,crc,shamir,raid,rs --bench gf     -- --noplot
+	$(CARGO) bench --features thread-rng,lfsr,crc,shamir,raid,rs --bench find-p -- --noplot
+	$(CARGO) bench --features thread-rng,lfsr,crc,shamir,raid,rs --bench lfsr   -- --noplot
+	$(CARGO) bench --features thread-rng,lfsr,crc,shamir,raid,rs --bench crc    -- --noplot
+	$(CARGO) bench --features thread-rng,lfsr,crc,shamir,raid,rs --bench shamir -- --noplot
+	$(CARGO) bench --features thread-rng,lfsr,crc,shamir,raid,rs --bench raid   -- --noplot
+	$(CARGO) bench --features thread-rng,lfsr,crc,shamir,raid,rs --bench rs     -- --noplot
 
 .PHONY: bench-no-xmul
 bench-no-xmul:
-	$(ENV) cargo +nightly bench --features nightly,no-xmul,thread-rng,lfsr,crc,shamir,raid,rs --bench xmul   -- --noplot
-	$(ENV) cargo +nightly bench --features nightly,no-xmul,thread-rng,lfsr,crc,shamir,raid,rs --bench gf     -- --noplot
-	$(ENV) cargo +nightly bench --features nightly,no-xmul,thread-rng,lfsr,crc,shamir,raid,rs --bench find-p -- --noplot
-	$(ENV) cargo +nightly bench --features nightly,no-xmul,thread-rng,lfsr,crc,shamir,raid,rs --bench lfsr   -- --noplot
-	$(ENV) cargo +nightly bench --features nightly,no-xmul,thread-rng,lfsr,crc,shamir,raid,rs --bench crc    -- --noplot
-	$(ENV) cargo +nightly bench --features nightly,no-xmul,thread-rng,lfsr,crc,shamir,raid,rs --bench shamir -- --noplot
-	$(ENV) cargo +nightly bench --features nightly,no-xmul,thread-rng,lfsr,crc,shamir,raid,rs --bench raid   -- --noplot
-	$(ENV) cargo +nightly bench --features nightly,no-xmul,thread-rng,lfsr,crc,shamir,raid,rs --bench rs     -- --noplot
+	$(CARGO) bench --features no-xmul,thread-rng,lfsr,crc,shamir,raid,rs --bench xmul   -- --noplot
+	$(CARGO) bench --features no-xmul,thread-rng,lfsr,crc,shamir,raid,rs --bench gf     -- --noplot
+	$(CARGO) bench --features no-xmul,thread-rng,lfsr,crc,shamir,raid,rs --bench find-p -- --noplot
+	$(CARGO) bench --features no-xmul,thread-rng,lfsr,crc,shamir,raid,rs --bench lfsr   -- --noplot
+	$(CARGO) bench --features no-xmul,thread-rng,lfsr,crc,shamir,raid,rs --bench crc    -- --noplot
+	$(CARGO) bench --features no-xmul,thread-rng,lfsr,crc,shamir,raid,rs --bench shamir -- --noplot
+	$(CARGO) bench --features no-xmul,thread-rng,lfsr,crc,shamir,raid,rs --bench raid   -- --noplot
+	$(CARGO) bench --features no-xmul,thread-rng,lfsr,crc,shamir,raid,rs --bench rs     -- --noplot
 
 .PHONY: clean
 clean:
-	$(ENV) cargo clean
+	$(CARGO) clean
diff --git a/README.md b/README.md
@@ -151,7 +151,7 @@ assert_ne!(a + a, gf256(2)*a);
 Finite-fields can be very useful for applying high-level math onto machine
 words, since machine words (`u8`, `u16`, `u32`, etc) are inherently finite.
 Normally we just ignore this until an integer overflow occurs and then we just
-waive our hands around wailing that math has failed us.
+wave our hands around wailing that math has failed us.
 
 In Rust this has the fun side-effect that the Galois-field types are incapable
 of overflowing, so Galois-field types don't need the set of overflowing
@@ -331,8 +331,6 @@ Carry-less multiplication:
 
 gf256 takes advantage of these instructions when possible. However, at the time
 of writing, `pmull` support in Rust is only available on [nightly][nightly].
-To take advantage of `pmull` on aarch64 you will need to use a nightly compiler
-and enable gf256's `nightly` feature ([tracking issue](https://github.com/rust-lang/rust/issues/48556)).
 
 ``` rust
 # use ::gf256::*;
@@ -358,7 +356,7 @@ let b = if gf256::HAS_XMUL {
 };
 ```
 
-gf256 also leverages the hardware accelrated [carry-less addition][xor]
+gf256 also leverages the hardware accelerated [carry-less addition][xor]
 instructions, sometimes called polynomial addition, or simply xor. But this
 is much less notable.
 
@@ -445,10 +443,6 @@ evaluated before use, and you use this library at your own risk.
 
 ## Features
 
-- `nightly` - Enable features only found on a nightly compiler
-
-  This is required to leverage `pmull` on aarch64 ([tracking issue](https://github.com/rust-lang/rust/issues/48556))
-
 - `no-xmul` - Disables carry-less multiplication instructions, forcing the use
   of naive bitwise implementations
 
@@ -459,7 +453,7 @@ evaluated before use, and you use this library at your own risk.
 
   This may be useful on memory constrained devices
 
-- `small-tables` - Limits lookup tables to "small tables", tables with <16
+- `small-tables` - Limits lookup tables to "small tables", tables with <=16
   elements
 
   This provides a compromise between full 256-byte tables and no-tables,

diff --git a/gf256-macros/src/common.rs b/gf256-macros/src/common.rs
@@ -36,7 +36,6 @@ pub(crate) fn xmul_predicate() -> TokenStream {
                     target_feature="pclmulqdq"
                 ),
                 all(
-                    feature="nightly",
                     target_arch="aarch64",
                     target_feature="neon"
                 )

diff --git a/src/crc.rs b/src/crc.rs
@@ -21,7 +21,7 @@
 //! [`examples/crc.rs`][crc-example]:
 //!
 //! ``` bash
-//! $ RUSTFLAGS="-Ctarget-cpu=native" cargo +nightly run --features nightly,thread-rng,lfsr,crc,shamir,raid,rs --example crc
+//! $ RUSTFLAGS="-Ctarget-cpu=native" cargo run --features thread-rng,lfsr,crc,shamir,raid,rs --example crc
 //!
 //! testing crc("Hello World!")
 //! naive_crc                => 0x1c291ca3

diff --git a/src/gf.rs b/src/gf.rs
@@ -511,7 +511,7 @@
 //! bit-width. Note you need a 5-bit irreducible polynomial for a 4-bit field:
 //!
 //! ``` bash
-//! $ RUSTFLAGS="-Ctarget-cpu=native" cargo +nightly run --release --features nightly --example find-p -- --width=5 -m=1
+//! $ RUSTFLAGS="-Ctarget-cpu=native" cargo run --release --example find-p -- --width=5 -m=1
 //! polynomial=0x13, generator=0x2
 //! polynomial=0x19, generator=0x2
 //! polynomial=0x1f, generator=0x3
@@ -521,7 +521,7 @@
 //! their generators up to 65-bits:
 //!
 //! ``` bash
-//! $ RUSTFLAGS="-Ctarget-cpu=native" cargo +nightly run --release --features nightly --example find-p -- --width=65 -m=1 -n=1
+//! $ RUSTFLAGS="-Ctarget-cpu=native" cargo --release --example find-p -- --width=65 -m=1 -n=1
 //! polynomial=0x1000000000000001b, generator=0x2
 //! ```
 //!

diff --git a/src/lfsr.rs b/src/lfsr.rs
@@ -29,7 +29,7 @@
 //! [`examples/lfsr.rs`][lfsr-example]:
 //!
 //! ``` bash
-//! $ RUSTFLAGS="-Ctarget-cpu=native" cargo +nightly run --features nightly,thread-rng,lfsr,crc,shamir,raid,rs --example lfsr
+//! $ RUSTFLAGS="-Ctarget-cpu=native" cargo run --features thread-rng,lfsr,crc,shamir,raid,rs --example lfsr
 //!
 //! testing lfsr64
 //! lfsr64_naive                        => 0000000000000001000000000000001b00000000000001450000000000001db7

diff --git a/src/lib.rs b/src/lib.rs
@@ -1,12 +1,6 @@
 #![doc=include_str!("../README.md")]
 
 
-// Enable stdsimd for pmull on aarch64
-#![cfg_attr(
-    all(not(feature="no-xmul"), feature="nightly", target_arch="aarch64"),
-    feature(stdsimd)
-)]
-
 // We don't really need std
 #![no_std]
 

diff --git a/src/p.rs b/src/p.rs
@@ -96,8 +96,7 @@
 //! implementation.
 //!
 //! Note that at the time of writing, aarch64 [`pmull`][pmull] support is only
-//! available on a [nightly][nightly] compiler, and requires gf256's `nightly`
-//! feature to be enabled ([tracking issue](https://github.com/rust-lang/rust/issues/48556)).
+//! available on a [nightly][nightly] compiler.
 //!
 //! gf256 also exposes the flag [`HAS_XMUL`], which can be used to choose
 //! algorithms based on whether or not hardware accelerated carry-less

diff --git a/src/raid.rs b/src/raid.rs
@@ -44,7 +44,7 @@
 //! [`examples/raid.rs`][raid-example]:
 //!
 //! ``` bash
-//! $ RUSTFLAGS="-Ctarget-cpu=native" cargo +nightly run --features nightly,thread-rng,lfsr,crc,shamir,raid,rs --example raid
+//! $ RUSTFLAGS="-Ctarget-cpu=native" cargo run --features thread-rng,lfsr,crc,shamir,raid,rs --example raid
 //!
 //! testing raid5("Hello World?")
 //! format  => Hello World?U)_<  48656c6c6f20576f726c643f55295f3c

diff --git a/src/rs.rs b/src/rs.rs
@@ -33,7 +33,7 @@
 //! [`examples/rs.rs`][rs-example]:
 //!
 //! ``` bash
-//! $ RUSTFLAGS="-Ctarget-cpu=native" cargo +nightly run --features nightly,thread-rng,lfsr,crc,shamir,raid,rs --example rs
+//! $ RUSTFLAGS="-Ctarget-cpu=native" cargo run --features thread-rng,lfsr,crc,shamir,raid,rs --example rs
 //!
 //! testing rs("Hello World!")
 //! dimension = (255,223), 16 errors, 32 erasures

diff --git a/src/shamir.rs b/src/shamir.rs
@@ -27,7 +27,7 @@
 //! [`examples/shamir.rs`][shamir-example]:
 //!
 //! ``` bash
-//! $ RUSTFLAGS="-Ctarget-cpu=native" cargo +nightly run --features nightly,thread-rng,lfsr,crc,shamir,raid,rs --example shamir
+//! $ RUSTFLAGS="-Ctarget-cpu=native" cargo run --features thread-rng,lfsr,crc,shamir,raid,rs --example shamir
 //!
 //! testing shamir("Hello World!")
 //! generate share1 => .....uT4.z.O.  019ddb829d755434f77ae84ffd

diff --git a/src/xmul.rs b/src/xmul.rs
@@ -34,7 +34,6 @@ pub const HAS_XMUL: bool = {
             ),
             all(
                 not(feature="no-xmul"),
-                feature="nightly",
                 target_arch="aarch64",
                 target_feature="neon"
             )
@@ -59,7 +58,6 @@ pub const HAS_XMUL: bool = {
     ),
     all(
         not(feature="no-xmul"),
-        feature="nightly",
         target_arch="aarch64",
         target_feature="neon"
     )
@@ -83,7 +81,6 @@ pub fn xmul8(a: u8, b: u8) -> (u8, u8) {
             }
         } else if #[cfg(all(
             not(feature="no-xmul"),
-            feature="nightly",
             target_arch="aarch64",
             target_feature="neon"
         ))] {
@@ -109,7 +106,6 @@ pub fn xmul8(a: u8, b: u8) -> (u8, u8) {
     ),
     all(
         not(feature="no-xmul"),
-        feature="nightly",
         target_arch="aarch64",
         target_feature="neon"
     )
@@ -133,7 +129,6 @@ pub fn xmul16(a: u16, b: u16) -> (u16, u16) {
             }
         } else if #[cfg(all(
             not(feature="no-xmul"),
-            feature="nightly",
             target_arch="aarch64",
             target_feature="neon"
         ))] {
@@ -159,7 +154,6 @@ pub fn xmul16(a: u16, b: u16) -> (u16, u16) {
     ),
     all(
         not(feature="no-xmul"),
-        feature="nightly",
         target_arch="aarch64",
         target_feature="neon"
     )
@@ -183,7 +177,6 @@ pub fn xmul32(a: u32, b: u32) -> (u32, u32) {
             }
         } else if #[cfg(all(
             not(feature="no-xmul"),
-            feature="nightly",
             target_arch="aarch64",
             target_feature="neon"
         ))] {
@@ -209,7 +202,6 @@ pub fn xmul32(a: u32, b: u32) -> (u32, u32) {
     ),
     all(
         not(feature="no-xmul"),
-        feature="nightly",
         target_arch="aarch64",
         target_feature="neon"
     )
@@ -234,7 +226,6 @@ pub fn xmul64(a: u64, b: u64) -> (u64, u64) {
             }
         } else if #[cfg(all(
             not(feature="no-xmul"),
-            feature="nightly",
             target_arch="aarch64",
             target_feature="neon"
         ))] {
@@ -260,7 +251,6 @@ pub fn xmul64(a: u64, b: u64) -> (u64, u64) {
     ),
     all(
         not(feature="no-xmul"),
-        feature="nightly",
         target_arch="aarch64",
         target_feature="neon"
     )
@@ -296,7 +286,6 @@ pub fn xmul128(a: u128, b: u128) -> (u128, u128) {
             }
         } else if #[cfg(all(
             not(feature="no-xmul"),
-            feature="nightly",
             target_arch="aarch64",
             target_feature="neon"
         ))] {
@@ -327,7 +316,6 @@ mod test {
         ),
         all(
             not(feature="no-xmul"),
-            feature="nightly",
             target_arch="aarch64",
             target_feature="neon"
         )