feat(sha2): use latest sha2-asm and enable M1

[dignifiedquire]

Depends on RustCrypto/asm-hashes#35

Benchmarks on Mac Mini M1:

Soft

test bench1_10    ... bench:          40 ns/iter (+/- 0) = 250 MB/s
test bench2_100   ... bench:         390 ns/iter (+/- 1) = 256 MB/s
test bench3_1000  ... bench:       3,869 ns/iter (+/- 8) = 258 MB/s
test bench4_10000 ... bench:      38,662 ns/iter (+/- 175) = 258 MB/s

ASM

test bench1_10    ... bench:           6 ns/iter (+/- 0) = 1666 MB/s
test bench2_100   ... bench:          46 ns/iter (+/- 0) = 2173 MB/s
test bench3_1000  ... bench:         463 ns/iter (+/- 2) = 2159 MB/s
test bench4_10000 ... bench:       4,636 ns/iter (+/- 12) = 2157 MB/s

[tarcieri]

It is. See the discussion here:

RustCrypto/utils#378 (comment)

Perhaps leave a TODO to use the cpufeatures crate for this detection, once ARM support lands?

[tarcieri]

Confirm the tests pass on my M1 Mac Mini:

% cargo test --features asm
    Updating git repository `https://github.com/dignifiedquire/asm-hashes`
    Updating crates.io index
  Downloaded hex-literal v0.2.1
  Downloaded version_check v0.9.3
  Downloaded opaque-debug v0.3.0
  Downloaded cc v1.0.67
  Downloaded proc-macro-hack v0.5.19
  Downloaded digest v0.9.0
  Downloaded cfg-if v1.0.0
  Downloaded block-buffer v0.9.0
  Downloaded blobby v0.1.2
  Downloaded byteorder v1.4.3
  Downloaded generic-array v0.14.4
  Downloaded typenum v1.13.0
  Downloaded hex-literal-impl v0.2.2
  Downloaded 13 crates (227.5 KB) in 0.44s
   Compiling version_check v0.9.3
   Compiling typenum v1.13.0
   Compiling proc-macro-hack v0.5.19
   Compiling cc v1.0.67
   Compiling byteorder v1.4.3
   Compiling opaque-debug v0.3.0
   Compiling cfg-if v1.0.0
   Compiling blobby v0.1.2
   Compiling generic-array v0.14.4
   Compiling sha2-asm v0.6.0 (https://github.com/dignifiedquire/asm-hashes?branch=arm-m1#8f5a9eb1)
   Compiling hex-literal-impl v0.2.2
   Compiling block-buffer v0.9.0
   Compiling digest v0.9.0
   Compiling sha2 v0.9.3 (/Users/tony/src/hashes/sha2)
   Compiling hex-literal v0.2.1
    Finished test [optimized + debuginfo] target(s) in 1m 39s
     Running /Users/tony/src/hashes/target/debug/deps/sha2-be3e015d4c750ddb

running 0 tests

test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s

     Running /Users/tony/src/hashes/target/debug/deps/lib-208bf8ba67780be5

running 8 tests
test sha224_main ... ok
test sha512_224_main ... ok
test sha256_main ... ok
test sha512_256_main ... ok
test sha384_main ... ok
test sha512_main ... ok
test sha256_1million_a ... ok
test sha512_1million_a ... ok

test result: ok. 8 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.02s

   Doc-tests sha2

running 1 test
test src/lib.rs - (line 22) ... ok

Edit: looks like they aren't passing on macOS/x86-64 in CI, though

[dignifiedquire]

Benchmarks on Mac Mini M1:

Soft

test bench1_10    ... bench:          40 ns/iter (+/- 0) = 250 MB/s
test bench2_100   ... bench:         390 ns/iter (+/- 1) = 256 MB/s
test bench3_1000  ... bench:       3,869 ns/iter (+/- 8) = 258 MB/s
test bench4_10000 ... bench:      38,662 ns/iter (+/- 175) = 258 MB/s

ASM

test bench1_10    ... bench:           6 ns/iter (+/- 0) = 1666 MB/s
test bench2_100   ... bench:          46 ns/iter (+/- 0) = 2173 MB/s
test bench3_1000  ... bench:         463 ns/iter (+/- 2) = 2159 MB/s
test bench4_10000 ... bench:       4,636 ns/iter (+/- 12) = 2157 MB/s