aboutsummaryrefslogtreecommitdiff
path: root/c/README.md
diff options
context:
space:
mode:
authorJack O'Connor <[email protected]>2020-01-08 23:19:07 -0500
committerJack O'Connor <[email protected]>2020-01-09 09:48:52 -0500
commitaf9b44c8810d651355db48bca5fee86f7b4c93e3 (patch)
treeddc1f386d22b01a49cde8c53cc1832642f822451 /c/README.md
parent88dcee7005be962a81516f7863e70009d9caa2c9 (diff)
replace the C code README with a "not yet ready" remark
Diffstat (limited to 'c/README.md')
-rw-r--r--c/README.md33
1 files changed, 6 insertions, 27 deletions
diff --git a/c/README.md b/c/README.md
index 3c98c9b..380c15f 100644
--- a/c/README.md
+++ b/c/README.md
@@ -1,28 +1,7 @@
-# BLAKE3-c [![Actions Status](https://github.com/veorq/BLAKE3-c/workflows/tests/badge.svg)](https://github.com/veorq/BLAKE3-c/actions)
+This is the C implementation of BLAKE3. It's tested, and parts of it are
+linked into the Rust implementation for AVX-512 and NEON support.
+However, it doesn't yet have a friendly public interface.
-A very rough initial implementation of BLAKE3 in C. SSE4.1, AVX2,
-AVX-512, and NEON are supported, using compile-time feature selection in
-the Makefile.
-
-This implementation is simpler than the [Rust
-implementation](https://github.com/veorq/BLAKE3). It doesn't support
-multithreading, and it doesn't parallelize parent hashes, so throughput
-is lower.
-
-TODO:
-- CI testing for AVX-512 and NEON.
-- Cross-platform build, e.g. Windows.
-- Dynamic CPU feature detection, at least for x86.
-
-Example usage:
-
-```bash
-$ make avx2
-$ head -c 1000000 /dev/urandom | ./blake3
-43f2cae3cfd7678bc3a3ebdbf170608d19d5ebaad23e9d06291dba3269853608
-$ head -c 1000000 /dev/urandom | ./blake3 --length 50
-4fc0ee74a60aa77fb699821997498fd93f1a98bd03eaf2a7969c4b35fb742c233a7a161fd2a431605f6e92dcf4cd7d052102
-$ head -c 1000000 /dev/urandom | ./blake3 --keyed 0123456789abcdef0123456789abcdef0123456789abcdef0123456789abcdef
-8aee87b232fe90b042bf9119591e24409763a268139ff157d20021003e314064
-
-```
+This implementation is simpler than the Rust implementation. It doesn't
+support multithreading, and it doesn't parallelize parent hashes, so
+throughput is lower.