RFR: 8322179: RISC-V: Implement SHA-1 intrinsic [v3]

[Hamlin Li]

On Wed, 10 Jan 2024 03:02:09 GMT, Fei Yang <fyang at openjdk.org> wrote:

> Simply ran `micro:java.security.MessageDigests` JMH on my Lichee-pi-4a board, seems there is a small regression for the `MessageDigests.getAndDigest` (length = 64) case:
> 
> ```
> Before:
> MessageDigests.digest                 SHA-1        64     DEFAULT  thrpt   15   417.311 ?  2.686  ops/ms
> MessageDigests.digest                 SHA-1     16384     DEFAULT  thrpt   15     5.206 ?  0.008  ops/ms
> MessageDigests.getAndDigest           SHA-1        64     DEFAULT  thrpt   15   404.769 ? 14.810  ops/ms
> MessageDigests.getAndDigest           SHA-1     16384     DEFAULT  thrpt   15     5.106 ?  0.046  ops/ms
> 
> After:
> MessageDigests.digest                 SHA-1        64     DEFAULT  thrpt   15   518.057 ?  5.935  ops/ms
> MessageDigests.digest                 SHA-1     16384     DEFAULT  thrpt   15     5.569 ?  0.009  ops/ms
> MessageDigests.getAndDigest           SHA-1        64     DEFAULT  thrpt   15   378.184 ? 37.425  ops/ms
> MessageDigests.getAndDigest           SHA-1     16384     DEFAULT  thrpt   15     5.515 ?  0.017  ops/ms
> ```

Hey, 
Thanks for testing.
The data is interesting, seems there is just a little improvement when data is `16384`, compared to my test result more than 7X speedup. I will try different test environment. At the same time, can you help to double check your env whether the sha-1 intrinsic is enabled/disabled as expected?

-------------

PR Comment: https://git.openjdk.org/jdk/pull/17130#issuecomment-1884624161