Crappy rng #148
Crappy rng #148
Conversation
| impl rand::RngCore for CrappyRng { | ||
|
|
||
| fn next_u32(&mut self) -> u32 { | ||
| self.next_u64() as u32 | ||
| } | ||
|
|
||
| fn next_u64(&mut self) -> u64 { | ||
| let mut x = self.0; | ||
| x ^= x << 13; | ||
| x ^= x >> 7; | ||
| x ^= x << 17; | ||
| self.0 = x; | ||
| x | ||
| } | ||
|
|
||
| fn fill_bytes(&mut self, dest: &mut [u8]) { | ||
| for chunk in dest.chunks_mut(8) { | ||
| let x = self.next_u64().to_be_bytes(); | ||
| chunk.copy_from_slice(&x[..chunk.len()]); | ||
|
|
||
| } | ||
| } | ||
|
|
||
| fn try_fill_bytes(&mut self, dest: &mut [u8]) -> Result<(), rand::Error> { | ||
| Ok(self.fill_bytes(dest)) |
There was a problem hiding this comment.
This code is duplicated in the 2 examples, I thought it to be better to duplicate the code instead of having it in the main lib
There was a problem hiding this comment.
I think we should put it in the main lib, but in a module which has #[cfg(test)] on it?
There was a problem hiding this comment.
From the examples, I cannot access something I put behind a #[cfg(test)] in the lib.
eg.
If I put #[cfg(test)] pub const CIAO: u8 = 0u8; in src/lib.rs
and I do cargo run --example pset_blind_coinjoin
I got
error[E0425]: cannot find value `CIAO` in crate `elements`
There was a problem hiding this comment.
Oh, derp, I missed that this was in examples rather than unit tests.
Now I'm unsure if we should be using a broken RNG given that this is "example" code ... we should probably go back to using rng_chacha :P
There was a problem hiding this comment.
Ok, but the wrong interpretation due the ambivalent nature of this examples/tests leave me the doubt of moving those in /tests
There was a problem hiding this comment.
I think for now let's just go with this, since it's blocking our move to rand 0.8. We can rearrange things later.
There was a problem hiding this comment.
Opened issue #149
|
|
||
|
|
||
| /// Xorshift | ||
| pub struct CrappyRng(u64); |
There was a problem hiding this comment.
I took a simple PRNG from Wikipedia,
for laziness it fail something like
rng = CrappyRng(1);
rng2 = CrappyRng(1);
assert_eq!( concat!( rng.fill_bytes( 2_bytes_buffer), rng.fill_bytes( 2_bytes_buffer)), rng2.fill_bytes(4_bytes_buffer))
I don't know if it matters here or in general
There was a problem hiding this comment.
Can you link the wikipedia page in the comments?
My impulse would've been to use a LCG, eg any of the params from https://en.wikipedia.org/wiki/Linear_congruential_generator#Parameters_in_common_use but if this one is on wikipedia somewhere I'm sure it's just as good :P.
I don't think that invariant matters at all .. is it supposed to hold for RngCore? Why would it be your responsibility to make it hold when you only implement one method?
There was a problem hiding this comment.
Can you link the wikipedia page in the comments?
To start I post it here https://en.wikipedia.org/wiki/Xorshift#Example_implementation
I don't think that invariant matters at all .. is it supposed to hold for RngCore?
No, I found in the doc:
many implementations of fill_bytes consume a whole number of u32 or u64 values and drop any remaining unused bytes
RCasatta
force-pushed
the
crappy_rng
branch
from
a1e8020 to
948416b
Compare
2 participants
As suggested in #147 (comment)