Jets for min, max, and peg
#117
Conversation
|
I rewrote |
|
@eamsden I applied your changes. There was a "wrong bitshift direction" bug that didn't get caught by the tests, because it would allocate too many words, so there was no actual problem (just memory inefficiency). I added tests for the case, but it looks like the bug is not detectable after the fact because the excessive size is resolved by normalization to atom. I'll leave the tests in, but I'm not sure how we would catch bugs like this with tests. |
Good catch! The test would have to be somewhat imprecise and check that the allocation didn't exceed what was reasonable. That wrong bitshift would have increased the allocation size from what was necessary by a factor of 4k, so it shouldn't be too hard to catch with thresholding. I'll make an issue to have tests like this. |
| @@ -64,16 +63,15 @@ pub fn jet_peg(context: &mut Context, subject: Noun) -> Result { | |||
| let b_bits = met(0, b); | |||
| let out_bits = a_bits + b_bits - 1; | |||
|
|
|||
| let out_words = (out_bits + 63) << 6; // bits to 8-byte words | |||
| let out_words = (out_bits + 63) >> 6; // bits to 8-byte words | |||
No description provided.