Bit slice #2176
Bit slice #2176
Conversation
There was a problem hiding this comment.
Thanks @Angelica-Schell! I left a few comments on the new bit manipulation implementation; let me know if you'd like me to clarify anything.
| @@ -106,6 +116,7 @@ fn convert( | |||
| convert_from: NumType, | |||
| convert_to: NumType, | |||
| exponent: i32, | |||
| bits: bool, | |||
There was a problem hiding this comment.
Since you're listing all the arguments in the doc comment, you should probably add bits. And incidentally, it looks like the descriptions of convert_{from,to} are out of date, as they refer to "references" that were removed.
| 11110100 |
There was a problem hiding this comment.
Is there a reason that this test is being overwritten instead of adding a new one? (Or, if these tests aren't meant to be part of a permanent test suite, then they probably shouldn't be committed.)
| /// # Panics | ||
| /// | ||
| /// This function will panic if the input string cannot be parsed as a binary number. | ||
| fn binary_to_fixed_bit_slice( |
There was a problem hiding this comment.
Perhaps I'm missing something, but I was under the impression that it was the fixed_to_binary function where we were going to try a bit manipulation implementation? Having a second implementation of binary_to_fixed would still work for testing your benchmarking setup, though.
data-conversion/src/main.rs
Outdated
| let frac = fractional_part; | ||
| for i in 0..shift_amount { | ||
| if (frac & (1 << i)) != 0 { | ||
| frac_part_value += 1.0 / (1 << (shift_amount - i)) as f64; |
There was a problem hiding this comment.
Maybe the following comment is misguided, since I was originally thinking about bit manipulation in the context of fixed_to_binary. But I would sort of imagine that the goal for a bit manipulation version of binary_to_fixed would be to assemble the floating-point bit representation directly from the fixed-point bit representation and the exponent, in order to avoid doing any expensive floating-point operations. I.e., we would try to avoid doing these floating-point additions and divisions here?
Adding binary to fixed using bit slicing instead of multiplication. Use the -b flag to use this version instead