Timeline for Why is this extra MUX here when shifting mantissa as part of a FP add?
Current License: CC BY-SA 4.0
7 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Mar 2 at 20:50 | vote | accept | EE18 | ||
| Mar 2 at 20:50 | comment | added | EE18 | Understood, thank you once again! | |
| Mar 2 at 20:36 | history | edited | Dave Tweed | CC BY-SA 4.0 |
added 263 characters in body
|
| Mar 2 at 20:30 | comment | added | Dave Tweed | Yes, it's redundant. The barrel shifter can be (and probably is) designed to output zeros for shift amounts of 24 - 31. The AND gate is probably just there to reinforce the pedagogical point. | |
| Mar 2 at 20:09 | comment | added | EE18 | ...follow though. Why do we AND the [4] and [3] bits. Aren't these taken care of by the barrel shifter? That is, if they're both asserted then doesn't the barrel shifter do more than enough of a shift? This seems to be a redundant gate to me. | |
| Mar 2 at 20:09 | comment | added | EE18 | Ah I think I see, thanks so much! Just to confirm: the barrel shifter only accepts the low 5 (out of 8) bits of \$shamt\$, whence we see that it can do shifts of up to 31 bits (which, incidentally, is more than enough than is required for a mantissa of 23 bits). We cannot use less than this for the barrel shifter since \$2^4 = 16 < 23\$ wouldn't achieve all the shifts conceivably necessary. However, by omitting the shifts associated with more significant bits in \$shamt\$ being possible, if we didn't have the AND-OR logic then we wouldn't shift in these cases. There is one thing I don't... | |
| Mar 2 at 18:34 | history | answered | Dave Tweed | CC BY-SA 4.0 |