Fix/vzext

[Navaneeth-KunhiPurayil]

Vector integer extension operations have src EMUL smaller than LMUL and can have registers in the EMUL group. For example vzext.vf2 v2, v1 when LMUL=2 is a valid instruction.

Changelog

Fixed

• Assign lmul_vs2 with the appropriate src EMUL for illegal src operand check

Added

• Tests to check for integer extensions for different LMUL and SEW

Checklist

• Automated tests pass
• Changelog updated
• Code style guideline is observed

[mp-17]

Hi Navaneeth, thanks! This is probably better than before (at least it's more permissive and does not trap on legal code), but I am not sure it's fully complying with the specs req.

• the destination reg should be checked against the current lmul and sew
• source regs checked against emul and eew
• rules for overlapping are the following (look at the third requirement):

A destination vector register group can overlap a source vector register group only if one of the
following holds:
⚫ The destination EEW equals the source EEW.
⚫ The destination EEW is smaller than the source EEW and the overlap is in the lowest-numbered
part of the source register group (e.g., when LMUL=1, vnsrl.wi v0, v0, 3 is legal, but a destination
of v1 is not).
⚫ The destination EEW is greater than the source EEW, the source EMUL is at least 1, and the overlap
is in the highest-numbered part of the destination register group (e.g., when LMUL=8, vzext.vf4 v0,
v6 is legal, but a source of v0, v2, or v4 is not).

What happens if you try out the following?

With LMUL8
vzext.vf4 v0, v6 # ok
vzext.vf4 v0, v0 # trap
vzext.vf8 v4, v16 # trap

Do the last two cases trap?

Anyway, if the legal cases are now passing, we can merge this.

[Navaneeth-KunhiPurayil]

Hi Matteo,
Thanks for reviewing!

You are right, there are more cases to be considered. Here I just want to avoid trapping on legal code.
There is more to be done for src dst overlap checks.

For the 3 cases, you mentioned,
vzext.vf4 v0, v6 # ok
vzext.vf4 v0, v0 # trap
vzext.vf8 v4, v16 # trap

cases 1, 3 work as expected now.
For case 2, the overlap check needs to be added in general for narrowing/widening operations.

If you are okay with the changes, you can merge them.

[mp-17]

Great! Thanks again!