Add support for Zvksh extension #862
Conversation
model/riscv_insts_zvksh.sail
Outdated
| foreach (j from 16 to 23) | ||
| w[j] = zvk_sh_w(w[j - 16], w[j - 9], w[j - 3], w[j - 13], w[j - 6]); | ||
|
|
||
| write_velem_oct_vec(vd, SEW, vrev8([w[23], w[22], w[21], w[20], w[19], w[18], w[17], w[16]]), i); |
There was a problem hiding this comment.
Might be clearer to break the vrev8 out into its own line before this. Kind of easy for it to get lost otherwise when reading through this function.
There was a problem hiding this comment.
In regard to this discussion. Would it make sense to do something like:
| write_velem_oct_vec(vd, SEW, vrev8([w[23], w[22], w[21], w[20], w[19], w[18], w[17], w[16]]), i); | |
| write_velem_oct_vec(vd, SEW, | |
| vrev8([w[23], w[22], w[21], w[20], w[19], w[18], w[17], w[16]]), i); |
There was a problem hiding this comment.
It's definitely better. Does it really not work writing it to an intermediate variable?
There was a problem hiding this comment.
Unfortunately not
what works is:
let 'm = 4;
let w_out : vector(8, bits('m * 8)) = vrev8([w[23], w[22], w[21], w[20], w[19], w[18], w[17], w[16]]);
write_velem_oct_vec(vd, SEW, w_out, i);
|
@Timmmm yes some parts are from #248. I added @charmitro as a co-author in the first commit. Usually, I make a few commits during development, and then before merging, I squash everything into the first commit, with him listed as a co-author. |
nadime15
force-pushed
the
ext_zvksh
branch
2 times, most recently
from
ef87c0e to
c05d468
Compare
Co-authored-by: Charalampos Mitrodimas <[email protected]>
|
@pmundkur I just noticed that this PR has two approvals, so we could theoretically merge it unless there is a reason not to. The Rocq build is currently failing, I am not sure whether this needs to be resolved before merging or if it is acceptable for now. Let me know what you think. |
|
The Rocq output is failing because it isn't as clever as Sail about solving |
Add following instructions:
Tested with riscv-vector-tests.