Add c.sext.w and zext.w #290
Conversation
Signed-off-by: Afonso Oliveira <Afonso.Oliveira@synopsys.com>
AFOliveira
force-pushed
the
Instructions
branch
from
50a1135
to
27a708c
Compare
| @@ -1 +1,3 @@ | |||
| c.zext.w rd_rs1_p 1..0=1 15..13=4 12..10=7 6..5=3 4..2=4 | |||
|
|
|||
| $pseudo_op rv64_c::c.addiw c.sext.w rd_rs1_n0 15..13=1 12=0 6..2=0 1..0=1 | |||
There was a problem hiding this comment.
c.addiw is part of the C extension, not Zcb, so this should be moved to rv64_c.
There was a problem hiding this comment.
I might be wrong here, but c.addiw is identified as part of the c extension by rv64_c::c.addiw and c.sext.w pseudo instruction appears to only be available in rv64_zcb. This is the case in the binutils and in the ISA manual
There was a problem hiding this comment.
It looks like Binutils picks up sext.w as a pseudoinstruction for c.addiw when the C extension is enabled: https://github.com/bminor/binutils-gdb/blob/2754d75a11556577fcddf23718bfbff7d0489a3f/opcodes/riscv-opc.c#L580
But it only picks up c.sext.w as a pseudoinstruction for c.addiw when Zcb is enabled.
This all seems pretty illogical to me, but I guess we are stuck with the binutils behavior at this point for compatibility reasons, right @kito-cheng?
There was a problem hiding this comment.
Although it seems ilogical, my guess is binutils is doing that because the pseudo-instruction is only mentioned as a note on Zcb, maybe if the ISA clarified that, we could change all the others.
There was a problem hiding this comment.
@aswaterman Just a ping. Do we want to keep this on hold?
Technically, the PR is according to the rest of the sources. if anything changes in the meantime, I'll be happy to address those changes and PR any update.
I'm now looking on to binutils and the ISA and trying to fix gaps between all the sources.
c.sext.w
zext.w
I also found that this repo doesn't have the any instruction with acquire and release, like "amoswap.b.rl", is there a reason for these, or can i also commit them?