jameysharp commented on issue #5873:
@afonso360, you've been working on the sign-extension bits this PR relies on, as well as the
select
lowerings that this affects. What are your thoughts on this PR?I was wondering if it would be better to pass an
ExtendOp
tonormalize_cmp_value
, instead of anIntCC
. Most of the callers always want a zero-extend and some of them have to fake a condition code in this PR, so maybe passing the choice of sign-extending or zero-extending directly would be more clear.I don't know what the best way to do this is. I think I understand this PR though, and if it looks good to @afonso360 then it looks good to me.
yuyang-ok commented on issue #5873:
@jameysharp It is hard to decide use
IntCC
orExtendOp
as parameter tonormalize_cmp_value
too.
Because the function namednormalize_cmp_value
I thought maybe better pass theIntCC
.
ActualyIntCC.Equal
andIntCC.NotEqual
make no differrence.
yuyang-ok edited a comment on issue #5873:
@jameysharp It is hard to decide use
IntCC
orExtendOp
as parameter tonormalize_cmp_value
to me too.
Because the function namednormalize_cmp_value
I thought maybe better pass theIntCC
.
ActualyIntCC.Equal
andIntCC.NotEqual
make no differrence.
yuyang-ok commented on issue #5873:
@jameysharp Maybe a
convert
So we can acceptIntCC
andExtendOP
both.
yuyang-ok edited a comment on issue #5873:
@jameysharp Maybe a
convert
So we can acceptIntCC
andExtendOP
both.
Likeconvert Intcc ExtendOp convert_intcc_to_extend_op
andnormalize_cmp_value
acceptExtendOp
as parameter.
yuyang-ok edited a comment on issue #5873:
@jameysharp Maybe a
convert
So we can acceptIntCC
andExtendOP
both.
Like(convert Intcc ExtendOp convert_intcc_to_extend_op)
andnormalize_cmp_value
acceptExtendOp
as parameter.
yuyang-ok commented on issue #5873:
@jameysharp I have changed the parameter to
ExtendOp
.
yuyang-ok commented on issue #5873:
@jameysharp @afonso360 I think we are ready.
Last updated: Nov 22 2024 at 16:03 UTC