[PATCH] RISC-V : Support bitmanip-0.93 ZBA/ZBB/ZBC instructions

[Jim Wilson]

On Wed, Feb 24, 2021 at 12:18 PM Jim Wilson <jimw@sifive.com> wrote:

> On Wed, Feb 24, 2021 at 1:19 AM Jan Beulich <jbeulich@suse.com> wrote:
>
>> I notice the xor vs xnor testsuite aspect pointed out in my
>> post from Jan 18th was addressed, but the other questions
>> remain:
>>
>
> We had to rewrite the patch from scratch due to a copyright assignment
> problem.  So there could be no progress until we had the new patch.  The
> new patch doesn't repeat the xor mistake in the original patch.
>
>>
>> For consistency with other insns taking immediate operands,
>> shouldn't slli.uw one have a sll.uw alias?
>>
>
> Yes, that makes sense.  We can add that in a follow on patch.
>
>>
>> Also two perhaps more spec related questions: Why does slli.uw
>> allow for 7-bit wide shamt? Any shift count 32 and up is not
>> in need of this new insn, as slli will yield the same result.
>> (I could see the need in RV128, where counts up to 95 would be
>> needed, but right now talk is - I take it - mainly of RV32 and
>> RV64.)
>>
>
> RV128 needs 7 bits for shift counts.  So we have to reserve 7 bits in the
> encoding for future expansion.  slli.uw only uses 5 bits of that field.
> slli.uw uses the same instruction format as other instructions that can use
> the entire 7 bits of the shift count, which is why the field needs to be 7
> bits.  This is done to reduce the number of instruction formats that need
> to be decoded.  But a slli.uw with either of the upper 2 bits set should
> probably be a reserved encoding.  Or maybe use a different encoding.  This
> should be raised with the bitmanip committee.
>
>>
>> The current draft also doesn't say anything about hints; one
>> could assume destination being x0 may get treated the same as
>> in the base spec, but in the absence of this being said
>> explicitly one could as well imply these are all simply
>> sort-of-nop-s, or even illegal.
>>
>
> They are all nops.  And all hints use nop encodings.  A hint is supposed
> to be executed as a nop, except when it is a hint.  This makes it safe to
> use a hint anywhere.  Since it will be executed as a nop if the hardware
> doesn't support the hint.  But this is another issue for the bitmanip
> committee, whether the new nop encodings should be allowed to be used for
> hints.
>

I created 3 issues in github.com/riscv/riscv-bitmanip, one for each of the
3 questions you asked.

Jim