> On Jul 12, 2017, at 10:28 AM, David Miller wrote:
>
> From: Qing Zhao
> Date: Wed, 12 Jul 2017 08:49:52 -0500
>
>> and it also clearly mentioned that “specially aligned memory might
>> use this constraint”.
>
> It guarantees the achieve the
From: Qing Zhao
Date: Wed, 12 Jul 2017 08:49:52 -0500
> and it also clearly mentioned that “specially aligned memory might
> use this constraint”.
It guarantees the achieve the opposite of what you are trying to do.
That is, it can be used to guarantee that something is
> Actually, My major question is whether the current handling of
> special_memory_constraint in lra_constraints.c is correct or NOT based on
> GCC internal documentation. I thought that’s independent from this
> misaligned insns generation for M8, but looks I was wrong.
The answer is yes, the
From: Eric Botcazou
Date: Wed, 12 Jul 2017 01:19:03 +0200
>> we add this new constraint as:
>>
>> ;; We need a special memory constraint for the misaligned memory access
>> ;; This is only for TARGET_MISALIGN target
>> (define_special_memory_constraint "B"
>> "Memory
> we add this new constraint as:
>
> ;; We need a special memory constraint for the misaligned memory access
> ;; This is only for TARGET_MISALIGN target
> (define_special_memory_constraint "B"
> "Memory reference whose address is misaligned"
> (and (match_code "mem")
> (match_test
> On Jul 11, 2017, at 4:24 PM, Eric Botcazou wrote:
>
>> we need to generate misaligned load/store insns ONLY for misaligned memory
>> access, therefore need a new constraints for misaligned address.
>
> Why? What happens exactly if the memory access turns out to be
> we need to generate misaligned load/store insns ONLY for misaligned memory
> access, therefore need a new constraints for misaligned address.
Why? What happens exactly if the memory access turns out to be aligned?
--
Eric Botcazou
> On Jul 11, 2017, at 2:00 PM, Eric Botcazou wrote:
>
>> the problem I had is:
>>
>> 1. we added a new special_memory_constraint for misaligned memory access,
>> one important requirement for this new special_memory_constraint is, the
>> address of the memory access is
> the problem I had is:
>
> 1. we added a new special_memory_constraint for misaligned memory access,
> one important requirement for this new special_memory_constraint is, the
> address of the memory access is misaligned.
OK, it's the other way around from the usage of
thanks for the replying.
> On Jul 11, 2017, at 2:44 AM, Eric Botcazou wrote:
>
>> From the above doc, the major difference between a memory_constraint and a
>> special_memory_constraint is: whether "reload can or cannot make them match
>> by reloading the address".
>
>
> From the above doc, the major difference between a memory_constraint and a
> special_memory_constraint is: whether "reload can or cannot make them match
> by reloading the address".
Right, i.e. by just changing the form of the address (instead of the address
itself).
> For memory_constraint,
Hi,team:
The following doc for memory_constraint and special_memory_constraint seems
imply that the handling of the special_memory_constraint in lra-constraints.c
is NOT correct:
(https://gcc.gnu.org/onlinedocs/gccint/Define-Constraints.html#Define-Constraints)
MD Expression:
12 matches
Mail list logo