llvm dev - Apr 2019 - How to reliably get memory operand size

Hi All,

I'm working on a MIR pass that tries to duplicate memory operation
instructions. For example, if there is a move instruction "movq (%rax),
%rbx", I will insert another MIR instruction before it. For other
instructions, like add, sub, I also insert a mov instruction.

Now my question is, is there any way to obtain the size of the memory
operation? in the given example, the size should be 8, while for "movslq
(%rax), %tbx)", the size should be 4.

I have tried memoperand size, but apparently the memoperand is optional and
is not necessarily always there. So other than parsing the opcode, do we
have any reliably way to get the memory operation size?

Thanks in advance.

Regards,
Hong Hu
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20190422/b23e67e5/attachment.html>

Hi,
> On Apr 22, 2019, at 1:21 PM, Hong Hu via llvm-dev <llvm-dev at
lists.llvm.org> wrote:
> 
> Hi All,
> 
> I'm working on a MIR pass that tries to duplicate memory operation
instructions. For example, if there is a move instruction "movq (%rax),
%rbx", I will insert another MIR instruction before it. For other
instructions, like add, sub, I also insert a mov instruction.
> 
> Now my question is, is there any way to obtain the size of the memory
operation? in the given example, the size should be 8, while for "movslq
(%rax), %tbx)", the size should be 4.
As far as I know, every target has its own way of doing this. For AArch64, we
have AArch64InstrInfo::getMemOpInfo which can give you the width. Maybe on X86
you can use things from llvm/lib/Target/X86/MCTargetDesc/X86BaseInfo.h for this.

+Craig for X86.
> 
> I have tried memoperand size, but apparently the memoperand is optional and
is not necessarily always there. So other than parsing the opcode, do we have
any reliably way to get the memory operation size?
Right, the memoperands are not always there.

Sorry for the very late answer,

— 
Francis
> 
> Thanks in advance. 
> 
> Regards,
> Hong Hu
> _______________________________________________
> LLVM Developers mailing list
> llvm-dev at lists.llvm.org
> https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev