[llvm-commits] [PATCH] Remove redundant register copies when inserting to Q registers on ARM

[James Molloy]

Hi Jakob,

That would be neat, and indeed makes sense and works for i32 set-lanes:

def : Pat<(insertelt (v4i32 QPR:$src1), GPR:$src2, imm:$lane),
          (v4i32 (INSERT_SUBREG QPR:$src1,
                  GPR:$src2,
                  (SSubReg_f32_reg imm:$lane)))>;

But this doesn't work for i16 or i8; lane widths that are smaller than
the smallest subregister (ssubx).

My patch as sent supports set-lane operations for all integer bitwidths.

Cheers,

James

On Fri, 2012-08-31 at 15:58 +0100, Jakob Stoklund Olesen wrote:
> On Aug 31, 2012, at 1:27 AM, James Molloy <James.Molloy at arm.com> wrote:
>
> > Hi,
> >
> > The attached patch improves code generation for NEON when setting one
> > (<= 32bit) lane of a Q (128 bit) register.
> >
> > There is no such instruction to implement this natively, because the
> > vmov.32 instructions work on D registers, not Q registers.
> >
> > Previously a pattern was used for this, something like
> >
> >  (INSERT_SUBREG (<set lane> (EXTRACT_SUBREG ... ) ) )
>
> Couldn't you simply insert_subreg the GPR directly into an ssubx sub-register of the QPR?
>
> Cross class inserts are fine as long as copyPhysReg can do it.
>
>
>
> > But this caused inefficient code to be generated because when it got to
> > the machineinstr stage, the live region for the base Q register and the
> > mutated D subregister overlapped, so they couldn't be coalesced and you
> > ended up with a pointless copy:
> >
> >  mov r0, #255
> >  vorr d19, d21, d21
> >  vmov.32 d19[1], r0
> >  vorr d21, d19, d19
> >
> > This patch adds pseudo instructions that set a lane of a Q register
> > instead. This is similar to how loads and stores are handled.
> >
> > Please review!
> >
> > Cheers,
> >
> > James
> >
> > -- IMPORTANT NOTICE: The contents of this email and any attachments are confidential and may also be privileged. If you are not the intended recipient, please notify the sender immediately and do not disclose the contents to any other person, use it for any purpose, or store or copy the information in any medium.  Thank you.
> > <remove-redundant-vorr.diff>
> > _______________________________________________
> > llvm-commits mailing list
> > llvm-commits at cs.uiuc.edu
> > http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
>


-- IMPORTANT NOTICE: The contents of this email and any attachments are confidential and may also be privileged. If you are not the intended recipient, please notify the sender immediately and do not disclose the contents to any other person, use it for any purpose, or store or copy the information in any medium.  Thank you.