llvm dev - Jul 2017 - [GlobalISel] G_LOAD/G_STORE i64/f64 handling

Hi all,
I am working on enabling X86 using  GLobalIsel framework.
I have 32bit platform + float/double  configuration (-mtriple=i386-linux-gnu
-mattr=+sse2 )
               load i64, i64* %p1                   -   illegal, require
narrowScalar action
               load double, double * %p1   -    legal

What is the best approach to Legalize this case ? Should I mark G_LOAD/G_STORE
s64 as Custom?

Regards,
Igor Breger
---------------------------------------------------------------------
Intel Israel (74) Limited

This e-mail and any attachments may contain confidential material for
the sole use of the intended recipient(s). Any review or distribution
by others is strictly prohibited. If you are not the intended
recipient, please contact the sender and delete all copies.
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20170702/88f190e1/attachment.html>

On 7/2/2017 3:50 AM, Breger, Igor via llvm-dev wrote:
>
> Hi all,
>
> I am working on enabling X86 using  GLobalIsel framework.
>
> I have 32bit platform + float/double  configuration 
> (-mtriple=i386-linux-gnu -mattr=+sse2 )
>
>                load i64, i64* %p1             -   illegal, require 
> narrowScalar action
>
>                load double, double * %p1 -    legal
>
> What is the best approach to Legalize this case ? Should I mark 
> G_LOAD/G_STORE s64 as Custom?
>
In the world of GlobalISel, i64 and double are both s64, so "load i64"
and "load double" are the same operation; if one is Legal, both are 
Legal.  This should just work, barring any bugs.

The resulting code will probably be low-quality because there isn't any 
good way to lower the resulting transfers between SSE and integer 
registers, but you can put off dealing with that for now.

-Eli

-- 
Employee of Qualcomm Innovation Center, Inc.
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, a Linux
Foundation Collaborative Project

-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20170703/4667581b/attachment.html>

On Mon, Jul 3, 2017 at 4:28 PM, Friedman, Eli via llvm-dev <
llvm-dev at lists.llvm.org> wrote:
> On 7/2/2017 3:50 AM, Breger, Igor via llvm-dev wrote:
>
> Hi all,
>
> I am working on enabling X86 using  GLobalIsel framework.
>
> I have 32bit platform + float/double  configuration
> (-mtriple=i386-linux-gnu -mattr=+sse2 )
>
>                load i64, i64* %p1                   -   illegal, require
> narrowScalar action
>
>                load double, double * %p1   -    legal
>
>
>
> What is the best approach to Legalize this case ? Should I mark
> G_LOAD/G_STORE s64 as Custom?
>
>
>
>
> In the world of GlobalISel, i64 and double are both s64, so "load
i64" and
> "load double" are the same operation; if one is Legal, both are
Legal.
> This should just work, barring any bugs.
>
> The resulting code will probably be low-quality because there isn't any
> good way to lower the resulting transfers between SSE and integer
> registers, but you can put off dealing with that for now.
>
>
Ok but for 32-bit target, will GlobalISel handle the splitting of
i64/double on 2 32-bit registers automatically like the SelectionDAG is
currently doing?
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20170703/682ccc16/attachment.html>