[Mesa-dev] [PATCH 05/12] i965/fs: Use offset a lot more places

[Matt Turner]

On Mon, Sep 22, 2014 at 1:42 PM, Matt Turner <mattst88 at gmail.com> wrote:
> On Mon, Sep 22, 2014 at 1:40 PM, Jason Ekstrand <jason at jlekstrand.net> wrote:
>> On Fri, Sep 19, 2014 at 5:34 PM, Matt Turner <mattst88 at gmail.com> wrote:
>>> On Fri, Sep 19, 2014 at 1:10 PM, Jason Ekstrand <jason at jlekstrand.net>
>>> > diff --git a/src/mesa/drivers/dri/i965/brw_fs.cpp
>>> > b/src/mesa/drivers/dri/i965/brw_fs.cpp
>>> > index af8c087..ea91705 100644
>>> > --- a/src/mesa/drivers/dri/i965/brw_fs.cpp
>>> > +++ b/src/mesa/drivers/dri/i965/brw_fs.cpp
>>> > @@ -310,8 +310,8 @@ fs_visitor::VARYING_PULL_CONSTANT_LOAD(const fs_reg
>>> > &dst,
>>> >           inst->mlen = 1 + dispatch_width / 8;
>>> >     }
>>> >
>>> > -   vec4_result.reg_offset += (const_offset & 3) * scale;
>>> > -   instructions.push_tail(MOV(dst, vec4_result));
>>> > +   fs_reg result = offset(vec4_result, (const_offset & 3) * scale);
>>> > +   instructions.push_tail(MOV(dst, result));
>>>
>>> Isn't this going to cause us to copy an fs_reg twice, rather than just
>>> setting .reg_offset?
>>>
>>> I'd like to check the generated code.
>>
>> What's your concern there?  Just that we're useing more CPU?
>
> Yeah, that we're now potentially copying an fs_reg twice when our
> purpose is just to set a single integer.

Ignore this feedback for now. You've got a bunch of patches that would
have to be rebased if we changed this and it should be really trivial
to fix it up after the fact.