On Mon, Sep 21, 2015 at 2:16 PM, Jason Ekstrand <[email protected]> wrote:
> On Mon, Sep 21, 2015 at 2:04 PM, Matt Turner <[email protected]> wrote:
>> With NIR:
>>
>> instructions in affected programs: 111508 -> 109193 (-2.08%)
>> helped: 507
>
> *sigh* Any idea who's emitting these MOV's? If it's vec_to_movs, we
> should probably fix it one of these days. In any case,
No idea.
I found it by looking at a shader like
team-fortress-2/2200.shader_test -- which contains a series of no-op
moves inside an if/endif.
>> Without NIR:
>>
>> instructions in affected programs: 28763 -> 28474 (-1.00%)
>> helped: 186
>> ---
>> src/mesa/drivers/dri/i965/brw_vec4.cpp | 21 +++++++++++++++++++++
>> 1 file changed, 21 insertions(+)
>>
>> diff --git a/src/mesa/drivers/dri/i965/brw_vec4.cpp
>> b/src/mesa/drivers/dri/i965/brw_vec4.cpp
>> index ed49cd3..d09a8dd 100644
>> --- a/src/mesa/drivers/dri/i965/brw_vec4.cpp
>> +++ b/src/mesa/drivers/dri/i965/brw_vec4.cpp
>> @@ -1021,6 +1021,27 @@ vec4_visitor::opt_register_coalesce()
>> inst->src[0].abs || inst->src[0].negate || inst->src[0].reladdr)
>> continue;
>>
>
> Could we add a quick comment here that we're detecting a no-op MOV?
Sure. I added
/* Remove no-op MOVs */
above the if statement.
> With that,
>
> Reviewed-by: Jason Ekstrand <[email protected]>
Thanks!
_______________________________________________
mesa-dev mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/mesa-dev
