[Mesa-dev] [PATCH 3/2] intel/nir: Call nir_lower_io_to_scalar_early

[Jason Ekstrand]

On Tue, Jul 31, 2018 at 9:55 AM Nils Wallménius <nils.wallmenius at gmail.com>
wrote:

> Hi, sorry for the drive-by but a comment in line below
>
> Den tis 31 juli 2018 16:09Jason Ekstrand <jason at jlekstrand.net> skrev:
>
>> Shader-db results on Kaby Lake:
>>
>>     total instructions in shared programs: 15166953 -> 15073611 (-0.62%)
>>     instructions in affected programs: 2390284 -> 2296942 (-3.91%)
>>     helped: 16469
>>     HURT: 505
>>
>>     total loops in shared programs: 4954 -> 4951 (-0.06%)
>>     loops in affected programs: 3 -> 0
>>     helped: 3
>>     HURT: 0
>> ---
>>  src/intel/compiler/brw_nir.c | 17 ++++++++++++-----
>>  1 file changed, 12 insertions(+), 5 deletions(-)
>>
>> diff --git a/src/intel/compiler/brw_nir.c b/src/intel/compiler/brw_nir.c
>> index 99ef6bec77f..a6e3e502cdd 100644
>> --- a/src/intel/compiler/brw_nir.c
>> +++ b/src/intel/compiler/brw_nir.c
>> @@ -713,6 +713,18 @@ brw_nir_link_shaders(const struct brw_compiler
>> *compiler,
>>     nir_validate_shader(*producer);
>>     nir_validate_shader(*consumer);
>>
>> +   const bool p_is_scalar =
>> +      compiler->scalar_stage[(*producer)->info.stage];
>> +   const bool c_is_scalar =
>> +      compiler->scalar_stage[(*producer)->info.stage];
>>
>
> Shouldn't the above index be s/producer/consumer/ ? else both booleans are
> the same.
>

Good catch!  That's a pre-existing bug.  I've fixed it the v2 I just sent.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.freedesktop.org/archives/mesa-dev/attachments/20180731/266c9533/attachment-0001.html>