[Mesa-dev] [PATCH 12/20] i965/fs: Handle explicit flag sources in flags_read()

[Matt Turner]

On Fri, Jul 14, 2017 at 5:52 PM, Francisco Jerez <currojerez at riseup.net> wrote:
> Matt Turner <mattst88 at gmail.com> writes:
>
>> The implementations of the ARB_shader_ballot intrinsics will explicitly
>> read the flag as a source register.
>> ---
>>  src/intel/compiler/brw_fs.cpp | 18 ++++++++++++++----
>>  1 file changed, 14 insertions(+), 4 deletions(-)
>>
>> diff --git a/src/intel/compiler/brw_fs.cpp b/src/intel/compiler/brw_fs.cpp
>> index 97908a4563..de3aafafb1 100644
>> --- a/src/intel/compiler/brw_fs.cpp
>> +++ b/src/intel/compiler/brw_fs.cpp
>> @@ -862,9 +862,6 @@ namespace {
>>  unsigned
>>  fs_inst::flags_read(const gen_device_info *devinfo) const
>>  {
>> -   /* XXX - This doesn't consider explicit uses of the flag register as source
>> -    *       region.
>> -    */
>>     if (predicate == BRW_PREDICATE_ALIGN1_ANYV ||
>>         predicate == BRW_PREDICATE_ALIGN1_ALLV) {
>>        /* The vertical predication modes combine corresponding bits from
>> @@ -875,7 +872,20 @@ fs_inst::flags_read(const gen_device_info *devinfo) const
>>     } else if (predicate) {
>>        return flag_mask(this);
>>     } else {
>> -      return 0;
>> +      unsigned mask = 0;
>> +      for (int i = 0; i < sources; i++) {
>> +         if (src[i].file == ARF) {
>> +            if (src[i].nr == BRW_ARF_FLAG + 0 && src[i].subnr == 0)
>> +               mask |= 0b0001;
>> +            if (src[i].nr == BRW_ARF_FLAG + 0 && src[i].subnr == 1)
>> +               mask |= 0b0010;
>> +            if (src[i].nr == BRW_ARF_FLAG + 1 && src[i].subnr == 0)
>> +               mask |= 0b0100;
>> +            if (src[i].nr == BRW_ARF_FLAG + 1 && src[i].subnr == 1)
>> +               mask |= 0b1000;
>> +         }
>> +      }
>> +      return mask;
>
> This has the same problems as PATCH 5.  The solution I suggested there
> should work here too.

Thanks. Done locally.