[Mesa-dev] [PATCH v2 1/8] intel: devinfo: store number of EUs per subslice

[Lionel Landwerlin]

On 19/03/18 23:34, Kenneth Graunke wrote:
> On Wednesday, March 14, 2018 10:19:07 AM PDT Lionel Landwerlin wrote:
>> This will be reused to store values reported by the kernel. The main
>> use case will be for use as the input values of the metric sets
>> equations for the INTEL_performance_queries extension. By storing this
>> information in the gen_device_info we make this non GL specific so
>> this can be reused by Vulkan if we ever have an equivalent extension.
>>
>> Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin at intel.com>
>> ---
>>   src/intel/dev/gen_device_info.c | 35 +++++++++++++++++++++++++++++++++--
>>   src/intel/dev/gen_device_info.h |  5 +++++
>>   2 files changed, 38 insertions(+), 2 deletions(-)
> Nice job tracking these all down, even in the Gen4-5 PRMs, and all the
> crazy fusing stuff.
>
>> @@ -260,6 +267,7 @@ static const struct gen_device_info gen_device_info_byt = {
>>      GEN7_FEATURES, .is_baytrail = true, .gt = 1,
>>      .num_slices = 1,
>>      .num_subslices = { 1, },
>> +   .num_eu_per_subslice = 4,
>>      .num_thread_per_eu = 8,
>>      .l3_banks = 1,
>>      .has_llc = false,
> I guess there aren't really notions of subslices prior to Haswell, but
> we're pretending to have 1 subslice, so num_eu_per_subslice is just the
> total number of EUs.  Seems fine...
>
>> @@ -468,6 +482,7 @@ static const struct gen_device_info gen_device_info_chv = {
>>      .has_integer_dword_mul = false,
>>      .num_slices = 1,
>>      .num_subslices = { 2, },
>> +   .num_eu_per_subslice = 8,
> You're probably going to have to update this to 6 based on the
> Cherryview fusing information.  Look for sseu in i965. :(

Yeah, so I've put the max number here.
Once queried by the kernel we update this number.

>
> So far, we've been putting the larger number (for the 2x8 model) in
> gen_device_info and correcting it down for the 2x6 model.
>
> We'll need to sort that out.  In the meantime, this looks good.
>
> Reviewed-by: Kenneth Graunke <kenneth at whitecape.org>