On 2/1/2024 1:14 PM, Zbigniew Kempczyński wrote:
> On Thu, Feb 01, 2024 at 10:30:16AM +0000, Matthew Auld wrote:
>> On Xe2 compression is handled via PAT indexes and flat CCS, even on
>> igpu. Ensure the device has flat CCS support on such hardware, to
>> determine if these tests are valid.
>>
>> v2: (Zbigniew)
>>   - There is at least one uncompressed test also, which is valid
>>     regardless of CCS. Move the skip deeper.
>>
>> Signed-off-by: Matthew Auld<matthew.auld@intel.com>
>> Cc: Akshata Jahagirdar<akshata.jahagirdar@intel.com>
>> Cc: Zbigniew Kempczyński<zbigniew.kempczynski@intel.com>
>> ---
>>   tests/intel/xe_ccs.c | 4 ++++
>>   1 file changed, 4 insertions(+)
>>
>> diff --git a/tests/intel/xe_ccs.c b/tests/intel/xe_ccs.c
>> index 7d0e8ed7a..3430db2d2 100644
>> --- a/tests/intel/xe_ccs.c
>> +++ b/tests/intel/xe_ccs.c
>> @@ -532,6 +532,7 @@ static void block_copy_test(int xe,
>>   			    struct igt_collection *set,
>>   			    enum copy_func copy_function)
>>   {
>> +	uint16_t dev_id = intel_get_drm_devid(xe);
>>   	struct drm_xe_engine_class_instance inst = {
>>   		.engine_class = DRM_XE_ENGINE_CLASS_COPY,
>>   	};
>> @@ -540,6 +541,9 @@ static void block_copy_test(int xe,
>>   	uint32_t vm, exec_queue;
>>   	int tiling;
>>   
>> +	if (AT_LEAST_GEN(dev_id, 20) && config->compression)
>> +		igt_require(HAS_FLATCCS(xe));
> Just to make sure - above will lead to skip on xe2 if there's
> no flatccs.
>
>> +
>>   	if (config->compression && !blt_block_copy_supports_compression(xe))
>>   		return;
> This will check compression either, but indirectly via .cmds_info
> set to xe2_cmds_info in which we use BLT_CMD_SUPPORTS_COMPRESSION
> flag for block-copy command.

on a side note, IIUC the BLT_CMD_SUPPORTS_COMPRESSION flag doesn't make 
sense anymore due to xe2+ unified compression?

- Akshata

> Do we have any Xe2 without compression on block-copy?
>
> --
> Zbigniew
>
>>   
>> -- 
>> 2.43.0
>>