Re: [Intel-xe] [PATCH v4 3/9] drm/xe/xe2: Updates on XY_CTRL_SURF_COPY_BLT

["Ghimiray, Himal Prasad" <himal.prasad.ghimiray@intel.com>]

[-- Attachment #1: Type: text/plain, Size: 3551 bytes --]


On 07-12-2023 04:52, Matt Roper wrote:
> On Wed, Dec 06, 2023 at 10:01:20AM +0530, Himal Prasad Ghimiray wrote:
>> - The XY_CTRL_SURF_COPY_BLT instruction operationg on ccs data expects
>> size in pages of main memory for which CCS data should be copied.
>> - The bitfield representing copy size in XY_CTRL_SURF_COPY_BLT has
>> shifted one bit higher in the instruction.
>>
>> v2:
>>   - Fix the num_pages for ccs size calculation.
>>   - Address nits (Thomas)
>>
>> Reviewed-by: Thomas Hellström<thomas.hellstrom@linux.intel.com>
>> Cc: Thomas Hellström<thomas.hellstrom@linux.intel.com>
>> Signed-off-by: Himal Prasad Ghimiray<himal.prasad.ghimiray@intel.com>
>> ---
>>   drivers/gpu/drm/xe/regs/xe_gpu_commands.h |  1 +
>>   drivers/gpu/drm/xe/xe_migrate.c           | 21 +++++++++++++++------
>>   2 files changed, 16 insertions(+), 6 deletions(-)
>>
>> diff --git a/drivers/gpu/drm/xe/regs/xe_gpu_commands.h b/drivers/gpu/drm/xe/regs/xe_gpu_commands.h
>> index 1f9c32e694c6..6ff6c7aaa648 100644
>> --- a/drivers/gpu/drm/xe/regs/xe_gpu_commands.h
>> +++ b/drivers/gpu/drm/xe/regs/xe_gpu_commands.h
>> @@ -13,6 +13,7 @@
>>   #define   DST_ACCESS_TYPE_SHIFT		20
>>   #define   CCS_SIZE_MASK			0x3FF
>>   #define   CCS_SIZE_SHIFT		8
>> +#define   XE2_CCS_SIZE_SHIFT		9
> It would be preferred if we just removed the shifts and used
> FIELD_PREP() to put values into proper field masks instead.
>
>>   #define   XY_CTRL_SURF_MOCS_MASK	GENMASK(31, 26)
>>   #define   XE2_XY_CTRL_SURF_MOCS_INDEX_MASK	GENMASK(31, 28)
>>   #define   NUM_CCS_BYTES_PER_BLOCK	256
>> diff --git a/drivers/gpu/drm/xe/xe_migrate.c b/drivers/gpu/drm/xe/xe_migrate.c
>> index e8b567708ac0..7ef068451b59 100644
>> --- a/drivers/gpu/drm/xe/xe_migrate.c
>> +++ b/drivers/gpu/drm/xe/xe_migrate.c
>> @@ -526,21 +526,30 @@ static void emit_copy_ccs(struct xe_gt *gt, struct xe_bb *bb,
>>   	struct xe_device *xe = gt_to_xe(gt);
>>   	u32 *cs = bb->cs + bb->len;
>>   	u32 num_ccs_blks;
>> +	u32 num_pages;
>> +	u32 ccs_copy_size;
>>   	u32 mocs;
>>   
>> -	num_ccs_blks = DIV_ROUND_UP(xe_device_ccs_bytes(gt_to_xe(gt), size),
>> -				    NUM_CCS_BYTES_PER_BLOCK);
>> -	xe_gt_assert(gt, num_ccs_blks <= NUM_CCS_BLKS_PER_XFER);
>> +	if (GRAPHICS_VERx100(xe) >= 2000) {
>> +		num_pages = DIV_ROUND_UP(size, XE_PAGE_SIZE);
>> +		xe_gt_assert(gt, num_pages <= 1024);
> Rather than using a magic number here, FIELD_FIT() might be better.  We
> can do the same for the other branch as well and get rid of
> NUM_CCS_BLKS_PER_XFER too.

Sure. Using FIELD_FIT is much better. Will update in next version.

BR

Himal

>
> Matt
>
>>   
>> -	if (GRAPHICS_VERx100(xe) >= 2000)
>> +		ccs_copy_size = ((num_pages - 1) & CCS_SIZE_MASK) << XE2_CCS_SIZE_SHIFT;
>>   		mocs = FIELD_PREP(XE2_XY_CTRL_SURF_MOCS_INDEX_MASK, gt->mocs.uc_index);
>> -	else
>> +
>> +	} else {
>> +		num_ccs_blks = DIV_ROUND_UP(xe_device_ccs_bytes(gt_to_xe(gt), size),
>> +					    NUM_CCS_BYTES_PER_BLOCK);
>> +		xe_gt_assert(gt, num_ccs_blks <= NUM_CCS_BLKS_PER_XFER);
>> +
>> +		ccs_copy_size = ((num_ccs_blks - 1) & CCS_SIZE_MASK) << CCS_SIZE_SHIFT;
>>   		mocs = FIELD_PREP(XY_CTRL_SURF_MOCS_MASK, gt->mocs.uc_index);
>> +	}
>>   
>>   	*cs++ = XY_CTRL_SURF_COPY_BLT |
>>   		(src_is_indirect ? 0x0 : 0x1) << SRC_ACCESS_TYPE_SHIFT |
>>   		(dst_is_indirect ? 0x0 : 0x1) << DST_ACCESS_TYPE_SHIFT |
>> -		((num_ccs_blks - 1) & CCS_SIZE_MASK) << CCS_SIZE_SHIFT;
>> +		ccs_copy_size;
>>   	*cs++ = lower_32_bits(src_ofs);
>>   	*cs++ = upper_32_bits(src_ofs) | mocs;
>>   	*cs++ = lower_32_bits(dst_ofs);
>> -- 
>> 2.25.1
>>

[-- Attachment #2: Type: text/html, Size: 4702 bytes --]