bybrooklyn/linux
1
0
Fork 0
mirror of https://github.com/torvalds/linux.git synced 2026-05-04 22:43:04 -04:00
Code Issues Packages Projects Releases Wiki Activity

drm/amdkfd: Add CU-masking ioctl to KFD

Browse Source 
CU-masking allows a KFD client to control the set of CUs used by a
user mode queue for executing compute dispatches. This can be used
for optimizing the partitioning of the GPU and minimize conflicts
between concurrent tasks.

Signed-off-by: Flora Cui <flora.cui@amd.com>
Signed-off-by: Kent Russell <kent.russell@amd.com>
Signed-off-by: Eric Huang <JinHuiEric.Huang@amd.com>
Signed-off-by: Felix Kuehling <Felix.Kuehling@amd.com>
Acked-by: Oded Gabbay <oded.gabbay@gmail.com>
Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
This commit is contained in:
Felix Kuehling
2018-07-14 19:05:59 -04:00
committed by Oded Gabbay
parent a7fe68a1e8
commit 39e7f33186
9 changed files with 219 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

41
drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.c
View File

@@ -21,7 +21,7 @@
*
*/
#include "kfd_priv.h"
#include "kfd_mqd_manager.h"
struct mqd_manager *mqd_manager_init(enum KFD_MQD_TYPE type,
struct kfd_dev *dev)
@@ -48,3 +48,42 @@ struct mqd_manager *mqd_manager_init(enum KFD_MQD_TYPE type,
return NULL;
}
void mqd_symmetrically_map_cu_mask(struct mqd_manager *mm,
const uint32_t *cu_mask, uint32_t cu_mask_count,
uint32_t *se_mask)
{
struct kfd_cu_info cu_info;
uint32_t cu_per_sh[4] = {0};
int i, se, cu = 0;
mm->dev->kfd2kgd->get_cu_info(mm->dev->kgd, &cu_info);
if (cu_mask_count > cu_info.cu_active_number)
cu_mask_count = cu_info.cu_active_number;
for (se = 0; se < cu_info.num_shader_engines; se++)
for (i = 0; i < 4; i++)
cu_per_sh[se] += hweight32(cu_info.cu_bitmap[se][i]);
/* Symmetrically map cu_mask to all SEs:
* cu_mask[0] bit0 -> se_mask[0] bit0;
* cu_mask[0] bit1 -> se_mask[1] bit0;
* ... (if # SE is 4)
* cu_mask[0] bit4 -> se_mask[0] bit1;
* ...
*/
se = 0;
for (i = 0; i < cu_mask_count; i++) {
if (cu_mask[i / 32] & (1 << (i % 32)))
se_mask[se] |= 1 << cu;
do {
se++;
if (se == cu_info.num_shader_engines) {
se = 0;
cu++;
}
} while (cu >= cu_per_sh[se] && cu < 32);
}
}