On 2022-11-01 04:52, Ma Jun wrote:
For some GPUs with more CUs, the original sibling_map[32]
in struct crat_subtype_cache is not enough
to save the cache information when create the VCRAT table,
so skip filling the struct crat_subtype_cache info instead
fill struct kfd_cache_properties directly to fix this problem.
v4:
- Fix the bug when get cache info
v3:
- Drop processor id calc function
v2:
- Remove unnecessary sys interface "cache_ext"
Signed-off-by: Ma Jun <Jun.Ma2@xxxxxxx>
One more small comment inline. With that fixed, the patch is
Reviewed-by: Felix Kuehling <Felix.Kuehling@xxxxxxx>
---
drivers/gpu/drm/amd/amdkfd/kfd_crat.c | 312 +++-------------------
drivers/gpu/drm/amd/amdkfd/kfd_crat.h | 12 +
drivers/gpu/drm/amd/amdkfd/kfd_topology.c | 245 ++++++++++++++++-
drivers/gpu/drm/amd/amdkfd/kfd_topology.h | 5 +-
4 files changed, 282 insertions(+), 292 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_crat.c b/drivers/gpu/drm/amd/amdkfd/kfd_crat.c
index 3c7f0d49da59..f4a9d0e49072 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_crat.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_crat.c
[snip]
-/* kfd_fill_gpu_cache_info - Fill GPU cache info using kfd_gpu_cache_info
- * tables
- *
- * @kdev - [IN] GPU device
- * @gpu_processor_id - [IN] GPU processor ID to which these caches
- * associate
- * @available_size - [IN] Amount of memory available in pcache
- * @cu_info - [IN] Compute Unit info obtained from KGD
- * @pcache - [OUT] memory into which cache data is to be filled in.
- * @size_filled - [OUT] amount of data used up in pcache.
- * @num_of_entries - [OUT] number of caches added
- */
-static int kfd_fill_gpu_cache_info(struct kfd_dev *kdev,
- int gpu_processor_id,
- int available_size,
- struct kfd_cu_info *cu_info,
- struct crat_subtype_cache *pcache,
- int *size_filled,
- int *num_of_entries)
+int get_gpu_cache_info(struct kfd_dev *kdev, struct kfd_gpu_cache_info **pcache_info)
This function name should have a kfd_ prefix.
Regards,
Felix
