On 3/25/2024 10:49 AM, Rameshkumar Sundaram wrote: > On 3/22/2024 2:34 AM, Jeff Johnson wrote: >> On 3/20/2024 12:09 PM, Rameshkumar Sundaram wrote: >>> + struct ath12k_vif_cache cache; >> >> in my v4 comment I had suggested that this cache be a pointer. >> instead you chose to add pointers to the subordinate records within the cache. >> why take that approach? >> > Thought of making the actual cache objects as pointers, instead of > allocating memory for all type(key,tx) of caches during vdev create itself. > So that well allocate required cache alone if needed. Thoughts ? There is a tradeoff here. This cache is temporary, but the allocation exists for the lifetime of the struct. So by having a single pointer here you only waste one pointer, but at the expense that for each object type you have to see if the cache is already present, and if not, then allocate the cache. I'm OK either way, but just wanted to point out that the current approach wastes more memory when the cache isn't needed. /jeff
