Which of the following statements about the cache blocking optimization in the DGEMM matrix multiplication is correct?

Group of answer choices

In the cache blocked version of DGEMM, the do$\_$block function is inlined by the compiler, eliminating overhead associated with function calls.

The performance improvement from cache blocking is greater for smaller matrices compared to larger matrices, as smaller matrices fit entirely in the L$1$ cache.

The fully optimized DGEMM code with cache blocking runs at the same performance level as the original unoptimized C version for all matrix sizes.

Cache blocking increases the number of floating$-$point operations performed per matrix element, making it less efficient for small matrices.