Buffer allocation code could be put in separate function

[johnramsden]

A large part of populate_mram is dedicated to allocating and reallocating buffers. For readability and maintainability it should (or could) be put in a separate function which returns 0 if allocation was successful.

We could then focus on loading the dpu in populate_mram.

[nzarif]

What do you have in mind for the prototype of the function?

[johnramsden]

/**
 * @brief Allocate buffers
 * @param table_id   embedding table number.
 * @param nr_rows    number of rows of the embedding table
 * @param nr_cols    number of columns of the embedding table
 * @param table_data a pointer of the size nr_rows*nr_cols containing table's data
 * @return 0 on success, ENOMEM on allocation failure
 */
static int
alloc_buffers(uint32_t table_id, uint64_t nr_rows, uint64_t nr_cols, const int32_t *table_data);

[nzarif]

I see! so every time before calling populate_mram on each table this function will be called. So then populate_mram() will only be copying data to buffers and allocating dpus when needed and copying to dpus?