-
Aditya A authored
PROBLEM ------- More memory usage was seen due to increase in size of ha_innobase object . The two character arrays srch_key_val* in ha_innobase class each contribute approximately 6Kbytes ,so when large databases with many partitions are used ,memory usage increases. FIX --- MySQL stores the INTs datatypes in little endian format where as innodb stores them in big endian format.The srch_key_val* arrays are only needed when we are converting INTs datatypes in the search key from little endian format to big endian format. So we allocate memory equivalent to number of INTs in each index of the table ,which will reduce memory footprint. Also we have moved the allocation of the memory to ha_innobase::open() so that the memory is only allocated during opening of table handler. [Approved by Mattiasj and Jimmy #rb5565]
Aditya A authoredPROBLEM ------- More memory usage was seen due to increase in size of ha_innobase object . The two character arrays srch_key_val* in ha_innobase class each contribute approximately 6Kbytes ,so when large databases with many partitions are used ,memory usage increases. FIX --- MySQL stores the INTs datatypes in little endian format where as innodb stores them in big endian format.The srch_key_val* arrays are only needed when we are converting INTs datatypes in the search key from little endian format to big endian format. So we allocate memory equivalent to number of INTs in each index of the table ,which will reduce memory footprint. Also we have moved the allocation of the memory to ha_innobase::open() so that the memory is only allocated during opening of table handler. [Approved by Mattiasj and Jimmy #rb5565]
Loading