Messages in this thread | | | From | fred thompson <> | Date | Wed, 24 Jan 2018 16:38:51 -0800 | Subject | Using very large multidimensional arrays to buffer IO/possibly used for the LM scheduler |
| |
I beleive that multidimensional arrays to a high power tied to a binary tree with a hash could help IO, at the cost of CPU power and memory. I would like this to be added to the BTRFS source. Here is an example.. before I start i just want to say I'm not that great at the Linux kernel, and I havent been writing software in a while:
#define MDA_SIZE 9999999999
void *move(char *direction); void *delet(); node1 search; // right? lol
struct node1 { double long ID; double long DATA[MDA][MDA][MDA]; // etc, i do not posses the math skills to make this work well char hash[]; // used as a hash, when accessed it should provide the location of a node with the same string
// navigating BT node1 *up; node1 *down; node1 *left; node1 *right; node1 *direction;
} *NODE1;
struct IO { node1 *data; node1 *temp; node1 *up; node1 *down; node1 *left; node1 *right; } *IO;
void *malloc(); // overload malloc for customization // node2/3/4/etc for other dimensions/directions
int main(){ IO file[MDA][MDA][MDA][MDA]; // etc // use *malloc() to assign memory to data, algorithms i cant help you wih // access/search/move/delete data through functions or a class if using cpp
return 0; }
I will port this to cpp some day but it took a long time just to come up with what i call a neural variable..
Thank you for reading, good luck, and for legal reasons, may the schwartz be with you
| |