t********y
发帖数: 47 | 1 I have an account on a SunOS server using sparc processor.
THe matrix is with 3000 row x 80000 cols,
I want to calculate A x B, and I think seperate A , B to two or more
sub matrix helps to speed up, since the server has multiple processors.
but how?
I don't have root account, just a normal user.
I tried multi-thread, it did work for 2 threads, but no improvement
for 3, 4... I believe a single process can only use 2 threads since
sparc processor or the system may have 2 CPU or 2 core for each bo | j**u
发帖数: 6059 | 2
问系统管理员最快最直接最准确。
【在 t********y 的大作中提到】
: I have an account on a SunOS server using sparc processor.
: THe matrix is with 3000 row x 80000 cols,
: I want to calculate A x B, and I think seperate A , B to two or more
: sub matrix helps to speed up, since the server has multiple processors.
: but how?
: I don't have root account, just a normal user.
: I tried multi-thread, it did work for 2 threads, but no improvement
: for 3, 4... I believe a single process can only use 2 threads since
: sparc processor or the system may have 2 CPU or 2 core for each bo
| l******n
发帖数: 9344 | 3 yes
【在 j**u 的大作中提到】
:
: 问系统管理员最快最直接最准确。
|
|
