3

I'm a new on cuda programming.

I have to GPU profiling using the nvprof about my application.

I find a metric l2_subp0_write_sector_misses that means number of write requests sent to DRAM from slice 0 of L2 cache.

But I don't know what does a slice mean.

Just as I guess, a slice is a sm because Gtx750 have 4 sm and there are 4 slices in events(l2_subp0_write_sector_misses, l2_subp1_write_sector_misses, l2_subp2_write_sector_misses, l3_subp0_write_sector_misses)

However, when I tested an application using 1 thread, all write_misses events have value.

I think that if a slice is a sm, just one write_misses has value and the others are 0.

What is a slice?

And Does it wrong that 1 thread uses 1 sm?

kh.chung
  • 53
  • 1
  • 4
  • @CygnusX1: I don't think that is a very appropriate duplicate given the age and vagueness of the answers it attracted. – talonmies Feb 24 '17 at 13:01
  • It is a bit dated, but effectively asks for exactly the same thing: what are slices and how do they relate to SM. I actually found that question when googling for possible answers for this one. I agree that a concrete answer - here or there - would help, and then the other question could be redirected as a duplicate. – CygnusX1 Feb 24 '17 at 17:43

0 Answers0