Cache L1 block size

Autonomous Machines Jetson & Embedded Systems Jetson TX2
#1

Hi Nvidia’s people.

I have a question for you.

I am using a Jetson Tx2.

Why, when i am running a p_chase like algorithm on local memory and global memory, i got two different results ?
Indeed, on local window the cache block size appears to be 4 bytes, and on global window it appears to be 32 bytes.

Maybe, it’s because the coalescing mechanism ? or possibly, since Pascal L1 cache is split into two different region, maybe one is for local data and the other for global data (in extension with a separated constant cache and separated shared memory )?

#2

Hi Michael,

Your question is better suited for the Jetson TX2 category.
I will move it over for you.

Best,
Tom K

#5

Hi Michael,

Can you give more context?
the L1/L2 cacheline size is 64 bytes for all the cores in TX2.
What is this local and global window?

thanks
Bibek

#6

Hi Bibek

I should have give more details. My problem was on the GPU. But i figure out why by myself later.
Thank you anyway.

Michaël

#8

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.