Comments (6)
Well I would say the warp size shall be the number of physical ALUs. That's basically it. The number of warps is a side product of the register counting test. Please refer to Section 3.1 - The number of registers in our Romou paper for detail. It's already well illustrated. I think that's the most I could help. :)
from archprobe.
Refer from the paper, I think the number of ALUs shall be how many threads can run parallel that calculating by multiplying the warp size and the number of parallel warps. Is this right?
My discussion is based on this theory I understood from your paper. But the current problem is that I cannot find the number of parallel warps, so that I cannot calculate the number of ALUs.
You said that the warp size shall be the number of physical ALUs, which I cannot understand and agree based on the knowledge of gpu programming and your paper.
I have understood that we can detect the number of ALUs by plotting the figure of relationship between latency and the number of registers per work item and I will try it. But how to reproduce the method of Section 3.1 - The number of ALUs and Figure 8 which I think is interesting?
Can I have your Wechat number?
from archprobe.
I think you obscured the idea of physical threads and logical threads. They are all threads but the ALU count equals the number of physical threads. If you mean the total number of ALUs on chip, that's ALU count multiplies the number of SMs (GPU cores).
I prefer to keep this conversation open here in case anyone run into similar issue and they can have something to refer to. But sure, if you want to, it's the same as as my Github ID.
from archprobe.
I agree. But please forgive me that my reply may not be quick and keep this issue open for a while. I will learn more and try more to detect the number of ALUs. Thanks for your reply. And this is an amazing work I would like to say it again, which help my research a lot.
from archprobe.
Thanks for that. I'll keep this conversation for a while then. If you have questions other than the ALU count please feel free to open another issue so it could be tracked in a cleaner. :)
from archprobe.
Closing for inactivity.
from archprobe.
Related Issues (9)
- Get stuck when run ./ArchProbe on Adreno 640 on Meizu 16s HOT 2
- ImageCachelineSize causes memory object allocation failure on Mali-G78 HOT 1
- Global Memory to GPU Bandwidth HOT 2
- Does this contains the code to reproduce or generate the figures or data of the document? HOT 2
- Values between in GPU hierarchy images and output files HOT 3
- How to calculate ALUs? HOT 4
- question about "number of registers" HOT 4
- Is rooting required? HOT 4
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from archprobe.