Comments (6)
Hi, we would like to know if there are any solutions for the above mentioned issue. Thanks.
from sycl-bench.
I cannot reproduce the issue with hipSYCL. Your output indicates that somehow a block size of 0 enters the benchmark. This value is derived from the local size. I had a quick look at the code paths, and I don't understand how this could happen - it does not for me.
There's an assert that checks that the block size is non-zero. Can you check what happens when compiling with debug assertions enabled?
from sycl-bench.
Hi, we are not working with hipSYCL. The issue that we are facing is occurring during runtime. The test case is failing to execute when we are not passing the local parameter (as in, when it is taking the value of local parameter as 256 by default).
Command being used to execute - ./blocked_transform --device=gpu
However, it is working fine when we are explicitly defining the local parameter to 256 during runtime.
Command being used to execute - ./blocked_transform --device=gpu --local=256
We are not sure as to why this issue is occurring.
Thanks.
from sycl-bench.
Hi, we are not working with hipSYCL. The issue that we are facing is occurring during runtime.
I'm aware of this. But I don't have an installation of the DPC++ SYCL implementation with CUDA backend here. I'm just saying I cannot reproduce this with my setup. And I don't understand why DPC++ or hipSYCL would behave differently here anyway. The error does not seem to be related to SYCL specific functionality.
The test case is failing to execute when we are not passing the local parameter (as in, when it is taking the value of local parameter as 256 by default).
Command being used to execute - ./blocked_transform --device=gpu
However, it is working fine when we are explicitly defining the local parameter to 256 during runtime.
Command being used to execute - ./blocked_transform --device=gpu --local=256
I understood this. As I've said I cannot reproduce here. Command line option handling is the same for DPC++ and hipSYCL. For further investigation into the issue, I asked you the following:
There's an assert that checks that the block size is non-zero. Can you check what happens when compiling with debug assertions enabled?
i.e. make sure that the NDEBUG
macro is not set when building.
from sycl-bench.
Hi, as suggested, I've added the following in the blocked_transform.cpp code and I've rebuilt it again.
#include<assert.h>
#define NDEBUG
It seems that by default, the value of local size is being taken as 1024 (please see attached screenshot below).
However, when I am defining '--local' to be either 256 (default value) or 1024 explicitly, it is working fine.
Command being used:
./blocked_transform --device=gpu --local=256
./blocked_transform --device=gpu --local=1024
Could this be a bug in the code?
Thanks.
from sycl-bench.
Hi, is there any update regarding this issue? Thanks.
from sycl-bench.
Related Issues (17)
- Add references to original implementations of benchmarks
- Use ndrange of hierarchical parallel for in pattern_shared
- More special treatment for implementations in CMakeLists.txt
- Some thoughts on DRAM throughput benchmarking HOT 7
- Fix the run-suite brommy.bmp not found issue HOT 2
- Investigate nbody on ComputeCpp HOT 2
- Let run-suite read test profiles from yaml or json
- Questions about the single kernel set HOT 2
- use of undeclared identifier 'device_selector' HOT 2
- Race condition in scalar prod HOT 1
- Problem in compilation stage with. computecpp 2.0.0 HOT 3
- build procedure HOT 1
- blocked_transform is broken due to SYCL 2020 offset semantics HOT 1
- `emitResults` has invalid memory accesses when used with `--warmup-run` & the first run fails HOT 1
- Decide on future of sycl2020 branch - make default branch or merge into main? HOT 2
- Runtime failure for the DGEMM application HOT 2
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 sycl-bench.