You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Some of my omnitrace proto files do not have the complete information on all memory operations. For example, if there is a hipMemcpy at the end from device to host, it does not show that information. I am using the following database call to find all instances:
SELECT caller_slice.name AS caller_name, caller_slice.ts AS caller_ts, caller_slice.dur AS caller_dur, caller_slice.id AS caller_id, caller_slice.name AS caller_name, caller_slice.track_id AS caller_track_id, callee_slice.ts AS callee_ts, callee_slice.dur AS callee_dur, callee_slice.id AS callee_id, callee_slice.name AS callee_name, callee_slice.track_id AS callee_track_id FROM (slice AS caller_slice JOIN flow ON caller_slice.id = flow.slice_out) JOIN slice AS callee_slice ON callee_slice.id = flow.slice_in
I think this means some of the memory operations do not have a caller or callee. I have checked the corresponding rocTracer file and it does have this information. I think something is not being picked up or lost. Or maybe it is simpler and I just need a different SQL query?
I cannot attach the proto file, but have attached the roctracer file where you can see that the hipMemcpy is in there. I will take a screenshot of what I am seeing in Perfetto. Hopefully you can see in Perfetto screen capture that the very last hipMemcpy does not link anywhere while the hipMemset does.
Some of my omnitrace proto files do not have the complete information on all memory operations. For example, if there is a hipMemcpy at the end from device to host, it does not show that information. I am using the following database call to find all instances:
SELECT caller_slice.name AS caller_name, caller_slice.ts AS caller_ts, caller_slice.dur AS caller_dur, caller_slice.id AS caller_id, caller_slice.name AS caller_name, caller_slice.track_id AS caller_track_id, callee_slice.ts AS callee_ts, callee_slice.dur AS callee_dur, callee_slice.id AS callee_id, callee_slice.name AS callee_name, callee_slice.track_id AS callee_track_id FROM (slice AS caller_slice JOIN flow ON caller_slice.id = flow.slice_out) JOIN slice AS callee_slice ON callee_slice.id = flow.slice_in
I think this means some of the memory operations do not have a caller or callee. I have checked the corresponding rocTracer file and it does have this information. I think something is not being picked up or lost. Or maybe it is simpler and I just need a different SQL query?
I cannot attach the proto file, but have attached the roctracer file where you can see that the hipMemcpy is in there. I will take a screenshot of what I am seeing in Perfetto. Hopefully you can see in Perfetto screen capture that the very last hipMemcpy does not link anywhere while the hipMemset does.
Thank you.
![hipmemcpy](https://private-user-images.githubusercontent.com/142359348/272011382-dfd382c4-e21b-45b1-9807-b55d314fee44.PNG?jwt=eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.eyJpc3MiOiJnaXRodWIuY29tIiwiYXVkIjoicmF3LmdpdGh1YnVzZXJjb250ZW50LmNvbSIsImtleSI6ImtleTUiLCJleHAiOjE3MzkyNzUxNTcsIm5iZiI6MTczOTI3NDg1NywicGF0aCI6Ii8xNDIzNTkzNDgvMjcyMDExMzgyLWRmZDM4MmM0LWUyMWItNDViMS05ODA3LWI1NWQzMTRmZWU0NC5QTkc_WC1BbXotQWxnb3JpdGhtPUFXUzQtSE1BQy1TSEEyNTYmWC1BbXotQ3JlZGVudGlhbD1BS0lBVkNPRFlMU0E1M1BRSzRaQSUyRjIwMjUwMjExJTJGdXMtZWFzdC0xJTJGczMlMkZhd3M0X3JlcXVlc3QmWC1BbXotRGF0ZT0yMDI1MDIxMVQxMTU0MTdaJlgtQW16LUV4cGlyZXM9MzAwJlgtQW16LVNpZ25hdHVyZT1iZTlhN2E4OTZhMmVlMzg5MzMyZGYwY2I1OTYyMjAwMzMyZmI0M2E5Yzg4MjAzYTA3ZDQ1OGRhZTJlMDBjYmRjJlgtQW16LVNpZ25lZEhlYWRlcnM9aG9zdCJ9.0eWI_yLlIAgiDhV6omYMxudrrOaPEFuXUvwBB5SPs_8)
![hipmemset](https://private-user-images.githubusercontent.com/142359348/272011385-f620ee0f-e3ee-454c-90ef-e362e07db168.PNG?jwt=eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.eyJpc3MiOiJnaXRodWIuY29tIiwiYXVkIjoicmF3LmdpdGh1YnVzZXJjb250ZW50LmNvbSIsImtleSI6ImtleTUiLCJleHAiOjE3MzkyNzUxNTcsIm5iZiI6MTczOTI3NDg1NywicGF0aCI6Ii8xNDIzNTkzNDgvMjcyMDExMzg1LWY2MjBlZTBmLWUzZWUtNDU0Yy05MGVmLWUzNjJlMDdkYjE2OC5QTkc_WC1BbXotQWxnb3JpdGhtPUFXUzQtSE1BQy1TSEEyNTYmWC1BbXotQ3JlZGVudGlhbD1BS0lBVkNPRFlMU0E1M1BRSzRaQSUyRjIwMjUwMjExJTJGdXMtZWFzdC0xJTJGczMlMkZhd3M0X3JlcXVlc3QmWC1BbXotRGF0ZT0yMDI1MDIxMVQxMTU0MTdaJlgtQW16LUV4cGlyZXM9MzAwJlgtQW16LVNpZ25hdHVyZT01NzIwMjNkMjUxOGJlNTI0MmU3ZWM0ODIzNjNiZGM1YjllMjc4NTg0OGIzMzJhMjY5NzZlNjFmOGMxNjMxYjYwJlgtQW16LVNpZ25lZEhlYWRlcnM9aG9zdCJ9.qyYG4TnEw4arYMbUJigRLN52H0P1cs2zhhauVw4uFSU)
roctracer-61388.txt
The text was updated successfully, but these errors were encountered: