-
Notifications
You must be signed in to change notification settings - Fork 33
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Omniperf is not displaying VGPRs in results. #117
Labels
Comments
coleramos425
added a commit
that referenced
this issue
May 15, 2023
Signed-off-by: coleramos425 <colramos@amd.com>
Looks like this counter was updated in ROCm 5.3 Note that analysis of ROCm 5.2.x workload data will now yield missing values. Conditional counter-naming based on detected ROCm gets a bit tricky, eventually, we'll bump minimum ROCm up to 5.3 (base) [omniperf-pub]$ ./src/omniperf analyze -p workloads/vcopy_rocm_5_2_3/mi200/ -b 7.1
--------
Analyze
--------
--------------------------------------------------------------------------------
0. Top Stat
╒════╤══════════════════════════════════════════╤═════════╤═══════════╤════════════╤══════════════╤════════╕
│ │ KernelName │ Count │ Sum(ns) │ Mean(ns) │ Median(ns) │ Pct │
╞════╪══════════════════════════════════════════╪═════════╪═══════════╪════════════╪══════════════╪════════╡
│ 0 │ vecCopy(double*, double*, double*, int, │ 1.00 │ 26400.00 │ 26400.00 │ 26400.00 │ 100.00 │
│ │ int) [clone .kd] │ │ │ │ │ │
╘════╧══════════════════════════════════════════╧═════════╧═══════════╧════════════╧══════════════╧════════╛
--------------------------------------------------------------------------------
7. Wavefront
7.1 Wavefront Launch Stats
╒═════════╤═════════════════════╤═══════════╤═══════════╤═══════════╤════════════╕
│ Index │ Metric │ Avg │ Min │ Max │ Unit │
╞═════════╪═════════════════════╪═══════════╪═══════════╪═══════════╪════════════╡
│ 7.1.0 │ Grid Size │ 1048576.0 │ 1048576.0 │ 1048576.0 │ Work items │
├─────────┼─────────────────────┼───────────┼───────────┼───────────┼────────────┤
│ 7.1.1 │ Workgroup Size │ 256.0 │ 256.0 │ 256.0 │ Work items │
├─────────┼─────────────────────┼───────────┼───────────┼───────────┼────────────┤
│ 7.1.2 │ Total Wavefronts │ 16384.0 │ 16384.0 │ 16384.0 │ Wavefronts │
├─────────┼─────────────────────┼───────────┼───────────┼───────────┼────────────┤
│ 7.1.3 │ Saved Wavefronts │ 0.0 │ 0.0 │ 0.0 │ Wavefronts │
├─────────┼─────────────────────┼───────────┼───────────┼───────────┼────────────┤
│ 7.1.4 │ Restored Wavefronts │ 0.0 │ 0.0 │ 0.0 │ Wavefronts │
├─────────┼─────────────────────┼───────────┼───────────┼───────────┼────────────┤
│ 7.1.5 │ VGPRs │ │ │ │ Registers │
├─────────┼─────────────────────┼───────────┼───────────┼───────────┼────────────┤
│ 7.1.6 │ AGPRs │ │ │ │ Registers │
├─────────┼─────────────────────┼───────────┼───────────┼───────────┼────────────┤
│ 7.1.7 │ SGPRs │ 24.0 │ 24.0 │ 24.0 │ Registers │
├─────────┼─────────────────────┼───────────┼───────────┼───────────┼────────────┤
│ 7.1.8 │ LDS Allocation │ 0.0 │ 0.0 │ 0.0 │ Bytes │
├─────────┼─────────────────────┼───────────┼───────────┼───────────┼────────────┤
│ 7.1.9 │ Scratch Allocation │ 0.0 │ 0.0 │ 0.0 │ Bytes │
╘═════════╧═════════════════════╧═══════════╧═══════════╧═══════════╧════════════╛
|
coleramos425
added
bug
Something isn't working
metric definition
Related to the definition of a metric in Omniperf
labels
May 15, 2023
feizheng10
pushed a commit
to feizheng10/omniperf
that referenced
this issue
Dec 6, 2023
Signed-off-by: coleramos425 <colramos@amd.com> Signed-off-by: fei.zheng <fei.zheng@amd.com>
feizheng10
pushed a commit
to feizheng10/omniperf
that referenced
this issue
Dec 20, 2023
Signed-off-by: coleramos425 <colramos@amd.com> Signed-off-by: fei.zheng <fei.zheng@amd.com>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Labels
Reported at a recent customer hackathon, but I've noticed it as well.
Most likely the cause is that rocprof (correctly) now reports arch_vgprs and accum_vgprs (see: ROCm/rocprofiler@5fd1c7e).
Omniperf likely just needs to read the new fields, and report them accordingly (VGPRs and AGPRs)
The text was updated successfully, but these errors were encountered: