AI/rocm-systems
2
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
Files
376a86d2d0fbf0cba537671aae4ac2dbf18e03df
rocm-systems/projects/rocprofiler-compute/src/docs-2.x/profiling.md
T
David Galiffi 376a86d2d0 Updated links in documentation. (#328) 
Updated to reflect new GitHub organization.
Fixed broken links to GitHub pages.

Signed-off-by: David Galiffi <David.Galiffi@amd.com>

[ROCm/rocprofiler-compute commit: f5712875aa]
2024-03-21 10:14:37 -05:00

396 lines
19 KiB
Markdown
Raw Blame History

# Profile Mode
```eval_rst
.. toctree::
:glob:
:maxdepth: 5
```
The [Omniperf](https://github.com/ROCm/omniperf) repository
includes source code for a sample GPU compute workload,
__vcopy.cpp__. A copy of this file is available in the `share/sample`
subdirectory after a normal Omniperf installation, or via the
`$OMNIPERF_SHARE/sample` directory when using the supplied modulefile.
A compiled version of this workload is used throughout the following
sections to demonstrate the use of Omniperf in MI GPU performance
analysis. Unless otherwise noted, the performance analysis is done on
the MI200 platform.
## Workload Compilation
**vcopy compilation:**
```shell-session
$ hipcc vcopy.cpp -o vcopy
$ ls
vcopy vcopy.cpp
$ ./vcopy -n 1048576 -b 256
vcopy testing on GCD 0
Finished allocating vectors on the CPU
Finished allocating vectors on the GPU
Finished copying vectors to the GPU
sw thinks it moved 1.000000 KB per wave
Total threads: 1048576, Grid Size: 4096 block Size:256, Wavefronts:16384:
Launching the kernel on the GPU
Finished executing kernel
Finished copying the output vector from the GPU to the CPU
Releasing GPU memory
Releasing CPU memory
```
## Omniperf Profiling
The *omniperf* script, available through the Omniperf repository, is used to aquire all necessary performance monitoring data through analysis of compute workloads.
**omniperf help:**
```shell-session
$ omniperf profile --help
usage:
omniperf profile --name <workload_name> [profile options] [roofline options] -- <profile_cmd>
---------------------------------------------------------------------------------
Examples:
omniperf profile -n vcopy_all -- ./vcopy -n 1048576 -b 256
omniperf profile -n vcopy_SPI_TCC -b SQ TCC -- ./vcopy -n 1048576 -b 256
omniperf profile -n vcopy_kernel -k vecCopy -- ./vcopy -n 1048576 -b 256
omniperf profile -n vcopy_disp -d 0 -- ./vcopy -n 1048576 -b 256
omniperf profile -n vcopy_roof --roof-only -- ./vcopy -n 1048576 -b 256
---------------------------------------------------------------------------------
Help:
-h, --help show this help message and exit
General Options:
-v, --version show program's version number and exit
-q, --quiet Run in quiet mode.
-V, --verbose Increase output verbosity (use multiple times for higher levels)
-s, --specs Print system specs.
Profile Options:
-n , --name Assign a name to workload.
-p , --path Specify path to save workload.
-k [ ...], --kernel [ ...] Kernel filtering.
-d [ ...], --dispatch [ ...] Dispatch ID filtering.
-b [ ...], --block [ ...] Hardware block filtering:
SQ
SQC
TA
TD
TCP
TCC
SPI
CPC
CPF
--join-type Choose how to join rocprof runs: (DEFAULT: grid)
kernel (i.e. By unique kernel name dispatches)
grid (i.e. By unique kernel name + grid size dispatches)
--no-roof Profile without collecting roofline data.
-- [ ...] Provide command for profiling after double dash.
Standalone Roofline Options:
--roof-only Profile roofline data only.
--sort Overlay top kernels or top dispatches: (DEFAULT: kernels)
kernels
dispatches
-m [ ...], --mem-level [ ...] Filter by memory level: (DEFAULT: ALL)
HBM
L2
vL1D
LDS
--device Target GPU device ID. (DEFAULT: ALL)
--kernel-names Include kernel names in roofline plot.
```
- The `-k` \<kernel> flag allows for kernel filtering, which is compatible with the current rocProf utility.
- The `-d` \<dispatch> flag allows for dispatch ID filtering, which is compatible with the current rocProf utility.
- The `-b` \<ipblocks> allows system profiling on one or more selected hardware components to speed up the profiling process. One can gradually include more hardware components, without overwriting performance data acquired on other hardware components.
The following sample command profiles the *vcopy* workload.
**vcopy profiling:**
```shell-session
$ omniperf profile --name vcopy -- ./vcopy -n 1048576 -b 256
___ _ __
/ _ \ _ __ ___ _ __ (_)_ __ ___ _ __ / _|
| | | | '_ ` _ \| '_ \| | '_ \ / _ \ '__| |_
| |_| | | | | | | | | | | |_) | __/ | | _|
\___/|_| |_| |_|_| |_|_| .__/ \___|_| |_|
|_|
Omniperf version: 2.0.0
Profiler choice: rocprofv1
Path: /home/auser/repos/omniperf/sample/workloads/vcopy/MI200
Target: MI200
Command: ./vcopy -n 1048576 -b 256
Kernel Selection: None
Dispatch Selection: None
IP Blocks: All
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Collecting Performance Counters
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
[profiling] Current input file: /home/auser/repos/omniperf/sample/workloads/vcopy/MI200/perfmon/SQ_IFETCH_LEVEL.txt
|-> [rocprof] RPL: on '240312_174329' from '/opt/rocm-5.2.1' in '/home/colramos/GitHub/omniperf'
|-> [rocprof] RPL: profiling '""./vcopy -n 1048576 -b 256""'
|-> [rocprof] RPL: input file '/home/auser/repos/omniperf/sample/workloads/vcopy/MI200/perfmon/SQ_IFETCH_LEVEL.txt'
|-> [rocprof] RPL: output dir '/tmp/rpl_data_240312_174329_692890'
|-> [rocprof] RPL: result dir '/tmp/rpl_data_240312_174329_692890/input0_results_240312_174329'
|-> [rocprof] ROCProfiler: input from "/tmp/rpl_data_240312_174329_692890/input0.xml"
|-> [rocprof] gpu_index =
|-> [rocprof] kernel =
|-> [rocprof] range =
|-> [rocprof] 6 metrics
|-> [rocprof] GRBM_COUNT, GRBM_GUI_ACTIVE, SQ_WAVES, SQ_IFETCH, SQ_IFETCH_LEVEL, SQ_ACCUM_PREV_HIRES
|-> [rocprof] vcopy testing on GCD 0
|-> [rocprof] Finished allocating vectors on the CPU
|-> [rocprof] Finished allocating vectors on the GPU
|-> [rocprof] Finished copying vectors to the GPU
|-> [rocprof] sw thinks it moved 1.000000 KB per wave
|-> [rocprof] Total threads: 1048576, Grid Size: 4096 block Size:256, Wavefronts:16384:
|-> [rocprof] Launching the kernel on the GPU
|-> [rocprof] Finished executing kernel
|-> [rocprof] Finished copying the output vector from the GPU to the CPU
|-> [rocprof] Releasing GPU memory
|-> [rocprof] Releasing CPU memory
|-> [rocprof]
|-> [rocprof] ROCPRofiler: 1 contexts collected, output directory /tmp/rpl_data_240312_174329_692890/input0_results_240312_174329
|-> [rocprof] File '/home/auser/repos/omniperf/sample/workloads/vcopy/MI200/SQ_IFETCH_LEVEL.csv' is generating
|-> [rocprof]
[profiling] Current input file: /home/auser/repos/omniperf/sample/workloads/vcopy/MI200/perfmon/SQ_INST_LEVEL_LDS.txt
...
[roofline] Checking for roofline.csv in /home/auser/repos/omniperf/sample/workloads/vcopy/MI200
[roofline] No roofline data found. Generating...
Empirical Roofline Calculation
Copyright © 2022 Advanced Micro Devices, Inc. All rights reserved.
Total detected GPU devices: 4
GPU Device 0: Profiling...
99% [||||||||||||||||||||||||||||||||||||||||||||||||||||||||||| ]
HBM BW, GPU ID: 0, workgroupSize:256, workgroups:2097152, experiments:100, traffic:8589934592 bytes, duration:6.2 ms, mean:1388.0 GB/sec, stdev=3.1 GB/sec
99% [||||||||||||||||||||||||||||||||||||||||||||||||||||||||||| ]
L2 BW, GPU ID: 0, workgroupSize:256, workgroups:8192, experiments:100, traffic:687194767360 bytes, duration:136.5 ms, mean:5020.8 GB/sec, stdev=16.5 GB/sec
99% [||||||||||||||||||||||||||||||||||||||||||||||||||||||||||| ]
L1 BW, GPU ID: 0, workgroupSize:256, workgroups:16384, experiments:100, traffic:26843545600 bytes, duration:2.9 ms, mean:9229.5 GB/sec, stdev=2.9 GB/sec
99% [||||||||||||||||||||||||||||||||||||||||||||||||||||||||||| ]
LDS BW, GPU ID: 0, workgroupSize:256, workgroups:16384, experiments:100, traffic:33554432000 bytes, duration:1.9 ms, mean:17645.6 GB/sec, stdev=20.1 GB/sec
99% [||||||||||||||||||||||||||||||||||||||||||||||||||||||||||| ]
Peak FLOPs (FP32), GPU ID: 0, workgroupSize:256, workgroups:16384, experiments:100, FLOP:274877906944, duration:13.078 ms, mean:20986.9 GFLOPS, stdev=310.8 GFLOPS
99% [||||||||||||||||||||||||||||||||||||||||||||||||||||||||||| ]
Peak FLOPs (FP64), GPU ID: 0, workgroupSize:256, workgroups:16384, experiments:100, FLOP:137438953472, duration:6.7 ms, mean:20408.029297.1 GFLOPS, stdev=2.7 GFLOPS
99% [||||||||||||||||||||||||||||||||||||||||||||||||||||||||||| ]
Peak MFMA FLOPs (BF16), GPU ID: 0, workgroupSize:256, workgroups:16384, experiments:100, FLOP:2147483648000, duration:12.6 ms, mean:170280.0 GFLOPS, stdev=22.3 GFLOPS
99% [||||||||||||||||||||||||||||||||||||||||||||||||||||||||||| ]
Peak MFMA FLOPs (F16), GPU ID: 0, workgroupSize:256, workgroups:16384, experiments:100, FLOP:2147483648000, duration:13.0 ms, mean:164733.6 GFLOPS, stdev=24.3 GFLOPS
99% [||||||||||||||||||||||||||||||||||||||||||||||||||||||||||| ]
Peak MFMA FLOPs (F32), GPU ID: 0, workgroupSize:256, workgroups:16384, experiments:100, FLOP:536870912000, duration:13.0 ms, mean:41399.6 GFLOPS, stdev=4.1 GFLOPS
99% [||||||||||||||||||||||||||||||||||||||||||||||||||||||||||| ]
Peak MFMA FLOPs (F64), GPU ID: 0, workgroupSize:256, workgroups:16384, experiments:100, FLOP:268435456000, duration:6.5 ms, mean:41379.2 GFLOPS, stdev=4.4 GFLOPS
99% [||||||||||||||||||||||||||||||||||||||||||||||||||||||||||| ]
Peak MFMA IOPs (I8), GPU ID: 0, workgroupSize:256, workgroups:16384, experiments:100, IOP:2147483648000, duration:12.9 ms, mean:166281.9 GOPS, stdev=2495.9 GOPS
GPU Device 1: Profiling...
...
GPU Device 2: Profiling...
...
GPU Device 3: Profiling...
...
```
You will notice two main stages in *default* Omniperf profiling. The first stage collects all the counters needed for Omniperf analysis (omitting any filters you have provided). The second stage collects data for the roofline analysis (this stage can be disabled using `--no-roof`)
In this document, we use the term System on Chip (SoC) to refer to a particular family of accelerators. At the end of profiling, all resulting csv files should be located in a SoC specific target directory, e.g.:
- "mi200" for the AMD Instinct (tm) MI200 family of accelerators
- "mi100" for the AMD Instinct (tm) MI100 family of accelerators
etc. The SoC names are generated as a part of Omniperf, and do not necessarily distinguish between different accelerators in the same family (e.g., an AMD Instinct (tm) MI210 vs an MI250)
> Note: Additionally, you will notice a few extra files. An SoC parameters file, *sysinfo.csv*, is created to reflect the target device settings. All profiling output is stored in *log.txt*. Roofline specific benchmark results are stored in *roofline.csv*.
```shell-session
$ ls workloads/vcopy/MI200/
total 112
total 60
drwxr-xr-x 1 auser agroup 0 Mar 1 15:15 perfmon
-rw-r--r-- 1 auser agroup 26175 Mar 1 15:15 pmc_perf.csv
-rw-r--r-- 1 auser agroup 1708 Mar 1 15:17 roofline.csv
-rw-r--r-- 1 auser agroup 519 Mar 1 15:15 SQ_IFETCH_LEVEL.csv
-rw-r--r-- 1 auser agroup 456 Mar 1 15:15 SQ_INST_LEVEL_LDS.csv
-rw-r--r-- 1 auser agroup 474 Mar 1 15:15 SQ_INST_LEVEL_SMEM.csv
-rw-r--r-- 1 auser agroup 474 Mar 1 15:15 SQ_INST_LEVEL_VMEM.csv
-rw-r--r-- 1 auser agroup 599 Mar 1 15:15 SQ_LEVEL_WAVES.csv
-rw-r--r-- 1 auser agroup 650 Mar 1 15:15 sysinfo.csv
-rw-r--r-- 1 auser agroup 399 Mar 1 15:15 timestamps.csv
```
### Filtering
To reduce profiling time and the counters collected one may use profiling filters. Profiling filters and their functionality depend on the underlying profiler being used. While Omniperf is profiler agnostic, we have provided a detailed description of profiling filters available when using Omniperf with [rocProf](https://rocm.docs.amd.com/projects/rocprofiler/en/latest/rocprof.html) below.
Filtering Options:
- The `-k` \<kernel> flag allows for kernel filtering. Useage is equivalent with the current rocProf utility ([see details below](#kernel-filtering)).
- The `-d` \<dispatch> flag allows for dispatch ID filtering. Useage is equivalent with the current rocProf utility ([see details below](#dispatch-filtering)).
- The `-b` \<ipblocks> allows system profiling on one or more selected hardware components to speed up the profiling process. One can gradually include more hardware components, without overwriting performance data acquired on other hardware components.
```{note}
Be cautious while combining different profiling filters in the same call. Conflicting filters may result in error.
i.e. filtering dispatch X, but dispatch X does not match your kernel name filter
```
#### Hardware Component Filtering
One can profile specific hardware components to speed up the profiling process. In Omniperf, we use the term IP block to refer to a hardware component or a group of hardware components. All profiling results are accumulated in the same target directory, without overwriting those for other hardware components, hence enabling the incremental profiling and analysis.
The following example only gathers hardware counters for the Shader Sequencer (SQ) and L2 Cache (TCC) components, skipping all other hardware components:
```shell-session
$ omniperf profile --name vcopy -b SQ TCC -- ./vcopy -n 1048576 -b 256
___ _ __
/ _ \ _ __ ___ _ __ (_)_ __ ___ _ __ / _|
| | | | '_ ` _ \| '_ \| | '_ \ / _ \ '__| |_
| |_| | | | | | | | | | | |_) | __/ | | _|
\___/|_| |_| |_|_| |_|_| .__/ \___|_| |_|
|_|
fname: pmc_cpc_perf: Skipped
fname: pmc_spi_perf: Skipped
fname: pmc_cpf_perf: Skipped
fname: pmc_tcp_perf: Skipped
fname: pmc_sq_perf4: Added
fname: pmc_tcc_perf: Added
fname: pmc_sq_perf8: Added
fname: pmc_ta_perf: Skipped
fname: pmc_sq_perf1: Added
fname: pmc_sq_perf3: Added
fname: pmc_td_perf: Skipped
fname: pmc_tcc2_perf: Skipped
fname: pmc_sqc_perf1: Skipped
fname: pmc_sq_perf6: Added
fname: pmc_sq_perf2: Added
Omniperf version: 2.0.0
Profiler choice: rocprofv1
Path: /home/auser/repos/omniperf/sample/workloads/vcopy/MI200
Target: MI200
Command: ./vcopy -n 1048576 -b 256
Kernel Selection: None
Dispatch Selection: None
IP Blocks: ['sq', 'tcc']
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Collecting Performance Counters
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
...
```
#### Kernel Filtering
Kernel filtering is based on the name of the kernel(s) you would like to isolate. Use a kernel name substring list to isolate desired kernels.
The following example demonstrates profiling isolating the kernel matching substring "vecCopy":
```shell-session
$ omniperf profile --name vcopy -k vecCopy -- ./vcopy -n 1048576 -b 256
___ _ __
/ _ \ _ __ ___ _ __ (_)_ __ ___ _ __ / _|
| | | | '_ ` _ \| '_ \| | '_ \ / _ \ '__| |_
| |_| | | | | | | | | | | |_) | __/ | | _|
\___/|_| |_| |_|_| |_|_| .__/ \___|_| |_|
|_|
Omniperf version: 2.0.0
Profiler choice: rocprofv1
Path: /home/auser/repos/omniperf/sample/workloads/vcopy/MI200
Target: MI200
Command: ./vcopy -n 1048576 -b 256
Kernel Selection: ['vecCopy']
Dispatch Selection: None
IP Blocks: All
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Collecting Performance Counters
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
...
```
#### Dispatch Filtering
Dispatch filtering is based on the *global* dispatch index of kernels in a run.
The following example profiles only the 0th dispatched kernel in execution of the application:
```shell-session
$ omniperf profile --name vcopy -d 0 -- ./vcopy -n 1048576 -b 256
___ _ __
/ _ \ _ __ ___ _ __ (_)_ __ ___ _ __ / _|
| | | | '_ ` _ \| '_ \| | '_ \ / _ \ '__| |_
| |_| | | | | | | | | | | |_) | __/ | | _|
\___/|_| |_| |_|_| |_|_| .__/ \___|_| |_|
|_|
Omniperf version: 2.0.0
Profiler choice: rocprofv1
Path: /home/auser/repos/omniperf/sample/workloads/vcopy/MI200
Target: MI200
Command: ./vcopy -n 1048576 -b 256
Kernel Selection: None
Dispatch Selection: ['0']
IP Blocks: All
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Collecting Performance Counters
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
...
```
### Standalone Roofline
If you are only interested in generating roofline analysis data try using `--roof-only`. This will only collect counters relevant to roofline, as well as generate a standalone .pdf output of your roofline plot.
Standalone Roofline Options:
- The `--sort` \<desired_sort> allows you to specify whether you would like to overlay top kernel or top dispatch data in your roofline plot.
- The `-m` \<cache_level> allows you to specify specific level(s) of cache you would like to include in your roofline plot.
- The `--device` \<gpu_id> allows you to specify a device id to collect performace data from when running our roofline benchmark on your system.
- If you would like to distinguish different kernels in your .pdf roofline plot use `--kernel-names`. This will give each kernel a unique marker identifiable from the plot's key.
#### Roofline Only
The following example demonstrates profiling roofline data only:
```shell-session
$ omniperf profile --name vcopy --roof-only -- ./vcopy -n 1048576 -b 256
...
[roofline] Checking for roofline.csv in /home/auser/repos/omniperf/sample/workloads/vcopy/MI200
[roofline] No roofline data found. Generating...
Checking for roofline.csv in /home/auser/repos/omniperf/sample/workloads/vcopy/MI200
Empirical Roofline Calculation
Copyright © 2022 Advanced Micro Devices, Inc. All rights reserved.
Total detected GPU devices: 4
GPU Device 0: Profiling...
99% [||||||||||||||||||||||||||||||||||||||||||||||||||||||||||| ]
...
Empirical Roofline PDFs saved!
```
An inspection of our workload output folder shows .pdf plots were generated successfully
```shell-session
$ ls workloads/vcopy/MI200/
total 48
-rw-r--r-- 1 auser agroup 13331 Mar 1 16:05 empirRoof_gpu-0_fp32.pdf
-rw-r--r-- 1 auser agroup 13136 Mar 1 16:05 empirRoof_gpu-0_int8_fp16.pdf
drwxr-xr-x 1 auser agroup 0 Mar 1 16:03 perfmon
-rw-r--r-- 1 auser agroup 1101 Mar 1 16:03 pmc_perf.csv
-rw-r--r-- 1 auser agroup 1715 Mar 1 16:05 roofline.csv
-rw-r--r-- 1 auser agroup 650 Mar 1 16:03 sysinfo.csv
-rw-r--r-- 1 auser agroup 399 Mar 1 16:03 timestamps.csv
```
A sample *empirRoof_gpu-ALL_fp32.pdf* looks something like this:
![Sample Standalone Roof Plot](images/sample-roof-plot.png)
Reference in New Issue View Git Blame Copy Permalink