AI/rocm-systems
2
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

SWDEV-406042: Updated readme and changelog

Browse Source 
Change-Id: If67e96f898f7eae041c879ad0bee278badc8681f


[ROCm/rocprofiler commit: a1eb64603b]
This commit is contained in:
Giovanni LB
2023-06-18 18:09:33 -03:00
committed by Giovanni Baraldi
parent ff89ec020a
commit 62d3c0018c
2 changed files with 11 additions and 10 deletions
Download Patch File Download Diff File Expand all files Collapse all files
projects/rocprofiler/CHANGELOG.md
+1 -4
View File
@@ -221,10 +221,7 @@ The resulting `a.out` will depend on
## ROCprofiler for rocm 5.7.0
### Navi support
Rocprofiler for ROCm 5.7 added support for counter collection (PMC) and advanced thread tracing (ATT) for Navi21 and Navi31 GPUs.
- On Navi, specially Navi31, counter collection requires the GPU to be in a stable power state. See README.md for instructions.
- Navi does not support streaming SQ counters and ATT at the same time, unlike GFX9.
- On Navi ATT, "att: target_cu" indexes the WGP and the SIMD_MASK parameter is actually the SIMD_ID, in the range [0,3].
- HIP RT in ATT not yet supported.
- On Navi3x, counter collection requires the GPU to be in a stable power state. See README.md for instructions. HIP RT in ATT not yet supported.
### Changed
- ATT analysis will not run by default. For ATT to have the same behaviour as 5.5, use --plugin att <as.s> --mode network
### Optimized
projects/rocprofiler/README.md
+10 -6
View File
@@ -236,15 +236,19 @@ The user has two options for building:
# app_relative_path is the path for the application binary
# Mode:
# - Network: opens the server with the browser UI.
# att needs 2 ports opened (8000, 18000), In case the browser is running on a different machine.
# att needs 2 ports available (e.g. 8000, 18000). There is an option (default: --ports "8000,18000") option to change these.
# In case the browser is running on a different machine, port forwarding can be done with ssh -L 8000:localhost:8000 <user@IP>.
# - File: dumps the json files to disk, it can be used to quickly verify if there is anything wrong with the data.
# - Off runs collection but not analysis/parsing. So it can be later used on another system to be viewed.
# Run python3 httpserver.py from within the generated ui/ folder to view the trace. The folder can be copied to another machine, and will run without rocm.
# - Off runs collection but not analysis/parsing. So it can be later viewed another time and/or system.
# input.txt gives flexibility to to target the compute unit and provide filters.
# input.txt contents:
# TARGET_CU=1 // or some other CU [0,15]
# SE_MASK=0x1 // bitmask of shader engines. The fewer, the easier on the hardware. Default enables all 24 because SE_MASK code is recent.
# SIMD_MASK=0xF // bitmask of SIMDs, there are four in GFX9.
# samples/att.txt is having an example on how to right input file for ATT
# TARGET_CU=1 // or some other CU [0,15] - WGP for Navi
# SE_MASK=0x1 // bitmask of shader engines. The fewer, the easier on the hardware. Default enables all shader engines.
# SIMD_MASK=0xF // There are four SIMDs. GFX9: bitmask of SIMDs. Navi: SIMD Index [0-3].
# PERFCOUNTERS_COL_PERIOD=0x3 // Multiplier period for counter collection [0~31]. GFX9 only.
# PERFCOUNTER=<counter_name> // Add a SQ counter to be collected with ATT; period defined by PERFCOUNTERS_COL_PERIOD. GFX9 only.
# samples/att.txt is the simplest input file for ATT
```
- Plugin Support: We have a template for adding new plugins. New plugins can be written on top of rocprofv2 to support the desired output format using include/rocprofiler/v2/rocprofiler_plugins.h header file. These plugins are modular in nature and can easily be decoupled from the code based on need. E.g.