* Add regex for undefined behavior to ROCPROFILER_DEFAULT_FAIL_REGEX
- add UBSAN_OPTIONS to setup-sanitizer-env.sh
* Improve ROCPROFILER_DEFAULT_FAIL_REGEX
* Use -fno-sanitize-recover=undefined flag
- this compiler flag causes all undefined behavior errors to exit
* Revert ROCPROFILER_DEFAULT_FAIL_REGEX
* fix for shift overflow
---------
Co-authored-by: Jonathan R. Madsen <jonathanrmadsen@gmail.com>
Co-authored-by: Manjunath-Jakaraddi <manjunath.jakaraddi@amd.com>
* Add example for synchronous reading of device counters
We already have test cases for this use case but this a sample
such that our collaborators can have a place to quickly pull
code from for use on their end (and to serve as a working example).
* Formatting fix
* Formatting fix
* Minor change for testing
---------
Co-authored-by: Benjamin Welton <ben@amd.com>
Set execute permission for HSA allocated memory
We need execute permission for HSA memory (req for IB buffers).
Enforcement is upcoming which will break counter collection (see ticket).
Change-Id: Id939bbdd8686915407bc60d0dfcd5110b0e91e90
Co-authored-by: Benjamin Welton <bewelton@amd.com>
Co-authored-by: Bhardwaj, Gopesh <Gopesh.Bhardwaj@amd.com>
1) When `clang` is used as system compiler, libraries were built without respecting LDFLAGS. For example, this affected LTO flags, if any (and it only affected clang, not gcc).
2) Linker flags are registered as CXX flags, which produces warnings during compilation:
```
clang++: warning: -Wl,-z,noexecstack: 'linker' input unused [-Wunused-command-line-argument]
clang++: warning: -Wl,-znoexecheap: 'linker' input unused [-Wunused-command-line-argument]
clang++: warning: -Wl,-z,relro: 'linker' input unused [-Wunused-command-line-argument]
clang++: warning: -Wl,-z,now: 'linker' input unused [-Wunused-command-line-argument]
```
3) Clang does not support `-Wtrampolines` flag:
```
warning: unknown warning option '-Wtrampolines' [-Wunknown-warning-option]
```
4) No linkers support `noexecheap` anymore. `noexecheap` linker flag was a part of PaX patches to GNU ld, (which were dropped in 2017)[https://www.gentoo.org/support/news-items/2017-08-19-hardened-sources-removal.html]. Now ld/ld.lld/ld.gold don't support it and protection of heap is managed by NX bit. Therefore every compiler produces this warning:
```
ld.lld: warning: unknown -z value: noexecheap
```
Closes#210.
Co-authored-by: Sv. Lockal <lockalsash@gmail.com>
* Adding New HIP APIs
* Format Fix
* Format Fix
* Removing changes from ostream and moving it to format
* Addressing Code Review Comments
* Versioning the new hip calls formatting
---------
Co-authored-by: Ammar ELWazir <aelwazir@amd.com>
* [SWDEV-509876] Remove buffer requirement from device counting service
No longer require a buffer to be given when setting up device counting
service. This is to reduce performance overhead in cases where immediate
return of counting samples is being used (synchronous mode).
* Missed file
* Update source/include/rocprofiler-sdk/device_counting_service.h
Co-authored-by: Madsen, Jonathan <Jonathan.Madsen@amd.com>
* Update source/lib/rocprofiler-sdk/counters/controller.cpp
Co-authored-by: Madsen, Jonathan <Jonathan.Madsen@amd.com>
* Update source/lib/rocprofiler-sdk/counters/device_counting.cpp
Co-authored-by: Madsen, Jonathan <Jonathan.Madsen@amd.com>
* Fixes for build
---------
Co-authored-by: Benjamin Welton <ben@amd.com>
Co-authored-by: Madsen, Jonathan <Jonathan.Madsen@amd.com>
Co-authored-by: Benjamin Welton <bewelton@amd.com>
* Fix async copy validation test
- make the async copy tracing test work regardless of however many HSA memory copies the HIP memory copy decomposes into
* Fix rocprofv3 memory copy tests
* Fix compilation support for hipGraphBatchMemOpNodeGetParams
* Fix rocprofv3-test-summary-*-validate
---------
Co-authored-by: Jonathan R. Madsen <jonathanrmadsen@gmail.com>
Fix HIP data type stringify
- when ROCPROFILER_CI is not defined, provide default for case statements
- Add support for hipGraphNodeTypeBatchMemOp when HIP version is >= 6.4.0
Co-authored-by: Jonathan R. Madsen <jonathanrmadsen@gmail.com>
* Integrating amd-smi into rocprofiler-systems due to rocm-smi deprecation.
* No functionality changes to users other than naming conventions.
* New tracks available in perfetto- gpu busy percentage metrics now splits gfx busy into separate gfx, umc, and mm engine measurements.
---------
Signed-off-by: Carrie Fallows <Carrie.Fallows@amd.com>
Co-authored-by: David Galiffi <David.Galiffi@amd.com>
BUILD_SHARED_LIBS is a global flag so we don't need to set a default
option for it in both libhsakmt and hsa-runtime, only the top level
CMakeLists file. Also updated README to reflect that libhsakmt is
always built statically and gets linked to libhsa-runtime.
Change-Id: I1511f68a268032bec9758bc731d8074f33ec980f
Luna Nova
Madsen, Jonathan
Welton, Benjamin
U, Srihari
Nagaraj, Sriraksha
Elwazir, Ammar
Jonathan Luu
xuchen-amd
Bhardwaj, Gopesh
Galantsev, Dmitrii
Jakaraddi, Manjunath
Choudhary, Rahul
Williams, Justin
AravindanC
ajanicijamd
Vignesh Edithal
David Galiffi
Young Hui - AMD
cfallows-amd
dependabot[bot]
Kandula, Venkateshwar reddy
Kuricheti, Mythreya
sonadeem
Pryor, Adam