* Add HasExpertSchedMode device prop
* Add unit tests for HasExpertSchedMode
* Add gfx12 check for HasExpertSchedMode prop
* Update gfx major version check and test for ExpertSchedMode
* Minor fix and ROCr version bump
* Update projects/rocr-runtime/runtime/hsa-runtime/inc/hsa_ext_amd.h
* Update projects/rocr-runtime/runtime/hsa-runtime/inc/hsa_ext_amd.h
* Apply suggestion from @dayatsin-amd
* Apply suggestion from @dayatsin-amd
---------
Co-authored-by: Stefan Sokolovic <stefan.sokolovic2@amd.com>
Co-authored-by: David Yat Sin <77975354+dayatsin-amd@users.noreply.github.com>
* hsakmt: Expose CWSR and Control stack sizes
This is better than hardcoding values and hoping that they align with
KFD's definitions
Signed-off-by: Kent Russell <kent.russell@amd.com>
* hsakmt: Use CwsrSize and CtlStackSize if available
If KFD is providing the CwsrSize and CtlStackSize, use the maximum
of those and the old calculations for the ctx_save_restore_size
and ctl_stack_size defined in the queue
Signed-off-by: Kent Russell <kent.russell@amd.com>
* hsakmt: Add warning when ABI<1.20 on GFX1151
CwsrSize and CtlStackSize are reported by KFD ABI 1.20. GFX1151
specifically may have some issues if these regions are misaligned, so
report a strong warning during topology initialization if the system is
GFX1151 but is using KFD ABI < 1.20
Signed-off-by: Kent Russell <kent.russell@amd.com>
---------
Signed-off-by: Kent Russell <kent.russell@amd.com>
Although the value is correct; there is no source of truth between
kernel and userspace. This leads to problems if the kernel has strict
restrictions (such as kernel 6.17 or earlier). The restrictions were
lifted in 6.17.9 and and 6.18, but there is no guarantee userspace is
using this.
So short term this value will be wrong. But on newer kernels the kernel
will communicate the right size and rocr-runtime will be adjusted to
use that.
Link: https://github.com/ROCm/TheRock/pull/2505
Signed-off-by: Mario Limonciello (AMD) <superm1@kernel.org>
* rocrtst: Updated CMakeFiles to find_package instead of hardcoded
This is to support TheROCK build environment
* rocrtst: Fix CMake to use find_package() instead of hardcoded ENV paths
Fixed CMake style issues from previos first commit's code review
* rocrtst: Fix rocrtst NUMA dependency detection to use find_package
Also added handling of missing headers
* rocrtst: Fix NUMA and hwloc detection for cross-platform builds
---------
Co-authored-by: Shweta Khatri <shweta.khatri@amd.com>
* SWDEV-558848 - vmm api support for rocr on windows
* Fixes to VMM handle Map/Unmap Set/Get Access
* Fix GetShareableHandle to use pointer for shareable handle
* Update os specific map/unmap memory calls
* clang format update
* Minor syntax fixes from code review
Co-authored-by: Yiannis Papadopoulos <102817138+ypapadop-amd@users.noreply.github.com>
---------
Co-authored-by: Rahul Manocha <rmanocha@amd.com>
Co-authored-by: Yiannis Papadopoulos <102817138+ypapadop-amd@users.noreply.github.com>
* Run pre-commit's whitespace related hooks on projects/rocr-runtime
In order for pre-commit to be useful, everything needs to meet a common
baseline.
Signed-off-by: Mario Limonciello (AMD) <superm1@kernel.org>
* Add missing semicolon which would block compilation on big endian CPUs
Signed-off-by: Mario Limonciello (AMD) <superm1@kernel.org>
---------
Signed-off-by: Mario Limonciello (AMD) <superm1@kernel.org>
* Fixing the copy back to the original buffer malformed packets
* Addressing Copilot Comments
* Addressing Review comments
* Adjust staging buffer size allocation
Change staging buffer size to match the number of packets.
This patch enhances compatibility for DXG environments by introducing conditional
checks for DRM operations, particularly around buffer object metadata handling
in IPC scenarios. These changes improve robustness in DXG IPC memory management
without impacting existing functionality in standard Linux environments.
Signed-off-by: Horatio Zhang <Hongkun.Zhang@amd.com>
WSL uses the call just for the thread wake-up, however under Windows
KMD needs the actual value (SWDEV-568592). The interface is changed
to avoid programming of a modified write_ptr value, which somewhat
changes the client's logic.
Changed ipc_sock_server_conns_ map's value type to size_t. Previous
type of int caused allocations of sizes greater than 2GB to overflow,
causing the message len to be stored as a negative value, preventing the
IPC server from exporting dmabuf file descriptors, which lead to hangs.
Signed-off-by: Sunday Clement <Sunday.Clement@amd.com>
* kfdtest: Replace pthread with std::thread
Modify concurrent kfdtest to use std::thread
instead of pthread, eventually modify KFDTestLaunch
to take in a member function of test instance
instead of static function.
Convert KFDQMTest to pass in member function for
multi-gpu kfdtest.
* kfdtest: Convert KFDPerfCountersTest to use std::thread
Convert KFDPerfCountersTest to use std::thread for
multi-gpu kfdtest.
* kfdtest: Convert KFDGraphicsInterop to use std::thread
Convert KFDGraphicsInterop to use std::thread for
multi-gpu kfdtest.
* kfdtest: Convert KFDGWSTest to use std::thread
Convert KFDGWSTest to use std::thread for
multi-gpu kfdtest.
* kfdtest: Convert KFDCWSRTest to use std::thread
Convert KFDCWSRTest to use std::thread for
multi-gpu kfdtest.
* kfdtest: Convert KFDEventTest to use std::thread
Convert KFDEventTest to use std::thread for
multi-gpu kfdtest.
* kfdtest: Convert KFDExceptionTest to use std::thread
Convert KFDExceptionTest to use std::thread for
multi-gpu kfdtest.
* kfdtest: Convert KFDLocalMemoryTest to use std::thread
Convert KFDLocalMemoryTest to use std::thread for
multi-gpu kfdtest.
* kfdtest: Convert KFDMemoryTest to use std::thread
Convert KFDMemoryTest to use std::thread for
multi-gpu kfdtest.
* kfdtest: Convert KFDSVMRangeTest to use std::thread
Convert KFDSVMRangeTest to use std::thread for
multi-gpu kfdtest.
* kfdtest: Convert KFDHWSTest to use std::thread
Convert KFDHWSTest to use std::thread for
multi-gpu kfdtest.
* kfdtest: Remove pthread multigpu test structure
Remove older multi-gpu test framework which
uses pthread.
Core dumps are not supporetd for gfx110x, but should be possible for
gfx115x. The current code disables core dumps completly for all gfx11xx
agents, relax this to allow gfx115x.
Update the format script to use absolute path for clang-format-diff.py
instead of relative path. This ensures the script works correctly
regardless of the current working directory when executed.
- Change from './clang-format-diff.py' to '${root}/projects/rocr-runtime/clang-format-diff.py'
- Improves script reliability and portability
Signed-off-by: Honglei Huang <honghuan@amd.com>
* libhsakmt/virtio: change shmem size to 80
Some DGPU props have a lot of information,
so it is necessary to increase the size of shmem.
Signed-off-by: Honglei Huang <honghuan@amd.com>
* libhsakmt/virtio: use BO handle instead of pointer in memory registration
Change vhsakmt_map_to_gpu() return type from void* to vhsakmt_bo_handle
to properly handle buffer object information. This allows access to
both the host address and resource ID needed for memory registration.
Signed-off-by: Honglei Huang <honghuan@amd.com>
* libhsakmt/virtio: Improve memory mapping logic
- Update vhsakmt_mappable() to check NoAddress flag and require HostAccess
- Remove mappable checks in cpu_map/unmap to allow all BOs to be mapped
- Set BO flags properly in vhsakmt_alloc_memory and scratch memory creation
- Ensure scratch memory is correctly flagged for proper handling
Signed-off-by: Honglei Huang <honghuan@amd.com>
* libhsakmt/virtio: add no svm mode for libhsakmt virtio
Add no svm mode for libhsakmt virtio driver, in no svm mode userptrs
need UMD to manage, so add interval tree to manage them.
New Features:
- Add augmented red-black tree based interval tree implementation
* Implement RB-tree insertion, deletion, and color balancing
* Provide interval query for fast overlapping range lookup
* Based on Linux kernel's augmented rbtree implementation
- Improve userptr memory management
* Use interval tree to efficiently track userptr memory regions
* Support finding registered memory within given address ranges
* Optimize memory mapping and unmapping performance
Signed-off-by: Honglei Huang <honghuan@amd.com>
---------
Signed-off-by: Honglei Huang <honghuan@amd.com>
* rocr: Fix exception on AsyncEventControl init
Fix exception on init when compiling with in release mode.
* rocr: Fix crash when interrupts are disabled
Fix segfault due to assert for signal->EopEvent() being false when
HSA_ENABLE_INTERRUPT=0. Use Signal::WaitMultiple(..) when interrupt is
disabled.
---------
Co-authored-by: JeniferC99 <150404595+JeniferC99@users.noreply.github.com>
Fix race condition when SetAsyncSignalHandler for the first time because
async_events_thread_ could be null and launched twice.
Refactored async-events to use lazy_pointer.
* Introduce HsaKFDContext structure and infrastructure for multiple KFD contexts, enabling
independent contexts within a single process.
* Refactor core components (queue, event, FMM, topology) to be context-aware,
using explicit HsaKFDContext parameters instead of global state.
* Replace global hsakmt_kfd_fd with context-specific file descriptors, ensuring full context isolation.
* Maintain backward compatibility by redirecting legacy APIs to use the primary context.
This refactoring establishes a foundation for multi-context support while preserving existing functionality.
Signed-off-by: Junhua Shen <Junhua.Shen@amd.com>
* Support Windows HANDLE in interop_map_buffer
* Refactored Windows HANDLE in interop_map_buffer
* ROCr System Dependent Handle Type
* Fix for ROCr Handle Conversion Bug
* Remove Windows Header
Remove libamdhsacode/win32/elf.h due to license restrictions.
Separate Linux coredump implementation because we do not have the ELF
definitions on Windows.
Co-authored-by: JeniferC99 <150404595+JeniferC99@users.noreply.github.com>
Enable image build in Windows.
Remove some useless codes that fail building in Windows.
Some minor improvement.
Temporarily exclude mipmap test files.
Prevent negative tests affect some tests.
Move some catch info log codes into failed cases.
* kfdtest: Enable GPU selection via CLI for multi-GPU tests
Replaced environment variable-based GPU selection with
GPU selection via command-line parameter --concurrentnodes (-c)
Modified g_TestGPUsNum to be passed in via command-line
parameter --testnodenum (t)
Signed-off-by: Alysa Liu <Alysa.Liu@amd.com>
* kfdtest: Enable GPU selection via CLI for multi-GPU tests
Replaced environment variable-based GPU selection with
GPU selection via command-line parameter --concurrentnodes (-c)
Modified g_TestGPUsNum to be passed in via command-line
parameter --testnodenum (t)
---------
Signed-off-by: Alysa Liu <Alysa.Liu@amd.com>
Co-authored-by: Alysa Liu <Alysa.Liu@amd.com>
Maneesh Gupta
David Yat Sin
Filip Jankovic
Kent Russell
randyh62
Rahul Manocha
Mario Limonciello
shwetakhatri-amd
Jin Jung
Mario Limonciello
Ammar ELWazir
hongkzha-amd
German Andryeyev
Alysa Liu
Sunday Clement
lancesix
Jaydeep
Honglei Huang
jokim-amd
cfreeamd
jonatluu
andmar-amd
JeniferC99
Tim Huang
Junhua Shen
Yiannis Papadopoulos
Xie, AlexBin
Apurv Mishra
MachineTom
systems-assistant[bot]
arvindcheru