Ensure that the write index and ring buffer contents are visible
to the HW before sending the doorbell. The latter is a write-combined
MMIO store and must be ordered with prior cacehable non-MMIO stores.
Also be more explicit about memory semantics for doorbell stores.
Change-Id: Ie4d96a7ee2a507237a8dbe7705fdf234d62ce9ba
[ROCm/ROCR-Runtime commit: d5b4078072]
If we issue too many copy commands without syncing and wrapping happens,
we need to wait for the blits to be done before moving forward otherwise
we will overwrite the kernel args of the blits in flight.
Change-Id: I9a21e31ce07f8e8157ca38e96dc264ff47fd3639
[ROCm/ROCR-Runtime commit: 5519c96b74]
Introducing tiling format for images, still using LINEAR for now.
Using the new KFD/Thunk API hsaKmtGetTileConfig API for the address library.
Change-Id: Ic0677429dd320eef09ab62dddaf9b2dd94c4f904
[ROCm/ROCR-Runtime commit: 538736a660]
C11 atomics are not statically guaranteed to be lock free and so
may not be atomic with respect to atomic operations originating
outside the standard library, such as platform atomics.
C11 macros to statically discover always lock free operations
(ATOMIC_*_LOCK_FREE) do not cover uint64_t in GCC and
std::atomic<uint64_t> is not a type alias of any covered type.
All use of __atomic by atomic_helpers.h is statically checked to be
always lock free.
GCC builtin fencing does not appear to be strong enough for WC memory.
Added an option (enabled) to enforce consistency for WC memory on x64.
__sync builtin's were not used as they were declared legacy by GCC.
Added a strongly conservative option (ALWAYS_CONSERVATIVE) to enable
use of full memory fences in place of partial fences and compiler
driven processor specific optimization.
Change-Id: Id7aaaca626144070f58759f6a348cbee4612bbc0
[ROCm/ROCR-Runtime commit: 1bc15bbf79]
Change hsa_code_object_serialize and hsa_code_object_deserialize to use memcpy instead of hsa_memory_copy since it is system->system copy
Change-Id: I329e270ae4e2fc25e177dc8080d93662ffb261ab
[ROCm/ROCR-Runtime commit: 73ed2116d5]
Route all device-visible system memory allocations through system_allocator.
Change-Id: I5e90a1bf491e432678a6d8ab1f9f3770734cbda1
[ROCm/ROCR-Runtime commit: 74f5aca93d]
- Includes Sean's latest changes
- Cleanups/improvements
- Fixes for few bugs that crept over from previous releases
Change-Id: I839dc4895bf13ebd0afc8843424387a9fef667b0
[ROCm/ROCR-Runtime commit: c2c993e0d8]
The PM4 IB must have executable permission.
A second part of this fix concerns robustness when this is not the case.
This remains under investigation.
This fix will shortly be cleaned up in a refactoring pass to consolidate
calls to hsaKmtAllocMemory.
Change-Id: I326fe01949a77669e0b07c3cadc9fd44b8065055
[ROCm/ROCR-Runtime commit: f71de56c79]
- Doxygenify comments
- Match order of implementation with order of declaration
Change-Id: I3c7e486c4dd3616f4b10b2f3e69532a4b5fb9e8e
[ROCm/ROCR-Runtime commit: 01dc3a8ff3]
Due to a misinterpretation of the HSA specification the microcode has,
until now, been responsible for ensuring a coherent view of the
amd_kernel_code_t object when acquire_fence_scope is set to agent or system.
To correct this the runtime must instead assume this responsibility.
Introduce GpuAgentInt::InvalidateCodeCaches to perform this operation
on-demand. Invoke this after code object allocation. Extend the Queue
implementations to support PM4 command submission, through which the
PM4 command ACQUIRE_MEM can be submitted to perform cache invalidation.
Submit through a runtime-managed queue shared with the blit implementation.
This change depends on microcode support and this is checked against the
running version. Older microcode builds will perform cache invalidation
themselves, so it is acceptable for this change to do nothing in that case.
Change-Id: I268dd2b83af3decdd9ad07430a81df8a2ecb6bd2
[ROCm/ROCR-Runtime commit: f76577ae43]
The default optimization level may interfere with debugging.
Change-Id: Ie694ef35b05e4cf2bf4f68bc346e8d60a2d27bc8
[ROCm/ROCR-Runtime commit: d2a4629c55]
This option was disabled by default to address issues writing to stderr
in Windows applications. The lack of an error message for memory access
faults is confusing to users, however.
Enable the error message by default on Linux only.
Change-Id: I1f44ba42362f8874abdc7c8e63ddd54a855b5394
[ROCm/ROCR-Runtime commit: acc5f15e4c]
The runtime needs a queue on which to submit cache management commands.
Device-to-device blit copy already creates a queue unconditionally.
We can share this queue for both purposes.
This change restructures the BlitKernel interface to accept, rather than
create, a queue. GpuAgent creates queues as needed for both cache
management and blit compute.
Fix queue full detection in AcquireWriteIndex (<= vs <).
Change-Id: I61d0c6b9d04f2dba74872f0676ad791435778ba4
[ROCm/ROCR-Runtime commit: f7ab361347]
This is the first part of transitioning to the LLVM-based assembler.
SP3 is deprecated and all references to the library are removed.
Pending LLVM support, relevant shaders have been precompiled.
Change-Id: I7d44cef5ded1836c4a74b77881af5bea8803d2c1
[ROCm/ROCR-Runtime commit: 712ea75377]
On multi-node systems only the first CPU node was recognized in the
signal consumer list, causing fallback to non-interrupt signals.
Change-Id: I9bd0706bafbe046be9d7f210d05fa4cf1fcd16fa
[ROCm/ROCR-Runtime commit: b44417043b]
Before this change, runtime hard code the device name, in this commit,
we will query the name from KFD. Will use codecvt to do UTF-16 to
UTF-8 transfer after GCC supports it.
Change-Id: I7c4dc32ef857296296c810d083888c5ba1c808b6
[ROCm/ROCR-Runtime commit: 88708b8e5a]
Sean Keely
Konstantin Zhuravlyov
Chris Freehill
Jay Cornwall
James Edwards
Christophe Paquot
Ramesh Errabolu
Fan Cao