This is a hack to allow the Runtime to allocate system memory with
PreferredNode=0 on a dGPU system. We allocated it from Node 1
instead so that the node 1 GPU can map the memory. A proper fix
will be implemented together with multi-GPU support.
Change-Id: Ieb52599e5275781c04ee34405ea850bf782c523a
[ROCm/ROCR-Runtime commit: 590c8e522c]
Try to reserve as much SVM address space as GPUVM can address.
Implement a fallback scheme to smaller sizes if larger allocations
fail or are not addressable by the GPU, down to an (arbitrary)
minimum of 4GB.
Change-Id: I770177834cc9e6ddd6ef4f20d789eab63c8055cb
[ROCm/ROCR-Runtime commit: 39bde26c9b]
When 'make deb' is run create a libhsakmt.deb archive that installs
libhsakmt into the appropriate folder on the target where the dymanic
linker can find it.
Change-Id: I32de7198975f7831e509a67371e78456982b5c42
[ROCm/ROCR-Runtime commit: 0df346aaf9]
Kernel ioctl AMDKFD_IOC_GET_PROCESS_APERTURES returns process apertures
only for GPU nodes. The current implementation assumed that this list of
GPU nodes returned by the ioctl has one to one correpondence to sysfs
topology nodes. This fails when non-GPU nodes exist in topology as in
case of Intel + gfx802
Fix this by using gpu_id (./sys/.../kfd/topology/nodes/1/gpu_id) to map
information obtained from kernel ioctl call.
Change-Id: I4ab8ae5354f12cf0b6609fc4b24182b82eb3677f
[ROCm/ROCR-Runtime commit: 5cc56a2647]
Fix TONGA_PAGE_SIZE value and move it to libhsakmt.h for usiing it
consistently in all places that require the same alignment for the
same reason. Create a generic alignment helper macro to replace some
incorrect hand-coded size alignments.
Move virtual address and size alignments down into aperture management
functions. Alignment is a per-aperture property that is set during
fmm_init_process_apertures. Doing the alignment there ensures that
all allocations in the same aperture are aligned the same way. Finding
objects by size and address can take the alignment into account.
Also align the size of physical allocations to back aligned virtual
address allocations. CPU mappings do not need to be aligned.
Map anonymous pages over released memory mappings to allow the
backing pages to be released, while keeping the address space
reserved.
Add alignment parameter to free_exec_aligned_memory_gpu to match the
interface of allocate_exec_aligned_memory_cpu. It doesn't make sense
to allow an alignment parameter in one but assume a specific
alignment in the other.
Change-Id: I74226ca6938f4948f643e5aee1d474720cd89e78
[ROCm/ROCR-Runtime commit: 6a5ca4bc5a]
Create new device_info and add device ID. Add helper macros to
identify chip families (VI, discrete). For now gfx803 behaves like
gfx802. But if necessary we can have gfx802 or gfx803-specific
code paths or workarounds in the future.
Change-Id: I61b4ffef7dd7796bb34cb01fbff0089bd49507bb
[ROCm/ROCR-Runtime commit: 0fc0a5b526]
hsa_gfxip_table lists only (supported) GPUs. So assert fail only when a
non-supported GPU is detected.
Change-Id: I6207dc7cd55860c8b3348b6a4ca6102131975722
[ROCm/ROCR-Runtime commit: 758824db17]
The default is non-coherent access for better performance on dGPU.
Disabled hsaKmtSetMemoryPolicy function on dGPU to prevent app from
overriding the APE1 settings at runtime.
Fixed dGPU VM aperture limit to be inclusive.
Change-Id: I378ff74a654f533572775c0c97c19779a56bc6d9
[ROCm/ROCR-Runtime commit: 8e836f8183]
1. Add IOCTL defines to set trap handler
2. Add control stack size information on create queue argument.
3. Increase the total save&restore area size for carrizo to include the control stack size.
Signed-off-by: Shaoyun Liu <Shaoyun.liu@amd.com>
Change-Id: Iccf15e073b7db2519e96e7f7b46a89d57ab9a4df
[ROCm/ROCR-Runtime commit: 2d63ee7b8f]
HSA_ENGINE_ID in Perforce added ui32 to the typedef while in Git it doesn't.
This causes conflicts to RT applications. Decision being made is to change Git
to match Perforce.
Change-Id: I7e9c6437b023bb23ec9578737f8534e9453589b9
[ROCm/ROCR-Runtime commit: 082f8314c4]
Currently, Kernel imposes a limit on VM. Thunk should be aware of it.
This fix is required till Kernel VM limit is sorted.
For now both "Host Access" memory and "Local Memory" share the same VM range.
Change-Id: I5a9220face20df9ede2b78bd6201a01dd2ea70e0
Signed-off-by: Harish Kasiviswanathan <Harish.Kasiviswanathan@amd.com>
[ROCm/ROCR-Runtime commit: 1438f15fd0]
Memory size is 64-bit. So use HSAuint64 instead of uint32.
Change-Id: Iaa607dec9c1a1c5ac46ea442fd482210ea550b45
Signed-off-by: Harish Kasiviswanathan <Harish.Kasiviswanathan@amd.com>
[ROCm/ROCR-Runtime commit: 4b768872c0]
Add GFXIP version 8.0.2(major.minor.stepping) for gfx802 and 8.0.3 for gfx803.
Change-Id: Icc7cac6b2e8a78d9cff4105aeb2bfcd2c7759027
[ROCm/ROCR-Runtime commit: f7fffdc2be]
Patch submitted by Besar Wicaksono
1. Bug on detecting local memory size interpreted as 32 bit value
instead of 64. The bug causes thunk to go into an infinite loop trying
to reserve virtual address range for dgpu system memory.
2. SIMD count in the node property is 0. Runtime use this attribute to
find a gpu device.
Regarding other attributes of intel+tonga topology, Harish started a
discussion on August iirc, could you please share an update ?
This would help me progress with more tests such as scratch memory,
which require the scratch aperture information in order to construct a
buffer srd in gpuvm space.
3. Bug on releasing memory via fmm_release, where no actual release is
being done. The vm_object can't be found because the memory size does
not match due to the allocation padded the size with 32KB.
4. Pointer arithmetic on vm_area allocation/release. The value of
vm_area_t::end seems to be interpreted inconsistently whether it is
(start + size -1) or (start + size).
One example of potential issue I see is the logic could report
larger size of the hole in the vm area list.
5. Resource cleanup on multiple library load/unload within a single
process.
- Any memory allocation on subsequent library load will result
an error "va above limit". To my understanding this is due to
the reserved memory for the system memory not being released on unload.
- The static variable events_page needs to be invalidated
appropriately on library unload so the next load could
reinitialize it.
6. Could you please update if AQL queue is ready to test with the stg
kfd/kmt ?
7. The system memory allocation with size larger than 32KB seems to be
padded by an extra 32KB. I was wondering if we could remove this
overhead.
Change-Id: I039988d36637525089c7569dc3b77e58750e2121
[ROCm/ROCR-Runtime commit: ee08f537a7]
Makefile currently sends build output a default location.
Allow choice of build output location if so desired
using a variable.
Signed-off-by: David Ogbeide <davidboyowa.ogbeide@amd.com>
[ROCm/ROCR-Runtime commit: 8a01cd1212]
- Replace HSAuint32 with HSA_ENGINE_ID for EngineId type so it explicitely
presents version information for ucode and GfxIP
- Created a GfxIP lookup table to pass the version information. This lookup
searches for matching device ID.
Signed-off-by: Amber Lin <Amber.Lin@amd.com>
Acked-by: John Bridgman <John.Bridgman@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
[ROCm/ROCR-Runtime commit: a3925a3a19]
including Cherry pick of CL#1166690 from SC Stg and update for RT samples. It contains:
Change 1166280 on 2015/06/30 by bolek@bolek-common2
EPR #092474 - Fix missing integrations
Change 1164156 on 2015-06-23 by nhaustov
ECR #010005 - HSA Finalizer: Add missing tests.
Change 1164232 on 2015-06-23 by nhaustov
ECR #333756 - HSA Finalizer: Implement reading of notes.
Change 1166268 on 2015/06/30 by bolek@bolek-laser
EPR #010001 - Promotion of the Shader Compiler (SC)
Release SC Library version 0001.IL01-02.0339 Date: June 30, 2015
Changelist (stg/sc): 1165197
Change 1163976 on 2015-06-22 by sashao
EPR #373149 - OpenGL ES 3.0 Development
Change 1164122 on 2015-06-23 by nhaustov
ECR #033756 - HSA Finalizer: Fix compilation warnings.
Change 1164150 on 2015-06-23 by efinger
EPR #092474 - bugzilla 10829 - optimize out useless V_PERM_B32 feeding packed math op, in early expansion, rather than late expansion, and do it regardless of whether the packed op will be split or not.
Change 1164187 on 2015-06-23 by efinger
EPR #092474 - Add and use GetUAVInfo() and GetNumUAVs()
Change 1164194 on 2015-06-23 by rgottlie
EPR #092474 - Fix Linux Build Issue for SC_OPEN_SOURCE
Change 1164204 on 2015-06-23 by mbedy
EPR #092474 - Update Open Source build - now working from SC stg.
Change 1164216 on 2015-06-23 by rouellet
EPR #092474 - Add directive to do what -il_interpreter does.
Change 1164232 on 2015-06-23 by nhaustov
ECR #333756 - HSA Finalizer: Implement reading of notes.
Change 1164239 on 2015-06-23 by nhaustov
ECR #333756 - HSA Finalizer: Fix OpenCL build problem.
Change 1164275 on 2015-06-23 by nhaustov
ECR #333756 - HSA Finalizer: Fix Linux build errors.
Change 1164365 on 2015-06-23 by efinger
EPR #092474 - Cleanup UAV Atomic handling
Change 1164393 on 2015-06-23 by kzhuravl
EPR #333756 - Finalizer/Loader fixes
Change 1164654 on 2015-06-24 by dpreobra
ECR #333753 - HSA HLC: SPB_ASM: TestGen improvements
Change 1164727 on 2015-06-24 by bolek
EPR #092474 - Enable function level linking, COMDAT folding and unused function removal optimizations in Dev release builds. This saves about 2.6MB in code size on 64-bit Dev.
Change 1164760 on 2015-06-24 by rgottlie
EPR #422210 - Fix problem with TransformScratch heuristics
Change 1164761 on 2015-06-24 by rgottlie
EPR #422181 - Fix handling of sub-dword load instructions in propagation of immediates from store to load in RefineMemory
Change 1164764 on 2015-06-24 by bolek
EPR #092474 - Add missing const
Change 1164769 on 2015-06-24 by efinger
EPR #092474 - Cleanup GDS atomics
Change 1164776 on 2015-06-24 by efinger
EPR #092474 - Fix linux build
Change 1164799 on 2015-06-24 by mbedy
EPR #092474 - Improve alignment for 2 DWORD instructions by more closely
Change 1164803 on 2015-06-24 by efinger
EPR #092474 - Open Source Cleanup
Change 1164809 on 2015-06-24 by bfavela
EPR #092474 - Escape an infinite loop in shader during the build of a DAG when a block is visited twice
Change 1164814 on 2015-06-24 by bolek
EPR #092474 - Add Dev command line option to disable individual peephole patterns (blame Chris for this one).
Change 1164827 on 2015-06-24 by bfavela
EPR #092474 - Adding small change to CL 1164809 as suggested by creeve to remove superfluous if()
Change 1164842 on 2015-06-24 by gujin
EPR #092474 - Prevent moving exit-loop checking to the end of loop if there is a branch in the loop that is optimized with a target replacement bypassing the loop end. This is to fix an OpenGL hull shader conformance test fail (bug 10859).
Change 1164876 on 2015-06-24 by rgottlie
EPR #092474 - Only allow memory merging if no memory scope or order is specified
Change 1164883 on 2015-06-24 by kdintino
EPR #092474 - Add HSAIL files to the AMD -> LLVM copyright replacement loop.
Change 1165060 on 2015-06-25 by efinger
EPR #092474 - Open Source Cleanup - Copyright
Change 1165077 on 2015-06-25 by efinger
EPR #092474 - Cleanup LDS atomics - part 1 (groundwork)
Change 1165080 on 2015-06-25 by bfavela
EPR #092474 - Extension to SUPPRESS_PI_REDUCE_F32 for TAN (TAN_F16 is already handled by expansion)
Change 1165189 on 2015-06-25 by efinger
EPR #092474 - Cleanup LDS atomics - part 2
Change 1165196 on 2015-06-25 by bolek
EPR #092474 - Add syntax to the peephole pattern language to specify SCInst flag values or wildcards.
Change 1165197 on 2015-06-25 by bolek
EPR #092474 - Allow the MulAddToMadF peephole pattern to modify instructcions marked as invariant (result should still be the same)
Change 1165438 on 2015/06/26 by bolek@bolek-common2
EPR #010001 - Promotion of the Shader Compiler (SC)
Release SC Library version 0001.IL01-02.0338 Date: June 26, 2015
Changelist (stg/sc): 1163954
Change 1161629 on 2015-06-15 by efinger
EPR #092474 - Move CFG:IL2IRProcessDeclare() to global scope
Change 1161633 on 2015-06-15 by rouellet
EPR #092474 - Bugzilla 10852 call ConverInstFields when translating COND_MOVE.
Change 1161643 on 2015-06-15 by rgottlie
EPR #092474 - Handle manually inserted wait state for SALU writing M0 followd by VINTERP
Change 1161718 on 2015-06-15 by lifpan
EPR #092474 - The "point size" in copy shader of GS
Change 1161721 on 2015-06-15 by xlji
EPR #092474 - Split DIV_F16 and DIV_PRECISE_F16
Change 1161850 on 2015-06-16 by kzhuravl
EPR #333756 - Change a few function names, general cleanup (no functional change)
Change 1161934 on 2015-06-16 by efinger
EPR #092474 - Fix linux compile warnings
Change 1161946 on 2015-06-16 by nhaustov
ECR #333756 - HSA Finalizer: Fix Linux build warnings.
Change 1161981 on 2015-06-16 by efinger
EPR #092474 - Open Source Cleanup
Change 1161991 on 2015-06-16 by efinger
EPR #092474 - Move CFG::IL2IRProcessSpecial() to global scope
Change 1161997 on 2015-06-16 by rgottlie
EPR #092474 - Fix compile warnings under Linux
Change 1162001 on 2015-06-16 by efinger
EPR #092474 - Fix linux build
Change 1162045 on 2015-06-16 by mherdeg
EPR #092474 - Comment out unused functions to fix linux compiler warnings.
Change 1162048 on 2015-06-16 by akasauro
EPR #092474 - SC: Some AMD OCL SDK tests (including BinomialOption) assert in SCInst.cpp. [on behalf of Atrem Tamazov]
Change 1162061 on 2015-06-16 by efinger
EPR #092474 - Rename NewIRInst to MakeIRInst and drop last (unused) arg.
Change 1162066 on 2015-06-16 by creeve
EPR #092474 - Linux build fixes for open source.
Change 1162067 on 2015-06-16 by creeve
EPR #092474 - Improve hash table grow and sanitize.
Change 1162072 on 2015-06-16 by creeve
EPR #092474 - Peephole |x| * |x| => x*x
Change 1162089 on 2015-06-16 by chfang
EPR #092474 - Fix linux compiler warnings in SCStructureAnalyzer.cpp.
Change 1162145 on 2015-06-16 by efinger
EPR #092474 - Improve interface to MakeInstOp[123]
Change 1162427 on 2015-06-17 by efinger
EPR #092474 - bugzilla 10862 - Back out changelist 1161549
Change 1162434 on 2015-06-17 by rgottlie
EPR #092474 - Only dump individual functions in each pass of Refine Memory
Change 1162436 on 2015-06-17 by kzhuravl
EPR #333756 - Integrate runtime independent loader from stg hsa + update project files
Change 1162442 on 2015-06-17 by efinger
EPR #092474 - Add and use CreateRegTemp()
Change 1162505 on 2015-06-17 by skolton
ECR #333756 - HSA Finalizer: Doorbell signals support
Change 1162527 on 2015-06-17 by kzhuravl
EPR #333756 - Always set dx10_clamp to true for hsa
Change 1162531 on 2015-06-17 by efinger
EPR #092474 - Fix linux compile warnings
Change 1162568 on 2015-06-17 by mbedy
EPR #092474 - Specify a newer DX9 SDK for SCDevUtil that correctly links with WDK n10136.
Change 1162623 on 2015-06-17 by mherdeg
EPR #092474 - Remove duplicate #include "SCHSAInterface.h". It confuses Intellisense in Visual Studio.
Change 1162905 on 2015-06-18 by rgottlie
EPR #092474 - Fix Linux Build Warnings
Change 1162930 on 2015-06-18 by nhaustov
ECR #333756 - HSA Finalizer: Cleanup amdhsafin command-line tool.
Change 1162938 on 2015-06-18 by nhaustov
ECR #333756 - HSA Finalizer: Fix build problem.
Change 1162944 on 2015-06-18 by rgottlie
EPR #092474 - Clean up bug descriptions as per Phil's suggestion
Change 1162951 on 2015-06-18 by skolton
ECR #333756 - HSA Finalizer: Bug fix for 1DB query image
Change 1163009 on 2015-06-18 by nhaustov
ECR #333756 - HSA Finalizer: build amdhsafin with WITH_LIBBRIGDWARF when needed.
Change 1163263 on 2015-06-19 by nhaustov
ECR #092474 - Fix patgen VS build by quoting %TMPDIR%.
Change 1163265 on 2015-06-19 by skolton
ECR #333756 - HSA Finalizer: Fix for doorbell signal store.
Change 1163310 on 2015-06-19 by nhaustov
ECR #333756 - HSA Finalizer: Introduce separate amdhsacode library.
Change 1163316 on 2015-06-19 by nhaustov
ECR #333756 - HSA Finalizer: Fix OpenCL build problem.
Change 1163320 on 2015-06-19 by nhaustov
ECR #333756 - HSA Finalizer: Fix another OpenCL build problem.
Change 1163331 on 2015-06-19 by mjared
EPR #092474 - Replace asin/acos 5th order minimax polynomial with a 6th order double locked (at 0 and 1) minimax polynomial
Change 1163353 on 2015-06-19 by efinger
EPR #092474 - Use normal temps (not expansion temps) for expansion template T regs.
Change 1163473 on 2015-06-19 by mjared
EPR #092474 - Improve accuracy of ATAN instruction by replacing rational approximation with a 17th order double locked minimax polynomial. Also increase degree of ASIN/ACOS double locked minimax polynomial to 7.
Change 1163475 on 2015-06-19 by creeve
EPR #092474 - Avoid putting partial write on export instruction. This feature existed before but only occurred if the output was point sprite. This change removed that restriction. Also fixed the implementation of //EsMode and //LsMode shader directi
Change 1163481 on 2015-06-19 by mjared
EPR #092747 - Misc. python scripts for working with transcendental functions. Includes fast implementation of remez minimax algorithm for absolute error and slower optimization-based remez for weighted/custom error reduction.
Change 1163528 on 2015-06-19 by creeve
EPR #092474 - Fix build issue.
Change 1163603 on 2015-06-21 by bolek
EPR #092474 - patgen makefile cleanup
Change 1163614 on 2015-06-21 by kzhuravl
EPR #333756 - Integrate runtime independent loader changes from stg hsa
Change 1163699 on 2015-06-22 by rouellet
EPR #092474 - bugzilla 10854 Get cb0[1] initialized with group dimensions for compute shaders on r800 and newer. Make IL and HW interpreter details and variable names more closely match. Flush denorms when doing cube mapped samples (the cb0[1] init
Change 1163713 on 2015-06-22 by efinger
EPR #092474 - Convert all usage of expansion temps to regular temps
Change 1163718 on 2015-06-22 by nhaustov
ECR #333756 - HSA Finalizer: Add loader (-loader option) to amdhsafin and update tests.
Change 1163732 on 2015-06-22 by nhaustov
ECR #333756 - HSA Finalizer: Implement images in amdhsafin loader and update tests.
Change 1163774 on 2015-06-22 by mbedy
EPR #092474 - Strip _DEV macros from open source. Fix issue in ifdef stripping.
Change 1163786 on 2015-06-22 by mbedy
EPR #092474 - Revert unintentionally submitted change.
Change 1163803 on 2015-06-22 by bolek
EPR #092474 - Peephole compile-time performance improvements
Change 1163832 on 2015-06-22 by efinger
EPR #092474 - bugzilla 10849 - fix copy propagation bug with SDWA
Change 1163916 on 2015-06-22 by efinger
EPR #092474 - Nuke support for expansion temps
Change 1163954 on 2015-06-22 by creeve
EPR #092474 - More code sanitization.
Change 1164740 on 2015/06/24 by vpykhtin@vpykhtin-SC
ECR #333753 - Cherrypicking CL1164641 from stg/sc (that is cherrypick of CL1164640 form stg/opencl)
Testing: TC PSDB
[git-p4: depot-paths = "//depot/stg/hsa/drivers/hsa/runtime/": change = 1167011]
[ROCm/ROCR-Runtime commit: 3bbf3c6a8b]
Ben Goz
Felix Kuehling
Andres Rodriguez
Harish Kasiviswanathan
shaoyunl
Amber Lin
David Ogbeide
Ding, Wei (xN/A) TX
Flora Cui
Moses Reuben
Andrey Kasaurov (xN/A) SP
Nikolay Haustov [TEXT]
Oded Gabbay
Zhuravlyov, Konstantin (x21446) MA
Besar Wicaksono (xN/A) TX [TEXT]
Xihan Zhang