İşleme Grafiği

287 İşleme

Yazar SHA1 Mesaj Tarih
Tao Sang b14d9e0a36 Deduce hops from numa distance
Rocr won't return real hops so that we have to deduce hops from
numa distance as a workaround. This will be subject to change
as driver team will provide a long term solution in rocm3.7

Change-Id: Ifb939ed848db190c3d544bb7f30a5821161921e6
2020-06-13 17:04:17 -04:00
Jason Tang 30e36a4925 Get gpu asicRevision_
Change-Id: I8afb53e027d912843655823eebaec29901c33d28
2020-06-12 14:51:52 -04:00
German Andryeyev c5afd5d412 Initial HMM support
- Expose ROCclr interfaces for HIP usage
- ROCr interfaces aren't available in staging, thus control the
build with AMD_HMM_SUPPORT define

Change-Id: Iadc2bcc230e78d3b0dc22b235189c8cc80843446
2020-06-12 09:06:07 -04:00
Saleel Kudchadker 2b771d2f5f Add logging support for AQL packet
Use AMD_LOG_LEVEL=4 and AMD_LOG_MASK=8 to print AQL log
explicitly
Change-Id: I4209d91b460e64be44261d3ab773580067e47c29
2020-06-10 14:04:47 -07:00
Aryan Salmanpour 5203c5271b reuse available queue with the same priority if queue creation fails
Change-Id: I1522935ce1a73c4724135668d039e88a9097710b
2020-06-09 17:56:18 -04:00
Aryan Salmanpour 1d938697a1 fix a build error on TeamCity due to order of initialization in rocm Device class
Change-Id: I2e45e71e53272b7a7813422d3b949e3835077ab0
2020-06-08 13:56:39 -04:00
Aryan Salmanpour b5552aa97f Add support for setting queue priority for ROCm backend
Change-Id: I67ed5a6868af79538f7f4522d8d11c043cdf3c1e
2020-06-04 20:16:32 -04:00
Saleel Kudchadker 2b6fea4348 Fix a potential mem leak
If a offset of the pointer is passed to free it may release
the mem object but may not release from MemObjMap. Erase the map
by getting the parent pointer.
Change-Id: I06b92548de2d49b4029efe6b511329225007cc55
2020-06-04 14:05:11 -04:00
Tao Sang fabfc42b68 Fix TC linux build issue due to previous Numa patch
Change-Id: I6068edaf38cac6fad187c8429707afdb727e8d41
2020-06-03 16:42:53 -04:00
Tao Sang aedb9590be Support Numa-aware cpu selection
Select cpu in terms of the smallest Numa distance for a GPU device.
This will improve performance of hipMemcpy in the mode of
hipMemcpyHostToDevice or hipMemcpyDeviceToHost for small buffer.
`

Change-Id: I2860f1f83b79be0dff7bf5e64cf68ab4448db0a1
2020-06-01 21:01:24 -04:00
Aryan Salmanpour fec4adfd19 check for valid queue before accessing cuMask()
Change-Id: I8d4b0dbcd097c2ec5c31dea5a3d0060f0864a7e8
2020-05-20 16:23:09 -04:00
Chauncey Hui 0af9c06968 Modified IpcDetach to return status instead of void.
Change-Id: I68ed94b93f0383babe25eb046b4047d249a0fdc1
2020-05-20 03:38:21 -04:00
Jason Tang cd2a713d63 Add major/minor/stepping to device layer
Change-Id: If82ea55a46b166b243a98089a6e9c40ccfdb479f
2020-05-17 12:57:34 -04:00
Aryan Salmanpour fed94b8604 Add support for setting CU mask on ROCclr for ROCm backend
Change-Id: I0dbe2eeb33467fc0f24b26929119c10e9b455da7
2020-05-15 14:23:43 -04:00
Saleel Kudchadker d10d691e76 Add env var to toggle large bar support in runtime
Use ROC_ENABLE_LARGE_BAR (0/1) to toggle. The support is
enabled by default.

Change-Id: I6cb93a46594cb6f5e90bf6057738330225efb553
2020-05-12 13:20:06 -04:00
Jason Tang b4f1239f34 device/rocm: split gfxVersion to major/minor/stepping
Change-Id: I1e437eaee30794147713d9516229211670f01d90
2020-05-12 12:17:13 -04:00
German Andryeyev ae4aceb55e Make sure the list of HSA agents is valid
If HIP_VISIBLE_DEVICES is active, then make sure the list of HSA
agents contains the valid agents

Change-Id: I584aad999a230ab7f88a0cfe20dcd0abe79c43a5
2020-05-11 15:49:30 -04:00
Christophe Paquot 3ed185307e Fix cooperative flag for hsa_queue creation in case they're not available
SWDEV-233766

Change-Id: If410ecfed61f2b3bb50b847cf2ededc573139494
2020-05-11 13:40:50 -04:00
Michael LIAO 503ef06555 Clear executable permission.
Change-Id: Ia0d363b1ba89d7947e5b5a55cb67edba86f0515e
2020-05-07 10:38:58 -04:00
Alex Xie bfbc8cd09b SWDEV-234684 - hipmemcpy optimization does not work in tests
Change-Id: I899d172c5b2af88c796fe9a36f97d15ac45caf94
2020-05-05 15:58:03 -04:00
Saleel Kudchadker 0fbc0a895b Disable small copy optimization for now
Change-Id: Ib7a4aa676bb60940e067c985eb19070bd63b2fc2
2020-05-05 11:52:42 -04:00
Alex Xie 6c5a42b33c SWDEV-232894 Port hipMemcpy optimizations from HCC to VDI
Apply the optimization to change for OpenCL too.
Clean up some unnecessary checks.

Change-Id: I840261fe35baeeadeba7388e86779d482f509aad
2020-04-30 11:06:28 -04:00
Christophe Paquot b54c3f7db9 Couple of cleanups.
Remove queue limitation since we loop through HW queues now.
Add a DevLogError if we fail to create the hsa_queue. A ticket showed a regression there.

Change-Id: I4f58e405f88e75600a762f6d6352838c969cdb5e
2020-04-29 09:18:07 -07:00
agodavar f149fe0803 P2PStating buffer allocation when P2P is not enabled between all GPUs
SWDEV-232580 & SWDEV-232580
Allocate p2p statging buffer when full P2P access is not available between all devices.
p2p staging buffer will eventually be used when required.

Change-Id: If8490ba7b1c52c432c1e942ae95421b9d2ec7097
2020-04-28 07:10:57 -04:00
Alex Xie 009d0b5f55 SWDEV-232894 Port hipMemcpy optimizations from HCC to VDI
Change-Id: I6bebe9ac503a9f80d067aeea8a848409ad210338
2020-04-27 14:53:58 -04:00
German Andryeyev 082cbfa1f5 Don't attempt to reuse the cooperative queue
Change-Id: I0e98e292a562715a7b395118f899af859f3e42bb
2020-04-27 09:18:05 -04:00
Michael LIAO 97f55b5c7f [vdi] Add device assertion support.
- Once device assertion occurs, abort the host execution as well.
- TODO: This's the initial support. As we need to drain hostcall queue
  to ensure device assertion message being flushed out, hostcall
  listener needs an interface to explicitly drain its queue.

Change-Id: I8a04400aa7109bfd054ae5777c41a4abbf0db4a9
2020-04-22 10:03:55 -04:00
kjayapra-amd 7458bf9964 SWDEV-229840 - Improve error messages on ROCCLR Layer.
Change-Id: Iab7d9156cdc206db86385aa05023a0095ed40f92
2020-04-19 20:01:49 -04:00
German Andryeyev 481d526859 SWDEV-184709 - support hipLaunchCooperativeKernel()
- Enable cooperative groups support, based on ROCr capability

Change-Id: I975bcea0af7865009eaed24454ce71d897ea8fc4
2020-04-01 12:13:33 -04:00
German Andryeyev 7ef8dfdfe7 SWDEV-184709 - support hipLaunchCooperativeKernel()
Add ROCr cooperative queue allocation

Change-Id: I1384482692f4080d31255b09e0f68a21ccad3da8
2020-03-30 16:09:09 -04:00
Vladislav Sytchenko 52046e41b2 SWDEV-224023
Correct typo.

Change-Id: I72131a6e0210e7b961e586cd0ae18608d21fc529
2020-03-15 16:37:25 -04:00
Vladislav Sytchenko e76d867740 SWDEV-224023
Each WGP consists of 2 CU, so the number of available SIMD units is doubled.

Change-Id: I43978a8a9139c33f5f776b344a36bee927cc187d
2020-03-06 13:43:36 -05:00
Saleel Kudchadker 0730b39adb Implement HIP_HIDDEN_FREE_MEM env var
Set value to 256Mb to reflect what HIP/HCC reserves
Change-Id: Icaadf79f60d3916965ac168da237d15b975b1fe4
2020-02-14 12:57:11 -05:00
Karthik Jayaprakash 7fb53890b8 SWDEV-210443 - For Numa nodes pick up the CPU that has Memory pool.
Change-Id: If52852b6f12053e4dfe8a83b8aa5743137c3d6dc
2020-02-13 20:48:37 -05:00
Laurent Morichetti d9d9c69399 Replace cl_* integral types with standard types.
cl_bool -> bool
cl_int -> int32_t
cl_uint -> uint32_t
cl_long -> int64_t
cl_ulong -> uint64_t
cl_float -> float
cl_double -> double
cl_bitfield -> uint64_t

Change-Id: I840c8993b55f98f5b745d21e27f5f28233647a58
2020-02-12 13:16:06 -08:00
Laurent Morichetti b4c6143a2f Update copyright info
Change-Id: Ia4f9ff0f5f873b4223a8cca154188bb0d2f1abba
2020-02-04 09:26:14 -08:00
Laurent Morichetti 20c7173849 Merge branch 'origin/pghafari/vdi-prototype' into lmoriche/amd-master
Change-Id: Id3b833d405596735becb3346f3b08c6da57033fe
2020-01-30 20:12:13 -08:00