AI/rocm-systems
2
0
Bifurcation 0
Code Tickets Demandes d'ajout Actions Paquets Projets Publications Wiki Activité
829 Révisions 3 Branches 0 Étiquette
694dcd168302aad932ab07d8e827ad2e4ac5ce41
Graphe des révisions

261 Révisions

Auteur SHA1 Message Date
Aditya Atluri 37dec666b2 added hipInit for cuInit nvcc path 
Change-Id: I594d08d936ac5d06f16e42c63062ac5776b65a0c


[ROCm/hip commit: dfc60a8359]
 2016-08-29 15:20:09 -05:00
Aditya Atluri dd8f451428 Changed module api to work with nvcc and hipcc path 
- Added cuda and cudart linking for nvcc path in hipcc
- Added hipMemcpyHtoD and hipMemcpyDtoH for nvcc path
- Changed hipDeviceptr to hipDeviceptr_t
- Fixed hipMemcpy*to* API signatues

Change-Id: I6ef076b507f92502efda475c83dcdcdd462afc37


[ROCm/hip commit: 4b377f63d0]
 2016-08-29 15:05:12 -05:00
Aditya Atluri 8e79c17ca6 Fixed nvcc path for module apis 
Change-Id: I43c7a6bc226f4f270c37f3c4af86b3b3efd0b175


[ROCm/hip commit: 9e21549139]
 2016-08-28 22:50:44 -05:00
Aditya Atluri 874d04291b Cleaned up module api 
- Moved HIP_INIT_API from internal to all public apis
- renamed hipLaunchModuleKernel to hipModuleLaunchKernel
- Changed tests according to the changed api

Change-Id: I822ff63c7c5b7dad340ece49456baf9d89428e9f


[ROCm/hip commit: 3d1bf4dbaf]
 2016-08-28 16:48:57 -05:00
Ben Sander 627a6d2f8f Remove deprecated message from hipLaunchModuleKernel 
Change-Id: I87675453ae4363e3340a9d1491bb00543fa8c6e0


[ROCm/hip commit: 685601f27c]
 2016-08-28 16:34:34 -05:00
Aditya Atluri 8185084ae8 added memset and memcpy device functions 
- Added memcpy and memset device functions
- Added test for memcpy and memset

Change-Id: Icd21a8dd964953b86d5e92889bf1664bee647219


[ROCm/hip commit: 8155f0e036]
 2016-08-27 11:13:56 -05:00
Aditya Atluri 2efa60a37e Added explicit memory copy direction apis 
- Fixed stale printf in context api
- Added 4 sync memcpy apis
  1. hipMemcpyHtoD
  2. hipMemcpyDtoH
  3. hipMemcpyDtoD
  4. hipMemcpyHtoH
- Added test for added apis

Change-Id: I4a9c382445b62631f8d0bcbb9a670322288b72b1


[ROCm/hip commit: 4152746e26]
 2016-08-26 13:11:01 -05:00
Aditya Atluri 34257c06e6 Added NVCC support and name changes 
- Added NVCC support for module APIs
- Changed hipFunction and hipModule data types to hipFunction_t and hipModule_t
- Created new intenal ihipModuleGetFunction as it is used twice
- Changed test to match with the new data types

Change-Id: I300a1c7fd40ed7065b1b8b9de97e3a06b96ed729


[ROCm/hip commit: 8b918b065a]
 2016-08-26 10:32:01 -05:00
Rahul Garg 58aa4f61c2 Added logic to update primary ctx when ctx stack is empty, updated hipCtxDestroy and ctxGetCurrent functions 
Change-Id: Ia0a8943c121bc1279788a1cfa9be59af614b04a6


[ROCm/hip commit: 1211cc931c]
 2016-08-26 19:03:23 +05:30
Rahul Garg c0a547b3fc Resolved errors due to hipCtxXXX APIs 
Change-Id: Iffac0095c4352864eca622ea318d2291571b5153


[ROCm/hip commit: ae77d4b6d7]
 2016-08-26 15:32:49 +05:30
Rahul Garg d6bdd0eafb NVCC path support for hipCtxXXX APIs 
Change-Id: Ic7dbfbdaee9d00c0de1363c50758e5e29a96a8b2


[ROCm/hip commit: 5108140087]
 2016-08-26 14:10:36 +05:30
Rahul Garg 376cc78852 Addition of hipCtxEnablePeerAccess and hipCtxDisablePeerAccess functions 
Change-Id: I381c8cbbde17eae7d9bb5d4cb1596cebf4bda039


[ROCm/hip commit: 524eb687d3]
 2016-08-26 13:51:33 +05:30
Aditya Atluri 69d6645be6 Changed how hipEvent_t is typedefed internall 
- Mapped hipEvent_t directly to ihipEvent_t* instead of a handle

Change-Id: I5a8bcca0ef962932e0738c03eb1fc914d23022ae


[ROCm/hip commit: 842553a6e1]
 2016-08-25 14:34:41 -05:00
Aditya Atluri ad9e04acf8 Added hipModuleGetGlobal and hipModuleLoadData 
Change-Id: Iaec873f7d86b72911b6ad32e067a4dfe3d552fe6


[ROCm/hip commit: 79e88a6af6]
 2016-08-25 14:16:53 -05:00
Aditya Atluri c40593fa5f changed internal structure of hipFunction and hipModule 
Change-Id: Ifa343782e29d7e056efc47e56253311013005093


[ROCm/hip commit: cb996c7b7a]
 2016-08-24 09:47:11 -05:00
Aditya Atluri 6842171207 Module test correction and hipModuleUnload API 
- Corrected the hipModule.cpp test to minimal code
- Added hipModuleUnload API
- Added hipModuleUnload API test

Change-Id: I9c40337043d7972a570b795e1bfc104bd2c4d8aa


[ROCm/hip commit: 2287af23a1]
 2016-08-23 14:19:15 -05:00
Aditya Atluri 138b5b83d2 Added stream synchronisation for hipLaunchModuleKernel 
- The module kernel launch is now in sync with commands in its stream
- Moved launch kernel inside ihipStream

Change-Id: Ic00cfcf4882bf81b6203c36881a52575ea68b529


[ROCm/hip commit: 8f0f97f8f9]
 2016-08-22 14:17:55 -05:00
Aditya Atluri 737fe67a25 Added nvcc path for hipComplex APIs 
- Changed from inline to static inline for hipComplex AMD APIs
- Added NVCC path for hipComplex APIs mapped to cuComplex APIs

Change-Id: I809cf3a11b5b1c8bbc7a57c5fbcc3dc6745ccb95


[ROCm/hip commit: 0806958a72]
 2016-08-22 10:29:46 -05:00
Rahul Garg 5a1cf6ccfd Added support for hipCtxSynchronize and hipCtxGetFlags,modified hipDeviceSynchronize 
Change-Id: If7bac667a262fa8c0cb3dc93e97f2534855acd07


[ROCm/hip commit: a498753041]
 2016-08-22 16:15:27 +05:30
Aditya Atluri 7d05468dc1 Added more complex apis and copyright 
- New header which redirects to CUDA/HIP path added for hipComplex.h
- Added more complex device api including fma
- Added copyright to new files

Change-Id: Iff0dece4c438e97d0ae33efa4312975d465a6464


[ROCm/hip commit: 24f6251b99]
 2016-08-19 23:02:04 -05:00
Aditya Atluri 66d76146d2 Added support for complex device functions 
- Added complex number arithmetic operation for float and double datatypes
- TODO: make them host functions and support half
- Added new function which is not in CUDA, hipCsqabs which is square of absolute value

Change-Id: Ib96e194ad45dc64fcba29eb19ad0376542e0591d


[ROCm/hip commit: 78b15bf062]
 2016-08-19 21:48:23 -05:00
Aditya Atluri 719e91aacf Added support for executable and symbols for data structures 
- symbol handle is added to hipFunction
- executable handle is added to hipModule
- This way, the APIs doesn't need to track the values

Change-Id: I7cf05329cf79fe946319d7746bd9f5503268fda4


[ROCm/hip commit: 921736782e]
 2016-08-19 08:49:34 -05:00
Aditya Atluri 784e9c6691 Added hipLaunchModuleKernel and new error codes 
- hipLaunchModuleKernel maps to cuLaunchKernel
- Whole lot of new error codes added for the use of driver api
 - KernelParams arguments is not yet supported
 - hipLaunchModuleKernel is a synchronous api (will change eventually)
 - All the commands in a stream will wait on host when hipLaunchModuleKernel is called on it

Change-Id: Ib4a4fae1db06fbb3a81d5a5575b026aa821264ed


[ROCm/hip commit: e51ce8fc09]
 2016-08-18 11:26:55 -05:00
Rahul Garg 962276c45a Added further hipCtxXXX Apis 
Change-Id: I286d962a06cee656c1c652b3f6b45078587fbc41


[ROCm/hip commit: 0962ca43d9]
 2016-08-17 16:28:22 +05:30
pensun e8870d2505 add occupancy support for NV path; fix hipPeekAtLastError on HCC path 
Change-Id: I26b0e1875c19d7c636ffcc18f1738926572ded81


[ROCm/hip commit: 2d228d8e0c]
 2016-08-16 16:25:03 -05:00
Aditya Atluri e582b27f7f Added kernel compilation driver apis 
1. Added 2 new driver apis, hipModuleLoad, hipModuleGetFunction

Change-Id: If464a7fad178121e3da791c7ac9e17ebc01a9cd0
Issues: When a sample written with them shows Aborted (core dumped) when exiting


[ROCm/hip commit: 3d27bbd3db]
 2016-08-16 14:36:25 -05:00
Evgeny Mankov 150e3acdad #define HIP_DYNAMIC_SHARED_ATTRIBUTE is added 
[ROCm/hip commit: 7015675983]
 2016-08-16 17:58:57 +03:00
Rahul Garg e33c5c07e3 Implementation of hipCtxGetDevice 
Change-Id: I067572e486323c3aad6f744a2c0c4997c8696af6


[ROCm/hip commit: eec9edef80]
 2016-08-13 01:17:46 +05:30
Rahul Garg 5e91fe9af3 First implementation of hipCtxXXX functions 
Change-Id: I4609cbe6bd90a1fff8655bff4fdd773864397aba


[ROCm/hip commit: 62d390da58]
 2016-08-13 00:09:08 +05:30
Jeffrey Poznanovic cbe87288c8 Adding hipblas include files 
Change-Id: I73064d410acd8f655dc62eaeb6f4bdefc5381e35


[ROCm/hip commit: 48491a3978]
 2016-08-12 11:59:25 +05:30
Ben Sander 0221323977 Context update. 
- Remove tls_deviceID.
- Add first passing test.

Change-Id: If3e2f254abf589028cfe4f9e6369745f04160de0


[ROCm/hip commit: 89164259ab]
 2016-08-10 08:59:47 -05:00
Rahul Garg e64ae4bda0 Changed StagingBuffer class to UnpinnedCopyEngine 
Change-Id: I1e212bfc8030dcf225ecf78fd7b23fda9b1de92f


[ROCm/hip commit: 2ac93c340d]
 2016-08-09 21:29:42 +05:30
Rahul Garg 99f4d4ebec Moved sync copy decision logic to staging buffer class 
Change-Id: I5c398772375fcc1f174a7597eea1215ce7bf80b4


[ROCm/hip commit: 023b1ecf33]
 2016-08-09 09:28:18 +05:30
Ben Sander db8c7e97bd Add initial context implementation. 
APIs: hipInit, hipCtxCreate.
Track TLS default ctx.  Set deviceID now changes the ctx.
Add first context test.

Change-Id: If1cb9989b5a04a36147e25e84904336c7b6f3d88


[ROCm/hip commit: 8f402132ba]
 2016-08-08 17:49:02 -05:00
Ben Sander 28db2f79ba Code cleanup, use camelCase where appropriate. 
Change-Id: I5a7ec50df8bbb3e7a3b313c0b12e2dd55ae4a09c


[ROCm/hip commit: ed0a2c02fe]
 2016-08-08 14:54:38 -05:00
Ben Sander e8757f8154 Move copy kernel templates into hip_memory.cpp 
Change-Id: I862529f3fa8232372c6bacaa5d36f035bbdd32a1


[ROCm/hip commit: 2a798152d4]
 2016-08-08 12:07:12 -05:00
Ben Sander 57df8a3967 Split ihipCtx_t into ihipCtx_t and ihipDevice_t . 
Major change to existing code base.
    Ctx holds streams, enables peers, and flags.
    Device holds accelerator, hsa-agent, device props.

Add hipCtx_t.

Add peer APIs that accept hipCtx_t (in addition to deviceId)

Compiles and passes directed tests.

Change-Id: Iddab1eb9edbf90caad2ef5959c6b811d658197f1


[ROCm/hip commit: cfdacab32f]
 2016-08-08 11:55:57 -05:00
Ben Sander ee356ad0d4 Change Device->Ctx 
Change ihipDevice_t -> ihipCtx_t (new)
Change ihipGetTlsDefaultDevice->ihipGetTlsDefaultCtx
Some other changes from device->ctx where appropriate.

Change-Id: I5c4ae93b2fd42c6303aa23d748eb166b7431925d


[ROCm/hip commit: 2dc3d3238b]
 2016-08-07 21:47:12 -05:00
Ben Sander 308578d520 Remove ihipStream_r::_device_index 
Replace with direct pointer to device.  Cleaner, and prep
for transition to contexts.

Change-Id: I0e550f34412923d46c541c0a14bb7d29c3fd4b11


[ROCm/hip commit: e7d7c5cbe8]
 2016-08-07 20:47:06 -05:00
Rahul Garg 63c4dc3273 Region based apis to pool based api changes 
Change-Id: If53019eebafe051ab4e811863995f78315297080


[ROCm/hip commit: fcb2fcce1e]
 2016-08-05 15:05:57 +05:30
Ben Sander 91368967bb Cleanup sync code. 
Remove dead depFutures, enqueueBarrier call.
Rename some parms to reflect usage.
Add comments to better explain tricky parts of sync code.

Change-Id: I763296421d9c2b3b58fc8cef5f010b12ab49553c


[ROCm/hip commit: 02dd7a7399]
 2016-07-27 18:31:11 -05:00
Aditya Atluri c2b4a3936c Signal Fix: Added signal limit to allocSignal 
1. Did not change the logic in allocSignal
2. Added guard to wait on signal limit

Change-Id: I78f29097e6a584b3c3d78319dac19869067bd1fe


[ROCm/hip commit: 1859c6e515]
 2016-07-27 13:48:49 -05:00
Aditya Atluri 6ff74b7780 Signal Fix: Moved kernel count to critical stream 
1. Added environment variable HIP_NUM_KERNELS_INFLIGHT
2. Moved kernelcount variable inside stream critical section

Change-Id: I51d24d0a2a109467209170de117a6d02ba4e308e


[ROCm/hip commit: 0a31b47e2e]
 2016-07-26 17:09:27 -05:00
Aditya Atluri eccab29360 Signal Fix: Changed global signal count to per stream signal count 
1. The number of kernels that can use signals are increased to 128
2. The kernel count is now specific to the stream

Change-Id: Ie6d1aa3f437aad8f08c3333fe48bd3f46e551e60


[ROCm/hip commit: 53d7629a85]
 2016-07-26 14:03:51 -05:00
Aditya Atluri 69aed2c6ea Added re-fix for memcpy kernel sync 
1. The patch uses HIP signal pools to sync between copy and kernel commands
2. The hsa_signal_create is removed
3. Left the redundant enqueueBarrier method just in case

Change-Id: I3dff3e8ee57fff3cd49bec802ff735ed128e5ca1


[ROCm/hip commit: 4bdf26a82e]
 2016-07-26 09:22:59 -05:00
Rahul Garg d506d58919 D2H and H2D unpinned memory transfer support 
Change-Id: If6d6c970f435e5d917d5cc6cddc2ee2918cd1c37

Conflicts:
	src/hip_hcc.cpp


[ROCm/hip commit: 42a3ed544c]
 2016-07-25 14:36:07 +05:30
Aditya Atluri 6f95ad0a8d Partial fix async after kernel launch signal issue 
Change-Id: Ib48d6564379160035bded9493b93663fba361710


[ROCm/hip commit: c756bb3398]
 2016-07-23 14:54:20 -05:00
pensun 2e5a6ab630 Add empty stubs for threadfence family routines, changes include: 
- stubs and documentation in include/hcc_details/hip_runtime.h
    - stubs with "no-op" in src/hip_memory.cpp
    - document update in hip_kernel_language.md, add suggestions to
    disable L1 and L2 caches when using the threadfence routines.

Change-Id: Ic0753170f802003055bca9d7476d7f48817b98b7


[ROCm/hip commit: f31668fee4]
 2016-07-22 10:40:58 -05:00
Maneesh Gupta f02201e8c3 Replace calls to ihipInit with use of HIP_INIT_API macro 
Change-Id: Iabf7df79f0238a8ddffea4607fe945df36642850


[ROCm/hip commit: 71d51170ef]
 2016-07-22 15:46:55 +05:30
Maneesh Gupta cfd8b8f0c4 Fix using ATP markers 
Change-Id: If2d04f80b580237426c569737551e2001a8cd35a


[ROCm/hip commit: b23fad53cc]
 2016-07-21 16:02:51 +05:30