30233cd2e0
Add experimental notes around p2p
2016-04-18 23:52:42 -05:00
3bcefdaa2d
Add USE_HCC_LOCK_API.
...
Default off, if set will compile assuming HCC lock APIs exist.
2016-04-18 23:50:52 -05:00
e020d68309
Fixes for P2P and hipDeviceReset
...
- devicereset would lose track of default stream and thus subsequent
synchronization calls might not actually sychronize.
- Also deviceReset now correctly frees streams.
- fix waits in P2P staging copy - first phase (Device0-to-Staging) must
wait for second phase (Staging to Device1) to finish draining the
buffer.
2016-04-18 20:49:33 -05:00
65abde6626
Move HIP_HCC define to CMake
2016-04-17 07:40:04 -05:00
49cc5aec91
Merge branch 'privatestaging' into p2p
...
Conflicts:
include/hcc_detail/hip_hcc.h
src/hip_hcc.cpp
2016-04-17 06:46:52 -05:00
846432a1c1
Merge branch 'privatestaging' of https://github.com/AMDComputeLibraries/HIP-privatestaging into privatestaging
2016-04-16 17:10:32 -05:00
8dc1bdcbe6
Corrected Memcpydefault
2016-04-16 17:10:13 -05:00
dcabc9dbf7
P2P Update.
...
- add P2P staging buffer copy.
- If copy device does not have sufficient access permissions, fall back
to staging buffer.
- improve docs for which copy device is used.
2016-04-16 10:18:56 -05:00
76f2670cef
Add __ldg
2016-04-16 21:12:09 +08:00
dc4c174a54
Bump min required hcc to workweek 16155
2016-04-16 14:49:10 +05:30
a0aef11da7
changed to guard from hc.hpp
2016-04-15 11:21:45 -05:00
2e80de6dcb
improved error returns for hipHostRegister/UnRegister functions
2016-04-15 10:42:31 -05:00
bdd26bd1f1
Merge branch 'privatestaging' of https://github.com/AMDComputeLibraries/HIP-privatestaging into privatestaging
2016-04-14 09:02:16 -05:00
facc7a39be
Merge branch 'privatestaging' of https://github.com/AMDComputeLibraries/HIP-privatestaging into privatestaging
2016-04-14 09:17:30 -05:00
573224e55f
Enabling make_* device code
2016-04-14 09:17:13 -05:00
234727c801
Add missing cudaMemsetAsync transformation
2016-04-14 09:02:02 -05:00
011a0335d4
Add hip soft-link to include for migrating to new include format
2016-04-14 16:37:27 +05:30
210ba67b84
Add missing USE_PEER_TO_PEER macro definition
2016-04-13 13:56:26 +08:00
4dfe27e8b5
Removed pragma once
2016-04-12 09:22:03 -05:00
d5feabfa1c
added C guard to hip_runtime.h
2016-04-11 13:01:02 -05:00
31dc13d2ec
Merge branch 'p2p' of https://github.com/AMDComputeLibraries/HIP-privatestaging into p2p
...
Conflicts:
RELEASE.md
include/hcc_detail/hip_hcc.h
samples/1_Utils/hipInfo/hipInfo.cpp
src/hip_hcc.cpp
src/hip_peer.cpp
2016-04-11 09:17:27 -05:00
1f53c55d3e
P2p checkpoint.
...
- set USE_PEER_TO_PEER=3 (requires HCC "am_memtracker_update_peers")
- when enabling peer, turn it on for previously allocated memory.
- hipDeviceCanAccessPeer is no longer self-ware (self does not qualify
as a peer)
- device peerlist always includes self, so when we call allow_access
we never remove self access.
- hipDeviceReset() removes old peer mappings.
2016-04-11 12:52:18 -05:00
b0529e04f1
Clean up disable.
...
Add USE_HCC_LOCK (disabled)
Disable USE_PEER_TO_PEER.
2016-04-11 09:09:36 -05:00
83f0de7806
P2p checkpoint.
...
- set USE_PEER_TO_PEER=3 (requires HCC "am_memtracker_update_peers")
- when enabling peer, turn it on for previously allocated memory.
- hipDeviceCanAccessPeer is no longer self-ware (self does not qualify
as a peer)
- device peerlist always includes self, so when we call allow_access
we never remove self access.
- hipDeviceReset() removes old peer mappings.
2016-04-11 07:58:59 -05:00
7a11a2178d
Use HCC get_is_pool, add USE_PEER_TO_PEER level 2, 3.
2016-04-11 07:58:58 -05:00
813b063888
For P2P, use the peer list when allocating Device memory or pinned host.
...
Each new allocation is automatically mapped into the address space of
all enabled peers.
2016-04-11 07:58:58 -05:00
f2aa470f7f
P2P checkpoint.
...
Maintain enabled peer tables for each device.
2016-04-11 07:58:58 -05:00
69f2469cbb
Checkpoint initial peer2peer implementation.
2016-04-11 07:58:58 -05:00
15b4b1f81f
Use HCC get_is_pool, add USE_PEER_TO_PEER level 2, 3.
2016-04-07 14:28:42 -05:00
41860cb90d
Merge branch 'privatestaging' of https://github.com/AMDComputeLibraries/HIP-privatestaging into privatestaging
2016-04-07 04:57:54 -05:00
91d1974204
C proofing hip
2016-04-07 04:57:31 -05:00
41f7317fb5
For P2P, use the peer list when allocating Device memory or pinned host.
...
Each new allocation is automatically mapped into the address space of
all enabled peers.
2016-04-06 16:44:31 -05:00
36926e6233
P2P checkpoint.
...
Maintain enabled peer tables for each device.
2016-04-06 15:50:47 -05:00
b02e9163ab
Checkpoint initial peer2peer implementation.
2016-04-06 15:50:47 -05:00
1dcdfc93f1
Added feature for cross-compilation and seperated hcc api from hip api
2016-04-06 14:18:51 -05:00
26985b746b
added device math functions
2016-03-31 12:29:19 -05:00
72c72d8748
moved cpp specific device code from headers to source
2016-03-31 04:45:00 -05:00
496da0aabd
C guarding the apis
2016-03-30 06:08:50 -05:00
da1d9caf6b
Moved flag to hip_runtime.h
2016-03-30 03:32:39 -05:00
39af5d9caf
Merge branch 'privatestaging' of https://github.com/AMDComputeLibraries/HIP-privatestaging into privatestaging
2016-03-29 11:18:09 -05:00
f8a4483a55
Inline some new string functions.
2016-03-29 17:28:27 -05:00
740b730cac
Add runtime switch to control HIP_ATP_MARKER
...
Only generate the function strings if requested at
compile-time && runtime.
2016-03-29 17:27:30 -05:00
fb7862dba3
Merge branch 'privatestaging' of https://github.com/AMDComputeLibraries/HIP-privatestaging into privatestaging
2016-03-29 14:43:35 -05:00
61cd2bb399
Calling intrinsics from global kernel
2016-03-29 11:17:55 -05:00
756c0edf9b
Merge branch 'privatestaging' of https://github.com/AMDComputeLibraries/HIP-privatestaging into privatestaging
...
Conflicts:
include/hcc_detail/trace_helper.h
2016-03-29 05:56:03 -05:00
4584fb4453
Merge branch 'privatestaging' of https://github.com/AMDComputeLibraries/HIP-privatestaging into privatestaging
2016-03-29 02:28:20 -05:00
f01810183b
Not using classes inside headers
2016-03-29 02:27:51 -05:00
8635863724
Tweak thread-safe implementation.
...
introduce LockedAccessor option so destructor does not unlock.
Allows locks to exist across function boundaries, required
for hipLaunchKernel macro which has several unusual requirements.
(including C comppatibility, must use variadic macro, more).
2016-03-28 21:41:47 -05:00
00ae9531a7
turn off ONE_OBJECT_FILE since we are building a static library
2016-03-28 12:56:15 -05:00
3f18bab2c7
Stream thread-safe checkpoint.
...
Moving data structures to critical / protected section.
2016-03-28 09:46:40 -05:00
Ben Sander
Aditya Atluri
streamhsa
Maneesh Gupta
pensun
Jack Chung
scchan