commit 9548493fa754b3bf5c31cbdc2211db1e73e8c07c
Author: Jack Chung <whchung@gmail.com>
Date: Mon May 23 11:57:23 2016 +0800
Rename hipExternShared test to hipDynamicShared
Change-Id: I180d9d539420fb69cfc121eceaa7db9da03483b2
commit 827081f8244a38f010789d556db0c4ff7b6422d8
Author: Jack Chung <whchung@gmail.com>
Date: Mon May 23 11:56:27 2016 +0800
Rename HIP_DECLARE_EXTERN_SHARED to HIP_DYNAMIC_SHARED
Change-Id: I22362d179812ac547e0f11ba4e2bb999050e08ae
commit 4c277228ed41af187739610fa17eab1fb144c947
Author: Jack Chung <whchung@gmail.com>
Date: Thu May 19 17:49:52 2016 +0800
Adopt new interface to get dynamic LDS in hc.hpp
Change-Id: I47b433b714633a4c97df87c40a0b1d3386429a00
commit 5a36117d777064113a528dc47b42e8c8413baa97
Author: Jack Chung <whchung@gmail.com>
Date: Thu May 19 11:29:24 2016 +0800
Add test patterns for regular expression to match "extern __shared__"
These test patterns should better be saved as an individual test case, but I'm
not familiar with HIP test structures so I leave them as comments in hipify as
of now.
Change-Id: I7fee89c89b9e73de2133357a226ec0c769733531
commit 1b26284168c7f5339f63338fd0149bed5d994656
Author: Jack Chung <whchung@gmail.com>
Date: Thu May 19 11:25:23 2016 +0800
Add one HIP unit test to use HIP_DECLARE_EXTERN_SHARED
Change-Id: I4d9907815920693a74ea9d575fe26e7c67636109
commit 77b816ee5972b13d829d5bbcf06fbfd07acea2af
Author: Jack Chung <whchung@gmail.com>
Date: Wed May 18 19:18:59 2016 +0800
Adopt HIP_ prefix for DECLARE_EXTERN_SHARED macro
Change-Id: I555ded16b449b67d2e20904013d86fe1ded6a2be
commit ef0997939c3578a9ae11621bf21c0416f04d2622
Author: Jack Chung <whchung@gmail.com>
Date: Wed May 18 17:42:04 2016 +0800
Modify hipify to support converting extern __shared__ to DECLARE_EXTERN_SHARED macro
Added regular expression to search & replace extern __shared__ declarations to
DECLARE_EXTERN_SHARED macro.
Limitation:
- Won't work if "extern __shared__" is declared at global scope
Sample Usages:
extern __shared__ double foo[];
extern __shared__ unsigned int foo[];
extern volatile __shared__ double foo[];
extern volatile __shared__ unsigned int sdata[];
extern __shared__ volatile unsigned int sdata[];
extern __shared__ T s[];
extern __shared__ T::type s[];
extern __shared__ blah<T>::type s[];
extern __shared__ typename mapper<Float>::type s_data[];
extern __attribute__((used)) __shared__ typename mapper<Float>::type s_data[];
Change-Id: I2be0b7039adeddb789f5a2b067d403a43fdc3e26
commit 93ff268724493aedfacdcd5a5aa9a100f4ebaed0
Author: Jack Chung <whchung@gmail.com>
Date: Wed May 18 15:13:09 2016 +0800
Introduce DECLARE_EXTERN_SHARED macro to encapsulate "extern __shared__" decls
Change-Id: I93b2d37c763195b0ca9fd0afee78605a1e3272db
commit cff9c95412de343cc6405158b5acc4f1029267ff
Author: Jack Chung <whchung@gmail.com>
Date: Wed May 18 12:53:54 2016 +0800
Add __get_dynamic_groupbaseptr() to point to dynamic LDS
Change-Id: I97b548d8a691488057617c551a8f331cad7afc77
Change-Id: I84e7875b76fa1f59e860e19c93bd4209cdd1fd2c
- 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.
- 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.
- 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.
- Ensure proper device is set before each command.
- Add command line switches:
+bool p_memcpyWithPeer = false; // use the peer device for the P2P copy
+bool p_mirrorPeers = false; // in addition to mapping current to peer
space, map peer to current space.
+int p_peerDevice = -1; // explicly specify which peer to use, else
use p_gpuDevice + 1.
Jack Chung
Maneesh Gupta
Rahul Garg
Ben Sander
Aditya Atluri
streamhsa
fdaniil