.TH "Peer Context Memory Access" 3 "24 Apr 2019" "Version 6.0" "Doxygen" \" -*- nroff -*-
.ad l
.nh
.SH NAME
Peer Context Memory Access \- 
.SS "Functions"

.in +1c
.ti -1c
.RI "\fBCUresult\fP \fBcuCtxDisablePeerAccess\fP (\fBCUcontext\fP peerContext)"
.br
.RI "\fIDisables direct access to memory allocations in a peer context and unregisters any registered allocations. \fP"
.ti -1c
.RI "\fBCUresult\fP \fBcuCtxEnablePeerAccess\fP (\fBCUcontext\fP peerContext, unsigned int Flags)"
.br
.RI "\fIEnables direct access to memory allocations in a peer context. \fP"
.ti -1c
.RI "\fBCUresult\fP \fBcuDeviceCanAccessPeer\fP (int *canAccessPeer, \fBCUdevice\fP dev, \fBCUdevice\fP peerDev)"
.br
.RI "\fIQueries if a device may directly access a peer device's memory. \fP"
.ti -1c
.RI "\fBCUresult\fP \fBcuDeviceGetP2PAttribute\fP (int *value, \fBCUdevice_P2PAttribute\fP attrib, \fBCUdevice\fP srcDevice, \fBCUdevice\fP dstDevice)"
.br
.RI "\fIQueries attributes of the link between two devices. \fP"
.in -1c
.SH "Detailed Description"
.PP 
\\brief direct peer context memory access functions of the low-level CUDA driver API (\fBcuda.h\fP)
.PP
This section describes the direct peer context memory access functions of the low-level CUDA driver application programming interface. 
.SH "Function Documentation"
.PP 
.SS "\fBCUresult\fP cuCtxDisablePeerAccess (\fBCUcontext\fP peerContext)"
.PP
Returns \fBCUDA_ERROR_PEER_ACCESS_NOT_ENABLED\fP if direct peer access has not yet been enabled from \fCpeerContext\fP to the current context.
.PP
Returns \fBCUDA_ERROR_INVALID_CONTEXT\fP if there is no current context, or if \fCpeerContext\fP is not a valid context.
.PP
\fBParameters:\fP
.RS 4
\fIpeerContext\fP - Peer context to disable direct access to
.RE
.PP
\fBReturns:\fP
.RS 4
\fBCUDA_SUCCESS\fP, \fBCUDA_ERROR_DEINITIALIZED\fP, \fBCUDA_ERROR_NOT_INITIALIZED\fP, \fBCUDA_ERROR_PEER_ACCESS_NOT_ENABLED\fP, \fBCUDA_ERROR_INVALID_CONTEXT\fP, 
.RE
.PP
\fBNote:\fP
.RS 4
Note that this function may also return error codes from previous, asynchronous launches.
.RE
.PP
\fBSee also:\fP
.RS 4
\fBcuDeviceCanAccessPeer\fP, \fBcuCtxEnablePeerAccess\fP, cudaDeviceDisablePeerAccess 
.RE
.PP

.SS "\fBCUresult\fP cuCtxEnablePeerAccess (\fBCUcontext\fP peerContext, unsigned int Flags)"
.PP
If both the current context and \fCpeerContext\fP are on devices which support unified addressing (as may be queried using \fBCU_DEVICE_ATTRIBUTE_UNIFIED_ADDRESSING\fP) and same major compute capability, then on success all allocations from \fCpeerContext\fP will immediately be accessible by the current context. See \fBUnified Addressing\fP for additional details.
.PP
Note that access granted by this call is unidirectional and that in order to access memory from the current context in \fCpeerContext\fP, a separate symmetric call to \fBcuCtxEnablePeerAccess()\fP is required.
.PP
There is a system-wide maximum of eight peer connections per device.
.PP
Returns \fBCUDA_ERROR_PEER_ACCESS_UNSUPPORTED\fP if \fBcuDeviceCanAccessPeer()\fP indicates that the \fBCUdevice\fP of the current context cannot directly access memory from the \fBCUdevice\fP of \fCpeerContext\fP.
.PP
Returns \fBCUDA_ERROR_PEER_ACCESS_ALREADY_ENABLED\fP if direct access of \fCpeerContext\fP from the current context has already been enabled.
.PP
Returns \fBCUDA_ERROR_TOO_MANY_PEERS\fP if direct peer access is not possible because hardware resources required for peer access have been exhausted.
.PP
Returns \fBCUDA_ERROR_INVALID_CONTEXT\fP if there is no current context, \fCpeerContext\fP is not a valid context, or if the current context is \fCpeerContext\fP.
.PP
Returns \fBCUDA_ERROR_INVALID_VALUE\fP if \fCFlags\fP is not 0.
.PP
\fBParameters:\fP
.RS 4
\fIpeerContext\fP - Peer context to enable direct access to from the current context 
.br
\fIFlags\fP - Reserved for future use and must be set to 0
.RE
.PP
\fBReturns:\fP
.RS 4
\fBCUDA_SUCCESS\fP, \fBCUDA_ERROR_DEINITIALIZED\fP, \fBCUDA_ERROR_NOT_INITIALIZED\fP, \fBCUDA_ERROR_PEER_ACCESS_ALREADY_ENABLED\fP, \fBCUDA_ERROR_TOO_MANY_PEERS\fP, \fBCUDA_ERROR_INVALID_CONTEXT\fP, \fBCUDA_ERROR_PEER_ACCESS_UNSUPPORTED\fP, \fBCUDA_ERROR_INVALID_VALUE\fP 
.RE
.PP
\fBNote:\fP
.RS 4
Note that this function may also return error codes from previous, asynchronous launches.
.RE
.PP
\fBSee also:\fP
.RS 4
\fBcuDeviceCanAccessPeer\fP, \fBcuCtxDisablePeerAccess\fP, cudaDeviceEnablePeerAccess 
.RE
.PP

.SS "\fBCUresult\fP cuDeviceCanAccessPeer (int * canAccessPeer, \fBCUdevice\fP dev, \fBCUdevice\fP peerDev)"
.PP
Returns in \fC*canAccessPeer\fP a value of 1 if contexts on \fCdev\fP are capable of directly accessing memory from contexts on \fCpeerDev\fP and 0 otherwise. If direct access of \fCpeerDev\fP from \fCdev\fP is possible, then access may be enabled on two specific contexts by calling \fBcuCtxEnablePeerAccess()\fP.
.PP
\fBParameters:\fP
.RS 4
\fIcanAccessPeer\fP - Returned access capability 
.br
\fIdev\fP - Device from which allocations on \fCpeerDev\fP are to be directly accessed. 
.br
\fIpeerDev\fP - Device on which the allocations to be directly accessed by \fCdev\fP reside.
.RE
.PP
\fBReturns:\fP
.RS 4
\fBCUDA_SUCCESS\fP, \fBCUDA_ERROR_DEINITIALIZED\fP, \fBCUDA_ERROR_NOT_INITIALIZED\fP, \fBCUDA_ERROR_INVALID_DEVICE\fP 
.RE
.PP
\fBNote:\fP
.RS 4
Note that this function may also return error codes from previous, asynchronous launches.
.RE
.PP
\fBSee also:\fP
.RS 4
\fBcuCtxEnablePeerAccess\fP, \fBcuCtxDisablePeerAccess\fP, cudaDeviceCanAccessPeer 
.RE
.PP

.SS "\fBCUresult\fP cuDeviceGetP2PAttribute (int * value, \fBCUdevice_P2PAttribute\fP attrib, \fBCUdevice\fP srcDevice, \fBCUdevice\fP dstDevice)"
.PP
Returns in \fC*value\fP the value of the requested attribute \fCattrib\fP of the link between \fCsrcDevice\fP and \fCdstDevice\fP. The supported attributes are:
.IP "\(bu" 2
\fBCU_DEVICE_P2P_ATTRIBUTE_PERFORMANCE_RANK\fP: A relative value indicating the performance of the link between two devices.
.IP "\(bu" 2
\fBCU_DEVICE_P2P_ATTRIBUTE_ACCESS_SUPPORTED\fP P2P: 1 if P2P Access is enable.
.IP "\(bu" 2
\fBCU_DEVICE_P2P_ATTRIBUTE_NATIVE_ATOMIC_SUPPORTED\fP: 1 if Atomic operations over the link are supported.
.IP "\(bu" 2
\fBCU_DEVICE_P2P_ATTRIBUTE_CUDA_ARRAY_ACCESS_SUPPORTED\fP: 1 if cudaArray can be accessed over the link.
.PP
.PP
Returns \fBCUDA_ERROR_INVALID_DEVICE\fP if \fCsrcDevice\fP or \fCdstDevice\fP are not valid or if they represent the same device.
.PP
Returns \fBCUDA_ERROR_INVALID_VALUE\fP if \fCattrib\fP is not valid or if \fCvalue\fP is a null pointer.
.PP
\fBParameters:\fP
.RS 4
\fIvalue\fP - Returned value of the requested attribute 
.br
\fIattrib\fP - The requested attribute of the link between \fCsrcDevice\fP and \fCdstDevice\fP. 
.br
\fIsrcDevice\fP - The source device of the target link. 
.br
\fIdstDevice\fP - The destination device of the target link.
.RE
.PP
\fBReturns:\fP
.RS 4
\fBCUDA_SUCCESS\fP, \fBCUDA_ERROR_DEINITIALIZED\fP, \fBCUDA_ERROR_NOT_INITIALIZED\fP, \fBCUDA_ERROR_INVALID_DEVICE\fP, \fBCUDA_ERROR_INVALID_VALUE\fP 
.RE
.PP
\fBNote:\fP
.RS 4
Note that this function may also return error codes from previous, asynchronous launches.
.RE
.PP
\fBSee also:\fP
.RS 4
\fBcuCtxEnablePeerAccess\fP, \fBcuCtxDisablePeerAccess\fP, \fBcuDeviceCanAccessPeer\fP, cudaDeviceGetP2PAttribute 
.RE
.PP

.SH "Author"
.PP 
Generated automatically by Doxygen from the source code.