Rework core for NVIDIA Trusted Computing
* Compress work structs so that they are shared between channels
* Utilize the full amount of kernel argument space permitted (4k)
before resorting to work fifo.
* Rework the task preprocessing phase.
* Use a separate abortDevFlag which is kept in sync with abortFlag
using cudaMemcpy operations.
* Rename src/include/align.h to src/include/bitops.h
Add lazy connection establishment for collective operations
* Move buffer allocation and connection establishment to the first
collective operation using that algorithm.
* Accelerate init time and reduce memory usage.
* Avoid allocating NVLS buffers if all calls are registered.
* Compute algo/proto in ncclLaunchCollTasksInfo early on.
* Connect peers in ncclCollPreconnectFunc if not connected already.
* Also move shared buffer creation to the first send/recv call.
Accelerate intra-node NVLink detection
* Make each rank only detect NVLinks attached to its GPU.
* Fuse XMLs to reconstruct the full NVLink topology
Add init profiling to report time spend in different init phases.
* Report timings of bootstrap, allgather, search, connect, etc.
* Add new "PROFILE" category for NCCL_DEBUG_SUBSYS.
Add support for PCI p2p on split PCI switches
* Detect split PCI switches through a kernel module exposing
switch information.
* Update the topology XML and graph to add those inter-switch
connections.
Add cost estimation API
* Add a new ncclGroupEndSimulate primitive to return the estimated
time a group would take.
Net/IB: Add separate traffic class for fifo messages
* Add NCCL_IB_FIFO_TC to control the traffic class of fifo messages
independently from NCCL_IB_TC.
Merges PR #1194
Net/IB: Add support for IB router
* Use flid instead of lid if subnets do not match
* Warn if flid is 0
Optimizations and fixes for device network offload (unpack)
* Double the default number of channels
* Cache netDeviceType
* Fix save/increment head logic to enable Tree support.
Support ncclGroupStart/End for ncclCommAbort/Destroy
* Allow Abort/Destroy to be called within a group when managing
multiple GPUs with a single process.
Improve Tuner API
* Provide to the plugin the original cost table so that the plugin
can leave unknown or disabled algo/proto combinations untouched.
* Remove nvlsSupport and collnetSupport.
Do not print version to stdout when using a debug file
* Also print version from all processes with INFO debug level.
Fixes issue #1271
Fix clang warnings in NVTX headers
* Update NVTX headers to the latest version
Fixes issue #1270
Disable port fusion in heterogeneous systems
* Do not fuse ports if a mix of multi-port and single port are detected.
Fix NVLS graphs search for dual NICs.
* Fix NVLS graph search when we have more than one NIC per GPU.
Fix crash with collnetDirect
* Add separate graph search for collnetDirect, testing alltoall paths
and working similarly to the NVLS search.
Fix hang when nodes have different CPU types
* Add the CPU type to the rank peer info.
* Align all ranks on the CPU type after the first allgather.
* Only use the aligned CPU type for all tuning operations.
Fixes issue #1136
Fixes issue #1184
Fix performance of registered send/recv operations
* Allow for single full size operations
* Add INFO to confirm the registration of send/recv buffers.
Move all sync ops to finalize stage
* Ensure ncclCommDestroy is non-blocking if ncclCommFinalize has
been called.
Improve error reporting during SHM segment creation
Improve support of various compilers
Merges PR #1177
Merges PR #1228
Allow net and tuner plugins to be statically linked
* Search for ncclNet or ncclTuner symbols in the main binary.
Merges PR #979
Plugin examples includes cleanup
* Harmonize err.h and common.h usage.
* Add mixed plugin with both net and tuner.
[ROCm/rccl commit: 178b6b7590]
Rework core for NVIDIA Trusted Computing
* Compress work structs so that they are shared between channels
* Utilize the full amount of kernel argument space permitted (4k)
before resorting to work fifo.
* Rework the task preprocessing phase.
* Use a separate abortDevFlag which is kept in sync with abortFlag
using cudaMemcpy operations.
* Rename src/include/align.h to src/include/bitops.h
Add lazy connection establishment for collective operations
* Move buffer allocation and connection establishment to the first
collective operation using that algorithm.
* Accelerate init time and reduce memory usage.
* Avoid allocating NVLS buffers if all calls are registered.
* Compute algo/proto in ncclLaunchCollTasksInfo early on.
* Connect peers in ncclCollPreconnectFunc if not connected already.
* Also move shared buffer creation to the first send/recv call.
Accelerate intra-node NVLink detection
* Make each rank only detect NVLinks attached to its GPU.
* Fuse XMLs to reconstruct the full NVLink topology
Add init profiling to report time spend in different init phases.
* Report timings of bootstrap, allgather, search, connect, etc.
* Add new "PROFILE" category for NCCL_DEBUG_SUBSYS.
Add support for PCI p2p on split PCI switches
* Detect split PCI switches through a kernel module exposing
switch information.
* Update the topology XML and graph to add those inter-switch
connections.
Add cost estimation API
* Add a new ncclGroupEndSimulate primitive to return the estimated
time a group would take.
Net/IB: Add separate traffic class for fifo messages
* Add NCCL_IB_FIFO_TC to control the traffic class of fifo messages
independently from NCCL_IB_TC.
Merges PR #1194
Net/IB: Add support for IB router
* Use flid instead of lid if subnets do not match
* Warn if flid is 0
Optimizations and fixes for device network offload (unpack)
* Double the default number of channels
* Cache netDeviceType
* Fix save/increment head logic to enable Tree support.
Support ncclGroupStart/End for ncclCommAbort/Destroy
* Allow Abort/Destroy to be called within a group when managing
multiple GPUs with a single process.
Improve Tuner API
* Provide to the plugin the original cost table so that the plugin
can leave unknown or disabled algo/proto combinations untouched.
* Remove nvlsSupport and collnetSupport.
Do not print version to stdout when using a debug file
* Also print version from all processes with INFO debug level.
Fixes issue #1271
Fix clang warnings in NVTX headers
* Update NVTX headers to the latest version
Fixes issue #1270
Disable port fusion in heterogeneous systems
* Do not fuse ports if a mix of multi-port and single port are detected.
Fix NVLS graphs search for dual NICs.
* Fix NVLS graph search when we have more than one NIC per GPU.
Fix crash with collnetDirect
* Add separate graph search for collnetDirect, testing alltoall paths
and working similarly to the NVLS search.
Fix hang when nodes have different CPU types
* Add the CPU type to the rank peer info.
* Align all ranks on the CPU type after the first allgather.
* Only use the aligned CPU type for all tuning operations.
Fixes issue #1136
Fixes issue #1184
Fix performance of registered send/recv operations
* Allow for single full size operations
* Add INFO to confirm the registration of send/recv buffers.
Move all sync ops to finalize stage
* Ensure ncclCommDestroy is non-blocking if ncclCommFinalize has
been called.
Improve error reporting during SHM segment creation
Improve support of various compilers
Merges PR #1177
Merges PR #1228
Allow net and tuner plugins to be statically linked
* Search for ncclNet or ncclTuner symbols in the main binary.
Merges PR #979
Plugin examples includes cleanup
* Harmonize err.h and common.h usage.
* Add mixed plugin with both net and tuner.
* Incremental Counter Profile Creation
Adds support for incremental counter creation. How this functions is the
behavior of rocprofiler_create_profile_config has been changed.
rocprofiler_create_profile_config(rocprofiler_agent_id_t agent_id,
rocprofiler_counter_id_t* counters_list,
size_t counters_count,
rocprofiler_profile_config_id_t* config_id)
The behavior of this function now allows an existing config_id to be
supplied via config_id. The counters contained in this config will be
copied over and used as a base for a new config along with any counters
supplied in counters_list. The new config id is returned via config_id
and can be used in future dispatch/agent counting sessions.
A new config is created over modifying an existing config since there
is no gaurentee that the existing config isn't already in use. While we
could add locks (or other mutual exclusion properties) to check if its
in use and reject an update, the benefit from doing so is minor in
comparison to just creating a new config. This also side steps a common
pattern a tool may use to add additional counters at some point later on
during execution. Now they can do that without destroying the existing
config.
---------
Co-authored-by: Benjamin Welton <ben@amd.com>
Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
[ROCm/rocprofiler-sdk commit: 81d1407565]
* Incremental Counter Profile Creation
Adds support for incremental counter creation. How this functions is the
behavior of rocprofiler_create_profile_config has been changed.
rocprofiler_create_profile_config(rocprofiler_agent_id_t agent_id,
rocprofiler_counter_id_t* counters_list,
size_t counters_count,
rocprofiler_profile_config_id_t* config_id)
The behavior of this function now allows an existing config_id to be
supplied via config_id. The counters contained in this config will be
copied over and used as a base for a new config along with any counters
supplied in counters_list. The new config id is returned via config_id
and can be used in future dispatch/agent counting sessions.
A new config is created over modifying an existing config since there
is no gaurentee that the existing config isn't already in use. While we
could add locks (or other mutual exclusion properties) to check if its
in use and reject an update, the benefit from doing so is minor in
comparison to just creating a new config. This also side steps a common
pattern a tool may use to add additional counters at some point later on
during execution. Now they can do that without destroying the existing
config.
---------
Co-authored-by: Benjamin Welton <ben@amd.com>
Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
Fix derived counters to have globals replicated across all architectures (that support them).
---------
Co-authored-by: Benjamin Welton <ben@amd.com>
[ROCm/rocprofiler-sdk commit: ab92bef1a4]
The provides tag is required when the package provides a virtual package.
Package name along with version will be provided by default and the provides tag is not required for this.
Using the tag for providing the name, but without version was resulting in package upgrade issues.
Change-Id: I74506d8c3bbd75d028bcdc03525c29541dce2b4c
[ROCm/rocm_smi_lib commit: d54bade574]
The provides tag is required when the package provides a virtual package.
Package name along with version will be provided by default and the provides tag is not required for this.
Using the tag for providing the name, but without version was resulting in package upgrade issues.
Change-Id: I74506d8c3bbd75d028bcdc03525c29541dce2b4c
1.Move global amd::monitor listenerLock before global
class runtime_tear_down as it will be referenced in
~RuntimeTearDown() after main(). It should be freed
later than runtime_tear_down.
2.Update Device::~Device() to SVM free coopHostcallBuffer_
before context_ is released and freed.
Change-Id: I1d21378ff463477d3238d71e5e2a1a7d6b9147ad
[ROCm/clr commit: 544c45364f]
1.Move global amd::monitor listenerLock before global
class runtime_tear_down as it will be referenced in
~RuntimeTearDown() after main(). It should be freed
later than runtime_tear_down.
2.Update Device::~Device() to SVM free coopHostcallBuffer_
before context_ is released and freed.
Change-Id: I1d21378ff463477d3238d71e5e2a1a7d6b9147ad
The provides tag is required when the package provides a virtual package.
Package name along with version will be provided by default and the provides tag is not required for this.
Using the tag for providing the name, but without version was resulting in package upgrade issues.
[ROCm/rocdecode commit: 7bf02a9606]
The provides tag is required when the package provides a virtual package.
Package name along with version will be provided by default and the provides tag is not required for this.
Using the tag for providing the name, but without version was resulting in package upgrade issues.
Ioannis Assiouras
Jatin Chaudhary
Branislav Brzak
gobhardw
Maisam Arif
Vladimir Indic
Jonathan R. Madsen
David Galiffi
Sylvain Jeaugey
Benjamin Welton
Istvan Kiss
Ranjith Ramakrishnan
Galantsev, Dmitrii
dependabot[bot]
taosang2
Rahul Manocha
raramakr
kjayapra-amd