AI/rocm-systems
2
0
Fork 0
Κώδικας Ζητήματα Pull Requests Δράσεις Πακέτα Έργα Κυκλοφορίες Wiki Δραστηριότητα

NCCL 2.28.3-1

Περιήγηση Στο Πηγαίο 
Device API (Experimental)
 * Introduces device-side APIs to integrate NCCL communication directly into application kernels.
 * Supports LSA (Load/Store Access) for CUDA P2P communication over NVLink and some PCIe platforms.
 * Supports Multimem for hardware multicast using NVLink SHARP.
 * Adds initial framework for GIN (GPU-Initiated Networking), currently under development.
 * Introduces device communicators created using ncclDevCommCreate.
 * Enables device-side communication operations with synchronization (ncclLsaBarrierSession) and memory accessors (ncclGetLsaPointer, ncclGetLsaMultimemPointer).
 * Experimental APIs - signatures and functionality may evolve in future releases.
 * No ABI compatibility is guaranteed — applications must be recompiled with each new NCCL release.

Symmetric memory improvements
 * Support for aggregating symmetric operations using ncclGroupStart/End APIs.
 * Reimplement symmetric kernels using device API.

New Host APIs
 * Introduce new host collective APIs: ncclAlltoAll, ncclScatter, ncclGather.

CE (Copy Engine) Collectives
 * Reduce SM utilization for alltoall, scatter, gather, and allgather within a single (MN)NVL domain.
 * Free up SM capacity for the application to do computation at the same time.
 * To enable the feature for ncclAllGather, ncclAlltoAll, ncclGather, ncclScatter, register buffers into symmetric windows and use the NCCL_CTA_POLICY_ZERO flag in the communicator config_t.

NCCL Inspector Plugin
 * Introduces an Inspector plugin for always-on performance monitoring.
 * Produces structured JSON output with metadata, execution time, bandwidth, and optional event traces for each NCCL operation.
 * Enables integration with analysis tools such as Performance Exporter to visualize NCCL performance bottlenecks.
 * Lightweight to enable via environment variables NCCL_PROFILER_PLUGIN and NCCL_INSPECTOR_ENABLE.

CMake support (Experiemental)
 * Adds a CMake build system as an alternative to existing Makefiles.
 * Known issues: pkg.build and Device API currently do not work with CMake.
 * The known issues will be addressed in a future release.

Decreased max CTA count from 32 to 16 on Blackwell
 * SM overhead is decreased by 50% with this improvement.
 * This may cause some perf drop on Blackwell because of the reduced SM usage.
 * If the extra SM capacity is not desired, two options are available to restore to previous behavior: 1) Setting NCCL_MIN_CTAS=32 NCCL_MAX_CTAS=32 environment variables; 2) setting communicator config to over-write max CTA count to 32.
 * Based on community feedback, future versions may consider different trade-offs between performance and SM overhead.

Plugins
 * Network
   * App-aware Network plugin. NCCL passes information about communication operations to be executed on the network end point. This allows for better tuning of network end points and their use in the plugins.
   * Improve handling of physical and virtual network devices and load/unload.
   * Network plugin version 11 - add explicit context and communication ID support for per communicator init/finalize.
   * Add Multi-Request Net API. Using this will help NCCL to anticipate multiple send/recv requests and optimize for it. See maxMultiRequestSize field in ncclNetProperties_v11_t.
 * Profiler
   * Add support for API events (group, collective, and p2p) and for tracking kernel launches in the profiler plugin.
   * Add Inspector Profiler Plugin (see section above).
   * Add a hook to Google’s CoMMA profiler on github.
 * Tuner
   * Expose NCCL tuning constants at tuner initialization via ncclTunerConstants_v5_t.
   * Add NVL Domain Information API.
 * Support multiple plugin types from a single shared object.

New Parameterization and ncclConfig changes:
 * Add new option NCCL_MNNVL_CLIQUE_ID=-2 which will use rack serial number to partition the MNNVL clique. This will limit NVLink domains to GPUs within a single rack.
 * Add NCCL_NETDEVS_POLICY to control how NET devices are assigned to GPUs. The default (AUTO) is the policy used in previous versions.
 * Add NCCL_SINGLE_PROC_MEM_REG_ENABLE control variable to enable NVLS UB registration in the “one process, multiple ranks” case as opt in.
 * Move nChannelsPerNetPeer into ncclConfig. NCCL_NCHANNELS_PER_NET_PEER can override the value in ncclConfig.
 * Enable PxN over C2C by default
   * PxN over C2C will improve performance for Grace-Blackwell platforms by allowing NCCL to leverage the NIC attached to a peer GPU over NVLINK, C2C, and PCIe.
   * This behavior can be overridden by setting NCCL_PXN_C2C=0.

Other Improvements:
 * Allow FP8 support for non-reductive operations on pre sm90 devices. (See https://github.com/pytorch/pytorch/pull/151594#discussion_r2135777776)
 * Fix NVLS+CollNet and temporarily disables COLLNET_CHAIN for >8 GPUs.
 * Only consider running interfaces for socket traffic. NCCL will not attempt to use interfaces that do not have the IFF_RUNNING bit. (https://github.com/NVIDIA/nccl/issues/1798)
 * Modernize mutex management. Convert to std::mutex and std::lock_guard.
 * Remove sm35 and sm50 GENCODE targets which have long been deprecated and were causing issues with the latest NCCL release builds.
 * Improved NVLS/NVLSTree tuning prediction to improve algorithm and protocol selection.
 * NVLSTree Tuning Fixes. Update tuning data for H100, GB200-NV72.
 * Respond better to RoCE link flaps. Instead of reporting an “unknown event” it will now report “GID table changed”.
 * Move libvirt bridge interface to the end of possible interfaces so that they are considered last. These interfaces are usually virtual bridges to relay traffic to containers running on the host and cannot be used for traffic to a remote node and are therefore unsuitable.
Αυτή η υποβολή περιλαμβάνεται σε:
Mark Santesson
2025-09-02 13:21:14 -07:00
γονέας 593de54e52
υποβολή f1308997d0
212 αρχεία άλλαξαν με 15535 προσθήκες και 2938 διαγραφές
Λήψη Αρχείου Patch Λήψη Αρχείου Diff Ανάπτυξη όλων των αρχείων Σύμπτυξη όλων των αρχείων
ext-profiler/example/README.md
+59 -121
Προβολή Αρχείου
@@ -13,8 +13,7 @@ change the size of the event window the profiler keeps track of.
## Building the profiler plugin
To use the example plugin, just type `make`. You will need a NCCL build's include directory present.
You can override `NCCL_HOME` to where the NCCL installation is on your system.
To build the example plugin shipped as part of NCCL, just type `make`.
## Using the profiler plugin
@@ -27,13 +26,13 @@ You can override `NCCL_HOME` to where the NCCL installation is on your system.
As an example, setting:
`NCCL_PROFILE_EVENT_MASK` to 1 (`ncclProfileGroup`) | 2 (`ncclProfileColl`) | 8 (`ncclProfileProxyOp`)
`NCCL_PROFILE_EVENT_MASK` to 256 (`ncclProfileGroupApi`) | 2 (`ncclProfileColl`) | 8 (`ncclProfileProxyOp`)
enables the profiling of the group, the collective and the proxy op events. The same events can be
enables the profiling of the group API, the collective and the proxy op events. The same events can be
expressed more concisely by setting `NCCL_PROFILE_EVENT_MASK` to 8 (`ncclProfileProxyOp`). Indeed,
in NCCL all the events above (in the event hierarchy) the one requested are also captured. The advantage
is that the profiler can easily correlate events that belong to the same NCCL operation and present
them accordingly.
them accordingly. Setting `NCCL_PROFILE_EVENT_MASK` to 4095 enables all events supported by the v5 profiler.
3. Set `NCCL_PROFILE_DUMP_FILE` to the name of the dump file for the collected traces. A file named
${NCCL_PROFILE_DUMP_FILE}-hostname-tid.txt is created. Profiler traces are saved using the chrome
@@ -57,11 +56,14 @@ The group, collective and p2p pools contain objects for the corresponding events
contains objects for `ProxyCtrl` events and the `ProxyDetach` pool contains objects for `ProxyOp` events
generated by remote proxies. A list of pools and their size is reported below:
- `NCCL_PROFILE_GROUP_POOL_SIZE` (16)
- `NCCL_PROFILE_COLL_POOL_SIZE` (16)
- `NCCL_PROFILE_P2P_POOL_SIZE` (1024)
- `NCCL_PROFILE_GROUP_API_POOL_SIZE` (256)
- `NCCL_PROFILE_COLL_API_POOL_SIZE` (256)
- `NCCL_PROFILE_P2P_API_POOL_SIZE` (256)
- `NCCL_PROFILE_KERNEL_LAUNCH_POOL_SIZE` (256)
- `NCCL_PROFILE_COLL_POOL_SIZE` (256)
- `NCCL_PROFILE_P2P_POOL_SIZE` (256)
- `NCCL_PROFILE_PROXY_CTRL_POOL_SIZE` (16)
- `NCCL_PROFILE_PROXY_DETACH_POOL_SIZE` (128)
- `NCCL_PROFILE_PROXY_DETACH_POOL_SIZE` (256)
Remote proxy operations are generated when PXN is in use. Refer to this article for more information
about PXN and how it works:
@@ -73,76 +75,58 @@ The example profiler generates traces using the json format. An example of trace
```
[
{"name": "Group", "cat": "GROUP", "ph": "b", "id": 0, "pid": 4157654, "tid": 1, "ts": 764234.611328, "args": {"groupId": 0}},
{"name": "AllReduce", "cat": "COLL", "ph": "b", "id": 0, "pid": 4157654, "tid": 1, "ts": 764237.294922, "args": {"SeqNum": 0, "CommHash": 673864846479792718, "Rank": 1, "Count": 32768, "Datatype": "ncclFloat32", "Algorithm": "RING", "Protocol": "LL", "nMaxChannels": 2}},
{"name": "Recv", "cat": "PROXY", "ph": "b", "id": 0, "pid": 4157654, "tid": 1, "ts": 768464.936523, "args": {"Channel": 0, "Peer": 0, "Steps": 14, "ChunkSize": 32768, "transSize": 229376, "POSTED": {"step": 14, "ts": 772020.300781}, "RECEIVED": {"step": 14, "ts": 772196.049805}, "TRANSMITTED": {"step": 14, "ts": 772197.326172}, "DONE": {"step": 14, "ts": 772201.538086}}},
{"name": "RecvBufferWait", "cat": "NET", "ph": "b", "id": 0, "pid": 4157654, "tid": 1, "ts": 768465.158203, "args": {"Step": 0}},
{"name": "RecvBufferWait", "cat": "NET", "ph": "e", "id": 0, "pid": 4157654, "tid": 1, "ts": 768477.924805},
{"name": "RecvWait", "cat": "NET", "ph": "b", "id": 0, "pid": 4157654, "tid": 1, "ts": 768477.924805, "args": {"Step": 0}},
{"name": "RecvWait", "cat": "NET", "ph": "e", "id": 0, "pid": 4157654, "tid": 1, "ts": 768547.197266},
{"name": "RecvFlushWait", "cat": "NET", "ph": "b", "id": 0, "pid": 4157654, "tid": 1, "ts": 768547.197266, "args": {"Step": 0}},
{"name": "RecvFlushWait", "cat": "NET", "ph": "e", "id": 0, "pid": 4157654, "tid": 1, "ts": 768564.174805},
{"name": "RecvGpuWait", "cat": "NET", "ph": "b", "id": 0, "pid": 4157654, "tid": 1, "ts": 768564.174805, "args": {"Step": 0}},
{"name": "RecvGpuWait", "cat": "NET", "ph": "e", "id": 0, "pid": 4157654, "tid": 1, "ts": 768568.276367},
{"name": "RecvBufferWait", "cat": "NET", "ph": "b", "id": 1, "pid": 4157654, "tid": 1, "ts": 768503.604492, "args": {"Step": 1}},
{"name": "RecvBufferWait", "cat": "NET", "ph": "e", "id": 1, "pid": 4157654, "tid": 1, "ts": 768504.549805},
{"name": "RecvWait", "cat": "NET", "ph": "b", "id": 1, "pid": 4157654, "tid": 1, "ts": 768504.549805, "args": {"Step": 1}},
{"name": "RecvWait", "cat": "NET", "ph": "e", "id": 1, "pid": 4157654, "tid": 1, "ts": 769994.490234},
{"name": "RecvFlushWait", "cat": "NET", "ph": "b", "id": 1, "pid": 4157654, "tid": 1, "ts": 769994.490234, "args": {"Step": 1}},
{"name": "RecvFlushWait", "cat": "NET", "ph": "e", "id": 1, "pid": 4157654, "tid": 1, "ts": 769995.012695},
{"name": "RecvGpuWait", "cat": "NET", "ph": "b", "id": 1, "pid": 4157654, "tid": 1, "ts": 769995.012695, "args": {"Step": 1}},
{"name": "RecvGpuWait", "cat": "NET", "ph": "e", "id": 1, "pid": 4157654, "tid": 1, "ts": 770006.914062},
{"name": "RecvBufferWait", "cat": "NET", "ph": "b", "id": 2, "pid": 4157654, "tid": 1, "ts": 768506.941406, "args": {"Step": 2}},
{"name": "RecvBufferWait", "cat": "NET", "ph": "e", "id": 2, "pid": 4157654, "tid": 1, "ts": 768507.435547},
{"name": "RecvWait", "cat": "NET", "ph": "b", "id": 2, "pid": 4157654, "tid": 1, "ts": 768507.435547, "args": {"Step": 2}},
{"name": "RecvWait", "cat": "NET", "ph": "e", "id": 2, "pid": 4157654, "tid": 1, "ts": 771452.536133},
{"name": "RecvFlushWait", "cat": "NET", "ph": "b", "id": 2, "pid": 4157654, "tid": 1, "ts": 771452.536133, "args": {"Step": 2}},
{"name": "RecvFlushWait", "cat": "NET", "ph": "e", "id": 2, "pid": 4157654, "tid": 1, "ts": 771453.060547},
{"name": "RecvGpuWait", "cat": "NET", "ph": "b", "id": 2, "pid": 4157654, "tid": 1, "ts": 771453.060547, "args": {"Step": 2}},
{"name": "RecvGpuWait", "cat": "NET", "ph": "e", "id": 2, "pid": 4157654, "tid": 1, "ts": 771468.458008},
{"name": "RecvBufferWait", "cat": "NET", "ph": "b", "id": 3, "pid": 4157654, "tid": 1, "ts": 768509.484375, "args": {"Step": 3}},
{"name": "RecvBufferWait", "cat": "NET", "ph": "e", "id": 3, "pid": 4157654, "tid": 1, "ts": 768510.250000},
{"name": "RecvWait", "cat": "NET", "ph": "b", "id": 3, "pid": 4157654, "tid": 1, "ts": 768510.250000, "args": {"Step": 3}},
{"name": "RecvWait", "cat": "NET", "ph": "e", "id": 3, "pid": 4157654, "tid": 1, "ts": 771904.499023},
{"name": "RecvFlushWait", "cat": "NET", "ph": "b", "id": 3, "pid": 4157654, "tid": 1, "ts": 771904.499023, "args": {"Step": 3}},
{"name": "RecvFlushWait", "cat": "NET", "ph": "e", "id": 3, "pid": 4157654, "tid": 1, "ts": 771904.991211},
{"name": "RecvGpuWait", "cat": "NET", "ph": "b", "id": 3, "pid": 4157654, "tid": 1, "ts": 771904.991211, "args": {"Step": 3}},
{"name": "RecvGpuWait", "cat": "NET", "ph": "e", "id": 3, "pid": 4157654, "tid": 1, "ts": 771910.500000},
{"name": "Send", "cat": "PROXY", "ph": "b", "id": 1, "pid": 4157654, "tid": 1, "ts": 768482.878906, "args": {"Channel": 0, "Peer": 2, "Steps": 14, "ChunkSize": 32768, "transSize": 229376, "POSTED": {"step": 14, "ts": 771995.675781}, "REM_FIFO_WAIT": {"step": 14, "ts": 772190.692383}, "TRANSMITTED": {"step": 14, "ts": 772191.516602}, "DONE": {"step": 14, "ts": 772208.473633}}},
{"name": "SendBufferWait", "cat": "NET", "ph": "b", "id": 14, "pid": 4157654, "tid": 1, "ts": 768483.019531, "args": {"Step": 0}},
{"name": "SendBufferWait", "cat": "NET", "ph": "e", "id": 14, "pid": 4157654, "tid": 1, "ts": 768483.300781},
{"name": "SendGpuWait", "cat": "NET", "ph": "b", "id": 14, "pid": 4157654, "tid": 1, "ts": 768483.300781, "args": {"Step": 0}},
{"name": "SendGpuWait", "cat": "NET", "ph": "e", "id": 14, "pid": 4157654, "tid": 1, "ts": 769594.615234},
{"name": "SendWait", "cat": "NET", "ph": "b", "id": 14, "pid": 4157654, "tid": 1, "ts": 769594.615234, "args": {"Step": 0}},
{"name": "SendWait", "cat": "NET", "ph": "e", "id": 14, "pid": 4157654, "tid": 1, "ts": 769618.889648},
{"name": "SendBufferWait", "cat": "NET", "ph": "b", "id": 15, "pid": 4157654, "tid": 1, "ts": 768505.083008, "args": {"Step": 1}},
{"name": "SendBufferWait", "cat": "NET", "ph": "e", "id": 15, "pid": 4157654, "tid": 1, "ts": 768505.163086},
{"name": "SendGpuWait", "cat": "NET", "ph": "b", "id": 15, "pid": 4157654, "tid": 1, "ts": 768505.163086, "args": {"Step": 1}},
{"name": "SendGpuWait", "cat": "NET", "ph": "e", "id": 15, "pid": 4157654, "tid": 1, "ts": 769610.555664},
{"name": "SendWait", "cat": "NET", "ph": "b", "id": 15, "pid": 4157654, "tid": 1, "ts": 769610.555664, "args": {"Step": 1}},
{"name": "SendWait", "cat": "NET", "ph": "e", "id": 15, "pid": 4157654, "tid": 1, "ts": 769622.517578},
{"name": "SendBufferWait", "cat": "NET", "ph": "b", "id": 16, "pid": 4157654, "tid": 1, "ts": 768507.937500, "args": {"Step": 2}},
{"name": "SendBufferWait", "cat": "NET", "ph": "e", "id": 16, "pid": 4157654, "tid": 1, "ts": 768508.017578},
{"name": "SendGpuWait", "cat": "NET", "ph": "b", "id": 16, "pid": 4157654, "tid": 1, "ts": 768508.017578, "args": {"Step": 2}},
{"name": "SendGpuWait", "cat": "NET", "ph": "e", "id": 16, "pid": 4157654, "tid": 1, "ts": 770002.129883},
{"name": "SendWait", "cat": "NET", "ph": "b", "id": 16, "pid": 4157654, "tid": 1, "ts": 770002.129883, "args": {"Step": 2}},
{"name": "SendWait", "cat": "NET", "ph": "e", "id": 16, "pid": 4157654, "tid": 1, "ts": 770013.848633},
{"name": "SendBufferWait", "cat": "NET", "ph": "b", "id": 17, "pid": 4157654, "tid": 1, "ts": 768510.742188, "args": {"Step": 3}},
{"name": "SendBufferWait", "cat": "NET", "ph": "e", "id": 17, "pid": 4157654, "tid": 1, "ts": 768510.822266},
{"name": "SendGpuWait", "cat": "NET", "ph": "b", "id": 17, "pid": 4157654, "tid": 1, "ts": 768510.822266, "args": {"Step": 3}},
{"name": "SendGpuWait", "cat": "NET", "ph": "e", "id": 17, "pid": 4157654, "tid": 1, "ts": 771461.563477},
{"name": "SendWait", "cat": "NET", "ph": "b", "id": 17, "pid": 4157654, "tid": 1, "ts": 771461.563477, "args": {"Step": 3}},
{"name": "SendWait", "cat": "NET", "ph": "e", "id": 17, "pid": 4157654, "tid": 1, "ts": 771469.171875},
{"name": "Group API", "cat": "GROUP_API", "ph": "b", "id": 0, "pid": 225798, "tid": 1, "ts": 3433.595001, "args": {"groupApiId": 0, "groupDepth":1}},
{"name": "KernelLaunch", "cat": "KERNEL_LAUNCH", "ph": "b", "id": 0, "pid": 225798, "tid": 1, "ts": 0.000000, "args": {"groupId": 0, "Stream": 0x5020000567d0}},
{"name": "KernelLaunch", "cat": "KERNEL_LAUNCH", "ph": "e", "id": 0, "pid": 225798, "tid": 1, "ts": 111991.558990},
{"name": "AllReduce", "cat": "COLL_API", "ph": "b", "id": 0, "pid": 225798, "tid": 1, "ts": 0.000000, "args": {"count": 262144, "datatype": ncclFloat32, "root": 0, "GraphCaptured":0, "Stream": 0x5020000567d0}},
{"name": "AllReduce", "cat": "COLL", "ph": "b", "id": 0, "pid": 225798, "tid": 1, "ts": 111994.477997, "args": {"SeqNum": 0, "CommHash": 1493613951195738943, "Rank": 0, "Count": 262144, "Datatype": "ncclFloat32", "Algorithm": "RING", "Protocol": "SIMPLE", "nChannels": 2}},
{"name": "KernelCh", "cat": "GPU", "ph": "b", "id": 0, "pid": 225798, "tid": 1, "ts": 119711.888000, "args": {"Channel": 0, "StartGpuClk": 1756135989724672000, "StopGpuClk": 1756135989732831232}},
{"name": "ScheduleRecv", "cat": "PROXY", "ph": "b", "id": 0, "pid": 225798, "tid": 1, "ts": 119652.709991, "args": {"Channel": 0, "Peer": 1, "Steps": 4, "ChunkSize": 4194304, "transSize": 524288}},
{"name": "ScheduleRecv", "cat": "PROXY", "ph": "e", "id": 0, "pid": 225798, "tid": 1, "ts": 119686.300995},
{"name": "ProgressRecv", "cat": "PROXY", "ph": "b", "id": 0, "pid": 225798, "tid": 1, "ts": 119686.300995, "args": {"Channel": 0, "Peer": 1, "Steps": 4, "ChunkSize": 4194304, "transSize": 524288}},
{“name": "RecvWait", "cat": "NET", "ph": "b", "id": 0, "pid": 225798, "tid": 1, "ts": 119707.677979, "args": {"Step": 0}},
{"name": "RecvWait", "cat": "NET", "ph": "e", "id": 0, "pid": 225798, "tid": 1, "ts": 119807.691986},
{"name": "RecvFlushWait", "cat": "NET", "ph": "b", "id": 0, "pid": 225798, "tid": 1, "ts": 119807.691986, "args": {"Step": 0}},
{"name": "RecvFlushWait", "cat": "NET", "ph": "e", "id": 0, "pid": 225798, "tid": 1, "ts": 119867.338989},
{"name": "RecvGpuWait", "cat": "NET", "ph": "b", "id": 0, "pid": 225798, "tid": 1, "ts": 119867.338989, "args": {"Step": 0}},
{"name": "RecvGpuWait", "cat": "NET", "ph": "e", "id": 0, "pid": 225798, "tid": 1, "ts": 120120.983002},
{"name": "RecvWait", "cat": "NET", "ph": "b", "id": 1, "pid": 225798, "tid": 1, "ts": 119733.647980, "args": {"Step": 1}},
{"name": "RecvWait", "cat": "NET", "ph": "e", "id": 1, "pid": 225798, "tid": 1, "ts": 119844.401001},
{"name": "RecvFlushWait", "cat": "NET", "ph": "b", "id": 1, "pid": 225798, "tid": 1, "ts": 119844.401001, "args": {"Step": 1}},
{"name": "RecvFlushWait", "cat": "NET", "ph": "e", "id": 1, "pid": 225798, "tid": 1, "ts": 119890.567993},
{"name": "RecvGpuWait", "cat": "NET", "ph": "b", "id": 1, "pid": 225798, "tid": 1, "ts": 119890.567993, "args": {"Step": 1}},
{"name": "RecvGpuWait", "cat": "NET", "ph": "e", "id": 1, "pid": 225798, "tid": 1, "ts": 120121.129974},
{"name": "RecvWait", "cat": "NET", "ph": "b", "id": 2, "pid": 225798, "tid": 1, "ts": 119753.023987, "args": {"Step": 2}},
{"name": "RecvWait", "cat": "NET", "ph": "e", "id": 2, "pid": 225798, "tid": 1, "ts": 120038.847992},
{"name": "RecvFlushWait", "cat": "NET", "ph": "b", "id": 2, "pid": 225798, "tid": 1, "ts": 120038.847992, "args": {"Step": 2}},
{"name": "RecvFlushWait", "cat": "NET", "ph": "e", "id": 2, "pid": 225798, "tid": 1, "ts": 120085.685974},
{"name": "RecvGpuWait", "cat": "NET", "ph": "b", "id": 2, "pid": 225798, "tid": 1, "ts": 120085.685974, "args": {"Step": 2}},
{"name": "RecvGpuWait", "cat": "NET", "ph": "e", "id": 2, "pid": 225798, "tid": 1, "ts": 120121.244995},
{"name": "RecvWait", "cat": "NET", "ph": "b", "id": 3, "pid": 225798, "tid": 1, "ts": 119772.510986, "args": {"Step": 3}},
{"name": "RecvWait", "cat": "NET", "ph": "e", "id": 3, "pid": 225798, "tid": 1, "ts": 120062.944977},
{"name": "RecvFlushWait", "cat": "NET", "ph": "b", "id": 3, "pid": 225798, "tid": 1, "ts": 120062.944977, "args": {"Step": 3}},
{"name": "RecvFlushWait", "cat": "NET", "ph": "e", "id": 3, "pid": 225798, "tid": 1, "ts": 120101.089996},
{"name": "RecvGpuWait", "cat": "NET", "ph": "b", "id": 3, "pid": 225798, "tid": 1, "ts": 120101.089996, "args": {"Step": 3}},
{"name": "RecvGpuWait", "cat": "NET", "ph": "e", "id": 3, "pid": 225798, "tid": 1, "ts": 120165.115997},
{"name": "ProgressRecv", "cat": "PROXY", "ph": "e", "id": 0, "pid": 225798, "tid": 1, "ts": 120165.356995},
{"name": "ScheduleSend", "cat": "PROXY", "ph": "b", "id": 1, "pid": 225798, "tid": 1, "ts": 119656.950989, "args": {"Channel": 0, "Peer": 1, "Steps": 4, "ChunkSize": 4194304, "transSize": 524288}},
{"name": "ScheduleSend", "cat": "PROXY", "ph": "e", "id": 1, "pid": 225798, "tid": 1, "ts": 119709.078979},
{"name": "ProgressSend", "cat": "PROXY", "ph": "b", "id": 1, "pid": 225798, "tid": 1, "ts": 119709.078979, "args": {"Channel": 0, "Peer": 1, "Steps": 4, "ChunkSize": 4194304, "transSize": 524288}},
{"name": "SendGpuWait", "cat": "NET", "ph": "b", "id": 4, "pid": 225798, "tid": 1, "ts": 119710.632996, "args": {"Step": 0}},
{"name": "SendGpuWait", "cat": "NET", "ph": "e", "id": 4, "pid": 225798, "tid": 1, "ts": 119808.636993},
{"name": "SendPeerWait", "cat": "NET", "ph": "b", "id": 4, "pid": 225798, "tid": 1, "ts": 119808.636993, "args": {"Step": 0}},
{"name": "SendPeerWait", "cat": "NET", "ph": "e", "id": 4, "pid": 225798, "tid": 1, "ts": 119818.972992},
... [ trace truncated for brevity ]
{"name": "AllReduce", "cat": "COLL", "ph": "e", "id": 0, "pid": 4157654, "tid": 1, "ts": 772209.317383},
{"name": "Group", "cat": "GROUP", "ph": "e", "id": 0, "pid": 4157654, "tid": 1, "ts": 772209.418945},
{"name": "AllReduce", "cat": "COLL", "ph": "e", "id": 17, "pid": 225798, "tid": 1, "ts": 170633.535980},
{"name": "AllReduce", "cat": "COLL_API", "ph": "e", "id": 17, "pid": 225798, "tid": 1, "ts": 170582.923981},
{"name": "Group API", "cat": "GROUP_API", "ph": "e", "id": 17, "pid": 225798, "tid": 1, "ts": 170637.582001},
{}]
```
Details about the fields used in the trace can be found at this link:
https://docs.google.com/document/d/1CvAClvFfyA5R-PhYUmn5OOQtYMH4h6I0nSsKchNAySU/preview?tab=t.0#heading=h.yr4qxyxotyw
The trace above is obtained by running a `ncclAllReduce` operation on 8 GPUs, communicating with each other through
The trace above is obtained by running a `ncclAllReduce` operation on 2 GPUs, communicating with each other through
the network interface. The `Group` event encloses all traces that are related to the single `ncclAllReduce` call.
(Note that for single collective invocations, where there are no explicit group calls, NCCL creates a group with only
one collective and this is what is presented in the traces above).
@@ -161,38 +145,17 @@ The `AllReduce` entry presents information about the `ncclAllReduce` operation.
- datatype : NCCL datatype
- algorithm : algorithm used to process the ncclAllReduce
- protocol : protocol used to process the ncclAllReduce
- nMaxChannels: max number of channels used to process the ncclAllReduce
- nChannels : Number of channels used to process the ncclAllReduce
If the proxy events are not active (e.g., the `ncclAllReduce` is intranode) the end timestamp will match the time
consumed by the CPU to launch the collective. For more details refer to `ext-profiler/README.md`, section `Profiling
of collective and p2p operations`.
### Proxy Send
The `Send` entry presents information about the `ProxyOp` processing in the progress thread. It contains the following
info in the args field:
- Channel : id of the channel used by this proxy operation to send data to the peer
- Peer : peer rank
- Steps : number of network steps required to transfer transSize bytes to the peer
- ChunkSize : chunk size used by NCCL to pipeline data through the proxy thread
- transSize : bytes transferred across the channel by this proxy operation
- POSTED : struct containing the number of buffer posts to the GPU and the time stamp for the last post
- REM_FIFO_WAIT: struct containing the number of remote buffer waits and the time stamp for the last wait
- TRANSMITTED : struct containing the number of network sends and the time stamp of the last send
- DONE : struct containing the number of network sends completed and the time stamp of the last send completed
In case of a network problem the POSTED, REM_FIFO_WAIT, TRANSMITTED and DONE might all have partially updated steps,
which could help identify at which point the network problem occurred.
The Proxy send trace gives a summary of the proxy progress thread activity for the channel. If more details are
needed, these can be obtained by enabling the proxy step event (`ncclProfileProxyStep`). In which case the trace
entries below are also reported by the profiler.
#### Proxy SendBufferWait
Presents, for every network step, the time the CPU proxy spends waiting for the channel staging buffer to become available.
#### Proxy SendGPUWait
#### Proxy SendGpuWait
Presents, for every network step, the time the CPU proxy spends waiting for the GPU to provide the data in the staging
buffer.
@@ -201,31 +164,6 @@ buffer.
Presents, for every network step, the time the CPU proxy spends waiting for the `isend` to complete
### Proxy Recv
The `Recv` entry presents information about the `ProxyOp` processing in the progress thread. It contains the following
info in the args field:
- Channel : id of the channel used by this proxy operation to recv data from the peer
- Peer : peer rank
- Steps : number of network steps required to transfer transSize bytes from the peer
- ChunkSize : chunk size used by NCCL to pipeline data through the proxy thread
- transSize : bytes transferred across the channel by this proxy operation
- POSTED : struct containing the number of recvs posted and the time stamp for the last recv posted
- RECEIVED : struct containing the number of recvs completed and the time stamp for the last recv completed
- TRANSMITTED: struct containing the number of recvs flushed to the GPU memory and the time stamp for the last recv flushed
- DONE : struct containing the number of flush completed and the time stamp for the last flush completed
The Proxy Recv trace gives a summary of the proxy progress thread activity for the channel. If more details are
needed, these can be obtained by enabling the proxy step event (`ncclProfileProxyStep`). In which case the trace
entries below are also reported by the profiler.
#### Proxy RecvBufferWait
Presents, for every network step, the time the CPU proxy spends waiting for the staging buffer for the channel to
become available.
#### Proxy RecvWait
Presents, for every network step, the time the CPU proxy spends waiting for a posted `irecv` to complete
@@ -234,6 +172,6 @@ Presents, for every network step, the time the CPU proxy spends waiting for a po
Presents, for every network step, the time the CPU proxy spends waitng for the recv data to be flushed to the GPU
#### Proxy RecvGPUWait
#### Proxy RecvGpuWait
Presents, for every network step, the time the CPU proxy spends waiting for the GPU to consume the recv data