# Motivation
This PR intends to extend `cuda_lazy_init` to `device_lazy_init` which is a device-agnostic API that can support any backend. And change `maybe_initialize_cuda` to `maybe_initialize_device` to support lazy initialization for CUDA while maintaining scalability.
# Design
We maintain a flag for each backend to manage the lazy initialization state separately.
# Additional Context
No need more UTs.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/118846
Approved by: https://github.com/malfet
<!--
copilot:summary
-->
### <samp>🤖 Generated by Copilot at 27084ed</samp>
This pull request simplifies and cleans up the code that uses the cuDNN library for convolution, batch normalization, CTC loss, and quantized operations. It removes the unnecessary checks and conditions for older cuDNN versions and the experimental cuDNN v8 API, and ~~replaces them with the stable `cudnn_frontend` API that requires cuDNN v8 or higher. It also adds the dependency and configuration for the `cudnn_frontend` library in the cmake and bazel files.~~ Correction: The v7 API will still be available with this PR, and can still be used, without any changes to the defaults. This change simply always _builds_ the v8 API, and removes the case where _only_ the v7 API is built.
This is a re-land of https://github.com/pytorch/pytorch/pull/91527
Pull Request resolved: https://github.com/pytorch/pytorch/pull/95722
Approved by: https://github.com/malfet, https://github.com/atalman
<!--
copilot:summary
-->
### <samp>🤖 Generated by Copilot at 27084ed</samp>
This pull request simplifies and cleans up the code that uses the cuDNN library for convolution, batch normalization, CTC loss, and quantized operations. It removes the unnecessary checks and conditions for older cuDNN versions and the experimental cuDNN v8 API, and ~~replaces them with the stable `cudnn_frontend` API that requires cuDNN v8 or higher. It also adds the dependency and configuration for the `cudnn_frontend` library in the cmake and bazel files.~~ Correction: The v7 API will still be available with this PR, and can still be used, without any changes to the defaults. This change simply always _builds_ the v8 API, and removes the case where _only_ the v7 API is built.
This is a re-land of https://github.com/pytorch/pytorch/pull/91527
Pull Request resolved: https://github.com/pytorch/pytorch/pull/95722
Approved by: https://github.com/malfet
Summary:
This patch prototypes a trace tracker callback mechanism based on existing TraceEntry records.
- It allows external of cache allocator to "attach" trace tracker callbacks.
- When a TraceEntry is recorded, it triggers all attached callbacks. Callbacks can selectively behave based on the trace action.
- **RISK**: The attached callback would be called within an allocator call stack (e.g., free during an allocate call). Potential deadlock may occur if other locks are called within the callback and has interdependency w/ the device allocator lock. It is the callback developer's responsibility to avoid any potential deadlock.
- **ADVICE**: The callback mechanism is designed **only for Pytorch internal use**. We should not expose it to Python layer due to Python GIL that would cause a deadlock.
See example in D50726970 that attaches NCCL register/deregister hooks via the trace tracker callback, so that all CUDA segments allocated by the allocator can be registered to NCCL communicators before any NCCL communication happens. This enables fast zero copy algorithms in NCCL.
Differential Revision: D50726971
Pull Request resolved: https://github.com/pytorch/pytorch/pull/112238
Approved by: https://github.com/zdevito
Printing just the device name is not helpful when investigating PyTorch issues filed for specific AMD GPUs, as the support/issue might depend on the gfx arch, which is part of the gcnArchName property.
`torch.cuda.get_device_properties(0).gcnArchName` will print the value of the `gcnArchName` property: eg.
```
>>> torch.cuda.get_device_properties(0).gcnArchName
'gfx906:sramecc+:xnack-'
```
```
root@6f064e3c19fb:/data/pytorch/test# python ../torch/utils/collect_env.py
...
GPU models and configuration: AMD Radeon Graphics(gfx906:sramecc+:xnack-)
```
Pull Request resolved: https://github.com/pytorch/pytorch/pull/107477
Approved by: https://github.com/albanD
Summary: This diff refactors the code by moving CUDAAllocatorConfig into the header file. This config refactoring is done so that we can use the same config code for CUDA pinned memory as well.
Test Plan: sandcastle
Differential Revision: D49653265
Pull Request resolved: https://github.com/pytorch/pytorch/pull/110123
Approved by: https://github.com/zdevito
Most `torch.cuda` ops (ex: `torch.cuda.synchronize`) do not release GIL in C++ land. This has the potential of causing deadlocks and freeze the python process. For example, `torch.cuda.synchronize` could hold GIL and get blocked on some operation. However, that operation might never complete in python land since GIL is held by `torch.cuda.synchronize`.
In this PR, I've tried to release GIL as much as possible in `torch.cuda` ops.
See https://github.com/pytorch/pytorch/issues/109074 for an example of how holding GIL causes a deadlock.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/109159
Approved by: https://github.com/ezyang
> capture_error_mode (str, optional): specifies the cudaStreamCaptureMode for the graph capture stream.
Can be "global", "thread_local" or "relaxed". During cuda graph capture, some actions, such as cudaMalloc,
may be unsafe. "global" will error on actions in other threads, "thread_local" will only error for
actions in the current thread, and "relaxed" will not error on these actions.
Inductor codegen is single-threaded, so it should be safe to enable "thread_local" for inductor's cuda graph capturing. We have seen errors when inductor cudagraphs has been used concurrently with data preprocessing in other threads.
Differential Revision: [D48656014](https://our.internmc.facebook.com/intern/diff/D48656014)
Pull Request resolved: https://github.com/pytorch/pytorch/pull/107407
Approved by: https://github.com/albanD, https://github.com/eqy
This change is to match the behavior of _record_memory_history which was
recently changed to enable history recording on all devices rather than
the current one. It prevents confusing situations where the observer
was registered before the device was set for the training run.
It also ensures the allocators have been initialized in the python binding just in case this is the first call to the CUDA API.
Fixes#107330
Pull Request resolved: https://github.com/pytorch/pytorch/pull/107399
Approved by: https://github.com/eellison
ghstack dependencies: #107171
Previously when we recorded a free action in a memory trace, we would provide
the stack for when the block was allocated. This is faster because we do not
have to record stacks for free, which would otherwise double the number of stacks
collected. However, sometimes knowing the location of a free is useful for
figuring out why a tensor was live. So this PR adds this behavior. If
performance ends up being a concern the old behavior is possible by passing
"alloc" to the context argument rather than "all".
Also refactors some of glue logic to be consistent across C++ and Python and
routes the Python API through the C++ version.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/106758
Approved by: https://github.com/albanD
PR #90689 replaces NVTX with NVTX3. However, the torch::nvtoolsext is created only when the third party NVTX is used.
This is clear a logical error. We now move the creation code out of the branch to cover all cases. This should fix the issues reported in the comments of #90689.
It would be better to move configurations of the failed FRL jobs to CI tests so that we can find such issues early before merging.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/97582
Approved by: https://github.com/peterbell10
We want to display the stack for the original cudaMalloc that created a segment.
Previously we could only report the last time the segment memory was used,
or the record of the segment_alloc could appear in the list of allocator actions.
This PR ensure regardless of whether we still have the segment_alloc action,
the context for a segment is still available. The visualizer is updated to
be able to incorporate this information.
This PR adds a new field to Block. However the previous stacked cleanup PR
removed a field of the same size, making the change to Block size-neutral.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/106113
Approved by: https://github.com/aaronenyeshi
For free blocks of memory in the allocator, we previously kept a linked list
of the stack frames of previous allocations that lived there. This was only
ever used in one flamegraph visualization and never proved useful at
understanding what was going on. When memory history tracing was added, it
became redundant, since we can see the history of the free space from recording
the previous actions anyway.
This patch removes this functionality and simplifies the snapshot format:
allocated blocks directly have a 'frames' attribute rather than burying stack frames in the history.
Previously the memory history tracked the real size of allocations before rounding.
Since history was added, 'requested_size' has been added directly to the block which records the same information,
so this patch also removes that redundancy.
None of this functionality has been part of a PyTorch release with BC guarentees, so it should be safe to alter
this part of the format.
This patch also updates our visualization tools to work with the simplified format. Visualization tools keep
support for the old format in `_legacy` functions so that during the transition old snapshot files can still be read.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/106079
Approved by: https://github.com/eellison
Summary:
Since NCCL 2.12.10, NCCL supports send/recv 0 byte: https://github.com/NVIDIA/nccl/issues/696. Therefore we don't have to skip.
One issue is that if a rank has 0 bytes to send and 0 bytes to recv, it'll skip send/recv completely. And it'll proceed to the next collective which it can send/recv something, making it confusing to the other ranks. Another solution is to add a barrier but that's very expensive.
Test Plan: will add a unit test
Differential Revision: D46507785
Pull Request resolved: https://github.com/pytorch/pytorch/pull/103140
Approved by: https://github.com/malfet, https://github.com/kwen2501
Changes the StreamID encoding to use the last bit to distinguish between external and internal streams, 4 bits for IdType (DEFAULT, EXT or user-created streams possibly with high priority), and 5 bits for index. This allows us to have more stream priorities exposed to user (I'm currently setting 4, but that's easy to change now). Note, we are pre-creating all 32 streams in the pool per each allowed priority, I don't know if it's a problem in practice. Currently cuda 11.8/A100 GPUs allow 6 different stream priorities, the number may be different for the different cards/different cuda versions.
Previous callsites explicitly requesting high prioity stream (`isHighPriority=true`) are now getting the highest priority stream.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/101956
Approved by: https://github.com/ezyang
Changes the StreamID encoding to use the last bit to distinguish between external and internal streams, 4 bits for IdType (DEFAULT, EXT or user-created streams possibly with high priority), and 5 bits for index. This allows us to have more stream priorities exposed to user (I'm currently setting 4, but that's easy to change now). Note, we are pre-creating all 32 streams in the pool per each allowed priority, I don't know if it's a problem in practice. Currently cuda 11.8/A100 GPUs allow 6 different stream priorities, the number may be different for the different cards/different cuda versions.
Previous callsites explicitly requesting high prioity stream (`isHighPriority=true`) are now getting the highest priority stream.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/101956
Approved by: https://github.com/ezyang
cudaGetLastError and hipGetLastError will clear any error value within CUDA and HIP, respectively. This is often done on purpose to clear benign errors. Discarding the return value should be indicated by casting to void and a nearby comment. This silences warnings from HIP:
warning: ignoring return value of function declared with 'nodiscard' attribute [-Wunused-result]
Performing an audit of pytorch sources found one use of cudaGetLastError that was incorrectly ignored in IndexKernel.cu.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/100488
Approved by: https://github.com/ezyang
When we run cudagraph trees we are not allowed to have permanent workspace allocations like in cublas because we might need to reclaim that memory for a previous cudagraph recording, and it is memory that is not accounted for in output weakrefs so it does not work with checkpointing. Previously, I would check that we didn't have any additional allocations through snapshotting. This was extremely slow so I had to turn it off.
This PR first does the quick checking to see if we are in an error state, then if we are does the slow logic of creating snapshot. Also turns on history recording so we get a stacktrace of where the bad allocation came from.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/99985
Approved by: https://github.com/zdevito
In this stack of PRs we adding caching to output tensors for cudagraph trees after we've done initial recording. On initial recording we do not cache tensor outputs because this prevents memory from being reclaimed. On subsequent exeuctions we do cache them to avoid overhead. However, because there is an extra reference around, this caused divergent recording & execution behavior in both autocast caching and autograd gradient stealing. Divergent recording & execution would keep on re-recording and eventually stabilize, but it's not what you want to see happen.
This pr makes the autocast cache and buffer stealing aware of the cudagraph static output tensors.
I will add this to the other cudagraph impl in another pr.
Not sure if this should be in autograd or in autocast since it affects both.. Or somewhere else
Pull Request resolved: https://github.com/pytorch/pytorch/pull/99368
Approved by: https://github.com/albanD, https://github.com/ezyang
A quick, trial fix for #99677.
My guess is that when the code instantiates an `AutoNcclGroup` object, it comes with an uninitialized random value for member `comm_nonblocking_`. Then `if (comm_nonblocking_)` evaluates to true, and `NCCL_CHECK_TIMEOUT` triggered.
This change is safe (and needed) anyway whether it indeed fixes#99677.
Cc @eqy
Pull Request resolved: https://github.com/pytorch/pytorch/pull/99679
Approved by: https://github.com/eqy, https://github.com/awgu
Caches output tensors for the common case when the output Tensor storage is unaliased for all graph outputs in all paths. For these persisted tensors we adjust the liveness tracking by also checking that the output tensor does not have an additional python reference.
I limit cached output tensors to be unaliased. If a descendent node discovers it has an alias of a prior output, then the aliased output will no longer be persisted in the ancestor.
The large majority of tensors are unaliased, and preserving aliased output tensors would add significant additional complexity with marginal gains. For instance, when do checkpointing and re-recordings, we need to remove the persisted tensors otherwise it would prevent memory from being reclaimed. If a single persisted tensor was present in multiple paths then that would create an inter-path dependence which adds complexity. Additionally, each further caching of the output would affect the reference count of the other caches, and that reference count would also need to be adjusted depending on if a node was checkpointed.
Still need to do a complete a run but for the models I tried makes the performance extremely close between trees and non trees impl.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/98944
Approved by: https://github.com/jansel, https://github.com/ngimel
Common advice we give for handling memory fragmentation issues is to
allocate a big block upfront to reserve memory which will get split up later.
For programs with changing tensor sizes this can be especially helpful to
avoid OOMs that happen the first time we see a new largest input and would
otherwise have to allocate new segments.
However the issue with allocating a block upfront is that is nearly impossible
to correctly estimate the size of that block. If too small, space in the block
will run out and the allocator will allocate separate blocks anyway. Too large,
and other non-PyTorch libraries might stop working because they cannot allocate
any memory.
This patch provides the same benefits as using a pre-allocating block but
without having to choose its size upfront. Using the cuMemMap-style APIs,
it adds the ability to expand the last block in a segment when more memory is
needed.
Compared to universally using cudaMallocAsync to avoid fragmentation,
this patch can fix this common fragmentation issue while preserving most
of the existing allocator behavior. This behavior can be enabled and disabled dynamically.
This should allow users to, for instance, allocate long-lived parameters and state in individual buffers,
and put temporary state into the large expandable blocks, further reducing
fragmentation.
See inline comments for information about the implementation and its limitations.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/96995
Approved by: https://github.com/eellison
Support for nonblocking NCCL communicators/fault tolerance/checking which was added in 2.14 as an experimental feature.
Enabled via the environment variable:
```
TORCH_NCCL_USE_COMM_NONBLOCKING=1
```
CC @ptrblck
Pull Request resolved: https://github.com/pytorch/pytorch/pull/95715
Approved by: https://github.com/kwen2501
Support for nonblocking NCCL communicators/fault tolerance/checking which was added in 2.14 as an experimental feature.
Enabled via the environment variable:
```
TORCH_NCCL_USE_COMM_NONBLOCKING=1
```
CC @ptrblck
Pull Request resolved: https://github.com/pytorch/pytorch/pull/95715
Approved by: https://github.com/kwen2501
Significantly reduces overhead of constructing Tensors and Storages and checking Storage Liveness. Removes the regression for HF models that I tested and removes 75% of overhead of the extremely overhead bound resnet50 training we have in torchbench. (.91x base commit, 1.02x torchinductor default, 1.16x this PR, 1.25 previous cudagraphs impl).
This PR takes care of all of the lower hanging fruit.
- Computes storage aliasing at record time instead of during at runtime. We no longer need to use a runtime storage cache, and can instead index directly into the existing alias if there is one, or construct a new Storage
- Moves the heavyweight C++ calls into a batch - getting storage weakrefs and constructing tensors
Pull Request resolved: https://github.com/pytorch/pytorch/pull/98529
Approved by: https://github.com/jansel, https://github.com/ngimel
This method has to be accessible from `c10` to enable CUDA-12 integration.
Implemented by providing private `c10::cuda:_internal::setHasPrimaryContext` that passes the pointer to the implementation (in `torch_cuda`) back to c10.
Use global class constructor/destructor to guarantee RAII.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/96800
Approved by: https://github.com/ngimel
CUDA Graph Trees
Design doc: https://docs.google.com/document/d/1ZrxLGWz7T45MSX6gPsL6Ln4t0eZCSfWewtJ_qLd_D0E/edit
Not currently implemented :
- Right now, we are using weak tensor refs from outputs to check if a tensor has dies. This doesn't work because a) aliasing, and b) aot_autograd detaches tensors (see note [Detaching saved tensors in AOTAutograd]). Would need either https://github.com/pytorch/pytorch/issues/91395 to land to use storage weak refs or manually add a deleter fn that does what I want. This is doable but theres some interactions with the caching allocator checkpointing so saving for a stacked pr.
- Reclaiming memory from the inputs during model recording. This isn't terribly difficult but deferring to another PR. You would need to write over the input memory during warmup, and therefore copy the inputs to cpu. Saving for a stacked pr.
- Warning on overwriting previous generation outputs. and handling nested torch.compile() calls in generation tracking
Differential Revision: [D43999887](https://our.internmc.facebook.com/intern/diff/D43999887)
Pull Request resolved: https://github.com/pytorch/pytorch/pull/89146
Approved by: https://github.com/ezyang
Number of OSS PR were reverted, because new signed-unsigned comparison warnings, which are treated as errors in some internal builds.
Not sure how those selective rules are applied, but this PR removes `-Wno-sign-compare` from PyTorch codebase.
The only tricky part in this PR, as making sure that non-ASCII character detection works for both signed and unsigned chars here:
6e3d51b08a/torch/csrc/jit/serialization/python_print.cpp (L926)
Exclude several files from sign-compare if flash attention is used, due to the violation in cutlass, to be fixed by https://github.com/NVIDIA/cutlass/pull/869
Do not try to fix sign compare violations in caffe2 codebase
Pull Request resolved: https://github.com/pytorch/pytorch/pull/96723
Approved by: https://github.com/albanD
Previously the allocator would query whether a stream was recording a graph,
and look up the pool associated with a graph. This change has the allocator
directly associate a stream with a mempool, decoupling "record this stream to a pool"
from the action of "record all actions to a cuda graph".
Pull Request resolved: https://github.com/pytorch/pytorch/pull/96542
Approved by: https://github.com/eellison
This refactors the stack trace facility specific to memory profiling
in python+cuda to make a generic facility to generate combined stack
traces.
The generic facility (combined_traceback.h) does not require
python to be around to work, but will return python stacks if it is
present.
This facility is then used to add support for stack trace gathering in memory profiling that
happens directly from C++.
It is also used to expose a python API for gathering and symbolizing
combineds stacks.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/95541
Approved by: https://github.com/ezyang
When we checkpoint the state of the private pool allocator, we will need to make sure that its current live allocated blocks will get properly cleaned up when the tensors they correspond to die. Return DataPtrs for these new allocated blocks that the callee can swap onto live Tensors.
The exact api for setting the checkpoint can be manipulated after this as the cudagraph implementation is built out, but this at least shows its sufficiently general.
This should be the last PR touching cuda caching allocator necessary for new cudagraphs integration.
Differential Revision: [D43999888](https://our.internmc.facebook.com/intern/diff/D43999888)
Pull Request resolved: https://github.com/pytorch/pytorch/pull/95020
Approved by: https://github.com/zdevito
Copying note from cuda caching allocator:
```
* Note [Checkpointing PrivatePoolState]
*
* Refer above to Note [Interaction with CUDA graph capture]. Allocations made
* during graph capture are made from a separate private pool. During graph
* capture allocations behave as usual. During graph replay the allocator
* state does not change even as new tensors are created. The private pool
* will not free its blocks to the main caching allocator until cuda graph use
* is finished to prevent an allocation from eager clobbering the memory from
* a live but unaccounted for tensor that was created during replay.
*
* `make_graphed_callables`, a series of separate callables chained in
* successive cuda graphs, can share a memory pool because after a cuda graph
* recording the allocations in the shared private pool exactly reflect the
* tensors that are allocated.
*
* We would like to extend callable chaining to support a graphed callable
* tree. In this scenario, we have a tree of callable chains which will be
* captured with cuda graphs. In the diagram below, we have a tree with four
* callables, A, B, C, and D. Suppose we have captured, and subsequently
* replayed, A, B, and C. Then on a new invocation, we replay A and B, but
* would now like to record D. At this point the private pool will not reflect
* any of the live tensors created during graph replay. Allocations made
* during a new recording with the pool could overwrite those live tensors.
*
* In order to record a new graph capture after replaying prior callables in
* the tree, we need the allocator to reflect the state of the live tensors.
* We checkpoint the state of the private after each recording, and then
* reapply it when we are starting a new recording chain. Additionally, we
* must free the allocations for any tensors that died between the end of our
* previous graph replaying and our new recording (TODO). All of the allocated
* segments that existed in the checkpointed state must still exist in the
* pool. There may also exist new segments, which we will free (TODO : link
* note [live tensors between iterations] when it exists).
*
*
* ---------------> A ---------------> B ---------------> C
* |
* |
* |
* |
* ---------------> D
```
A few TODOs:
- need to add logic for freeing tensors that have died between a last replay and current new recording
- Add logic for free that might be called on a pointer multiple times (because we are manually freeing live tensors)
The two scenarios above have not been exercised in the tests yet.
Differential Revision: [D43999889](https://our.internmc.facebook.com/intern/diff/D43999889)
Pull Request resolved: https://github.com/pytorch/pytorch/pull/94653
Approved by: https://github.com/zdevito
Adds the ability to quickly generate stack traces for C++,
and combine Python, TorchScript, and C++ frames into a single trace.
This makes it possible for the memory tracer to record allocations inside
C++ code (e.g. convolution temporaries, backward operators).
The unwinder code is ~10x faster than execinfo.h's backward because it
cache fast unwinder routines for instruction pointers that have already been seen.
It is also only 1.2--2x slower than copying the entire stack (the approach perf takes),
while using 2 orders of magnitude less space per stack.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/95357
Approved by: https://github.com/bertmaher
Summary:
The caching allocator can be configured to round memory allocations in order to reduce fragmentation. Sometimes however, the overhead from rounding can be higher than the fragmentation it helps reduce.
We have added a new stat to CUDA caching allocator stats to help track if rounding is adding too much overhead and help tune the roundup_power2_divisions flag:
- "requested_bytes.{current,peak,allocated,freed}": memory requested by client code, compare this with allocated_bytes to check if allocation rounding adds too much overhead
Test Plan: Added test case in caffe2/test/test_cuda.py
Differential Revision: D40810674
Pull Request resolved: https://github.com/pytorch/pytorch/pull/88575
Approved by: https://github.com/zdevito
This adds `torch.cuda._DeviceGuard` which is a stripped down version of
`torch.cuda.device` with lower overhead. To do this, it only accepts `int` as
the device so we don't need to call `_get_device_index` and is implemented
with a new C++ helper `torch._C._cuda_exchangeDevice` that allows
`_DeviceGuard.__enter__` to be just a single function call. On my machine,
I see a drop from 3.8us of overhead to 0.94 us with this simple benchmark:
```python
def set_device():
with torch.cuda.device(0):
pass
%timeit set_device()
```
Pull Request resolved: https://github.com/pytorch/pytorch/pull/91045
Approved by: https://github.com/ngimel, https://github.com/anijain2305
#75854
A naive attempt at working around the limitations of using a single 64-bit integer to pack `stream_id`, `device_index`, and `device_type`.
Stills needs sanity checks, testing, and minimization of BC-breaking changes.
Currently a Holder for the `StreamData3` struct is used for `IValue` compatibility. While doing this seems to work for `ivalue.h` and `ivalue_inl.h`, this doesn't seem to be naively working for the JIT CUDA stream wrapper? (Something about ambiguous calls if an `intrusive_ptr` to `c10::ivalue::StreamData3Holder` is used as the return type for `pack()`. It turns out that the methods required to access the fields for rematerializing a CUDA Stream are basically already present anyway, so `pack` is simply removed in the wrapper for now and the methods to access the required fields are called directly.
CC @ptrblck
Pull Request resolved: https://github.com/pytorch/pytorch/pull/81596
Approved by: https://github.com/ezyang
Fix the failure when building PyTorch from source code using CUDA 12
```
In file included from /home/jianyuhuang/Work/Github/pytorch/c10/cuda/CUDAFunctions.h:12,
from /home/jianyuhuang/Work/Github/pytorch/c10/cuda/CUDAStream.h:10,
from /home/jianyuhuang/Work/Github/pytorch/c10/cuda/CUDAGraphsC10Utils.h:3,
from /home/jianyuhuang/Work/Github/pytorch/aten/src/ATen/cuda/CUDAGraph.h:5,
from /home/jianyuhuang/Work/Github/pytorch/aten/src/ATen/cuda/CUDAGraph.cpp:2:
/home/jianyuhuang/Work/Github/pytorch/aten/src/ATen/cuda/CUDAGraph.cpp: In member function ‘void at::cuda::CUDAGraph::capture_end()’:
/home/jianyuhuang/Work/Github/pytorch/aten/src/ATen/cuda/CUDAGraph.cpp:168:75: warning: converting to non-pointer type ‘long long unsigned int’ from NULL [-Wconversion-null]
AT_CUDA_CHECK(cudaGraphInstantiate(&graph_exec_, graph_, NULL, NULL, 0));
^
/home/jianyuhuang/Work/Github/pytorch/c10/cuda/CUDAException.h:31:42: note: in definition of macro ‘C10_CUDA_CHECK’
C10_UNUSED const cudaError_t __err = EXPR; \
^~~~
/home/jianyuhuang/Work/Github/pytorch/aten/src/ATen/cuda/CUDAGraph.cpp:168:5: note: in expansion of macro ‘AT_CUDA_CHECK’
AT_CUDA_CHECK(cudaGraphInstantiate(&graph_exec_, graph_, NULL, NULL, 0));
^~~~~~~~~~~~~
/home/jianyuhuang/Work/Github/pytorch/aten/src/ATen/cuda/CUDAGraph.cpp:168:75: error: too many arguments to function ‘cudaError_t cudaGraphInstantiate(CUgraphExec_st**, cudaGraph_t, long long unsigned int)’
AT_CUDA_CHECK(cudaGraphInstantiate(&graph_exec_, graph_, NULL, NULL, 0));
^
/home/jianyuhuang/Work/Github/pytorch/c10/cuda/CUDAException.h:31:42: note: in definition of macro ‘C10_CUDA_CHECK’
C10_UNUSED const cudaError_t __err = EXPR; \
^~~~
/home/jianyuhuang/Work/Github/pytorch/aten/src/ATen/cuda/CUDAGraph.cpp:168:5: note: in expansion of macro ‘AT_CUDA_CHECK’
AT_CUDA_CHECK(cudaGraphInstantiate(&graph_exec_, graph_, NULL, NULL, 0));
^~~~~~~~~~~~~
In file included from /home/jianyuhuang/Work/Github/pytorch/c10/cuda/CUDAStream.h:6,
from /home/jianyuhuang/Work/Github/pytorch/c10/cuda/CUDAGraphsC10Utils.h:3,
from /home/jianyuhuang/Work/Github/pytorch/aten/src/ATen/cuda/CUDAGraph.h:5,
from /home/jianyuhuang/Work/Github/pytorch/aten/src/ATen/cuda/CUDAGraph.cpp:2:
/usr/local/cuda/include/cuda_runtime_api.h:11439:39: note: declared here
extern __host__ cudaError_t CUDARTAPI cudaGraphInstantiate(cudaGraphExec_t *pGraphExec, cudaGraph_t graph, unsigned long long flags __dv(0));
^~~~~~~~~~~~~~~~~~~~
ninja: build stopped: subcommand failed.
```
```
/home/jianyuhuang/Work/Github/pytorch/torch/csrc/cuda/shared/cudart.cpp: In function ‘void torch::cuda::shared::initCudartBindings(PyObject*)’:
/home/jianyuhuang/Work/Github/pytorch/torch/csrc/cuda/shared/cudart.cpp:34:13: error: ‘cudaOutputMode_t’ was not declared in this scope
py::enum_<cudaOutputMode_t>(
^~~~~~~~~~~~~~~~
/home/jianyuhuang/Work/Github/pytorch/torch/csrc/cuda/shared/cudart.cpp:34:13: note: suggested alternative: ‘cudaGraphNode_t’
py::enum_<cudaOutputMode_t>(
^~~~~~~~~~~~~~~~
cudaGraphNode_t
/home/jianyuhuang/Work/Github/pytorch/torch/csrc/cuda/shared/cudart.cpp:34:29: error: template argument 1 is invalid
py::enum_<cudaOutputMode_t>(
^
/home/jianyuhuang/Work/Github/pytorch/torch/csrc/cuda/shared/cudart.cpp:38:30: error: ‘cudaKeyValuePair’ was not declared in this scope
.value("KeyValuePair", cudaKeyValuePair)
^~~~~~~~~~~~~~~~
/home/jianyuhuang/Work/Github/pytorch/torch/csrc/cuda/shared/cudart.cpp:39:21: error: ‘cudaCSV’ was not declared in this scope
.value("CSV", cudaCSV);
^~~~~~~
/home/jianyuhuang/Work/Github/pytorch/torch/csrc/cuda/shared/cudart.cpp:39:21: note: suggested alternative: ‘cudart’
.value("CSV", cudaCSV);
^~~~~~~
cudart
/home/jianyuhuang/Work/Github/pytorch/torch/csrc/cuda/shared/cudart.cpp:99:7: error: ‘cudaProfilerInitialize’ was not declared in this scope
cudaProfilerInitialize);
^~~~~~~~~~~~~~~~~~~~~~
/home/jianyuhuang/Work/Github/pytorch/torch/csrc/cuda/shared/cudart.cpp:99:7: note: suggested alternative: ‘cudaProfilerStart’
cudaProfilerInitialize);
^~~~~~~~~~~~~~~~~~~~~~
cudaProfilerStart
ninja: build stopped: subcommand failed.
```
After these fixes, we can see CUDA 12 is successfully built with OSS PyTorch instructions.
USE_CUDA=1 python setup.py develop 2>&1 | tee compile.log
Pull Request resolved: https://github.com/pytorch/pytorch/pull/91118
Approved by: https://github.com/ngimel, https://github.com/brad-mengchi
Optimizes the nccl python bindings to reserve space when converting PythonObject* into Tensors. This should reduce the number of unnecessary allocations in the nccl bindings as the std::vector grows.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/88203
Approved by: https://github.com/ezyang
Fixes#43144
This uses the Backend system added by [82682](https://github.com/pytorch/pytorch/pull/82682) to change allocators dynamically during the code execution. This will allow us to use RMM, use CUDA managed memory for some portions of the code that do not fit in GPU memory. Write static memory allocators to reduce fragmentation while training models and improve interoperability with external DL compilers/libraries.
For example, we could have the following allocator in c++
```c++
#include <sys/types.h>
#include <cuda_runtime_api.h>
#include <iostream>
extern "C" {
void* my_malloc(ssize_t size, int device, cudaStream_t stream) {
void *ptr;
std::cout<<"alloc "<< size<<std::endl;
cudaMalloc(&ptr, size);
return ptr;
}
void my_free(void* ptr) {
std::cout<<"free "<<std::endl;
cudaFree(ptr);
}
}
```
Compile it as a shared library
```
nvcc allocator.cc -o alloc.so -shared --compiler-options '-fPIC'
```
And use it from PyTorch as follows
```python
import torch
# Init caching
# b = torch.zeros(10, device='cuda')
new_alloc = torch.cuda.memory.CUDAPluggableAllocator('alloc.so', 'my_malloc', 'my_free')
old = torch.cuda.memory.get_current_allocator()
torch.cuda.memory.change_current_allocator(new_alloc)
b = torch.zeros(10, device='cuda')
# This will error since the current allocator was already instantiated
torch.cuda.memory.change_current_allocator(old)
```
Things to discuss
- How to test this, needs compiling external code ...
Pull Request resolved: https://github.com/pytorch/pytorch/pull/86786
Approved by: https://github.com/albanD
This updates `wrap_pybind_function` to use `invoke` and adds the
`invoke_traits` object which is analogous to `function_traits` but
for member functions it includes the class as an explicit argument.
To test this is working properly, I've also applied it to the
`CUDAGraph` binding code.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/88932
Approved by: https://github.com/albanD
This replaces the manual function pointers, making it easier to write
new drop-in allocators.
Note that most allocation goes through the Allocator interface, which
CUDAAllocator inherits from, and this arrangement avoids adding and
additional layer of dispatch along this pathway compared to what existed before.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/87251
Approved by: https://github.com/wconstab
We currently can take snapshots of the state of the allocated cuda memory, but we do not have a way to correlate these snapshots with the actions the allocator that were taken between snapshots. This PR adds a simple fixed-sized buffer that records the major actions that the allocator takes (ALLOC, FREE, SEGMENT_ALLOC, SEGMENT_FREE, OOM, SNAPSHOT) and includes these with the snapshot information. Capturing period snapshots with a big enough trace buffer makes it possible to see how the allocator state changes over time.
We plan to use this functionality to guide how settings in the allocator can be adjusted and eventually have a more robust overall algorithm.
As a component of this functionality, we also add the ability to get a callback when the allocator will throw an OOM, primarily so that snapshots can be taken immediately to see why the program ran out of memory (most programs have some C++ state that would free tensors before the OutOfMemory exception can be caught).
This PR also updates the _memory_viz.py script to pretty-print the trace information and provide a better textual summary of snapshots distinguishing between internal and external fragmentation.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/86241
Approved by: https://github.com/ngimel
Sometimes the driving process want to save memory snapshots but isn't Python.
Add a simple API to turn it on without python stack traces. It still
saves to the same format for the vizualization and summary scripts, using
the C++ Pickler.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/86190
Approved by: https://github.com/ezyang
Summary:
- expose a python call to set the allocator settings, it uses the same format as the value for PYTORCH_CUDA_ALLOCATOR
- keep the implementation contained within the cpp file to avoid increasing build times, only expose a function to call the setting
- make some of the Allocator Config methods public, now it looks more like a singleton
Test Plan: added the unit test
Differential Revision: D39487522
Pull Request resolved: https://github.com/pytorch/pytorch/pull/84970
Approved by: https://github.com/zdevito
Record stack trace information for each allocated segment in the allocator.
It takes around 1.5us to record 50 stack frames of context.
Since invoking a Pytorch operator is around 8us, this adds minimal overhead but we still leave it disabled by default so that we can test it more on real workloads first.
Stack information is kept both for allocated blocks and the last allocation used inactive blocks. We could potential keep around the _first_ allocation that caused the block to get allocated from cuda as well.
Potential Followups:
* stack frame entries are small (16 bytes), but the list of Frames is not compressed eventhough most frames will share some entries. So far this doesn't produce huge dumps (7MB for one real workload that uses all memory on the GPU), but it can be much smaller through compression.
* Code to format the information is slow (a few seconds) because it uses python and FlameGraph.pl
* Things allocated during the backward pass have no stack frames because they are run on another C++ thread.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/82146
Approved by: https://github.com/albanD
### Description
cudaProfilerStart and cudaProfilerStop are deprecated but exposed by torch.cuda.cudart(). HIP has corresponding functions stubbed out, hipProfilerStart and hipProfilerStop, but they return hipErrorNotSupported. Profiling in HIP is supported, but not via these deprecated APIs.
See https://docs.nvidia.com/cuda/cuda-runtime-api/group__CUDART__PROFILER__DEPRECATED.html.
These functions are indirectly used by one or more unit tests that would otherwise pass if the non-functional HIP APIs were replaced with a dummy function.
### Testing
Unskipped a related unit test, run by ciflow/trunk.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/82778
Approved by: https://github.com/ezyang
We define specializations for pybind11 defined templates
(in particular, PYBIND11_DECLARE_HOLDER_TYPE) and consequently
it is important that these specializations *always* be #include'd
when making use of pybind11 templates whose behavior depends on
these specializations, otherwise we can cause an ODR violation.
The easiest way to ensure that all the specializations are always
loaded is to designate a header (in this case, torch/csrc/util/pybind.h)
that ensures the specializations are defined, and then add a lint
to ensure this header is included whenever pybind11 headers are
included.
The existing grep linter didn't have enough knobs to do this
conveniently, so I added some features. I'm open to suggestions
for how to structure the features better. The main changes:
- Added an --allowlist-pattern flag, which turns off the grep lint
if some other line exists. This is used to stop the grep
lint from complaining about pybind11 includes if the util
include already exists.
- Added --match-first-only flag, which lets grep only match against
the first matching line. This is because, even if there are multiple
includes that are problematic, I only need to fix one of them.
We don't /really/ need this, but when I was running lintrunner -a
to fixup the preexisting codebase it was annoying without this,
as the lintrunner overall driver fails if there are multiple edits
on the same file.
I excluded any files that didn't otherwise have a dependency on
torch/ATen, this was mostly caffe2 and the valgrind wrapper compat
bindings.
Note the grep replacement is kind of crappy, but clang-tidy lint
cleaned it up in most cases.
See also https://github.com/pybind/pybind11/issues/4099
Signed-off-by: Edward Z. Yang <ezyang@fb.com>
Pull Request resolved: https://github.com/pytorch/pytorch/pull/82552
Approved by: https://github.com/albanD
Summary:
The include ordering modified via clang-format prevented a
declaration of _wgetenv, declared and used in nvtxInit.h, which was
previously acquired transitively through the pybind include. Provide an
explicit platform include of wchar.h which defines it on Windows, prior
to including nvToolsExt.h
Test Plan: Windows msvc cuda11 build should pass.
Differential Revision: D37244689
Pull Request resolved: https://github.com/pytorch/pytorch/pull/79798
Approved by: https://github.com/kit1980
Resubmit of https://github.com/pytorch/pytorch/pull/77673, which was reverted due to Windows test failures: https://github.com/pytorch/pytorch/pull/77673#issuecomment-1130425845.
I suspect these failures happened because I don't explicitly set a side stream for graph capture in the new test.
Not setting a side stream explicitly is alright on Linux because cuda tests implicitly use a side stream.
I think Windows cuda tests implicitly use the default stream, breaking capture and leaving the backend in a bad state.
Other graphs tests explicitly set side streams and don't error in Windows builds, so i'm 95% sure doing the same for the new test will work.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/77789
Approved by: https://github.com/ezyang
In preparation of adopting future rocblas library options, it is necessary to track when the backward pass of training is executing. The scope-based helper class `BackwardPassGuard` is provided to toggle state.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/71881
Approved by: https://github.com/albanD
This PR allows user to author a CUDA kernel in python.
```
from torch.cuda.jiterator import create_jit_fn
code_string = "template <typename T> T my_kernel(T x, T y, T alpha) { return -x * y + x - y + alpha; }"
jitted_fn = create_jit_fn(code_string, alpha=0)
a = torch.rand(3, device='cuda')
b = torch.rand(3, device='cuda')
result = jitted_fn(a, b, alpha=1.0)
```
Limitations:
- Only supports elementwise kernel
- 1~8 tensor inputs (empty input, e.g. factory methods, is not supported)
- inputs tensors must live in cuda device
- cpu Scalar is not supported
- kwargs must be pre-declared when calling create_jit_fn
- kwargs must be convertible to at::Scalar, one of float64, int64_t, bool. (complex not support for now)
TODOs:
- [x] consolidate union and c10::variant implementation
- [x] plug into existing op testing framework
- [ ] rename files, place files in the right folder
- [ ] place util functions in the right file
- [x] enforce assumptions in python interface e.g <8 inputs, kwargs types
- [x] Add user-facing documentation
Pull Request resolved: https://github.com/pytorch/pytorch/pull/76394
Approved by: https://github.com/mruberry
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/72836
Replacing increment iterator loops with ranged loops. It allows loops such as for(int i=0;i<10;i++) to be expressed as for(const auto i : c10::irange(10)). This auto-types the loops and adds const-safety to the iteration variable.
Reviewed By: albanD
Differential Revision: D34136539
fbshipit-source-id: 760a70ad43ce6f05630ba8fea261d4dbb699e62e
(cherry picked from commit 0428408d88)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/70029
This PR implements NCCL scatter and add scatter to ProcessGroupNCCL.
NCCL doesn’t directly provide primitives for scatter, so we need to be implemented on top of NCCL’s send/recv API.
1. In ProcessGroupNCCL.cpp, the inputTensors are first flattened, then outputTensors and inputFlattened are passed by the collective class to scatter() function in nccl.cpp.
2. In nccl.cpp, scatter is implemented using ncclSend/ncclRecv: the root rank uses a for loop to send(distribute) the inputTensors to each rank, then all the ranks receive the inputTensor from the root rank.
ghstack-source-id: 147754837
Test Plan:
test_scatter_ops
test_scatter_stress
test_scatter_checks
Reviewed By: pritamdamania87
Differential Revision: D33154823
fbshipit-source-id: 4513e7eaf7d47a60eb67da99dc6c2e9a2882f3fd
(cherry picked from commit 93201f9d4a)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/66745
This PR implement NCCL gather and add gather to ProcessGroupNCCL using nccl send/recv api.
NCCL doesn’t directly provide primitives for gather, so we need to be implemented on top of NCCL’s send/recv API.
1. In ProcessGroupNCCL.cpp, the outputTensors are first flattened, then inputTensors and outputFlattened are passed by the collective class to gather() function in nccl.cpp.
1. In nccl.cpp, gather is implemented using ncclSend/ncclRecv: all the ranks send inputTensor to the root rank, and the root rank uses a for loop to receive these inputTensors.
ghstack-source-id: 147754838
Test Plan:
test_gather_ops
test_gather_checks
test_gather_stress
Reviewed By: pritamdamania87
Differential Revision: D29616361
fbshipit-source-id: b500d9b8e67113194c5cc6575fb0e5d806dc7782
(cherry picked from commit d560ee732e)
Summary:
This patch moves a CUDA-specific file, `CUDAGeneratorImpl.h` to `ATen/cuda` as the following TODO comment in `CUDAGeneratorImpl.h` suggests:
```
// TODO: this file should be in ATen/cuda, not top level
```
Pull Request resolved: https://github.com/pytorch/pytorch/pull/70650
Reviewed By: jianyuh, xw285cornell
Differential Revision: D33414890
Pulled By: shintaro-iwasaki
fbshipit-source-id: 4ff839205f4e4ea4c8767f164d583eb7072f1b8b
Summary:
Move TH<C>GenerateByteType includes into torch/csrc (the only place they are used), and we can remove TH folder altogether!
The only thing left in THC are includes left for bc compatibility.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/69929
Reviewed By: mruberry
Differential Revision: D33133013
Pulled By: ngimel
fbshipit-source-id: 78c87cf93d2d641631b0f71051ace318bf4ec3c1
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/69041
`TH_CONCAT_{N}` is still being used by THP so I've moved that into
it's own header but all the compiled code is gone.
Test Plan: Imported from OSS
Reviewed By: anjali411
Differential Revision: D32872477
Pulled By: ngimel
fbshipit-source-id: 06c82d8f96dbcee0715be407c61dfc7d7e8be47a
Summary:
Follow up to https://github.com/pytorch/pytorch/issues/68095
This also changes the files from the ATen folder to include c10's `Export.h` instead since they can't ever be exporting `TORCH_PYTHON_API`.
cc pietern mrshenli pritamdamania87 zhaojuanmao satgera rohan-varma gqchen aazzolini osalpekar jiayisuse SciPioneer H-Huang
Pull Request resolved: https://github.com/pytorch/pytorch/pull/69585
Reviewed By: mrshenli
Differential Revision: D32958594
Pulled By: albanD
fbshipit-source-id: 1ec7ef63764573fa2b486928955e3a1172150061
Summary:
Also fixes the documentation failing to appear and adds a test to validate that op works with multiple devices properly.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/69640
Reviewed By: ngimel
Differential Revision: D32965391
Pulled By: mruberry
fbshipit-source-id: 4fe502809b353464da8edf62d92ca9863804f08e
Summary:
These seem to not be needed and cause ninja to rebuild the files at every build.
(There also is THCStorage.cu, but hopefully this will go away with https://github.com/pytorch/pytorch/issues/68556 )
Pull Request resolved: https://github.com/pytorch/pytorch/pull/69024
Reviewed By: soulitzer
Differential Revision: D32705309
Pulled By: ngimel
fbshipit-source-id: 5255297f213fdcf36e7203de7460a71291f8c9a0
Summary:
Patch bfloat16 support in NCCL, PR https://github.com/pytorch/pytorch/issues/63260 adds bfloat16 support but is
still not complete to enable bfloat16 for allreduce in end-to-end training.
This patch does the followings:
* fix minimum NCCL version from 2.9.7 to 2.10, NCCL adds bf16 support in
v2.10.3-1 (commit 7e51592)
* update bfloat16 datatype flag in `csrc/cuda/nccl.cpp` so that NCCL
operations like all reduce can use it
* enable unit tests for bfloat16 datatype if possible
cc pietern mrshenli pritamdamania87 zhaojuanmao satgera rohan-varma gqchen aazzolini osalpekar jiayisuse SciPioneer H-Huang
Pull Request resolved: https://github.com/pytorch/pytorch/pull/67843
Reviewed By: H-Huang
Differential Revision: D32248132
Pulled By: mrshenli
fbshipit-source-id: 081e96e725af3b933dd65ec157c5ad11c6873525
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/62030
Remove dtype tracking from Python Storage interface, remove all the different `<type>Storage` classes except for `ByteStorage`, and update serialization accordingly, while maintaining as much FC/BC as possible
Fixes https://github.com/pytorch/pytorch/issues/47442
* **THE SERIALIZATION FORMAT IS FULLY FC/BC.** We worked very hard to make sure this is the case. We will probably want to break FC at some point to make the serialization structure of tensors make more sense, but not today.
* There is now only a single torch.ByteStorage class. Methods like `Tensor.set_` no longer check that the dtype of storage is appropriate.
* As we no longer know what dtype of a storage is, we've **removed** the size method from Storage, replacing it with nbytes. This is to help catch otherwise silent errors where you confuse number of elements with number of bytes.
* `Storage._new_shared` takes a `nbytes` kwarg and will reject previous positional only calls. `Storage._new_with_file` and `_set_from_file` require explicit element size arguments.
* It's no longer possible to convert storages to different types using the float/double/etc methods. Instead, do the conversion using a tensor.
* It's no longer possible to allocate a typed storage directly using FloatStorage/DoubleStorage/etc constructors. Instead, construct a tensor and extract its storage. The classes still exist but they are used purely for unpickling.
* The preexisting serialization format stores dtype with storage, and in fact this dtype is used to determine the dtype of the tensor overall.
To accommodate this case, we introduce a new TypedStorage concept that exists only during unpickling time which is used to temporarily store the dtype so we can construct a tensor. **If you overrode the handling of pickling/unpickling, you MUST add handling for TypedStorage** or your serialization code will degrade to standard file-based serialization.
Original pull request: https://github.com/pytorch/pytorch/pull/59671
Reviewed By: soulitzer, ngimel
Differential Revision: D29466819
Pulled By: ezyang
fbshipit-source-id: 4a14e5d3c2b08e06e558683d97f7378a3180b00e
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/65746
This also removes the cudaHostAllocator field on THCState, since there
doesn't seem to be an API anywhere for customizing it.
Test Plan: Imported from OSS
Reviewed By: mrshenli
Differential Revision: D31236630
Pulled By: ngimel
fbshipit-source-id: 2a8e756222ae70565e77f8e7139d60ec5be32276
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/65610
- Replace HIP_PLATFORM_HCC with USE_ROCM
- Dont rely on CUDA_VERSION or HIP_VERSION and use USE_ROCM and ROCM_VERSION.
- In the next PR
- Will be removing the mapping from CUDA_VERSION to HIP_VERSION and CUDA to HIP in hipify.
- HIP_PLATFORM_HCC is deprecated, so will add HIP_PLATFORM_AMD to support HIP host code compilation on gcc.
cc jeffdaily sunway513 jithunnair-amd ROCmSupport amathews-amd
Reviewed By: jbschlosser
Differential Revision: D30909053
Pulled By: ezyang
fbshipit-source-id: 224a966ebf1aaec79beccbbd686fdf3d49267e06
