Summary
A set of cudagraph-related tests across dynamo, inductor, and foreach test suites are blocked on XPU because the xpugraph feature is not yet implemented or equivalent to CUDA graph support. This is a feature gap tracking issue listing all affected test cases that need xpugraph to be enabled before they can pass.
Type
- Category: feature-gap
- Platform: xpu
- Related Components: xpugraph, test/dynamo/test_backends.py, test/dynamo/test_cudagraphs.py, test/dynamo/test_cudagraphs_expandable_segments.py, test/dynamo/test_structured_trace.py, test/test_foreach.py, test/inductor/test_cuda_repro.py
Objective
Implement or enable xpugraph support to unblock the listed cudagraph test cases on XPU.
Current Status
test/dynamo/test_backends.py | TestOptimizationsCUDA | test_aot_cudagraphs_cuda
test/dynamo/test_cudagraphs.py | TestAotCudagraphs | test_basic
test/dynamo/test_cudagraphs.py | TestAotCudagraphs | test_dead_fill
test/dynamo/test_cudagraphs.py | TestAotCudagraphs | test_dtoh
test/dynamo/test_cudagraphs.py | TestAotCudagraphs | test_factory
test/dynamo/test_cudagraphs.py | TestAotCudagraphs | test_htod
test/dynamo/test_cudagraphs.py | TestAotCudagraphs | test_mutate_constant
test/dynamo/test_cudagraphs.py | TestAotCudagraphs | test_mutate_input
test/dynamo/test_cudagraphs.py | TestAotCudagraphs | test_mutated_metadata
test/dynamo/test_cudagraphs_expandable_segments.py | TestAotCudagraphs | test_basic
test/dynamo/test_cudagraphs_expandable_segments.py | TestAotCudagraphs | test_dead_fill
test/dynamo/test_cudagraphs_expandable_segments.py | TestAotCudagraphs | test_dtoh
test/dynamo/test_cudagraphs_expandable_segments.py | TestAotCudagraphs | test_factory
test/dynamo/test_cudagraphs_expandable_segments.py | TestAotCudagraphs | test_htod
test/dynamo/test_cudagraphs_expandable_segments.py | TestAotCudagraphs | test_mutate_constant
test/dynamo/test_cudagraphs_expandable_segments.py | TestAotCudagraphs | test_mutate_input
test/dynamo/test_cudagraphs_expandable_segments.py | TestAotCudagraphs | test_mutated_metadata
test/dynamo/test_structured_trace.py | StructuredTraceTest | test_cudagraphs
test/test_foreach.py | TestForeachCUDA | test_big_num_tensors__foreach_max_use_cuda_graph_True_w_empty_False_cuda_float32
test/test_foreach.py | TestForeachCUDA | test_big_num_tensors__foreach_max_use_cuda_graph_True_w_empty_False_cuda_float64
test/test_foreach.py | TestForeachCUDA | test_big_num_tensors__foreach_max_use_cuda_graph_True_w_empty_True_cuda_float32
test/test_foreach.py | TestForeachCUDA | test_big_num_tensors__foreach_max_use_cuda_graph_True_w_empty_True_cuda_float64
test/test_foreach.py | TestForeachCUDA | test_big_num_tensors__foreach_norm_use_cuda_graph_True_w_empty_False_cuda_float32
test/test_foreach.py | TestForeachCUDA | test_big_num_tensors__foreach_norm_use_cuda_graph_True_w_empty_False_cuda_float64
test/test_foreach.py | TestForeachCUDA | test_big_num_tensors__foreach_norm_use_cuda_graph_True_w_empty_True_cuda_float32
test/test_foreach.py | TestForeachCUDA | test_big_num_tensors__foreach_norm_use_cuda_graph_True_w_empty_True_cuda_float64
test/inductor/test_cuda_repro.py | CudaReproTests | test_cpu_index
Context
Versions: main. Tests are blocked by xpugraph feature gap, not by a runtime error.
Root Cause Analysis
XPU does not have an equivalent implementation of CUDA graph support (xpugraph). The listed tests across dynamo, inductor, and foreach test suites require cudagraph/xpugraph functionality that has not yet been implemented for XPU. This is a feature gap, not a runtime bug.
Proposed Fix Strategy
Implement xpugraph support for XPU, analogous to CUDA graph support. This would require significant new feature development in torch-xpu-ops and/or pytorch for XPU graph capture/replay infrastructure.
Action Items
Original Issue
Original issue body
🐛 Describe the bug
The following cudagraph cases are blocked by xpugraph
| test/dynamo/test_backends.py |
TestOptimizationsCUDA |
test_aot_cudagraphs_cuda |
| test/dynamo/test_cudagraphs.py |
TestAotCudagraphs |
test_basic |
| test/dynamo/test_cudagraphs.py |
TestAotCudagraphs |
test_dead_fill |
| test/dynamo/test_cudagraphs.py |
TestAotCudagraphs |
test_dtoh |
| test/dynamo/test_cudagraphs.py |
TestAotCudagraphs |
test_factory |
| test/dynamo/test_cudagraphs.py |
TestAotCudagraphs |
test_htod |
| test/dynamo/test_cudagraphs.py |
TestAotCudagraphs |
test_mutate_constant |
| test/dynamo/test_cudagraphs.py |
TestAotCudagraphs |
test_mutate_input |
| test/dynamo/test_cudagraphs.py |
TestAotCudagraphs |
test_mutated_metadata |
| test/dynamo/test_cudagraphs_expandable_segments.py |
TestAotCudagraphs |
test_basic |
| test/dynamo/test_cudagraphs_expandable_segments.py |
TestAotCudagraphs |
test_dead_fill |
| test/dynamo/test_cudagraphs_expandable_segments.py |
TestAotCudagraphs |
test_dtoh |
| test/dynamo/test_cudagraphs_expandable_segments.py |
TestAotCudagraphs |
test_factory |
| test/dynamo/test_cudagraphs_expandable_segments.py |
TestAotCudagraphs |
test_htod |
| test/dynamo/test_cudagraphs_expandable_segments.py |
TestAotCudagraphs |
test_mutate_constant |
| test/dynamo/test_cudagraphs_expandable_segments.py |
TestAotCudagraphs |
test_mutate_input |
| test/dynamo/test_cudagraphs_expandable_segments.py |
TestAotCudagraphs |
test_mutated_metadata |
| test/dynamo/test_structured_trace.py |
StructuredTraceTest |
test_cudagraphs |
| test/test_foreach.py |
TestForeachCUDA |
test_big_num_tensors__foreach_max_use_cuda_graph_True_w_empty_False_cuda_float32 |
| test/test_foreach.py |
TestForeachCUDA |
test_big_num_tensors__foreach_max_use_cuda_graph_True_w_empty_False_cuda_float64 |
| test/test_foreach.py |
TestForeachCUDA |
test_big_num_tensors__foreach_max_use_cuda_graph_True_w_empty_True_cuda_float32 |
| test/test_foreach.py |
TestForeachCUDA |
test_big_num_tensors__foreach_max_use_cuda_graph_True_w_empty_True_cuda_float64 |
| test/test_foreach.py |
TestForeachCUDA |
test_big_num_tensors__foreach_norm_use_cuda_graph_True_w_empty_False_cuda_float32 |
| test/test_foreach.py |
TestForeachCUDA |
test_big_num_tensors__foreach_norm_use_cuda_graph_True_w_empty_False_cuda_float64 |
| test/test_foreach.py |
TestForeachCUDA |
test_big_num_tensors__foreach_norm_use_cuda_graph_True_w_empty_True_cuda_float32 |
| test/test_foreach.py |
TestForeachCUDA |
test_big_num_tensors__foreach_norm_use_cuda_graph_True_w_empty_True_cuda_float64 |
| test/inductor/test_cuda_repro.py |
CudaReproTests |
test_cpu_index |
Versions
main
triage log
**Verdict:** NEEDS_HUMAN
**Reason:** Feature gap issue — xpugraph is not yet implemented; requires substantial new feature development, not a single fixable bug.
Root Cause: XPU does not have an equivalent implementation of CUDA graph support (xpugraph). The listed tests across dynamo, inductor, and foreach test suites require cudagraph/xpugraph functionality that has not yet been implemented for XPU. This is a feature gap, not a runtime bug.
Fix Strategy: Implement xpugraph support for XPU, analogous to CUDA graph support. This would require significant new feature development in torch-xpu-ops and/or pytorch for XPU graph capture/replay infrastructure.
Tokens: model: claude-sonnet-4 | tokens: 37.6K | in: 3 | out: 383 | cache_read: 34.3K | cache_write: 2.9K | cost: $0.0270
Log: agent-issue-3080-triaging-20260511-232056.log
Summary
A set of cudagraph-related tests across dynamo, inductor, and foreach test suites are blocked on XPU because the xpugraph feature is not yet implemented or equivalent to CUDA graph support. This is a feature gap tracking issue listing all affected test cases that need xpugraph to be enabled before they can pass.
Type
Objective
Implement or enable xpugraph support to unblock the listed cudagraph test cases on XPU.
Current Status
test/dynamo/test_backends.py | TestOptimizationsCUDA | test_aot_cudagraphs_cuda
test/dynamo/test_cudagraphs.py | TestAotCudagraphs | test_basic
test/dynamo/test_cudagraphs.py | TestAotCudagraphs | test_dead_fill
test/dynamo/test_cudagraphs.py | TestAotCudagraphs | test_dtoh
test/dynamo/test_cudagraphs.py | TestAotCudagraphs | test_factory
test/dynamo/test_cudagraphs.py | TestAotCudagraphs | test_htod
test/dynamo/test_cudagraphs.py | TestAotCudagraphs | test_mutate_constant
test/dynamo/test_cudagraphs.py | TestAotCudagraphs | test_mutate_input
test/dynamo/test_cudagraphs.py | TestAotCudagraphs | test_mutated_metadata
test/dynamo/test_cudagraphs_expandable_segments.py | TestAotCudagraphs | test_basic
test/dynamo/test_cudagraphs_expandable_segments.py | TestAotCudagraphs | test_dead_fill
test/dynamo/test_cudagraphs_expandable_segments.py | TestAotCudagraphs | test_dtoh
test/dynamo/test_cudagraphs_expandable_segments.py | TestAotCudagraphs | test_factory
test/dynamo/test_cudagraphs_expandable_segments.py | TestAotCudagraphs | test_htod
test/dynamo/test_cudagraphs_expandable_segments.py | TestAotCudagraphs | test_mutate_constant
test/dynamo/test_cudagraphs_expandable_segments.py | TestAotCudagraphs | test_mutate_input
test/dynamo/test_cudagraphs_expandable_segments.py | TestAotCudagraphs | test_mutated_metadata
test/dynamo/test_structured_trace.py | StructuredTraceTest | test_cudagraphs
test/test_foreach.py | TestForeachCUDA | test_big_num_tensors__foreach_max_use_cuda_graph_True_w_empty_False_cuda_float32
test/test_foreach.py | TestForeachCUDA | test_big_num_tensors__foreach_max_use_cuda_graph_True_w_empty_False_cuda_float64
test/test_foreach.py | TestForeachCUDA | test_big_num_tensors__foreach_max_use_cuda_graph_True_w_empty_True_cuda_float32
test/test_foreach.py | TestForeachCUDA | test_big_num_tensors__foreach_max_use_cuda_graph_True_w_empty_True_cuda_float64
test/test_foreach.py | TestForeachCUDA | test_big_num_tensors__foreach_norm_use_cuda_graph_True_w_empty_False_cuda_float32
test/test_foreach.py | TestForeachCUDA | test_big_num_tensors__foreach_norm_use_cuda_graph_True_w_empty_False_cuda_float64
test/test_foreach.py | TestForeachCUDA | test_big_num_tensors__foreach_norm_use_cuda_graph_True_w_empty_True_cuda_float32
test/test_foreach.py | TestForeachCUDA | test_big_num_tensors__foreach_norm_use_cuda_graph_True_w_empty_True_cuda_float64
test/inductor/test_cuda_repro.py | CudaReproTests | test_cpu_index
Context
Versions: main. Tests are blocked by xpugraph feature gap, not by a runtime error.
Root Cause Analysis
XPU does not have an equivalent implementation of CUDA graph support (xpugraph). The listed tests across dynamo, inductor, and foreach test suites require cudagraph/xpugraph functionality that has not yet been implemented for XPU. This is a feature gap, not a runtime bug.
Proposed Fix Strategy
Implement xpugraph support for XPU, analogous to CUDA graph support. This would require significant new feature development in torch-xpu-ops and/or pytorch for XPU graph capture/replay infrastructure.
Action Items
🔍 Issue formatted (Discovery Agent)
Discovery log
[2026-05-10 23:15:53]
Summary: A set of cudagraph-related tests across dynamo, inductor, and foreach test suites are blocked on XPU because the xpugraph feature is not yet implemented or equivalent to CUDA graph support. This is a feature gap tracking issue listing all affected test cases that need xpugraph to be enabled before they can pass.
Failed tests: N/A
Dependency: N/A
Commit scope: N/A
Original Issue
Original issue body
🐛 Describe the bug
The following cudagraph cases are blocked by xpugraph
Versions
main
triage log
**Verdict:** NEEDS_HUMAN **Reason:** Feature gap issue — xpugraph is not yet implemented; requires substantial new feature development, not a single fixable bug.Root Cause: XPU does not have an equivalent implementation of CUDA graph support (xpugraph). The listed tests across dynamo, inductor, and foreach test suites require cudagraph/xpugraph functionality that has not yet been implemented for XPU. This is a feature gap, not a runtime bug.
Fix Strategy: Implement xpugraph support for XPU, analogous to CUDA graph support. This would require significant new feature development in torch-xpu-ops and/or pytorch for XPU graph capture/replay infrastructure.
Tokens: model: claude-sonnet-4 | tokens: 37.6K | in: 3 | out: 383 | cache_read: 34.3K | cache_write: 2.9K | cost: $0.0270
Log:
agent-issue-3080-triaging-20260511-232056.log