Reject invalid pointwise modules

[justinrosner]

Motivation

Some rocMLIR tuning configurations require splitting fused convolution/GEMM and pointwise modules. The resulting pointwise partition may contain unsupported shape operations, causing compilation failures.

Technical Details

• Validate split pointwise modules before aliasing operations are removed.
• Allow pointwise operations, literals, parameters, and returns.
• Report the first unsupported instruction in the error.

Changelog Category

Add a CHANGELOG.md entry for any option other than Not Applicable

• • Added: New functionality.
• • Changed: Changes to existing functionality.
• • Removed: Functionality or support that has been removed. (Compared to a previous release)
• • Optimized: Component performance that has been optimized or improved.
• • Resolved Issues: Known issues from a previous version that have been resolved.
• • Not Applicable: This PR is not to be included in the changelog.

Follow the LLVM AI Tool Use Policy for contributions using AI.

[Copilot]

Pull request overview

This PR adds an explicit validation step for rocMLIR-generated split pointwise modules on the GPU path, rejecting modules that contain unsupported non-pointwise (notably shape/aliasing) instructions and surfacing a clear error that points at the first offending instruction.

Changes:

• Introduces migraphx::gpu::validate_pointwise_module(const module&) to detect non-pointwise instructions (other than literals/params/returns) and throw with an actionable error message.
• Hooks validation into the MLIR pointwise compilation flow before aliasing shape ops are removed, so invalid split modules fail early with a clear diagnostic.
• Adds GPU tests that verify acceptance of pointwise + literal/param/return modules and rejection of representative shape operations.

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated no comments.

File	Description
test/gpu/mlir_pointwise.cpp	Adds GPU-side unit tests covering valid pointwise modules and rejection of shape ops.
src/targets/gpu/jit/mlir.cpp	Implements validate_pointwise_module and calls it prior to pointwise module creation/alias cleanup.
src/targets/gpu/include/migraphx/gpu/mlir.hpp	Exposes validate_pointwise_module as a GPU public API declaration.
CHANGELOG.md	Documents the improved validation/error reporting for invalid split pointwise modules.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[gh-app-migraphx-bot-pr-write]

Test	Batch	New Rate (858b0b)	Old Rate (3a503c)*	Diff	Status
torchvision-resnet50	64	3,317.27	3,264.92	1.60%	✅
torchvision-resnet50_fp16	64	7,871.85	7,548.67	4.28%	✅
torchvision-densenet121	32	2,488.05	2,483.99	0.16%	✅
torchvision-densenet121_fp16	32	5,022.11	5,004.24	0.36%	✅
torchvision-inceptionv3	32	2,068.55	2,058.51	0.49%	✅
torchvision-inceptionv3_fp16	32	4,455.34	4,416.99	0.87%	✅
cadene-inceptionv4	16	816.96	820.61	-0.44%	✅
cadene-resnext64x4	16	781.80	782.78	-0.12%	✅
slim-mobilenet	64	8,375.14	8,386.36	-0.13%	✅
slim-nasnetalarge	64	228.22	228.86	-0.28%	✅
slim-resnet50v2	64	3,237.49	3,180.91	1.78%	✅
bert-mrpc-onnx	8	1,169.58	1,168.84	0.06%	✅
bert-mrpc-tf	1	499.12	498.63	0.10%	✅
pytorch-examples-wlang-gru	1	484.01	473.35	2.25%	✅
pytorch-examples-wlang-lstm	1	404.25	384.83	5.05%	🔆
torchvision-resnet50_1	1	1,052.60	1,046.63	0.57%	✅
cadene-dpn92_1	1	453.21	437.32	3.63%	✅
cadene-resnext101_1	1	365.85	365.89	-0.01%	✅
onnx-taau-downsample	1	844.19	844.09	0.01%	✅
dlrm-criteoterabyte	1	32.26	32.42	-0.48%	✅
dlrm-criteoterabyte_fp16	1	51.63	51.80	-0.33%	✅
agentmodel	1	14,485.43	9,209.12	57.29%	🔆
unet_fp16	2	58.29	58.80	-0.87%	✅
resnet50v1_fp16	1	1,440.92	1,366.11	5.48%	🔆
resnet50v1_int8	1	1,794.24	1,883.96	-4.76%	✅
bert_base_cased_fp16	64	1,097.35	1,098.16	-0.07%	✅
bert_large_uncased_fp16	32	345.65	345.59	0.02%	✅
bert_large_fp16	1	206.75	206.59	0.08%	✅
distilgpt2_fp16	16	2,092.85	2,092.89	-0.00%	✅
yolov5s	1	566.25	558.33	1.42%	✅
tinyllama	1	45.78	45.83	-0.12%	✅
vicuna-fastchat	1	44.24	44.20	0.10%	✅
whisper-tiny-encoder	1	412.21	411.87	0.08%	✅
whisper-tiny-decoder	1	410.04	408.48	0.38%	✅
llama2_7b	1	20.85	20.84	0.08%	✅
qwen1.5-7b	1	23.66	23.58	0.32%	✅
phi3-3.8b	1	26.67	26.72	-0.16%	✅
llama3-8b	1	21.75	21.80	-0.23%	✅
whisper-large-encoder	1	10.16	10.18	-0.13%	✅
whisper-large-decoder	1	106.07	105.30	0.73%	✅
mistral-7b	1	23.78	23.78	0.03%	✅
FLUX.1-schnell	1	774.14	755.22	2.51%	✅

Check flagged results 🔆

* No develop baseline was found for this PR's branch point; compared against the latest available develop run instead.

[gh-app-migraphx-bot-pr-write]

Test	Status	Result
bert-mrpc-onnx	✅	PASSED: MIGraphX meets tolerance
bert-mrpc-tf	❌	ERROR - check error output traceback Traceback (most recent call last): File "/src/AMDMIGraphX/tools/accuracy/accuracy_checker.py", line 377, in main() File "/src/AMDMIGraphX/tools/accuracy/accuracy_checker.py", line 313, in main import tensorflow as tf File "/usr/local/lib/python3.12/dist-packages/tensorflow/init.py", line 40, in from tensorflow.python import pywrap_tensorflow as _pywrap_tensorflow # pylint: disable=unused-import ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/usr/local/lib/python3.12/dist-packages/tensorflow/python/pywrap_tensorflow.py", line 37, in self_check.preload_check() File "/usr/local/lib/python3.12/dist-packages/tensorflow/python/platform/self_check.py", line 63, in preload_check from tensorflow.python.platform import _pywrap_cpu_feature_guard ImportError: libnuma.so.1: cannot open shared object file: No such file or directory
pytorch-examples-wlang-gru	✅	PASSED: MIGraphX meets tolerance
pytorch-examples-wlang-lstm	✅	PASSED: MIGraphX meets tolerance
dlrm-criteoterabyte	✅	PASSED: MIGraphX meets tolerance
agentmodel	✅	PASSED: MIGraphX meets tolerance
unet	✅	PASSED: MIGraphX meets tolerance
resnet50v1	✅	PASSED: MIGraphX meets tolerance
bert_base_cased_fp16	✅	PASSED: MIGraphX meets tolerance
bert_large_uncased_fp16	🔴	FAILED: MIGraphX is not within tolerance - check verbose output
bert_large	✅	PASSED: MIGraphX meets tolerance
yolov5s	✅	PASSED: MIGraphX meets tolerance
tinyllama	✅	PASSED: MIGraphX meets tolerance
vicuna-fastchat	✅	PASSED: MIGraphX meets tolerance
whisper-tiny-encoder	✅	PASSED: MIGraphX meets tolerance
whisper-tiny-decoder	✅	PASSED: MIGraphX meets tolerance
distilgpt2_fp16	🔴	FAILED: MIGraphX is not within tolerance - check verbose output
llama2_7b	✅	PASSED: MIGraphX meets tolerance
qwen1.5-7b	✅	PASSED: MIGraphX meets tolerance
phi3-3.8b	✅	PASSED: MIGraphX meets tolerance
llama3-8b	✅	PASSED: MIGraphX meets tolerance
whisper-large-encoder	✅	PASSED: MIGraphX meets tolerance
whisper-large-decoder	✅	PASSED: MIGraphX meets tolerance
mistral-7b	✅	PASSED: MIGraphX meets tolerance
FLUX.1-schnell	✅	PASSED: MIGraphX meets tolerance

[justinrosner]

@pfultz2 are you able to merge this into develop when all the CI checks finish? It looks like I don't have write access to MIGraphX.