skip elimination when reshape_lazy

[weizhu12-amd]

Motivation

Prevent concat elimination by introducing a special operations checks. This issue caused YOLOv3 compilation to fail with: reshape_lazy on axis that is not packed

Technical Details

Concat elimination replaces an input's buffer with a non-packed slice of the super-buffer. reshape_lazy can only alias a packed input, so if the input feeds a reshape_lazy the elimination is illegal; force a copy instead.

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.

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

Test	Batch	New Rate (9f88e4)	Old Rate (3a503c)*	Diff	Status
torchvision-resnet50	64	3,326.70	3,264.92	1.89%	✅
torchvision-resnet50_fp16	64	7,881.40	7,548.67	4.41%	✅
torchvision-densenet121	32	2,491.47	2,483.99	0.30%	✅
torchvision-densenet121_fp16	32	5,035.82	5,004.24	0.63%	✅
torchvision-inceptionv3	32	2,076.12	2,058.51	0.86%	✅
torchvision-inceptionv3_fp16	32	4,497.17	4,416.99	1.82%	✅
cadene-inceptionv4	16	820.66	820.61	0.01%	✅
cadene-resnext64x4	16	782.90	782.78	0.02%	✅
slim-mobilenet	64	8,387.89	8,386.36	0.02%	✅
slim-nasnetalarge	64	228.79	228.86	-0.03%	✅
slim-resnet50v2	64	3,237.34	3,180.91	1.77%	✅
bert-mrpc-onnx	8	1,170.15	1,168.84	0.11%	✅
bert-mrpc-tf	1	497.32	498.63	-0.26%	✅
pytorch-examples-wlang-gru	1	530.44	473.35	12.06%	🔆
pytorch-examples-wlang-lstm	1	478.02	384.83	24.22%	🔆
torchvision-resnet50_1	1	1,051.96	1,046.63	0.51%	✅
cadene-dpn92_1	1	442.58	437.32	1.20%	✅
cadene-resnext101_1	1	364.61	365.89	-0.35%	✅
onnx-taau-downsample	1	845.10	844.09	0.12%	✅
dlrm-criteoterabyte	1	32.39	32.42	-0.08%	✅
dlrm-criteoterabyte_fp16	1	51.83	51.80	0.05%	✅
agentmodel	1	9,609.01	9,209.12	4.34%	✅
unet_fp16	2	58.71	58.80	-0.15%	✅
resnet50v1_fp16	1	1,372.07	1,366.11	0.44%	✅
resnet50v1_int8	1	1,827.65	1,883.96	-2.99%	✅
bert_base_cased_fp16	64	1,097.21	1,098.16	-0.09%	✅
bert_large_uncased_fp16	32	345.49	345.59	-0.03%	✅
bert_large_fp16	1	205.71	206.59	-0.43%	✅
distilgpt2_fp16	16	2,086.49	2,092.89	-0.31%	✅
yolov5s	1	558.36	558.33	0.01%	✅
tinyllama	1	45.82	45.83	-0.03%	✅
vicuna-fastchat	1	44.17	44.20	-0.07%	✅
whisper-tiny-encoder	1	412.59	411.87	0.18%	✅
whisper-tiny-decoder	1	408.12	408.48	-0.09%	✅
llama2_7b	1	20.82	20.84	-0.08%	✅
qwen1.5-7b	1	23.60	23.58	0.08%	✅
phi3-3.8b	1	26.72	26.72	0.02%	✅
llama3-8b	1	21.77	21.80	-0.14%	✅
whisper-large-encoder	1	10.18	10.18	-0.00%	✅
whisper-large-decoder	1	105.39	105.30	0.09%	✅
mistral-7b	1	23.75	23.78	-0.12%	✅
FLUX.1-schnell	1	778.79	755.22	3.12%	✅

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	✅	PASSED: MIGraphX meets tolerance
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
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

[TedThemistokleous]

Do you have the test case to verify this and the original model for us to look at? Where is that located? Feel free to reach out on teams

[weizhu12-amd]

Please download the YOLOv3 ONNX model from: https://github.com/onnx/models/raw/main/validated/vision/object_detection_segmentation/yolov3/model/yolov3-12.onnx,
Then run the following command:
migraphx-driver perf yolov3.onnx --input-dim @input_1 1 3 416 416 --input-dim @image_shape 1 2

The model runs successfully in my local environment with PRs #5103 and #5104 applied.

[codecov]

Codecov Report

❌ Patch coverage is 97.91667% with 1 line in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
src/eliminate_concat.cpp	97.92%	1 Missing ⚠️

Additional details and impacted files

@@           Coverage Diff            @@
##           develop    #5104   +/-   ##
========================================
  Coverage    93.26%   93.26%           
========================================
  Files          623      623           
  Lines        32969    33007   +38     
========================================
+ Hits         30747    30783   +36     
- Misses        2222     2224    +2     

Files with missing lines	Coverage Δ
src/eliminate_concat.cpp	96.75% <97.92%> (-0.72%)	⬇️

... and 1 file with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.