[AIMIGRAPHX-1229] Support ROCm10 TheRock packaging with alternative GPU dependencies

[kentqian]

Motivation

After reviewing the ROCm 7.13 packaging work in bumpci7_13 and comparing it with the ROCm 10 multi-arch package layout, we need a dependency model that supports both repository formats.

ROCm 7.13/7.14 repositories use GPU-family-suffixed packages, while ROCm 10 multi-arch repositories provide raw GPU-target packages together with architecture-neutral meta packages.

Generating one MIGraphX package per GPU architecture would duplicate the same MIGraphX files and introduce package conflicts. Weak dependencies such as Recommends or Suggests also do not guarantee that the required GPU library is installed.

This PR keeps one fat MIGraphX package set per build and expresses the GPU library requirement as a hard alternative dependency. An already installed device package satisfies the dependency, while the meta package remains a safe fallback.

Technical Details

• Derive package architectures from GPU_TARGETS, removing feature suffixes and duplicates.
• Keep MIGRAPHX_THEROCK_GPU_ARCH as an override for legacy GPU-family package names.
• Generate hard alternative dependencies for DNN and BLAS:
  • DEB: meta | arch1 | arch2
  • RPM: (arch1 or arch2 or meta)
• Fall back to the meta package when RPM rich dependencies are unavailable.
• Leave host dependencies and Python wheel packaging unchanged.

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]

Copilot encountered an error and was unable to review this pull request. You can try again by re-requesting a review.

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

Test	Batch	New Rate (23efe1)	Old Rate (3a503c)*	Diff	Status
torchvision-resnet50	64	1,750.33	3,264.92	-46.39%	🔴
torchvision-resnet50_fp16	64	3,382.13	7,548.67	-55.20%	🔴
torchvision-densenet121	32	1,775.36	2,483.99	-28.53%	🔴
torchvision-densenet121_fp16	32	1,597.33	5,004.24	-68.08%	🔴
torchvision-inceptionv3	32	1,173.19	2,058.51	-43.01%	🔴
torchvision-inceptionv3_fp16	32	4,211.97	4,416.99	-4.64%	✅
cadene-inceptionv4	16	775.59	820.61	-5.49%	🔴
cadene-resnext64x4	16	470.78	782.78	-39.86%	🔴
slim-mobilenet	64	7,202.73	8,386.36	-14.11%	🔴
slim-nasnetalarge	64	nan	228.86	nan	❌
slim-resnet50v2	64	1,461.49	3,180.91	-54.05%	🔴
bert-mrpc-onnx	8	1,007.46	1,168.84	-13.81%	🔴
bert-mrpc-tf	1	491.66	498.63	-1.40%	✅
pytorch-examples-wlang-gru	1	521.40	473.35	10.15%	🔆
pytorch-examples-wlang-lstm	1	386.28	384.83	0.38%	✅
torchvision-resnet50_1	1	1,050.50	1,046.63	0.37%	✅
cadene-dpn92_1	1	444.96	437.32	1.75%	✅
cadene-resnext101_1	1	282.74	365.89	-22.73%	🔴
onnx-taau-downsample	1	820.50	844.09	-2.79%	✅
dlrm-criteoterabyte	1	nan	32.42	nan	❌
dlrm-criteoterabyte_fp16	1	28.81	51.80	-44.39%	🔴
agentmodel	1	6,636.36	9,209.12	-27.94%	🔴
unet_fp16	2	14.95	58.80	-74.57%	🔴
resnet50v1_fp16	1	1,361.87	1,366.11	-0.31%	✅
resnet50v1_int8	1	263.55	1,883.96	-86.01%	🔴
bert_base_cased_fp16	64	534.96	1,098.16	-51.29%	🔴
bert_large_uncased_fp16	32	339.34	345.59	-1.81%	✅
bert_large_fp16	1	208.16	206.59	0.76%	✅
distilgpt2_fp16	16	1,429.55	2,092.89	-31.69%	🔴
yolov5s	1	443.17	558.33	-20.63%	🔴
tinyllama	1	21.36	45.83	-53.39%	🔴
vicuna-fastchat	1	44.12	44.20	-0.17%	✅
whisper-tiny-encoder	1	351.17	411.87	-14.74%	🔴
whisper-tiny-decoder	1	199.75	408.48	-51.10%	🔴
llama2_7b	1	20.85	20.84	0.03%	✅
qwen1.5-7b	1	14.61	23.58	-38.03%	🔴
phi3-3.8b	1	26.72	26.72	0.01%	✅
llama3-8b	1	19.87	21.80	-8.87%	🔴
whisper-large-encoder	1	10.18	10.18	0.02%	✅
whisper-large-decoder	1	105.23	105.30	-0.07%	✅
mistral-7b	1	23.70	23.78	-0.30%	✅
FLUX.1-schnell	1	467.79	755.22	-38.06%	🔴

Regressions detected 🔴

* 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
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	🔴	FAILED: MIGraphX is not within tolerance - check verbose output
whisper-large-decoder	✅	PASSED: MIGraphX meets tolerance
mistral-7b	✅	PASSED: MIGraphX meets tolerance
FLUX.1-schnell	✅	PASSED: MIGraphX meets tolerance