if miopen/hipblas/rocblas are not enabled, send gemms to rocmlir

[bdevorem]

Motivation

When external libraries are not enabled, gemm ops need to be sent to rocmlir so they are lowered properly.

Technical Details

Just adds a check for the libraries. If not enabled, then send relevant ops to rocmlir

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 updates the GPU MLIR fusion routing so that when external GPU libraries (MIOpen / rocBLAS / hipBLASLt) are disabled at build time, convolution and GEMM-like ops are routed through rocMLIR to ensure they are still lowered properly.

Changes:

• Extend fuse_mlir’s mode selection to force rocMLIR routing for convolution when MIOpen is disabled, and for dot/fused_dot when GEMM libraries are disabled.
• Update the dependency/source specification for rocMLIR in requirements.txt.
• Add a changelog entry documenting the behavior change.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 1 comment.

File	Description
src/targets/gpu/fuse_mlir.cpp	Forces rocMLIR routing for conv/dot under specific “library disabled” build configurations.
requirements.txt	Changes how rocMLIR is sourced for builds.
CHANGELOG.md	Documents the new routing behavior in the changelog.

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

Test	Batch	New Rate (5a7f29)	Old Rate (55ca93)	Diff	Status
torchvision-resnet50	64	3,151.33	3,169.20	-0.56%	✅
torchvision-resnet50_fp16	64	6,584.57	6,721.88	-2.04%	✅
torchvision-densenet121	32	2,569.27	2,712.99	-5.30%	🔴
torchvision-densenet121_fp16	32	2,479.06	4,578.90	-45.86%	🔴
torchvision-inceptionv3	32	1,771.36	1,805.02	-1.86%	✅
torchvision-inceptionv3_fp16	32	1,279.87	2,855.31	-55.18%	🔴
cadene-inceptionv4	16	515.40	824.13	-37.46%	🔴
cadene-resnext64x4	16	782.46	785.46	-0.38%	✅
slim-mobilenet	64	6,938.26	8,444.46	-17.84%	🔴
slim-nasnetalarge	64	145.55	229.50	-36.58%	🔴
slim-resnet50v2	64	3,093.28	3,185.35	-2.89%	✅
bert-mrpc-onnx	8	1,169.12	1,173.56	-0.38%	✅
bert-mrpc-tf	1	488.24	484.20	0.83%	✅
pytorch-examples-wlang-gru	1	542.20	472.27	14.81%	🔆
pytorch-examples-wlang-lstm	1	389.61	386.76	0.74%	✅
torchvision-resnet50_1	1	626.05	754.58	-17.03%	🔴
cadene-dpn92_1	1	315.96	444.58	-28.93%	🔴
cadene-resnext101_1	1	216.11	366.32	-41.00%	🔴
onnx-taau-downsample	1	87.57	402.99	-78.27%	🔴
dlrm-criteoterabyte	1	0.87	32.58	-97.32%	🔴
dlrm-criteoterabyte_fp16	1	35.24	52.65	-33.06%	🔴
agentmodel	1	9,257.72	7,784.58	18.92%	🔆
unet_fp16	2	5.76	57.48	-89.98%	🔴
resnet50v1_fp16	1	20.08	932.89	-97.85%	🔴
resnet50v1_int8	1	22.80	934.74	-97.56%	🔴
bert_base_cased_fp16	64	344.95	1,105.09	-68.79%	🔴
bert_large_uncased_fp16	32	150.02	347.51	-56.83%	🔴
bert_large_fp16	1	38.08	205.61	-81.48%	🔴
distilgpt2_fp16	16	1,243.42	2,104.89	-40.93%	🔴
yolov5s	1	69.55	559.63	-87.57%	🔴
tinyllama	1	13.65	46.13	-70.41%	🔴
vicuna-fastchat	1	22.20	44.16	-49.73%	🔴
whisper-tiny-encoder	1	193.75	415.45	-53.37%	🔴
whisper-tiny-decoder	1	61.27	412.20	-85.13%	🔴
llama2_7b	1	1.51	21.00	-92.79%	🔴
qwen1.5-7b	1	5.01	23.73	-78.90%	🔴
phi3-3.8b	1	19.42	26.83	-27.63%	🔴
llama3-8b	1	15.91	21.87	-27.28%	🔴
whisper-large-encoder	1	10.14	10.23	-0.93%	✅
whisper-large-decoder	1	14.58	104.89	-86.10%	🔴
mistral-7b	1	4.97	23.89	-79.21%	🔴
FLUX.1-schnell	1	294.19	760.54	-61.32%	🔴

Regressions detected 🔴

[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	✅	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