Remove rocmlir unsupported reduce types

[dhernandez0]

We are going to enable f16 for reduce in this rocmlir PR: ROCm/rocMLIR#1722

However, the code here assumes bf16 and fp8 are supported and that's not the case.

[umangyadav]

These would require tests in test/verify/ section

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 92.24%. Comparing base (720bc53) to head (bc8c548).

Additional details and impacted files

@@           Coverage Diff            @@
##           develop    #3766   +/-   ##
========================================
  Coverage    92.24%   92.24%           
========================================
  Files          519      519           
  Lines        22170    22170           
========================================
  Hits         20450    20450           
  Misses        1720     1720           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[migraphx-bot]

Test	Batch	Rate new bc8c54	Rate old 150105	Diff	Compare
torchvision-resnet50	64	3,254.65	3,254.33	0.01%	✅
torchvision-resnet50_fp16	64	6,927.69	6,927.70	-0.00%	✅
torchvision-densenet121	32	2,456.27	2,454.03	0.09%	✅
torchvision-densenet121_fp16	32	4,182.41	4,166.00	0.39%	✅
torchvision-inceptionv3	32	1,630.77	1,630.12	0.04%	✅
torchvision-inceptionv3_fp16	32	2,718.89	2,716.54	0.09%	✅
cadene-inceptionv4	16	763.85	763.13	0.09%	✅
cadene-resnext64x4	16	813.04	813.13	-0.01%	✅
slim-mobilenet	64	7,458.03	7,460.93	-0.04%	✅
slim-nasnetalarge	64	208.67	208.67	-0.00%	✅
slim-resnet50v2	64	3,446.42	3,446.18	0.01%	✅
bert-mrpc-onnx	8	1,150.08	1,146.26	0.33%	✅
bert-mrpc-tf	1	485.92	481.46	0.93%	✅
pytorch-examples-wlang-gru	1	489.18	470.60	3.95%	🔆
pytorch-examples-wlang-lstm	1	452.12	443.44	1.96%	✅
torchvision-resnet50_1	1	804.17	804.59	-0.05%	✅
cadene-dpn92_1	1	427.59	430.86	-0.76%	✅
cadene-resnext101_1	1	385.88	386.12	-0.06%	✅
onnx-taau-downsample	1	372.90	372.92	-0.01%	✅
dlrm-criteoterabyte	1	33.33	33.33	-0.02%	✅
dlrm-criteoterabyte_fp16	1	52.63	52.63	0.00%	✅
agentmodel	1	8,421.69	8,589.15	-1.95%	✅
unet_fp16	2	58.49	58.37	0.21%	✅
resnet50v1_fp16	1	1,004.27	1,036.76	-3.13%	🔴
resnet50v1_int8	1	1,027.73	1,039.07	-1.09%	✅
bert_base_cased_fp16	64	1,181.37	1,182.01	-0.05%	✅
bert_large_uncased_fp16	32	365.26	365.27	-0.00%	✅
bert_large_fp16	1	199.01	202.67	-1.81%	✅
distilgpt2_fp16	16	2,227.57	2,227.86	-0.01%	✅
yolov5s	1	530.50	523.22	1.39%	✅
tinyllama	1	43.64	43.57	0.16%	✅
vicuna-fastchat	1	175.55	173.92	0.94%	✅
whisper-tiny-encoder	1	416.90	418.33	-0.34%	✅
whisper-tiny-decoder	1	433.06	430.88	0.51%	✅

This build is not recommended to merge 🔴

[migraphx-bot]

     ✅ 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

     ✅ torchvision-resnet50_1: PASSED: MIGraphX meets tolerance

     ✅ cadene-dpn92_1: PASSED: MIGraphX meets tolerance

     ✅ cadene-resnext101_1: 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

[dhernandez0]

By the way, reduce f16 should not be enabled for navi. As it does not support atomic_add with packed_f16, so it would be slow. Leaving this comment here for whoever is assigned the PR. @causten

[CharlieL7]

What's the status on this PR? Is there additional work needed or is the removal of the allowed types the only thing that needs review?

[dhernandez0]

What's the status on this PR? Is there additional work needed or is the removal of the allowed types the only thing that needs review?

I thought this was all that needed to be done, so I created the PR. However, we need some tests and make sure reduce f16 is not enabled for Navi (except for gfx12). I think @causten will assign somebody from migraphx team to finish this.