Defaulting HIP_PLATFORM to amd on Windows

[apwojcik]

The HIP SDK on Windows does not detect the platform correctly. Before calling find_package(hip ...), it needs to be defaulted to "amd".

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 92.04%. Comparing base (ab5da3c) to head (47308a2).
Report is 3 commits behind head on develop.

Additional details and impacted files

@@           Coverage Diff            @@
##           develop    #3466   +/-   ##
========================================
  Coverage    92.04%   92.04%           
========================================
  Files          506      506           
  Lines        20872    20872           
========================================
  Hits         19212    19212           
  Misses        1660     1660           

Flag	Coverage Δ
	92.04% <ø> (ø)

Flags with carried forward coverage won't be shown. Click here to find out more.

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

[migraphx-bot]

Test	Batch	Rate new 47308a	Rate old 32a84a	Diff	Compare
torchvision-resnet50	64	3,247.97	3,250.93	-0.09%	✅
torchvision-resnet50_fp16	64	6,992.03	6,986.29	0.08%	✅
torchvision-densenet121	32	2,428.93	2,433.76	-0.20%	✅
torchvision-densenet121_fp16	32	4,100.59	4,110.78	-0.25%	✅
torchvision-inceptionv3	32	1,637.75	1,635.58	0.13%	✅
torchvision-inceptionv3_fp16	32	2,742.12	2,737.29	0.18%	✅
cadene-inceptionv4	16	779.39	775.61	0.49%	✅
cadene-resnext64x4	16	808.62	808.58	0.00%	✅
slim-mobilenet	64	7,455.26	7,452.87	0.03%	✅
slim-nasnetalarge	64	208.16	208.18	-0.01%	✅
slim-resnet50v2	64	3,433.62	3,434.69	-0.03%	✅
bert-mrpc-onnx	8	1,148.05	1,150.44	-0.21%	✅
bert-mrpc-tf	1	311.79	311.37	0.13%	✅
pytorch-examples-wlang-gru	1	421.11	426.16	-1.19%	✅
pytorch-examples-wlang-lstm	1	393.60	386.06	1.95%	✅
torchvision-resnet50_1	1	769.50	816.95	-5.81%	🔴
cadene-dpn92_1	1	403.35	401.62	0.43%	✅
cadene-resnext101_1	1	381.50	380.63	0.23%	✅
onnx-taau-downsample	1	344.33	345.07	-0.21%	✅
dlrm-criteoterabyte	1	35.04	35.06	-0.07%	✅
dlrm-criteoterabyte_fp16	1	58.02	58.01	0.01%	✅
agentmodel	1	7,988.31	7,975.82	0.16%	✅
unet_fp16	2	57.90	58.07	-0.29%	✅
resnet50v1_fp16	1	1,031.52	921.74	11.91%	🔆
resnet50v1_int8	1	967.66	953.21	1.52%	✅
bert_base_cased_fp16	64	1,152.84	1,152.80	0.00%	✅
bert_large_uncased_fp16	32	355.91	355.84	0.02%	✅
bert_large_fp16	1	211.86	211.95	-0.05%	✅
distilgpt2_fp16	16	2,152.14	2,157.57	-0.25%	✅
yolov5s	1	537.38	528.76	1.63%	✅
tinyllama	1	43.40	43.37	0.07%	✅
vicuna-fastchat	1	169.02	168.85	0.10%	✅
whisper-tiny-encoder	1	418.95	418.42	0.13%	✅
whisper-tiny-decoder	1	432.18	425.66	1.53%	✅

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

[pfultz2]

This should be fixed in hip, this is not a migraphx bug. As a workaround you can just pass -DHIP_PLATFORM=amd to the cmake command until hip fixes it.

[pfultz2]

@causten Why was this merged? My comments was not even addressed. We shouldn't put hip fixes in out repo, they should be fixed in hip. And there is a simple workaround that can already be used. I think I am going to revert this.

[Dnonmi]

@causten Why was this merged? My comments was not even addressed. We shouldn't put hip fixes in out repo, they should be fixed in hip. And there is a simple workaround that can already be used. I think I am going to revert this.

My question is why create a delay for fixes that were already deemed fine for other branches on the same software stack?
Namely MIOpen: ROCm/MIOpen#3263
I guess even here..

With this lack of consistency, It's not assuring the hip side of things will be fixed soon so other people can move on with their jobs.