add 'make generate' script for Windows

[apwojcik]

The PR from the series to enable MIGraphX on Windows.

[pfultz2]

This should be either a python or cmake script so it uses the same script for both windows and linux.

[codecov]

Codecov Report

Merging #1893 (ae2a183) into develop (697709a) will not change coverage.
The diff coverage is n/a.

❗ Current head ae2a183 differs from pull request most recent head 82f6680. Consider uploading reports for the commit 82f6680 to get more accurate results

@@           Coverage Diff            @@
##           develop    #1893   +/-   ##
========================================
  Coverage    91.37%   91.37%           
========================================
  Files          419      419           
  Lines        15544    15544           
========================================
  Hits         14204    14204           
  Misses        1340     1340           

[migraphx-bot]

Test	Batch	Rate new 82f668	Rate old 697709	Diff	Compare
torchvision-resnet50	64	895.43	895.75	-0.04%	✅
torchvision-resnet50_fp16	64	5,309.84	5,314.47	-0.09%	✅
torchvision-densenet121	32	1,126.36	1,126.59	-0.02%	✅
torchvision-densenet121_fp16	32	3,284.62	3,284.10	0.02%	✅
torchvision-inceptionv3	32	592.83	592.96	-0.02%	✅
torchvision-inceptionv3_fp16	32	2,512.33	2,517.46	-0.20%	✅
cadene-inceptionv4	16	329.05	329.93	-0.27%	✅
cadene-resnext64x4	16	394.60	393.50	0.28%	✅
slim-mobilenet	64	7,128.93	7,139.23	-0.14%	✅
slim-nasnetalarge	64	159.83	159.91	-0.05%	✅
slim-resnet50v2	64	1,089.34	1,089.62	-0.03%	✅
bert-mrpc-onnx	8	719.12	719.61	-0.07%	✅
bert-mrpc-tf	1	370.19	372.21	-0.54%	✅
pytorch-examples-wlang-gru	1	306.91	307.29	-0.12%	✅
pytorch-examples-wlang-lstm	1	306.90	313.39	-2.07%	✅
torchvision-resnet50_1	1	91.70	91.58	0.13%	✅
torchvision-inceptionv3_1	1	130.41	129.13	0.99%	✅
cadene-dpn92_1	1	334.39	335.84	-0.43%	✅
cadene-resnext101_1	1	236.97	236.90	0.03%	✅
slim-vgg16_1	1	53.39	53.37	0.03%	✅
slim-mobilenet_1	1	1,476.36	1,526.43	-3.28%	🔴
slim-inceptionv4_1	1	101.94	99.82	2.12%	✅
onnx-taau-downsample	1	316.20	316.86	-0.21%	✅
dlrm-criteoterabyte	1	21.44	21.44	0.03%	✅
dlrm-criteoterabyte_fp16	1	39.97	39.97	-0.00%	✅
agentmodel	1	5,808.42	5,625.18	3.26%	🔆
unet_fp16	2	52.72	52.81	-0.17%	✅

This build is not recommended to merge 🔴

[migraphx-bot]

    :white_check_mark:bert-mrpc-onnx: PASSED: MIGraphX meets tolerance

    :white_check_mark:bert-mrpc-tf: PASSED: MIGraphX meets tolerance

    :white_check_mark:pytorch-examples-wlang-gru: PASSED: MIGraphX meets tolerance

    :white_check_mark:pytorch-examples-wlang-lstm: PASSED: MIGraphX meets tolerance

    :white_check_mark:torchvision-resnet50_1: PASSED: MIGraphX meets tolerance

    :white_check_mark:torchvision-inceptionv3_1: PASSED: MIGraphX meets tolerance

🔴cadene-dpn92_1: FAILED: MIGraphX is not within tolerance - check verbose output

    :white_check_mark:cadene-resnext101_1: PASSED: MIGraphX meets tolerance

    :white_check_mark:slim-vgg16_1: PASSED: MIGraphX meets tolerance

    :white_check_mark:slim-mobilenet_1: PASSED: MIGraphX meets tolerance

    :white_check_mark:slim-inceptionv4_1: PASSED: MIGraphX meets tolerance

    :white_check_mark:dlrm-criteoterabyte: PASSED: MIGraphX meets tolerance

    :white_check_mark:agentmodel: PASSED: MIGraphX meets tolerance

    :white_check_mark:unet: PASSED: MIGraphX meets tolerance

[causten]

2023

[tperry-amd]

This should be either a python or cmake script so it uses the same script for both windows and linux.

I agree, a unified script for Linux and Windows is preferred. Our mandate for these initial changes was to minimally impact Linux. Are you okay if we generate a work item to address this after all the Windows changes have made it upstream? The reasoning is that this change is blocking Windows and the work for a new unified script has a testing requirement for both Windows and Linux that will take some time and is lower priority compared to enabling the larger Windows effort.

[pfultz2]

The reasoning is that this change is blocking Windows

make generate isnt needed to build migraphx either on windows or linux. Its only needed when a developer changes one of the autogenerated files, which can be done on linux at PR time. So I dont see how this could be a blocker, especially if all the changes has been upstreamed already.

[tperry-amd]

The reasoning is that this change is blocking Windows

make generate isnt needed to build migraphx either on windows or linux. Its only needed when a developer changes one of the autogenerated files, which can be done on linux at PR time. So I dont see how this could be a blocker, especially if all the changes has been upstreamed already.

Fair point. It is still additional work that affects Linux more directly. Our goal for the initial changes was to minimally impact Linux and follow your patterns. The pattern here was to solve this problem with a shell script and we solved it with a shell script for Windows. We can change the pattern and create a unified script that works on both OS but it could potentially affect Linux which increases the testing work.

I think we should consider a change in goal. A modest increase in Linux risk is okay if we create a unified solution. We'll pull the change, move it to another branch, and hold onto the branch until we have time to do the unified solution.