find_splits bugfix

[shivadbhavsar]

Solves #3218

Given a set of ops:
input: {2, 4, 4} -> slc1 {2,2,4}, slc2 {2,2,4} -> add(c1, slc1), add(c2, slc2)

Current logic assumes that the slices are always at the same input location in the binary op. This PR handles the case when graph has something like the following instead:
input: {2, 4, 4} -> slc1 {2,2,4}, slc2 {2,2,4} -> add(slc1, c1), add(c2, slc2)

In this case add is a commutative op so this simplification can still be done, so we force all the slices to be at input 0. In the case of non commutative ops, the matcher will return without performing a simplification.

[migraphx-bot]

Test	Batch	Rate new aad497	Rate old cbaa5b	Diff	Compare
torchvision-resnet50	64	1,749.80	1,741.54	0.47%	✅
torchvision-resnet50_fp16	64	4,180.17	4,066.82	2.79%	✅
torchvision-densenet121	32	1,465.84	1,461.16	0.32%	✅
torchvision-densenet121_fp16	32	2,548.89	2,522.45	1.05%	✅
torchvision-inceptionv3	32	888.56	885.32	0.37%	✅
torchvision-inceptionv3_fp16	32	1,492.13	1,478.16	0.95%	✅
cadene-inceptionv4	16	411.80	410.28	0.37%	✅
cadene-resnext64x4	16	419.31	417.37	0.46%	✅
slim-mobilenet	64	4,012.59	3,986.47	0.66%	✅
slim-nasnetalarge	64	100.98	100.58	0.40%	✅
slim-resnet50v2	64	1,679.81	1,671.88	0.47%	✅
bert-mrpc-onnx	8	616.53	612.70	0.63%	✅
bert-mrpc-tf	1	275.12	277.15	-0.73%	✅
pytorch-examples-wlang-gru	1	364.21	321.14	13.41%	🔆
pytorch-examples-wlang-lstm	1	294.68	294.00	0.23%	✅
torchvision-resnet50_1	1	471.15	473.62	-0.52%	✅
cadene-dpn92_1	1	246.88	246.57	0.13%	✅
cadene-resnext101_1	1	204.30	198.25	3.05%	🔆
onnx-taau-downsample	1	206.13	205.44	0.33%	✅
dlrm-criteoterabyte	1	22.89	22.82	0.28%	✅
dlrm-criteoterabyte_fp16	1	43.84	42.59	2.94%	✅
agentmodel	1	6,200.51	6,279.79	-1.26%	✅
unet_fp16	2	34.29	34.25	0.12%	✅
resnet50v1_fp16	1	583.96	585.19	-0.21%	✅
resnet50v1_int8	1	567.85	571.79	-0.69%	✅
bert_base_cased_fp16	64	646.65	642.42	0.66%	✅
bert_large_uncased_fp16	32	198.85	197.79	0.53%	✅
bert_large_fp16	1	116.81	116.84	-0.03%	✅
distilgpt2_fp16	16	1,211.08	1,205.29	0.48%	✅
yolov5s	1	301.66	299.73	0.64%	✅
tinyllama	1	23.31	23.22	0.37%	✅
vicuna-fastchat	1	133.74	132.49	0.94%	✅
whisper-tiny-encoder	1	243.77	243.14	0.26%	✅
whisper-tiny-decoder	1	256.00	255.32	0.27%	✅

Check results before 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

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 92.21%. Comparing base (88048b8) to head (96a4469).
Report is 1 commits behind head on develop.

Additional details and impacted files

@@             Coverage Diff             @@
##           develop    #3244      +/-   ##
===========================================
+ Coverage    92.20%   92.21%   +0.01%     
===========================================
  Files          493      493              
  Lines        19700    19725      +25     
===========================================
+ Hits         18164    18190      +26     
+ Misses        1536     1535       -1     

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