Dropout kernel OpenCL to HIP + gtest

[sgundabo]

Initial attempt at translating the Dropout OpenCL Kernel to HIP with a GTEST, with hardcoded PRNG matrices replaced with rocrand function calls.

[CAHEK7]

@CAHEK7 @amberhassaan

Please find the profiling results attached below.

@sgundabo Just for a reference - what king of gpu did you use to get those results? And I guess we need some bigger tensors.

[sgundabo]

@CAHEK7 @amberhassaan Please find the profiling results attached below.

@sgundabo Just for a reference - what king of gpu did you use to get those results? And I guess we need some bigger tensors.

These results are from a gfx90a.

[sgundabo]

PerfData

DropoutPerf_large_5.csv DropoutPerf_large_4.csv DropoutPerf_large_3.csv DropoutPerf_large_2.csv DropoutPerf_large_1.csv

HW tested: gfx90a

FP32 Perf

FP16 Perf

[sgundabo]

Raw Perf data with detailed kernel information DropoutPerf_FP16.zip DropoutPerf_FP32.zip

HW tested: gfx90a

FP32 Perf

FP16 Perf

[sgundabo]

RawPerf Data DropoutPerf_smalltensors.zip

Perf FP16

Perf FP32

[sgundabo]

FP32 Perf Analysis

Boxplot dropout0.50 mask0

Boxplot all dropouts and mask combinations

[CAHEK7]

@junliume could you merge this one, as there are few other PRs depend on it.

[junliume]

@sgundabo @CAHEK7 please feel free to ping me on any other channels. Also marking the urgency level of the PR would help for visibility. Thanks!