- Windows 10 laptop
- CPU 11th Gen Intel(R) Core(TM) i7-11375H @ 3.30GHz (cpu)
- Intel(R) Iris(R) Xe Graphics (iGPU)
- NVIDIA GeForce RTX 3060 Laptop GPU (gpu)
- vector sum
-
MatrixMultiplication project
-
Naive Matrix Multiplication (Completed)
- main.cpp with MatMul.cl
- No limit on input size
-
Matrix Multiplication Tiling in the local memoryv (Completed)
- main2.cpp with MatMul2.cl
- No limit on input size
-
Matrix Multiplication Tiling in the local memory with register level(Verification Required)
- main3.cpp with MatMul3.cl
-
Register Blocking Matrix Multiplication(Verification Required)
- main4.cpp with MatMul4.cl
- Time check for A[1024, 1024] * B[1024, 1024] = C[1024, 1024] (wo : without data transfer time for Device)
-
-
GEMM Convolution(Completed)
- process : im2col -> Matrix Multiplication -> col2im
- Matrix Multiplication can be changed to better logic
-
Naive Convolution(Completed)
-
FFT Convolution(Plan)
-
Winograd Convolution(Plan)
- Time Check (input[1, 3, 512, 512] kernel[3, 3, 3, 3] output[ 1, 3, 510, 510])
- GEMM Conv2d (gpu) : 1.76237 [msec]
- Naive Conv2d (gpu) : 0.22528 [msec]
- Naive Conv2d (cpu) : 22.32400 [msec]
- Time Check (input[1, 3, 512, 512] kernel[3, 3, 3, 3] output[ 1, 3, 510, 510])
- Bicubic Interpolation
- Add test code(validation results with pytorch Bicubic Interpolation)
- Nearnest Neighbor Interpolation
- Add test code(validation results with pytorch Nearnest Neighbor Interpolation)
- transformation image data format BGR to YCbC, NHWC->NCHW
- Add test code(validation results with python BGR2YCbC)
- transformation image data format YCbC to BGR, NCHW->NHWC
- Add test code(validation results with python YCbC2BGR)
- Concatenate two tensor for channal side
- Add test code(validation results with python concat)
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent