GPU-Accelerated Video Encoding - Nvidia

More documents

Recommendations

Info

ME – Diamond search • The majority of template methods are based on gradient descent and have certain application restrictions • When searching for a match for the block on the frame below, red initialization is not better than a random block selection, while green will likely converge • Used often in junction with some other methods as a refinement step 57 PRESENTED BY
ME – Candidate Set • Objects that move inside a frame are typically larger than a block, adjacent blocks have similar motion vectors • Candidate set is a list of motion vectors that have to be checked in first place. The list is usually filled with motion vectors of already found adjacent (spatially or temporarily) blocks • The majority of CPU-school ME algorithms fill candidate set with motion vectors found for left, top-left, top, top-right blocks (assuming block-linear YX-order processing) • This creates an explicit data dependency which hinders efficient porting of the algorithm to CUDA 58 PRESENTED BY
Page 1 and 2:
GPU-Accelerated Video Encoding NVID
Page 3 and 4:
Outline • Motivation • Video en
Page 5 and 6:
Trends • GPU encoding is gaining
Page 7 and 8: Outline • Motivation • Video en
Page 11 and 12: Hybrid CPU-GPU encoding pipeline
Page 13 and 14: GPU Encoding Use Cases • Video en
Page 15 and 16: GPU Encoding Use Cases Helper libra
Page 17 and 18: Algorithm primitives • Fundamenta
Page 19 and 20: Reduce • Use half of N threads
Page 21 and 22: Reduce sample: 8 numbers 5 14 9 5 2
Page 23 and 24: Reduce sample: 8 numbers 33 14 19 5
Page 25 and 26: Scan sample: 8 numbers 2 9 5 1 3 6
Page 31 and 32: Compact • Having a vector of numb
Page 33 and 34: Compact sample The algorithm makes
Page 35 and 36: Algorithm primitives discussion •
Page 37 and 38: Intra Prediction 36 PRESENTED BY
Page 39 and 40: Intra Prediction • Select the rul
Page 41 and 42: Intra Prediction Plane Constant Ver
Page 43 and 44: CUDA: Intra Prediction 16x16 1. Loa
Page 45 and 46: CUDA: Intra Prediction 16x16 3. __d
Page 47 and 48: CUDA: Intra Prediction 16x16 4. Cal
Page 49 and 50: SAD and SATD • Sum of Absolute Tr
Page 51 and 52: Motion Estimation 50 PRESENTED BY
Page 53 and 54: Motion Estimation 52 PRESENTED BY
Page 55 and 56: Motion Estimation Motion vectors fi
Page 57: ME - Diamond search • Template me
Page 61 and 62: ME - 3D-wave • 3D-wave treats tim
Page 63 and 64: ME - GPU-friendly candidate set •
Page 65 and 66: ME - Hierarchical Search Algorithm:
Page 69 and 70: Tesla and Fermi architectures • A
Page 71 and 72: Multi-GPU encoding • N GOPs for N
Page 75 and 76: Recommended GTC talks We have a ple
show all

GPU-Accelerated Video Encoding - Nvidia

Create successful ePaper yourself

Delete template?

Save as template?