Do Video Encoding Workloads Stress the Microarchitecture?
Bibliographic record
Abstract
Video encoding/decoding is an extremely relevant workload in our society today. Video accounts for a significant percentage of the world’s online traffic, which is expected only to be growing. Thus, it is important to understand these workloads to optimize the hardware to handle them better. There is significant interest in the royalty-free AV1 codec, but we identify that it consumes a significantly higher runtime than other popular codecs, such as H.264/AVC, H.265/HEVC, and VP9. However, the reasons for the slowdown of the AV1 codec are not well-understood by prior work to the best of our knowledge.This paper explores the reasons for the large runtimes taken by AV1 workloads. We first focus on profiling the microarchitectural characteristics of the SVT-AV1 encoder, which implements the AV1 codec, to identify acceleration opportunities with a wide spectrum of encoding parameters.We discover that the runtime of AV1 encoders is higher than other encoders because AV1 encoders require a larger number of instructions to encode the same video, rather than any significant microarchitectural inefficiencies. Among microarchitectural components, we observe branch misprediction to be the component with the most significant impact on performance. In light of this, we evaluate the performance of several different branch predictors using Championship Branch Prediction (CBP) frameworks. From this, we find that increasing the size of the branch predictor as well as using a TAGE branch predictor rather than Gshare both had a significant positive impact on branch predictor performance.We also compare the scaling of SVT-AV1 against other codecs by giving each encoder access to an increasing number of threads. Finally, we observe that SVT-AV1 contains the highest degree of parallelism of the tested encoders. Because of this, increasing concurrently running threads and optimizing branch prediction may help bridge the gap in runtime between SVT-AV1 and encoders implementing other codecs.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.002 | 0.001 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".