Skip to content

HWTODO.md

Latest commit

 

History

History
334 lines (262 loc) · 8.34 KB

HWTODO.md

File metadata and controls

334 lines (262 loc) · 8.34 KB

Hardware Testing TODO - M4 MacBook Air ↔ Arch PC (RTX 3070 Ti)

Tester: You (Local Hardware) Target Codecs: H.264, HEVC, AV1 Setup: M4 MacBook Air (VideoToolbox) ↔ Arch PC (NVIDIA 3070 Ti, NVENC)

🎯 Testing Matrix

macOS (M4 MacBook Air)

Codec Encoder Res FPS Target Bitrate Hardware Status
H.264 VideoToolbox 1080p 30 2.5 Mbps M4 Apple Silicon [ ] Test
H.264 VideoToolbox 1080p 60 5.0 Mbps M4 Apple Silicon [ ] Test
HEVC VideoToolbox 1080p 30 1.8 Mbps M4 Apple Silicon [ ] Test
HEVC VideoToolbox 1080p 60 3.5 Mbps M4 Apple Silicon [ ] Test
AV1 VideoToolbox 1080p 30 1.2 Mbps M4 Apple Silicon [ ] Test
AV1 VideoToolbox 1080p 60 2.5 Mbps M4 Apple Silicon [ ] Test

Linux (Arch PC - RTX 3070 Ti)

Codec Encoder Res FPS Target Bitrate Hardware Status
H.264 NVENC 1080p 30 2.5 Mbps RTX 3070 Ti [ ] Test
H.264 NVENC 1080p 60 5.0 Mbps RTX 3070 Ti [ ] Test
HEVC NVENC 1080p 30 1.8 Mbps RTX 3070 Ti [ ] Test
HEVC NVENC 1080p 60 3.5 Mbps RTX 3070 Ti [ ] Test
AV1 NVENC 1080p 30 1.2 Mbps RTX 3070 Ti [ ] Test
AV1 NVENC 1080p 60 2.5 Mbps RTX 3070 Ti [ ] Test

📋 Testing Procedure

Setup Phase (One-time)

  1. Build wavry for both platforms

    # On M4 MacBook
cargo build --release --workspace

# On Arch PC
cargo build --release --workspace

  2. Generate test certificates (for WebTransport)

    ./scripts/gen-wt-cert.sh

  3. Verify hardware acceleration detection

    # Check what codecs each platform detects
cargo run --bin wavry-server --release 2>&1 | grep -i "codec\|videoToolbox\|NVENC\|hardware"

Local Progress Snapshot (2026-02-12, Apple M4)

  • Ran ./scripts/av1-hardware-smoke.sh
  • Ran macOS probe tests (cargo test -p wavry-media mac_probe -- --nocapture)
  • Captured host startup capability output (Local encoder candidates: [Hevc, H264])
  • Wrote local results report (docs/HWTEST_RESULTS.md)
  • Cross-device bitrate/latency benchmarking (requires second host)

🧪 Test Case 1: H.264 Baseline (30fps, 1080p)

macOS (M4) - Server

RUST_LOG=info cargo run --bin wavry-server --release -- \
  --width 1920 \
  --height 1080 \
  --fps 30 \
  --codec h264 \
  --bitrate 2500

Monitor Output:

  • Encoder initialized (VideoToolbox)
  • Capture FPS: ~30
  • Bitrate maintaining ~2.5 Mbps
  • CPU usage: <30%
  • GPU memory: <200MB

Linux (Arch) - Client

RUST_LOG=info WAVRY_GATEWAY_URL=ws://192.168.x.x:3000/ws \
cargo run --bin wavry-client --release -- \
  --connect 192.168.x.x:5000

Monitor Output:

  • Decoder initialized (NVIDIA hardware)
  • Receiving ~30fps
  • RTT: <10ms (local network)
  • Jitter: <5ms
  • Frame drops: 0
  • CPU: <20%

🧪 Test Case 2: HEVC 60fps (1080p)

macOS (M4) - Server

RUST_LOG=info cargo run --bin wavry-server --release -- \
  --width 1920 \
  --height 1080 \
  --fps 60 \
  --codec hevc \
  --bitrate 3500

Monitor Output:

  • Encoder: HEVC/H.265 (VideoToolbox)
  • Encode latency: <16ms (for 60fps)
  • Bitrate: ±10% variance from 3.5 Mbps
  • GPU memory: <250MB
  • Thermal: Check if throttling

Linux (Arch) - Client

RUST_LOG=info WAVRY_GATEWAY_URL=ws://192.168.x.x:3000/ws \
cargo run --bin wavry-client --release -- \
  --connect 192.168.x.x:5000

Monitor Output:

  • Decoding 60 fps smoothly
  • Frame timing stability: ±2ms variance
  • No visible artifacts
  • GPU utilization: 40-60%

🎬 Test Case 3: AV1 Codec Validation

macOS (M4) - Server

RUST_LOG=info cargo run --bin wavry-server --release -- \
  --width 1920 \
  --height 1080 \
  --fps 30 \
  --codec av1 \
  --bitrate 1200

Verify:

  • AV1 hardware acceleration detected
  • Encoding latency: <33ms per frame (30fps)
  • Bitrate efficiency: 1.2 Mbps for comparable quality to H.264@2.5Mbps
  • Quality: Compare subjective quality vs H.264
  • Supported: Yes/No on M4

Linux (Arch) - Client

RUST_LOG=info WAVRY_GATEWAY_URL=ws://192.168.x.x:3000/ws \
cargo run --bin wavry-client --release -- \
  --connect 192.168.x.x:5000

Verify:

  • AV1 decoding works
  • Frame drops vs H.264: Compare
  • Visual quality: Assess artifacts
  • Supported: Yes/No on RTX 3070 Ti

📊 Performance Metrics to Capture

For each codec/FPS combination, record:

Codec: [H.264/HEVC/AV1]
FPS: [30/60/120]
Bitrate Target: [Mbps]
Resolution: [1920x1080]

=== ENCODER (macOS M4) ===
Actual Bitrate: _____ Mbps (variance: ±_____%)
Encode Latency: _____ ms
GPU Memory: _____ MB
CPU Load: _____ %
Thermal: OK / Throttling

=== NETWORK ===
RTT: _____ ms
Jitter: _____ ms
Loss: _____ %

=== DECODER (Arch/3070Ti) ===
Decode Latency: _____ ms
GPU Memory: _____ MB
CPU Load: _____ %
Frame Drops: _____
Visual Quality: Excellent / Good / Acceptable / Poor
Artifacts: None / [describe]

🔍 Quality Assessment (Subjective)

For each codec, evaluate:

Metric H.264 HEVC AV1 Notes
Detail preservation [ ] [ ] [ ] Text sharpness, edges
Motion smoothness [ ] [ ] [ ] Pan/scroll artifacts
Color accuracy [ ] [ ] [ ] Gradients, skin tones
Compression artifacts [ ] [ ] [ ] Banding, blockiness

🚀 Testing Sequence

Day 1: Baseline (H.264)

  1. M4 Server: H.264 30fps
  2. Arch Client: Receives & decodes
  3. Document baseline metrics
  4. Record: Bitrate, latency, quality

Day 2: High FPS (H.264 60fps + HEVC)

  1. M4 Server: H.264 60fps
  2. Arch Client: Receives 60fps
  3. M4 Server: HEVC 30fps
  4. Arch Client: HEVC decode

Day 3: AV1 & Reverse Test

  1. M4 Server: AV1 30fps
  2. Arch Client: AV1 decode
  3. Reverse: Arch PC as Server, M4 as Client
    • Run wavry-server on Arch (NVIDIA NVENC)
    • Run wavry-client on M4 (VideoToolbox decode)
  4. Test H.264/HEVC/AV1 from Arch → M4

📝 Results Template

## Test Run: [DATE] [CODEC] [FPS]fps [RESOLUTION]

### Server (macOS M4)
- Hardware Acceleration: [Yes/No]
- Encoder: [VideoToolbox]
- Target Bitrate: 2500 kbps
- Actual Bitrate: _______ kbps
- Bitrate Variance: ±_____%
- Encode Latency (min/avg/max): ___ / ___ / ___ ms
- GPU Memory: _______ MB
- CPU: _______ %
- Thermal: [OK/Throttling]

### Network
- RTT: _______ ms
- Jitter: _______ ms
- Loss: _______ %
- Bitrate Stability: ±_____%

### Client (Arch RTX 3070 Ti)
- Hardware Acceleration: [Yes/No]
- Decoder: [NVIDIA/CPU]
- Decode Latency (min/avg/max): ___ / ___ / ___ ms
- GPU Memory: _______ MB
- CPU: _______ %
- Frame Drops: _____
- Quality: [Excellent/Good/Acceptable/Poor]
- Issues: [None/describe]

### Analysis
- Hardware acceleration working: [Yes/No]
- Bitrate efficiency vs H.264: [Better/Same/Worse]
- Recommended bitrate: _______ kbps
- Ready for production: [Yes/No/Needs tuning]

🐛 Troubleshooting

M4 VideoToolbox Issues

# Check VideoToolbox availability
cargo run --bin wavry-server --release -- --list-codecs

# If AV1 not available:
# - Check macOS version (15.1+)
# - May require VideoToolbox API update

# GPU memory issues:
# - Check Activity Monitor for leaks
# - Reduce resolution/bitrate

Arch NVIDIA Issues

# Check NVIDIA driver
nvidia-smi

# Check NVENC availability
nvidia-smi -q | grep -i "nvenc\|video"

# If AV1 not available:
# - Driver version too old (need 525+)
# - GPU generation doesn't support AV1 (3070 Ti should support)

📤 Reporting Results

Once testing is complete:

  1. Update docs/AV1_VALIDATION.md with results
  2. Create HWTEST_RESULTS.md with full metrics
  3. Update TODO.md: Mark "AV1 Performance Validation" complete
  4. Note any codec recommendations in codebase

🎯 Success Criteria

  • H.264: Works 30fps & 60fps (baseline)
  • HEVC: Encoding confirmed, client decodes
  • AV1: Detects codec, encode/decode works
  • Bitrate variance: ±10% max
  • Frame drops: Zero on stable network
  • Latency: <50ms total (encode + network + decode)
  • Reverse test: Arch → M4 also works