cabana: Improve seeking and zooming (#32334)

* Improve seeking and zooming

* No repeated calculation of freq

* set min zoom seconds to 10ms
old-commit-hash: d72f000d98fc63f1ff7da44acedf95251c89faef

tools/replay/camera.cc

@@ -1,12 +1,13 @@
 #include "tools/replay/camera.h"
 
 #include <capnp/dynamic.h>
-
 #include <cassert>
 
 #include "third_party/linux/include/msm_media_info.h"
 #include "tools/replay/util.h"
 
+const int BUFFER_COUNT = 40;
+
 std::tuple<size_t, size_t, size_t> get_nv12_info(int width, int height) {
   int nv12_width = VENUS_Y_STRIDE(COLOR_FMT_NV12, width);
   int nv12_height = VENUS_Y_SCANLINES(COLOR_FMT_NV12, height);
@@ -36,10 +37,12 @@ CameraServer::~CameraServer() {
 void CameraServer::startVipcServer() {
   vipc_server_.reset(new VisionIpcServer("camerad"));
   for (auto &cam : cameras_) {
+    cam.cached_buf.clear();
+
     if (cam.width > 0 && cam.height > 0) {
       rInfo("camera[%d] frame size %dx%d", cam.type, cam.width, cam.height);
       auto [nv12_width, nv12_height, nv12_buffer_size] = get_nv12_info(cam.width, cam.height);
-      vipc_server_->create_buffers_with_sizes(cam.stream_type, YUV_BUFFER_COUNT, false, cam.width, cam.height,
+      vipc_server_->create_buffers_with_sizes(cam.stream_type, BUFFER_COUNT, false, cam.width, cam.height,
                                               nv12_buffer_size, nv12_width, nv12_width * nv12_height);
       if (!cam.thread.joinable()) {
         cam.thread = std::thread(&CameraServer::cameraThread, this, std::ref(cam));
@@ -50,13 +53,6 @@ void CameraServer::startVipcServer() {
 }
 
 void CameraServer::cameraThread(Camera &cam) {
-  auto read_frame = [&](FrameReader *fr, int frame_id) {
-    VisionBuf *yuv_buf = vipc_server_->get_buffer(cam.stream_type);
-    assert(yuv_buf);
-    bool ret = fr->get(frame_id, yuv_buf);
-    return ret ? yuv_buf : nullptr;
-  };
-
   while (true) {
     const auto [fr, event] = cam.queue.pop();
     if (!fr) break;
@@ -66,29 +62,41 @@ void CameraServer::cameraThread(Camera &cam) {
     auto eidx = capnp::AnyStruct::Reader(evt).getPointerSection()[0].getAs<cereal::EncodeIndex>();
     if (eidx.getType() != cereal::EncodeIndex::Type::FULL_H_E_V_C) continue;
 
-    const int id = eidx.getSegmentId();
-    bool prefetched = (id == cam.cached_id && eidx.getSegmentNum() == cam.cached_seg);
-    auto yuv = prefetched ? cam.cached_buf : read_frame(fr, id);
-    if (yuv) {
+    int segment_id = eidx.getSegmentId();
+    uint32_t frame_id = eidx.getFrameId();
+    if (auto yuv = getFrame(cam, fr, segment_id, frame_id)) {
       VisionIpcBufExtra extra = {
-          .frame_id = eidx.getFrameId(),
+          .frame_id = frame_id,
           .timestamp_sof = eidx.getTimestampSof(),
           .timestamp_eof = eidx.getTimestampEof(),
       };
-      yuv->set_frame_id(eidx.getFrameId());
       vipc_server_->send(yuv, &extra);
     } else {
-      rError("camera[%d] failed to get frame: %lu", cam.type, eidx.getSegmentId());
+      rError("camera[%d] failed to get frame: %lu", cam.type, segment_id);
     }
 
-    cam.cached_id = id + 1;
-    cam.cached_seg = eidx.getSegmentNum();
-    cam.cached_buf = read_frame(fr, cam.cached_id);
+    // Prefetch the next frame
+    getFrame(cam, fr, segment_id + 1, frame_id + 1);
 
     --publishing_;
   }
 }
 
+VisionBuf *CameraServer::getFrame(Camera &cam, FrameReader *fr, int32_t segment_id, uint32_t frame_id) {
+  // Check if the frame is cached
+  auto buf_it = std::find_if(cam.cached_buf.begin(), cam.cached_buf.end(),
+                             [frame_id](VisionBuf *buf) { return buf->get_frame_id() == frame_id; });
+  if (buf_it != cam.cached_buf.end()) return *buf_it;
+
+  VisionBuf *yuv_buf = vipc_server_->get_buffer(cam.stream_type);
+  if (fr->get(segment_id, yuv_buf)) {
+    yuv_buf->set_frame_id(frame_id);
+    cam.cached_buf.insert(yuv_buf);
+    return yuv_buf;
+  }
+  return nullptr;
+}
+
 void CameraServer::pushFrame(CameraType type, FrameReader *fr, const Event *event) {
   auto &cam = cameras_[type];
   if (cam.width != fr->width || cam.height != fr->height) {

tools/replay/camera.h

@@ -1,6 +1,7 @@
 #pragma once
 
 #include <memory>
+#include <set>
 #include <tuple>
 #include <utility>
 
@@ -26,12 +27,11 @@ protected:
     int height;
     std::thread thread;
     SafeQueue<std::pair<FrameReader*, const Event *>> queue;
-    int cached_id = -1;
-    int cached_seg = -1;
-    VisionBuf * cached_buf;
+    std::set<VisionBuf *> cached_buf;
   };
   void startVipcServer();
   void cameraThread(Camera &cam);
+  VisionBuf *getFrame(Camera &cam, FrameReader *fr, int32_t segment_id, uint32_t frame_id);
 
   Camera cameras_[MAX_CAMERAS] = {
       {.type = RoadCam, .stream_type = VISION_STREAM_ROAD},