Files
StarPilot/sunnypilot/modeld/models/commonmodel.cc
T
Jason Wen acd46aa94b modeld: retain SNPE and thneed drive model support (#555)
* modeld: Retain pre-20hz drive model support

* Method not available anymore on OP

* some fixes

* Revert "Long planner get accel: new function args (#34288)"

* Revert "Fix low-speed allow_throttle behavior in long planner (#33894)"

* Revert "long planner: allow throttle reflects usage (#33792)"

* Revert "Gate acceleration on model gas press predictions (#33643)"

* Reapply "Gate acceleration on model gas press predictions (#33643)"

This reverts commit 76b08e37cb8eb94266ad9f6fed80db227e7c3428.

* Reapply "long planner: allow throttle reflects usage (#33792)"

This reverts commit c75244ca4e9c48084b0205b7c871e1a4e0f4e693.

* Reapply "Fix low-speed allow_throttle behavior in long planner (#33894)"

This reverts commit b2b7d21b7b685a2785d1beede3d223f0bb954807.

* Reapply "Long planner get accel: new function args (#34288)"

This reverts commit 74dca2fccf4da59cc8ac62ba9c0ad10ba3fc264b.

* don't need

* retain snpe

* wrong

* they're symlinks

* remove

* put back into VCS

* add back

* don't include built

* Refactor model runner retrieval with caching support

Added caching for active model runner type via `ModelRunnerTypeCache` to enhance performance and avoid redundant checks. Introduced a `force_check` flag to bypass the cache when necessary. Updated related code to handle cache clearing during onroad transitions.

* Update model runner determination logic with caching fix

Enhances `get_active_model_runner` to utilize caching more effectively by ensuring type consistency and updating cache only when necessary. Also updates `is_snpe_model` to pass the `started` state to the runner determination function, improving behavior for dynamic checks.

* default to none

* enable in next PR

* more

---------

Co-authored-by: DevTekVE <devtekve@gmail.com>
2025-01-10 18:34:06 -05:00

51 lines
2.1 KiB
C++

#include "sunnypilot/modeld/models/commonmodel.h"
#include <cassert>
#include <cmath>
#include <cstring>
#include "common/clutil.h"
ModelFrame::ModelFrame(cl_device_id device_id, cl_context context) {
input_frames = std::make_unique<float[]>(buf_size);
q = CL_CHECK_ERR(clCreateCommandQueue(context, device_id, 0, &err));
y_cl = CL_CHECK_ERR(clCreateBuffer(context, CL_MEM_READ_WRITE, MODEL_WIDTH * MODEL_HEIGHT, NULL, &err));
u_cl = CL_CHECK_ERR(clCreateBuffer(context, CL_MEM_READ_WRITE, (MODEL_WIDTH / 2) * (MODEL_HEIGHT / 2), NULL, &err));
v_cl = CL_CHECK_ERR(clCreateBuffer(context, CL_MEM_READ_WRITE, (MODEL_WIDTH / 2) * (MODEL_HEIGHT / 2), NULL, &err));
net_input_cl = CL_CHECK_ERR(clCreateBuffer(context, CL_MEM_READ_WRITE, MODEL_FRAME_SIZE * sizeof(float), NULL, &err));
transform_init(&transform, context, device_id);
loadyuv_init(&loadyuv, context, device_id, MODEL_WIDTH, MODEL_HEIGHT);
}
float* ModelFrame::prepare(cl_mem yuv_cl, int frame_width, int frame_height, int frame_stride, int frame_uv_offset, const mat3 &projection, cl_mem *output) {
transform_queue(&this->transform, q,
yuv_cl, frame_width, frame_height, frame_stride, frame_uv_offset,
y_cl, u_cl, v_cl, MODEL_WIDTH, MODEL_HEIGHT, projection);
if (output == NULL) {
loadyuv_queue(&loadyuv, q, y_cl, u_cl, v_cl, net_input_cl);
std::memmove(&input_frames[0], &input_frames[MODEL_FRAME_SIZE], sizeof(float) * MODEL_FRAME_SIZE);
CL_CHECK(clEnqueueReadBuffer(q, net_input_cl, CL_TRUE, 0, MODEL_FRAME_SIZE * sizeof(float), &input_frames[MODEL_FRAME_SIZE], 0, nullptr, nullptr));
clFinish(q);
return &input_frames[0];
} else {
loadyuv_queue(&loadyuv, q, y_cl, u_cl, v_cl, *output, true);
// NOTE: Since thneed is using a different command queue, this clFinish is needed to ensure the image is ready.
clFinish(q);
return NULL;
}
}
ModelFrame::~ModelFrame() {
transform_destroy(&transform);
loadyuv_destroy(&loadyuv);
CL_CHECK(clReleaseMemObject(net_input_cl));
CL_CHECK(clReleaseMemObject(v_cl));
CL_CHECK(clReleaseMemObject(u_cl));
CL_CHECK(clReleaseMemObject(y_cl));
CL_CHECK(clReleaseCommandQueue(q));
}