fix: disp buffer sized by engine dtype but passed to float kernels

[andrewwhitecdw]

Problem

fix: disp buffer sized by engine dtype but passed to float kernels

Fix

Replace:

    const size_t disp_bytes = H * W * elementSize(engine_->getTensorDataType("disp"));
    cudaMallocChecked(reinterpret_cast<void**>(&d_disp_), disp_bytes, "d_disp_");

with:

    if (engine_->getTensorDataType("disp") != nvinfer1::DataType::kFLOAT) {
        throw std::runtime_error("[FFS single] disp tensor must be FP32");
    }
    const size_t disp_bytes = H * W * sizeof(float);
    cudaMallocChecked(reinterpret_cast<void**>(&d_disp_), disp_bytes, "d_disp_");

Files changed

• cpp/src/ffs_depth_single_tensorrt.cpp

[andrewwhitecdw]

Closing this sweep-generated PR: sole commit is missing a valid Signed-off-by trailer. It does not meet the sweep requirements (single signed-off commit).