Skip to content

O
openpose

shy_201992 / openpose
与 Fork 源项目一致

从无法访问的项目Fork

通知 7
Star 0
Fork 0
O
openpose

体验新版 GitCode，发现更多精彩内容 >>

提交 c047d13b 编写于 作者: G Gines
浏览文件
操作

Fixed CUDA 7 compatibility

上级 635dc580
隐藏空白更改
内联 并排
Showing with 12 addition and 2 deletion
src/openpose/core/maximumBase.cu
浏览文件 @ c047d13b
......@@ -19,7 +19,10 @@ namespace op
// __global__ void fillTargetPtrChannel(T* targetPtrOffsetted, const T* sourcePtrOffsetted, const int width, const int imageOffset)
// {
// const auto sourceThrustPtr = thrust::device_pointer_cast(sourcePtrOffsetted);
// const auto sourceIndexIterator = thrust::max_element(thrust::device, sourceThrustPtr, sourceThrustPtr + imageOffset);
// // Ideal option (not working for CUDA < 8)
// // const auto sourceIndexIterator = thrust::max_element(thrust::device, sourceThrustPtr, sourceThrustPtr + imageOffset);
// // Workaround to make it work for CUDA 7.5
// const auto sourceIndexIterator = thrust::max_element(sourceThrustPtr, sourceThrustPtr + imageOffset);
// const auto sourceIndex = (int)(sourceIndexIterator - sourceThrustPtr);
// targetPtrOffsetted[0] = sourceIndex % width;
// targetPtrOffsetted[1] = sourceIndex / width;
......@@ -40,6 +43,10 @@ namespace op
// const auto* const sourcePtrOffsetted = sourcePtr + (offsetChannel + part) * imageOffset;
// auto sourceThrustPtr = thrust::device_pointer_cast(sourcePtrOffsetted);
// const auto sourceIndexIterator = thrust::max_element(thrust::device, sourceThrustPtr, sourceThrustPtr + imageOffset);
// // Ideal option (not working for CUDA < 8)
// // const auto sourceIndexIterator = thrust::max_element(thrust::device, sourceThrustPtr, sourceThrustPtr + imageOffset);
// // Workaround to make it work for CUDA 7.5
// const auto sourceIndexIterator = thrust::max_element(sourceThrustPtr, sourceThrustPtr + imageOffset);
// const auto sourceIndex = (int)(sourceIndexIterator - sourceThrustPtr);
// targetPtrOffsetted[0] = sourceIndex % width;
// targetPtrOffsetted[1] = sourceIndex / width;
......@@ -81,7 +88,10 @@ namespace op
const auto* const sourcePtrOffsetted = sourcePtr + (offsetChannel + part) * imageOffset;
// Option a - 6.3 fps
const auto sourceThrustPtr = thrust::device_pointer_cast(sourcePtrOffsetted);
const auto sourceIndexIterator = thrust::max_element(thrust::device, sourceThrustPtr, sourceThrustPtr + imageOffset);
// Ideal option (not working for CUDA < 8)
// const auto sourceIndexIterator = thrust::max_element(thrust::device, sourceThrustPtr, sourceThrustPtr + imageOffset);
// Workaround to make it work for CUDA 7.5
const auto sourceIndexIterator = thrust::max_element(sourceThrustPtr, sourceThrustPtr + imageOffset);
const auto sourceIndex = (int)(sourceIndexIterator - sourceThrustPtr);
fillTargetPtrPart<<<1, 1>>>(targetPtrOffsetted, sourcePtrOffsetted, sourceIndex, sourceIndex % width, sourceIndex / width);
// // Option b - <1 fps
......
Markdown is supported
0% 或 .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑！
想要评论请 注册 或