張小凡

Please support new watch: FR235 and FR630

1

Please add code in src/HrWidgetView.mc and download last Connect IQ SDK release to new version, Thanks a lot~ code like this : else if(dc.getWidth() == 215 && dc.getHeight() == 180){...

Add destroy_gpu_instance() function new parameter need_wait about it will wait for all devices to be idle before destroy

2

书上说，须在销毁设备之前之前调用vkDeviceWaitIdle()，以等待所有设备可能存在的未完成的任何工作。 为了保持向前兼容性，所以默认参数设计成默认不启用。 ```cpp NCNN_EXPORT void destroy_gpu_instance(int need_wait = 0); ``` 引用出处： 

Add VK_EXT_pageable_device_local_memory support

2

Add VK_EXT_pageable_device_local_memory support ### Why do this change? * [https://registry.khronos.org/vulkan/specs/1.3-extensions/man/html/vkAllocateMemory.html](https://registry.khronos.org/vulkan/specs/1.3-extensions/man/html/vkAllocateMemory.html) * [https://registry.khronos.org/vulkan/specs/1.3-extensions/man/html/VkPhysicalDevicePageableDeviceLocalMemoryFeaturesEXT.html](https://registry.khronos.org/vulkan/specs/1.3-extensions/man/html/VkPhysicalDevicePageableDeviceLocalMemoryFeaturesEXT.html) If the [VkPhysicalDevicePageableDeviceLocalMemoryFeaturesEXT](https://registry.khronos.org/vulkan/specs/1.3-extensions/man/html/VkPhysicalDevicePageableDeviceLocalMemoryFeaturesEXT.html)::pageableDeviceLocalMemory feature is enabled, memory allocations made from a heap that includes VK_MEMORY_HEAP_DEVICE_LOCAL_BIT in [VkMemoryHeap](https://registry.khronos.org/vulkan/specs/1.3-extensions/man/html/VkMemoryHeap.html)::flags may...

Expect to add more x64 ISA about AI

1

### AVX512(Intel® Advanced Vector Extensions 512) > **AVX512_BITALG** (expect to add) > **AVX512_VPOPCNTDQ** (expect to add) > **AVX512_VP2INTERSECT** (expect to add) > > AVX512PF (Intel® Xeon Phi™ only.) > AVX512ER...

Add yolov8 ncnn example

1

# Add yolov8 ncnn example ReadMe Convert yolov8 model to ncnn model workflow: ## step 1: If you don't want to train the model yourself. You should go to the...

## error log | 日志或报错信息 | ログ 初始化ncnn时，在 [VulkanDevice::init_device_extension()](https://github.com/Tencent/ncnn/blob/86e1ed067a44c124812e056eec86eeef27ae468b/src/gpu.cpp#L3625C62-L3625C81) 中崩溃了。 以下是vc dump查看器的调用栈截图:   [以下来自目前最新版ncnn的代码片段，以及个人理解给的注释](https://github.com/Tencent/ncnn/blob/86e1ed067a44c124812e056eec86eeef27ae468b/src/gpu.cpp#L2654)： ```c++ VkResult ret = vkCreateDevice(info.physical_device(), &deviceCreateInfo, 0, &d->device); // 这一句失败了，d->device为无效值 if (ret != VK_SUCCESS) {...

## Added x86 'sm3/sm4' detection godbolt ISA test:  ## x86 ISA Document ### SM3  ### SM4 

## Added x86 'sha256/sha512' detection. x86 sha isa reference: [Intel_SHA_extensions Wikipedia](https://en.wikipedia.org/wiki/Intel_SHA_extensions)  So I rename x86 'sha' to 'sha1'. ## Test godbolt test: 

Improve yolov8 post-processing efficiency.

2

Using OpenMP to improve yolov8 post-processing efficiency. ```c++ #if NCNN_SIMPLEOMP #include "simpleomp.h" #else #include #endif ... static void parse_yolov8_detections( float* inputs, float confidence_threshold, int num_channels, int num_anchors, int num_labels, int...

add simplestl std::unordered_map

1

## simplestl std::unordered_map online compiler : https://coliru.stacked-crooked.com/a/a6c633f30aaed522 test code: ```cpp std::unordered_map map1; map1.insert(std::make_pair(1, 21)); map1.insert(std::make_pair(0, 10)); map1[0] = 20; map1.erase(1); map1.emplace(1, 11); map1.insert(std::make_pair(2, 12)); map1.emplace(3, 13); printf("map1 key %d has...