关于Lada在Amd的处理器上运行的报告之一。

gbywreyuerh@out

我的电脑是Amd 3600X+Amd RX 9700 GRE+Win11 Ltsc。

Lada默认启动是识别不到任何显卡的，只能用处理器转换，速度奇慢，20秒的1080P大概需要十几分钟。
然后试了用ZLUDA模拟N卡的CUDA，实测结果：

1，lada图形界面可以正常启动，并显示CUDA设备为A卡。
2，lada命令行界面可以识别到CUDA设备为A卡。

3，用第2条来处理视频，提示如下错误：
torch.AcceleratorError: CUDA error: named symbol not found
CUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect.
For debugging consider passing CUDA_LAUNCH_BLOCKING=1
Compile with `TORCH_USE_CUDA_DSA` to enable device-side assertions.

4，而如果用zluda分析程序来执行lada：zluda_with.exe --nvcuda C:\Users\Administrator\Desktop\lada-v0.8.1.zip\lada\zluda_trace.dll -- lada-cli.exe --device
则会出现一部分找不到CUDA设备的错误，这是所有出现的错误：
[ZLUDA_TRACE] cuGetProcAddress_v2(symbol: "", pfn: 0x0, cudaVersion: 0, flags: 0, symbolStatus: NULL) -> CUDA_ERROR_NOT_FOUND
[ZLUDA_TRACE] cuGetProcAddress_v2(symbol: "cuWGLGetDevice", pfn: 0x0, cudaVersion: 2020, flags: 0, symbolStatus: NULL) -> CUDA_ERROR_NOT_FOUND
[ZLUDA_TRACE] cuGetProcAddress_v2(symbol: "cuD3D11GetDevice", pfn: 0x0, cudaVersion: 3000, flags: 0, symbolStatus: NULL) -> CUDA_ERROR_NOT_FOUND
[ZLUDA_TRACE] cuGetProcAddress_v2(symbol: "cuD3D11GetDevices", pfn: 0x0, cudaVersion: 3020, flags: 0, symbolStatus: NULL) -> CUDA_ERROR_NOT_FOUND
[ZLUDA_TRACE] cuGetProcAddress_v2(symbol: "cuD3D11CtxCreateOnDevice", pfn: 0x0, cudaVersion: 3020, flags: 0, symbolStatus: NULL) -> CUDA_ERROR_NOT_FOUND
[ZLUDA_TRACE] cuGetProcAddress_v2(symbol: "cuGraphicsD3D11RegisterResource", pfn: 0x0, cudaVersion: 3000, flags: 0, symbolStatus: NULL) -> CUDA_ERROR_NOT_FOUND
[ZLUDA_TRACE] cuGetProcAddress_v2(symbol: "cuD3D11GetDirect3DDevice", pfn: 0x0, cudaVersion: 3020, flags: 0, symbolStatus: NULL) -> CUDA_ERROR_NOT_FOUND
[ZLUDA_TRACE] cuGetProcAddress_v2(symbol: "cuD3D10GetDevice", pfn: 0x0, cudaVersion: 2010, flags: 0, symbolStatus: NULL) -> CUDA_ERROR_NOT_FOUND
[ZLUDA_TRACE] cuGetProcAddress_v2(symbol: "cuD3D10GetDevices", pfn: 0x0, cudaVersion: 3020, flags: 0, symbolStatus: NULL) -> CUDA_ERROR_NOT_FOUND
[ZLUDA_TRACE] cuGetProcAddress_v2(symbol: "cuD3D10CtxCreateOnDevice", pfn: 0x0, cudaVersion: 3020, flags: 0, symbolStatus: NULL) -> CUDA_ERROR_NOT_FOUND
[ZLUDA_TRACE] cuGetProcAddress_v2(symbol: "cuD3D10GetDirect3DDevice", pfn: 0x0, cudaVersion: 3020, flags: 0, symbolStatus: NULL) -> CUDA_ERROR_NOT_FOUND
[ZLUDA_TRACE] cuGetProcAddress_v2(symbol: "cuGraphicsD3D10RegisterResource", pfn: 0x0, cudaVersion: 3000, flags: 0, symbolStatus: NULL) -> CUDA_ERROR_NOT_FOUND
[ZLUDA_TRACE] cuGetProcAddress_v2(symbol: "cuD3D10RegisterResource", pfn: 0x0, cudaVersion: 2010, flags: 0, symbolStatus: NULL) -> CUDA_ERROR_NOT_FOUND
[ZLUDA_TRACE] cuGetProcAddress_v2(symbol: "cuD3D10UnregisterResource", pfn: 0x0, cudaVersion: 2010, flags: 0, symbolStatus: NULL) -> CUDA_ERROR_NOT_FOUND
[ZLUDA_TRACE] cuGetProcAddress_v2(symbol: "cuD3D10MapResources", pfn: 0x0, cudaVersion: 2010, flags: 0, symbolStatus: NULL) -> CUDA_ERROR_NOT_FOUND
[ZLUDA_TRACE] cuGetProcAddress_v2(symbol: "cuD3D10UnmapResources", pfn: 0x0, cudaVersion: 2010, flags: 0, symbolStatus: NULL) -> CUDA_ERROR_NOT_FOUND
[ZLUDA_TRACE] cuGetProcAddress_v2(symbol: "cuD3D10ResourceSetMapFlags", pfn: 0x0, cudaVersion: 2010, flags: 0, symbolStatus: NULL) -> CUDA_ERROR_NOT_FOUND
[ZLUDA_TRACE] cuGetProcAddress_v2(symbol: "cuD3D10ResourceGetSurfaceDimensions", pfn: 0x0, cudaVersion: 3020, flags: 0, symbolStatus: NULL) -> CUDA_ERROR_NOT_FOUND
[ZLUDA_TRACE] cuGetProcAddress_v2(symbol: "cuD3D10ResourceGetMappedArray", pfn: 0x0, cudaVersion: 2010, flags: 0, symbolStatus: NULL) -> CUDA_ERROR_NOT_FOUND
[ZLUDA_TRACE] cuGetProcAddress_v2(symbol: "cuD3D10ResourceGetMappedPointer", pfn: 0x0, cudaVersion: 3020, flags: 0, symbolStatus: NULL) -> CUDA_ERROR_NOT_FOUND
[ZLUDA_TRACE] cuGetProcAddress_v2(symbol: "cuD3D10ResourceGetMappedSize", pfn: 0x0, cudaVersion: 3020, flags: 0, symbolStatus: NULL) -> CUDA_ERROR_NOT_FOUND
[ZLUDA_TRACE] cuGetProcAddress_v2(symbol: "cuD3D10ResourceGetMappedPitch", pfn: 0x0, cudaVersion: 3020, flags: 0, symbolStatus: NULL) -> CUDA_ERROR_NOT_FOUND
[ZLUDA_TRACE] cuGetProcAddress_v2(symbol: "cuD3D9GetDevice", pfn: 0x0, cudaVersion: 2000, flags: 0, symbolStatus: NULL) -> CUDA_ERROR_NOT_FOUND
[ZLUDA_TRACE] cuGetProcAddress_v2(symbol: "cuD3D9GetDevices", pfn: 0x0, cudaVersion: 3020, flags: 0, symbolStatus: NULL) -> CUDA_ERROR_NOT_FOUND
[ZLUDA_TRACE] cuGetProcAddress_v2(symbol: "cuD3D9CtxCreateOnDevice", pfn: 0x0, cudaVersion: 3020, flags: 0, symbolStatus: NULL) -> CUDA_ERROR_NOT_FOUND
[ZLUDA_TRACE] cuGetProcAddress_v2(symbol: "cuD3D9GetDirect3DDevice", pfn: 0x0, cudaVersion: 2000, flags: 0, symbolStatus: NULL) -> CUDA_ERROR_NOT_FOUND
[ZLUDA_TRACE] cuGetProcAddress_v2(symbol: "cuGraphicsD3D9RegisterResource", pfn: 0x0, cudaVersion: 3000, flags: 0, symbolStatus: NULL) -> CUDA_ERROR_NOT_FOUND
[ZLUDA_TRACE] cuGetProcAddress_v2(symbol: "cuD3D9RegisterResource", pfn: 0x0, cudaVersion: 2000, flags: 0, symbolStatus: NULL) -> CUDA_ERROR_NOT_FOUND
[ZLUDA_TRACE] cuGetProcAddress_v2(symbol: "cuD3D9UnregisterResource", pfn: 0x0, cudaVersion: 2000, flags: 0, symbolStatus: NULL) -> CUDA_ERROR_NOT_FOUND
[ZLUDA_TRACE] cuGetProcAddress_v2(symbol: "cuD3D9MapResources", pfn: 0x0, cudaVersion: 2000, flags: 0, symbolStatus: NULL) -> CUDA_ERROR_NOT_FOUND
[ZLUDA_TRACE] cuGetProcAddress_v2(symbol: "cuD3D9UnmapResources", pfn: 0x0, cudaVersion: 2000, flags: 0, symbolStatus: NULL) -> CUDA_ERROR_NOT_FOUND
[ZLUDA_TRACE] cuGetProcAddress_v2(symbol: "cuD3D9ResourceSetMapFlags", pfn: 0x0, cudaVersion: 2000, flags: 0, symbolStatus: NULL) -> CUDA_ERROR_NOT_FOUND
[ZLUDA_TRACE] cuGetProcAddress_v2(symbol: "cuD3D9ResourceGetSurfaceDimensions", pfn: 0x0, cudaVersion: 3020, flags: 0, symbolStatus: NULL) -> CUDA_ERROR_NOT_FOUND
[ZLUDA_TRACE] cuGetProcAddress_v2(symbol: "cuD3D9ResourceGetMappedArray", pfn: 0x0, cudaVersion: 2010, flags: 0, symbolStatus: NULL) -> CUDA_ERROR_NOT_FOUND
[ZLUDA_TRACE] cuGetProcAddress_v2(symbol: "cuD3D9ResourceGetMappedPointer", pfn: 0x0, cudaVersion: 3020, flags: 0, symbolStatus: NULL) -> CUDA_ERROR_NOT_FOUND
[ZLUDA_TRACE] cuGetProcAddress_v2(symbol: "cuD3D9ResourceGetMappedSize", pfn: 0x0, cudaVersion: 3020, flags: 0, symbolStatus: NULL) -> CUDA_ERROR_NOT_FOUND
[ZLUDA_TRACE] cuGetProcAddress_v2(symbol: "cuD3D9ResourceGetMappedPitch", pfn: 0x0, cudaVersion: 3020, flags: 0, symbolStatus: NULL) -> CUDA_ERROR_NOT_FOUND
[ZLUDA_TRACE] cuGetProcAddress_v2(symbol: "cuD3D9Begin", pfn: 0x0, cudaVersion: 2000, flags: 0, symbolStatus: NULL) -> CUDA_ERROR_NOT_FOUND
[ZLUDA_TRACE] cuGetProcAddress_v2(symbol: "cuD3D9RegisterVertexBuffer", pfn: 0x0, cudaVersion: 2000, flags: 0, symbolStatus: NULL) -> CUDA_ERROR_NOT_FOUND
[ZLUDA_TRACE] cuGetProcAddress_v2(symbol: "cuD3D9UnregisterVertexBuffer", pfn: 0x0, cudaVersion: 2000, flags: 0, symbolStatus: NULL) -> CUDA_ERROR_NOT_FOUND
[ZLUDA_TRACE] cuGetProcAddress_v2(symbol: "cuD3D9MapVertexBuffer", pfn: 0x0, cudaVersion: 3020, flags: 0, symbolStatus: NULL) -> CUDA_ERROR_NOT_FOUND
[ZLUDA_TRACE] cuGetProcAddress_v2(symbol: "cuD3D9UnmapVertexBuffer", pfn: 0x0, cudaVersion: 2000, flags: 0, symbolStatus: NULL) -> CUDA_ERROR_NOT_FOUND
[ZLUDA_TRACE] cuGetProcAddress_v2(symbol: "cuD3D9End", pfn: 0x0, cudaVersion: 2000, flags: 0, symbolStatus: NULL) -> CUDA_ERROR_NOT_FOUND

而其他CUDA参数都正常，会显示：
CUDA_SUCCESS



综上，目前来说Lada应该不可能运行在Amd的显卡之上。
或者zluda能修复一下？看上面的错误都是无法识别基础D3D设备的错误，说白了就是能够zluda能够调用到以A卡模拟CUDA的核心功能，但是却模拟不出一张N卡。
说人话就是……类似你可以听到对方的声音，但是看不到对方人在哪里。

我猜是这样吧~留给其它大神分析了。
我试了下自己喜欢的影片，不得不说……
Lada真牛！

ux22

a卡只能用来玩游戏

badhpanda

你该明白CUDA是N卡的技术，A卡类似的技术是ROCm和HIP，但是没什么应用生态。

hejack178

一直想试试国内品牌摩尔线程的显卡。据说他家的显卡可以兼容CUDA。有网友测试AI生产方面速度还很快。不知道用在这些软件上怎么样。

malegit

感谢分享

unitranscn

rx5600能不能用，有没有整合版。谢谢

xcooi

感谢分享