完成InfiniCore的Opencl版本算子，并且调通InfiniLM推理#3

Open

TRM-coding wants to merge 47 commits intoma-hang:openclfrom

TRM-coding:opencl-trm

Commits on Sep 15, 2025

issue/450: change indexToReducedOffset() to indexToOffset in elementwise framework on CPU, NVIDIA, Cambricon, Metax, Moore, and Kunlun
Ziminli
committed
issue/450: remove indexToReducedOffset() in all platforms
Ziminli
committed
issue/450: add the testcases that pinpoint the issue in infiniop-test
Ziminli
committed

Commits on Sep 16, 2025

Commits on Sep 17, 2025

issue/436: support kunlun rope U32

xgqdut2016
authored and
zhangyue207
committed
issue/436: 支持9g7b 4b模型
 zhangyue207
committed

Commits on Sep 19, 2025

fix: disable NVIDIA-dequantize on Iluvatar GPU via ENABLE_NVIDIA_API macro
spike-zhu
committed

Commits on Sep 23, 2025

Commits on Sep 24, 2025

Commits on Sep 26, 2025

Commits on Sep 28, 2025

可以运行推理的opencl算子
 TRM-coding
committed

Commits on Oct 1, 2025

update gemm add sub group
TRM-coding
committed

Commits on Oct 3, 2025

update
TRM-coding
committed

Commits on Oct 10, 2025