feat(ascend): add Ascend framework layer — runtime, type mapping, bui…#46

Merged

voltjia merged 19 commits intomasterfrom

feat/ascend-framework

Apr 14, 2026

Commits on Apr 13, 2026

feat(ascend): add Ascend framework layer — runtime, type mapping, build integration
zhangyue
committed
style(ascend): apply `clang-format` to framework headers
zhangyue
committed
fix(ascend): adapt `Memcpy`/`Memset` arity, assert workspace alloc, remove missing include
zhangyue
committed
feat(ascend): add GEMM kernel, NPU test infra, and example integration
zhangyue
committed
fix(ascend): move `aclrtMalloc` out of `assert()` in `WorkspacePool`
zhangyue
committed
fix(nvidia): restore `CUDA::cublasLt` link dependency
zhangyue
committed
feat(test): add `--devices` option to pytest for platform-name filtering
zhangyue
committed
fix(nvidia): add missing include and work around NVCC `std::forward` bug
zhangyue
committed
fix(ci): upgrade NVIDIA CI image to 25.12 and restore `std::forward`

zhangyue207
authored and
zhangyue
committed
fix: add explicit narrowing casts in `RotaryEmbedding` initializer list
zhangyue
committed
style: fix lint issues from PR review
zhangyue
committed
style: fix lint issues in `feat/ascend-framework`
zhangyue
committed
fix: address PR #46 review feedback
zhangyue
committed
docs(base): add vLLM interface references to `FlashAttention`, `ReshapeAndCache`, `RotaryEmbedding`
zhangyue
committed
fix(base): rename matmul.h to mat_mul.h to match MatMul class name
zhangyue
committed
fix(nvidia): revert std::forward workaround for NVCC 12.6 bug
zhangyue
committed
Revert "fix(nvidia): revert std::forward workaround for NVCC 12.6 bug"
zhangyue
committed

Commits on Apr 14, 2026