# AscendC passes
* [`-ascendc-allocate-tensor`](#ascendc-allocate-tensor)
* [`-ascendc-compute-memory-consumption`](#ascendc-compute-memory-consumption)
* [`-ascendc-declare-py-struct`](#ascendc-declare-py-struct)
* [`-ascendc-define-cube-only`](#ascendc-define-cube-only)
* [`-ascendc-detect-enable-debug`](#ascendc-detect-enable-debug)
* [`-ascendc-detect-kernel-type`](#ascendc-detect-kernel-type)
* [`-ascendc-erase-sync`](#ascendc-erase-sync)
* [`-ascendc-fill-asc-operands`](#ascendc-fill-asc-operands)
* [`-ascendc-fuse-bufid-sync`](#ascendc-fuse-bufid-sync)
* [`-ascendc-generate-boilerplate`](#ascendc-generate-boilerplate)
* [`-ascendc-hoist-que-bind`](#ascendc-hoist-que-bind)
* [`-ascendc-hoist-ub-allocation`](#ascendc-hoist-ub-allocation)
* [`-ascendc-input-output-tensor`](#ascendc-input-output-tensor)
* [`-ascendc-insert-bufid-sync`](#ascendc-insert-bufid-sync)
* [`-ascendc-insert-sync`](#ascendc-insert-sync)
* [`-ascendc-legalize-kernel-args`](#ascendc-legalize-kernel-args)
* [`-ascendc-lower-to-l0`](#ascendc-lower-to-l0)
* [`-ascendc-materialize-tensor`](#ascendc-materialize-tensor)
* [`-ascendc-noop`](#ascendc-noop)
* [`-ascendc-privatize-func`](#ascendc-privatize-func)
* [`-ascendc-reuse-ub-allocation`](#ascendc-reuse-ub-allocation)
* [`-ascendc-unify-pipe`](#ascendc-unify-pipe)
* [`-ascendc-verify-sync`](#ascendc-verify-sync)
<!-- Autogenerated by mlir-tblgen; don't manually edit -->
## `-ascendc-allocate-tensor`

_Allocate tensors for LocalTensorAuto_


## `-ascendc-compute-memory-consumption`

_Compute memory consumption for kernel functions_


## `-ascendc-declare-py-struct`

_Insert emitasc.declare_py_struct_


## `-ascendc-define-cube-only`

_Insert emitc.define if CUBE-ONLY_


## `-ascendc-detect-enable-debug`

_Check whether the kernel is using debug utils_


## `-ascendc-detect-kernel-type`

_Check whether the kernel is vector-only or mixed_


## `-ascendc-erase-sync`

_Erase intra-core synchronization operations_


## `-ascendc-fill-asc-operands`

_Filling AscendC operands_


## `-ascendc-fuse-bufid-sync`

_Erase unnecessary get_buf and rls_buf_


## `-ascendc-generate-boilerplate`

_Insert emitc.include and additional boilerplate code_


## `-ascendc-hoist-que-bind`

_Hoist TQueBind, TQue, TBuf initialization operations_


## `-ascendc-hoist-ub-allocation`

_Hoist tensor allocations to the function root_



### Options
```
-exclude-in-out : Prevent input/output tensors from hoisting
```
## `-ascendc-input-output-tensor`

_Set input and output operand for local_tensor_auto_


## `-ascendc-insert-bufid-sync`

_Insert get_buf and rls_buf synchronization_


## `-ascendc-insert-sync`

_Insert intra-core synchronization operations_


## `-ascendc-legalize-kernel-args`

_Attach emitasc.kernel_arg attributes and insert operations_



### Options
```
-set-ffts-addr : Append ffts_addr to kernel arguments and call set_ffts_addr
```
## `-ascendc-lower-to-l0`

_Lowering L2 to L0 Ascend C operations_


## `-ascendc-materialize-tensor`

_Insert ascendc.tbuf, ascendc.queue and ascendc.alloca for local_tensor_auto_



### Options
```
-always-buf : Always use TBuf to materialize tensors
```
## `-ascendc-noop`

_This pass does nothing_


## `-ascendc-privatize-func`

_Mark functions without ascendc.global attribute as private_


## `-ascendc-reuse-ub-allocation`

_Perform analysis and reuse temporary UB allocations_



### Options
```
-reuse-in-out : Allow input/output tensors to be reusable
```
## `-ascendc-unify-pipe`

_Unify pipe opertation_


## `-ascendc-verify-sync`

_Verify TQue synchronization_