BlasTransposedMatrixBatchTmpl

Namespace: SymTensor.Compiler.Cuda
Parent Module: ArgTemplates

BLAS view of ArrayND. The ArrayND is implicitly transposed and exposed as a "(float *)[]". All but the last two dimensions are exposed as batches.

Constructor	Description
`new(manikin, memAllocator)` Signature: (manikin:ArrayNDManikinT * memAllocator:MemAllocatorT) -> BlasTransposedMatrixBatchTmpl	CompiledName: `.ctor`
`new(...)` Signature: (manikin:ArrayNDManikinT * ptrAryDevMem:MemManikinT * ptrAryHostMem:MemManikinT) -> BlasTransposedMatrixBatchTmpl	CompiledName: `.ctor`

Instance member	Description
`Columns` Signature: int64	CompiledName: `get_Columns`
`GetColumnsForOp(op)` Signature: op:Operation -> int64
`GetPointerArrayDevice(env)` Signature: env:CudaExecEnvT -> CudaDeviceVariable<CUdeviceptr>
`GetPointerArrayHost(env)` Signature: env:CudaExecEnvT -> RegHostMemT
`GetPointerArrayValues(env)` Signature: env:CudaExecEnvT -> CUdeviceptr []
`GetRowsForOp(op)` Signature: op:Operation -> int64
`LeadingDimension` Signature: int64	CompiledName: `get_LeadingDimension`
`Manikin` Signature: ArrayNDManikinT	CompiledName: `get_Manikin`
`NSamples` Signature: int64	CompiledName: `get_NSamples`
`PointerArrayCacheKey(env)` Signature: env:CudaExecEnvT -> CUdeviceptr * int64
`PointerArrayCacheKeyOnDevice()` Signature: unit -> (CUdeviceptr * int64) option	CompiledName: `set_PointerArrayCacheKeyOnDevice`
`PointerArrayCacheKeyOnDevice()` Signature: unit -> unit	CompiledName: `get_PointerArrayCacheKeyOnDevice`
`Rows` Signature: int64	CompiledName: `get_Rows`