59 #if defined(CUTLASS_ARCH_WMMA_SM72_ENABLED)    61   using ElementA = int8_t;
    62   using LayoutA = LayoutA_;
    63   using ElementB = int8_t;
    64   using LayoutB = LayoutB_;
    65   using ElementC = int32_t;
    66   using LayoutC = LayoutC_;
    67   using Operator = cutlass::arch::OpMultiplyAdd;
    74     "Supported list of wmma operator shape for s8 multiplicands are: 16x16x16, 8x328x16, and 32x8x16");
    78   using FragmentA = nvcuda::wmma::fragment<
    79           nvcuda::wmma::matrix_a,
    83           typename CutlassToWmmaDataType<ElementA>::Type,
    84           typename CutlassToWmmaLayout<LayoutA>::Layout>;
    86   using FragmentB = nvcuda::wmma::fragment<
    87           nvcuda::wmma::matrix_b,
    91           typename CutlassToWmmaDataType<ElementB>::Type,
    92           typename CutlassToWmmaLayout<LayoutB>::Layout>;
    94   using FragmentC = nvcuda::wmma::fragment<
    95           nvcuda::wmma::accumulator,
    99           typename CutlassToWmmaDataType<ElementC>::Type>;
   107     FragmentC 
const &C)
 const {
   109       nvcuda::wmma::mma_sync(D, A, B, C);
   113     static_assert(
false, 
"wmma.mma.sync interger type multiplicands is avialable only for SM72 and beyond");
   139 #if defined(CUTLASS_ARCH_WMMA_SM72_ENABLED)   140   using Shape = Shape_;
   141   using ElementA = uint8_t;
   142   using LayoutA = LayoutA_;
   143   using ElementB = uint8_t;
   144   using LayoutB = LayoutB_;
   145   using ElementC = int32_t;
   146   using LayoutC = LayoutC_;
   147   using Operator = cutlass::arch::OpMultiplyAdd;
   154     "Supported list of wmma operator shape for u8 multiplicands are: 16x16x16, 8x328x16, and 32x8x16");
   157   using FragmentA = nvcuda::wmma::fragment<
   158           nvcuda::wmma::matrix_a,
   162           typename CutlassToWmmaDataType<ElementA>::Type,
   163           typename CutlassToWmmaLayout<LayoutA>::Layout>;
   165   using FragmentB = nvcuda::wmma::fragment<
   166           nvcuda::wmma::matrix_b,
   170           typename CutlassToWmmaDataType<ElementB>::Type,
   171           typename CutlassToWmmaLayout<LayoutB>::Layout>;
   173   using FragmentC = nvcuda::wmma::fragment<
   174           nvcuda::wmma::accumulator,
   178           typename CutlassToWmmaDataType<ElementC>::Type>;
   186     FragmentC 
const &C)
 const {
   188       nvcuda::wmma::mma_sync(D, A, B, C);
   192     static_assert(
false, 
"wmma.mma.sync interger type multiplicands is avialable only for SM72 and beyond");
 Definition: aligned_buffer.h:35
Shape of a matrix multiply-add operation. 
Definition: include/cutlass/gemm/gemm.h:57
Defines layout functions used by TensorRef and derived classes.