TDIVS

Instruction Diagram

Introduction

Performs element-wise division with a scalar (tile/scalar or scalar/tile).

Mathematical Semantics

For each element (i, j) in the valid region:

• Tile/scalar form:

  $$\mathrm{dst}_{i,j} = \frac{\mathrm{src}_{i,j}}{\mathrm{scalar}}$$

• Scalar/tile form:

  $$\mathrm{dst}_{i,j} = \frac{\mathrm{scalar}}{\mathrm{src}_{i,j}}$$

Assembly Syntax

Tile/scalar form:

%dst = tdivs %src, %scalar : !pto.tile<...>, f32

Scalar/tile form:

%dst = tdivs %scalar, %src : f32, !pto.tile<...>

AS Level 1 (SSA)

%dst = pto.tdivs %src, %scalar : (!pto.tile<...>, dtype) -> !pto.tile<...>
%dst = pto.tdivs %scalar, %src : (dtype, !pto.tile<...>) -> !pto.tile<...>

AS Level 2 (DPS)

pto.tdivs ins(%src, %scalar : !pto.tile_buf<...>, dtype) outs(%dst : !pto.tile_buf<...>)
pto.tdivs ins(%scalar, %src : dtype, !pto.tile_buf<...>) outs(%dst : !pto.tile_buf<...>)

C++ Built-in APIs

Declared in include/pto/common/pto_instr.hpp:

The public include header is <pto/pto-inst.hpp>, and the internal declaration is located in pto/common/pto_instr.hpp.

template <auto PrecisionType = DivAlgorithm::DEFAULT, typename TileDataDst, typename TileDataSrc,
          typename... WaitEvents>
PTO_INST RecordEvent TDIVS(TileDataDst &dst, TileDataSrc &src0, typename TileDataSrc::DType scalar,
                           WaitEvents &... events);

template <auto PrecisionType = DivAlgorithm::DEFAULT, typename TileDataDst, typename TileDataSrc,
          typename... WaitEvents>
PTO_INST RecordEvent TDIVS(TileDataDst &dst, typename TileDataDst::DType scalar, TileDataSrc &src0,
                           WaitEvents &... events)

PrecisionType can specify the following values:

• DivAlgorithm::DEFAULT: normal algorithm with high speed but low precision.
• DivAlgorithm::HIGH_PRECISION: high-precision algorithm with lower speed.

Constraints

• Implementation check (Atlas A2/A3 training products/Atlas A2/A3 inference products) (two overloads):
  • TileData::DType must be one of the following: int32_t, int, int16_t, half, float16_t, float, float32_t.
  • The Tile position must be a vector (TileData::Loc == TileType::Vec).
  • Static valid bounds: TileData::ValidRow <= TileData::Rows and TileData::ValidCol <= TileData::Cols.
  • Runtime: src0.GetValidRow() == dst.GetValidRow() and src0.GetValidCol() == dst.GetValidCol().
  • The tile layout must be row-major (TileData::isRowMajor).
• Implementation check (Ascend 950PR/Ascend 950DT) (two overloads):
  • TileData::DType must be one of the following: uint8_t, int8_t, uint16_t, int16_t, uint32_t, int32_t, half, float.
  • The tile position must be a vector (TileData::Loc == TileType::Vec).
  • Static valid bounds: TileData::ValidRow <= TileData::Rows and TileData::ValidCol <= TileData::Cols.
  • Runtime: src0.GetValidRow() == dst.GetValidRow() and src0.GetValidCol() == dst.GetValidCol().
  • The tile layout must be row-major (TileData::isRowMajor).
• Valid region:
  • The operation uses dst.GetValidRow()/dst.GetValidCol() as the iteration domain.
• Division by zero:
  • The behavior is target-defined; on Ascend 950PR/Ascend 950DT, the tile/scalar form maps to multiplication by the reciprocal, and uses 1/0 -> +inf for scalar == 0.
• High-precision algorithm
  • Valid only on Ascend 950PR/Ascend 950DT; the PrecisionType option is ignored on Atlas A3 training products/Atlas A3 inference products.

Examples

Automatic

#include <pto/pto-inst.hpp>

using namespace pto;

void example_auto() {
  using TileT = Tile<TileType::Vec, float, 16, 16>;
  TileT src, dst;
  TDIVS(dst, src, 2.0f);
  TDIVS<DivAlgorithm::HIGH_PRECISION>(dst, src, 2.0f);
}

Manual

#include <pto/pto-inst.hpp>

using namespace pto;

void example_manual() {
  using TileT = Tile<TileType::Vec, float, 16, 16>;
  TileT src, dst;
  TASSIGN(src, 0x1000);
  TASSIGN(dst, 0x2000);
  TDIVS(dst, 2.0f, src);
  TDIVS<DivAlgorithm::HIGH_PRECISION>(dst, 2.0f, src);
}

ASM Examples

Automatic Mode

# Automatic mode: the compiler/runtime handles resource placement and scheduling.
%dst = pto.tdivs %src, %scalar : (!pto.tile<...>, dtype) -> !pto.tile<...>

Manual Mode

# Manual mode: bind resources explicitly first, then issue the instruction.
# Optional (when the instruction contains a tile operand):
# pto.tassign %arg0, @tile(0x1000)
# pto.tassign %arg1, @tile(0x2000)
%dst = pto.tdivs %src, %scalar : (!pto.tile<...>, dtype) -> !pto.tile<...>

PTO Assembly Form

%dst = pto.tdivs %src, %scalar : (!pto.tile<...>, dtype) -> !pto.tile<...>
# AS Level 2 (DPS)
pto.tdivs ins(%src, %scalar : !pto.tile_buf<...>, dtype) outs(%dst : !pto.tile_buf<...>)