TCOLEXPANDDIV

Instruction Diagram

Introduction

Column broadcast division: divides each column by a per-column scalar vector.

Mathematical Semantics

Assume R = dst.GetValidRow() and C = dst.GetValidCol(). Assume that s_j is the per-column scalar obtained from src1 (one value per column).

For 0 <= i < R and 0 <= j < C:

\mathrm{dst}_{i,j} = \mathrm{src0}_{i,j} / s_j

Assembly Syntax

Synchronous form:

%dst = tcolexpanddiv %src0, %src1 : !pto.tile<...>, !pto.tile<...> -> !pto.tile<...>

AS Level 1 (SSA)

%dst = pto.tcolexpanddiv %src0, %src1 : !pto.tile<...>, !pto.tile<...> -> !pto.tile<...>

AS Level 2 (DPS)

pto.tcolexpanddiv ins(%src0, %src1 : !pto.tile_buf<...>, !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 <typename TileDataDst, typename TileDataSrc0, typename TileDataSrc1, typename... WaitEvents>
PTO_INST RecordEvent TCOLEXPANDDIV(TileDataDst &dst, TileDataSrc0 &src0, TileDataSrc1 &src1, WaitEvents &... events);

Constraints

TileDataDst::DType and TileDataSrc1::DType must be one of the following: half, float for Atlas A2 training products/Atlas A2 inference products, Atlas A3 training products/Atlas A3 inference products, and Ascend 950PR/Ascend 950DT; int16, int32, uint16, uint32, bfloat16_t, int8, uint8 for Ascend 950PR/Ascend 950DT.
Tile shape/layout constraint (compile time): TileDataDst::isRowMajor.
src1 is expected to provide one scalar per column (that is, its effective shape must cover C values).
The exact layout/fractal constraints are destination-specific; see the backend header files under include/pto/npu/*/TColExpand*.hpp.

Examples

See the relevant examples in docs/isa/ and docs/coding/tutorials/.

ASM Examples

Automatic Mode

# Automatic mode: the compiler/runtime is responsible for resource placement and scheduling.
%dst = pto.tcolexpanddiv %src0, %src1 : !pto.tile<...>, !pto.tile<...> -> !pto.tile<...>

Manual Mode

# Manual mode: explicitly bind resources first, then issue the instruction.
# Optional (when the instruction contains tile operands):
# pto.tassign %arg0, @tile(0x1000)
# pto.tassign %arg1, @tile(0x2000)
%dst = pto.tcolexpanddiv %src0, %src1 : !pto.tile<...>, !pto.tile<...> -> !pto.tile<...>

PTO Assembly Form

%dst = tcolexpanddiv %src0, %src1 : !pto.tile<...>, !pto.tile<...> -> !pto.tile<...>
# AS Level 2 (DPS)
pto.tcolexpanddiv ins(%src0, %src1 : !pto.tile_buf<...>, !pto.tile_buf<...>) outs(%dst : !pto.tile_buf<...>)