TCMP

Instruction Diagram

Introduction

Compares two tiles and writes a packed predicate mask.

Mathematical Semantics

Conceptually, for each element (i, j) in the valid region, a predicate is defined:

$$p_{i,j} = \left(\mathrm{src0}_{i,j}\ \mathrm{cmpMode}\ \mathrm{src1}_{i,j}\right)$$

The predicate mask is stored in dst using an implementation-defined packed layout.

Assembly Syntax

Synchronous form:

%dst = tcmp %src0, %src1 {cmpMode = #pto.cmp<EQ>} : !pto.tile<...> -> !pto.tile<...>

AS Level 1 (SSA)

%dst = pto.tcmp %src0, %src1{cmpMode = #pto<cmp xx>}: (!pto.tile<...>, !pto.tile<...>) -> !pto.tile<...>

AS Level 2 (DPS)

pto.tcmp ins(%src0, %src1{cmpMode = #pto<cmp xx>}: !pto.tile_buf<...>, !pto.tile_buf<...>) outs(%dst : !pto.tile_buf<...>)

C++ Built-in APIs

Declared in include/pto/common/pto_instr.hpp and include/pto/common/type.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 TileDataSrc, typename... WaitEvents>
PTO_INST RecordEvent TCMP(TileDataDst &dst, TileDataSrc &src0, TileDataSrc &src1, CmpMode cmpMode, WaitEvents &... events);

Constraints

• Implementation check (Atlas A2/A3 training products/Atlas A2/A3 inference products):
  • The input type must be one of the following: int32_t, half, float.
  • The output type must be uint8_t.
  • The src0/src1/dst tile position must be TileType::Vec.
  • Static valid bounds: TileDataSrc::ValidRow <= TileDataSrc::Rows and TileDataSrc::ValidCol <= TileDataSrc::Cols.
  • Runtime: src0.GetValidRow() == dst.GetValidRow() and src0.GetValidCol() == dst.GetValidCol().
  • Note: The shape/validity of src1 is not verified through explicit runtime assertions in this implementation.
  • For TileDataSrc::DType == int32_t, the implementation uses the EQ comparison path regardless of cmpMode.
• Implementation check (Ascend 950PR/Ascend 950DT):
  • The input type must be one of the following: uint32_t, int32_t, uint16_t, int16_t, uint8_t, int8_t, float, half, bfloat16_t.
  • The output type must be uint32_t.
  • Implemented (see include/pto/npu/a5/TCmp.hpp).
  • The Ascend 950PR/Ascend 950DT implementation uses dst.GetValidRow() / dst.GetValidCol() as the iteration domain and writes the packed predicate mask to dst (the destination-defined packing scheme).
• Mask encoding:
  • The mask tile is interpreted as packed predicate bits in the destination-defined layout.

Examples

Automatic

#include <pto/pto-inst.hpp>

using namespace pto;

void example_auto() {
  using SrcT = Tile<TileType::Vec, float, 16, 16>;
  using MaskT = Tile<TileType::Vec, uint8_t, 16, 32, BLayout::RowMajor, -1, -1>;
  SrcT src0, src1;
  MaskT mask(16, 2);
  TCMP(mask, src0, src1, CmpMode::GT);
}

Manual

#include <pto/pto-inst.hpp>

using namespace pto;

void example_manual() {
  using SrcT = Tile<TileType::Vec, float, 16, 16>;
  using MaskT = Tile<TileType::Vec, uint8_t, 16, 32, BLayout::RowMajor, -1, -1>;
  SrcT src0, src1;
  MaskT mask(16, 16);
  TASSIGN(src0, 0x1000);
  TASSIGN(src1, 0x2000);
  TASSIGN(mask, 0x3000);
  TCMP(mask, src0, src1, CmpMode::GT);
}

ASM Examples

Automatic Mode

# Automatic mode: the compiler/runtime handles resource placement and scheduling.
%dst = pto.tcmp %src0, %src1{cmpMode = #pto<cmp xx>}: (!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.tcmp %src0, %src1{cmpMode = #pto<cmp xx>}: (!pto.tile<...>, !pto.tile<...>) -> !pto.tile<...>

PTO Assembly Form

%dst = tcmp %src0, %src1 {cmpMode = #pto.cmp<EQ>} : !pto.tile<...> -> !pto.tile<...>
# AS Level 2 (DPS)
pto.tcmp ins(%src0, %src1{cmpMode = #pto<cmp xx>}: !pto.tile_buf<...>, !pto.tile_buf<...>) outs(%dst : !pto.tile_buf<...>)