CANN/PTO乘加运算指令

# TMULADDDST

【免费下载链接】pto-isaParallel Tile Operation (PTO) is a virtual instruction set architecture designed by Ascend CANN, focusing on tile-level operations. This repository offers high-performance, cross-platform tile operations across Ascend platforms.项目地址: https://gitcode.com/cann/pto-isa

Tile Operation Diagram

Introduction

Elementwise operation:src0 * src1 + dst.

Math Interpretation

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

$$ \mathrm{dst}{i,j} = \mathrm{src0}{i,j} * \mathrm{src1}{i,j} + \mathrm{dst}{i,j} $$

Assembly Syntax

Synchronous form:

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

AS Level 1 (SSA)

%dst = pto.tmuladddst %src0, %src1 : (!pto.tile<...>, !pto.tile<...>) -> !pto.tile<...>

AS Level 2 (DPS)

pto.tmuladddst ins(%src0, %src1 : !pto.tile_buf<...>, !pto.tile_buf<...>) outs(%dst : !pto.tile_buf<...>)

C++ Intrinsic

Declared ininclude/pto/common/pto_instr.hpp:

template <typename TileDataDst, typename TileDataSrc0, typename TileDataSrc1, typename... WaitEvents> PTO_INST RecordEvent TMULADDDST(TileDataDst &dst, TileDataSrc0 &src0, TileDataSrc1 &src1, WaitEvents &...events);

Constraints

• Implementation checks:
  • TileData::DTypemust be one of:float,half.
  • Tile layout must be row-major (TileData::isRowMajor).
• Valid region:
  • The op usesdst.GetValidRow()/dst.GetValidCol()as the iteration domain;src0/src1are assumed to be compatible (not validated by explicit runtime checks in this op).
• The op iterates overdst.GetValidRow()/dst.GetValidCol().

Examples

#include <pto/pto-inst.hpp> using namespace pto; void example() { using TileT = Tile<TileType::Vec, float, 16, 16>; TileT a, b, out; TMULADDDST(out, a, b); }

ASM Form Examples

Auto Mode

# Auto mode: compiler/runtime-managed placement and scheduling. %dst = pto.tmuladddst %src0, %src1 : (!pto.tile<...>, !pto.tile<...>) -> !pto.tile<...>

Manual Mode

# Manual mode: resources must be bound explicitly before issuing the instruction. # Optional for tile operands: # pto.tassign %arg0, @tile(0x1000) # pto.tassign %arg1, @tile(0x2000) %dst = pto.tmuladddst %src0, %src1 : (!pto.tile<...>, !pto.tile<...>) -> !pto.tile<...>

PTO Assembly Form

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

【免费下载链接】pto-isaParallel Tile Operation (PTO) is a virtual instruction set architecture designed by Ascend CANN, focusing on tile-level operations. This repository offers high-performance, cross-platform tile operations across Ascend platforms.项目地址: https://gitcode.com/cann/pto-isa