TQUANT¶
指令示意图¶
简介¶
量化 Tile(例如 FP32 到 FP8),生成指数/缩放/最大值输出。
数学语义¶
除非另有说明, semantics are defined over the valid region and target-dependent behavior is marked as implementation-defined.
汇编语法¶
PTO-AS 形式:参见 PTO-AS Specification.
AS Level 1 (SSA)¶
%dst = pto.tquant %src, %qp : (!pto.tile<...>, !pto.tile<...>) -> !pto.tile<...>
AS Level 2 (DPS)¶
pto.tquant ins(%src, %qp : !pto.tile_buf<...>, !pto.tile_buf<...>) outs(%dst : !pto.tile_buf<...>)
AS Level 1(SSA)¶
%dst = pto.tquant %src, %qp : (!pto.tile<...>, !pto.tile<...>) -> !pto.tile<...>
AS Level 2(DPS)¶
pto.tquant ins(%src, %qp : !pto.tile_buf<...>, !pto.tile_buf<...>) outs(%dst : !pto.tile_buf<...>)
C++ 内建接口¶
声明于 include/pto/common/pto_instr.hpp:
template <auto quant_type, typename TileDataOut, typename TileDataSrc, typename TileDataExp, typename TileDataMax,
typename... WaitEvents>
PTO_INST RecordEvent TQUANT(TileDataOut &dst, TileDataSrc &src, TileDataExp *exp, TileDataMax *max, TileDataSrc *scaling, WaitEvents &... events);
template <auto quant_type, auto store_mode, typename TileDataOut, typename TileDataSrc, typename TileDataExp,
typename TileDataMax, typename TileDataIdx, typename... WaitEvents>
PTO_INST RecordEvent TQUANT(TileDataOut &dst, TileDataSrc &src, TileDataExp *exp, TileDataMax *max, TileDataSrc *scaling, TileDataExp *exp_zz, TileDataIdx *vgather_idx, WaitEvents &... events);
template <auto quant_type, typename TileDataOut, typename TileDataSrc, typename TileDataPara, typename... WaitEvents>
PTO_INST RecordEvent TQUANT(TileDataOut &dst, TileDataSrc &src, TileDataPara &scale, TileDataPara *offset = nullptr, WaitEvents &... events);
约束¶
- This instruction is currently implemented for specific targets (see
include/pto/npu/*/TQuant.hpp). - Input type requirements and output tile types are mode/target-dependent.
示例¶
See related examples in docs/isa/ and docs/coding/tutorials/.