MGATHER¶
Tile Operation Diagram¶
Introduction¶
Gather-load elements from global memory into a tile using per-element indices.
Math Interpretation¶
For each element (i, j) in the destination valid region:
\[ \mathrm{dst}_{i,j} = \mathrm{mem}[\mathrm{idx}_{i,j}] \]
Assembly Syntax¶
PTO-AS form: see PTO-AS Specification.
Synchronous form:
%dst = mgather %mem, %idx : !pto.memref<...>, !pto.tile<...> -> !pto.tile<...>
AS Level 1 (SSA)¶
%dst = pto.mgather %mem, %idx : (!pto.partition_tensor_view<MxNxdtype>, pto.tile<...>)
-> !pto.tile<loc, dtype, rows, cols, blayout, slayout, fractal, pad>
AS Level 2 (DPS)¶
pto.mgather ins(%mem, %idx : !pto.partition_tensor_view<MxNxdtype>, !pto.tile_buf<...>) outs(%dst : !pto.tile_buf<...>)
C++ Intrinsic¶
Declared in include/pto/common/pto_instr.hpp:
template <typename TileDst, typename GlobalData, typename TileInd, typename... WaitEvents>
PTO_INST RecordEvent MGATHER(TileDst &dst, GlobalData &src, TileInd &indexes, WaitEvents &... events);
Constraints¶
- Supported data types:
dst/srcelement type must be one of:uint8_t,int8_t,uint16_t,int16_t,uint32_t,int32_t,half,bfloat16_t,float.- On AICore targets,
float8_e4m3_tandfloat8_e5m2_tare also supported. indexeselement type must beint32_toruint32_t.
- Tile and memory types:
dstmust be a vector tile (TileType::Vec).indexesmust be a vector tile (TileType::Vec).dstandindexesmust use row-major layout.srcmust be aGlobalTensorin GM memory.srcmust useNDlayout.
- Shape constraints:
dst.Rows == indexes.Rows.indexesmust be shaped as[N, 1]for row-indexed gather or[N, M]for element-indexed gather.dstrow width must be 32-byte aligned, that is,dst.Cols * sizeof(DType)must be a multiple of 32.srcstatic shape must satisfyShape<1, 1, 1, TableRows, RowWidth>.
- Index interpretation:
- Index interpretation is target-defined. The CPU simulator treats indices as linear element indices into
src.data(). - The CPU simulator does not enforce bounds checks on
indexes.
- Index interpretation is target-defined. The CPU simulator treats indices as linear element indices into
Examples¶
See related examples in docs/isa/ and docs/coding/tutorials/.
ASM Form Examples¶
Auto Mode¶
# Auto mode: compiler/runtime-managed placement and scheduling.
%dst = pto.mgather %mem, %idx : (!pto.partition_tensor_view<MxNxdtype>, pto.tile<...>)
Manual Mode¶
# Manual mode: bind resources explicitly before issuing the instruction.
# Optional for tile operands:
# pto.tassign %arg0, @tile(0x1000)
# pto.tassign %arg1, @tile(0x2000)
%dst = pto.mgather %mem, %idx : (!pto.partition_tensor_view<MxNxdtype>, pto.tile<...>)
PTO Assembly Form¶
%dst = mgather %mem, %idx : !pto.memref<...>, !pto.tile<...> -> !pto.tile<...>
# AS Level 2 (DPS)
pto.mgather ins(%mem, %idx : !pto.partition_tensor_view<MxNxdtype>, !pto.tile_buf<...>) outs(%dst : !pto.tile_buf<...>)