spdecompress GPU Virtual ISA NVIDIA

Data Movement and Conversion Instructions

spdecompress.elemsize.idxsize.spfactor.num data, mdata, cdata;

The spdecompress instruction copies elements from structured sparse compressed vector cdata to form the dense vector data.

Encoding

PTX is a virtual instruction set. It has no single, stable native binary encoding - the compiler lowers this instruction to different native machine code depending on the selected NVIDIA target architecture (compute capability). This page intentionally shows no bit-diagram; see the target/version requirements below for what governs how this instruction compiles.
PTX ISA Version Introduced PTX ISA 9.4
Minimum Target sm_107a

Syntax Forms

One mnemonic covers many type / state-space / scope / modifier combinations - each row below is an independently valid form.

Syntax Data Types State Space(s) Modifiers Min. Target Description
spdecompress.elemsize.idxsize.spfactor.num data, mdata, cdata; sm_107a The spdecompress instruction copies elements from structured sparse compressed vector cdata to form the dense vector data. (see the official PTX ISA docs for the full description)

Operands

  • data
    Operand
  • mdata
    Operand
  • cdata
    Operand

At a Glance

Data Types -

Related

More in Data Movement and Conversion Instructions

Reference

NVIDIA PTX ISA

Description

The spdecompress instruction copies elements from structured sparse compressed vector cdata to form the dense vector data. The metadata information corresponding to the structured sparse compressed vector cdata is given by the vector operand mdata. The decompression factor is derived from the qualifier .spfactor. The qualifier is of form .sp::num_src:num_target implying instruction decompresses each sequence of num_src elements so as to form the sequence comprising num_target many elements. Given decompress operation, it naturally follows that the num_src value must be strictly numerically smaller than the num_target value. (see the official PTX ISA docs for the full description)

Semantics

for N in .num: for S in num_source: data[mdata[N][S]] = cdata[N][S];

Examples

.reg .b32 mdata_<4>, cdata_<32>, data_<32>;

spdecompress.b8.b2.sp::2:4.x32 {data_0, data_1, data_2, data_3, data_4, data_5, data_6, data_7, data_8,
                                data_9, data_10, data_11, data_12, data_13, data_14, data_15, data_17,
                                data_18, data_19, data_20, data_21, data_22, data_23, data_24, data_25,
                                data_26, data_27, data_28, data_29, data_30, data_31},
// (truncated - see the official PTX ISA docs for the full example)

Reproduced from NVIDIA's official PTX ISA documentation for technical accuracy.

Sources