cp.async GPU Virtual ISA NVIDIA

cp.async Data Movement and Conversion Instructions

cp.async.ca.shared{::cta}.global{.level::cache_hint}{.level::prefetch_size}

cp.async is a non-blocking instruction which initiates an asynchronous copy operation of data from the location specified by source address operand src to the location specified by destination address operand dst.

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 7.0
Minimum Target sm_80

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
cp.async.ca.shared{::cta}.global{.level::cache_hint}{.level::prefetch_size} sm_80 cp.async is a non-blocking instruction which initiates an asynchronous copy operation of data from the location specified by source address operand src to the location specified by destination address… (see the official PTX ISA docs for the full description)

Operands

At a Glance

Data Types -

Related

More in Data Movement and Conversion Instructions

Reference

NVIDIA PTX ISA

Description

cp.async is a non-blocking instruction which initiates an asynchronous copy operation of data from the location specified by source address operand src to the location specified by destination address operand dst. Operand src specifies a location in the global state space and dst specifies a location in the shared state space. Operand cp-size is an integer constant which specifies the size of data in bytes to be copied to the destination dst. cp-size can only be 4, 8 and 16. Instruction cp.async allows optionally specifying a 32-bit integer operand src-size. Operand src-size represents the size of the data in bytes to be copied from src to dst and must be less than cp-size. (see the official PTX ISA docs for the full description)

Examples

cp.async.ca.shared.global  [shrd],    [gbl + 4], 4;
cp.async.ca.shared::cta.global  [%r0 + 8], [%r1],     8;
cp.async.cg.shared.global  [%r2],     [%r3],     16;

cp.async.cg.shared.global.L2::64B   [%r2],      [%r3],     16;
cp.async.cg.shared.global.L2::128B  [%r0 + 16], [%r1],     16;
// (truncated - see the official PTX ISA docs for the full example)

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

Sources