cvt GPU Virtual ISA NVIDIA

Convert Data Movement and Conversion Instructions

cvt.dtype.atype d, a;

Convert a value between integer and/or floating-point types with an explicit rounding mode.

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 1.0
Minimum Target sm_10

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
cvt.dtype.atype d, a; s8, s16, s32, s64, u8, u16, u32, u64, f16, f32, f64 sm_10 Type conversion between the source (atype) and destination (dtype) types.
cvt.rn.f32.f64 d, a; f32, f64 rn, rz, rm, rp sm_10 Floating-point narrowing/widening conversion with explicit IEEE rounding mode.

Operands

  • d
    Destination register
  • a
    Source operand

At a Glance

Data Types f16, f32, f64, s16, s32, s64, s8, u16, u32, u64, u8
Modifiers rm, rn, rp, rz

Related AMDGPU Concepts

Type Conversion ↗
equivalent with restrictions
v_cvt_f32_i32 (AMDGPU)

Related

More in Data Movement and Conversion Instructions

Reference

NVIDIA PTX ISA

Description

Convert between different types and sizes. For.f16x2 and.bf16x2 instruction type, two inputs a and b of.f32 type are converted into.f16 or.bf16 type and the converted values are packed in the destination register d, such that the value converted from input a is stored in the upper half of d and the value converted from input b is stored in the lower half of d For.f16x2 instruction type, destination operand d has.f16x2 or.b32 type. For.bf16 instruction type, operand d has.b16 type. For.bf16x2 instruction type, operand d has.b32 type. For.tf32 instruction type, operand d has.b32 type. When converting to.e4m3x2 /.e5m2x2 data formats, the destination operand d has.b16 type. (see the official PTX ISA docs for the full description)

Semantics

d = convert(a, atype -> dtype) using the specified rounding mode; saturating forms clamp on overflow.

Examples

cvt.f32.s32 f,i;
cvt.s32.f64 j,r;     // float-to-int saturates by default
cvt.rni.f32.f32 x,y; // round to nearest int, result is fp
cvt.f32.f32 x,y;     // note .ftz behavior for sm_1x targets
cvt.rn.relu.f16.f32      b, f;        // result is saturated with .relu saturation mode
cvt.rz.f16x2.f32         b1, f, f1;   // convert two fp32 values to packed fp16 outputs
// (truncated - see the official PTX ISA docs for the full example)

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

Sources