rocSOLVER precision support

rocSOLVER precision support#

This section provides an overview of the numerical precision types supported by the rocSOLVER library.

This page lists the data types supported by the library itself and does not indicate hardware support. A type listed here is only usable if the GPU architecture also supports it; otherwise it is unsupported. For data type support across the other ROCm libraries and by GPU architecture, see the Data types and precision support page.

Supported data types overview#

The following table summarizes the input and output data types supported by rocSOLVER. For the precision-prefix naming convention used across the API, see the sections that follow.

Icon

Definition

✅

Fully supported as both an input and output type.

⚠️

Partially supported as an input or output type.

Data types not listed in the table below are not supported.

Data type

Support

float32

✅

float64

✅

cfloat32

✅

cfloat64

✅

Supported precision types#

rocSOLVER supports four primary precision types across its functions:

Type prefix

C++ type

Description

s

float

Single-precision real (32-bit)

d

double

Double-precision real (64-bit)

c

rocblas_float_complex

Single-precision complex (32-bit real, 32-bit imaginary)

z

rocblas_double_complex

Double-precision complex (64-bit real, 64-bit imaginary)

Function naming convention#

rocSOLVER follows the standard LAPACK naming convention where the first letter of the function name indicates the precision type:

  • Functions beginning with rocsolver_s operate on single-precision real data.

  • Functions beginning with rocsolver_d operate on double-precision real data.

  • Functions beginning with rocsolver_c operate on single-precision complex data.

  • Functions beginning with rocsolver_z operate on double-precision complex data.

For example, the LU factorization function getrf is implemented as:

  • rocsolver_sgetrf - For single-precision real matrices

  • rocsolver_dgetrf - For double-precision real matrices

  • rocsolver_cgetrf - For single-precision complex matrices

  • rocsolver_zgetrf - For double-precision complex matrices

In the documentation, these are often represented generically as rocsolver_<type>getrf(), where <type> is a placeholder for the precision type prefix.

Understanding precision in function signatures#

In the function signatures throughout the documentation, precision information is indicated directly in the parameter types. For example:

rocblas_status rocsolver_slarfb(rocblas_handle handle, /* ... */
                                float *v, /* ... */
                                float *t, /* ... */
                                float *a, /* ... */)

The parameter types (float, double, rocblas_float_complex, or rocblas_double_complex) correspond to the function prefix and indicate the precision used by that specific function variant.

Real versus complex precision#

Some LAPACK functions have different behaviors or names when operating on real versus complex data:

  • Functions for symmetric matrices (prefix sy) use the same name for both real precision types.

  • Functions for Hermitian matrices (prefix he) are used for complex precision types.

  • Some auxiliary routines might be specific to real or complex precision types.

For example, rocsolver_ssytrd and rocsolver_dsytrd handle real symmetric matrices, while rocsolver_chetrd and rocsolver_zhetrd handle complex Hermitian matrices.