Struct ec_gpu_gen::SourceBuilder

pub struct SourceBuilder { /* private fields */ }

Builder to create the source code of a GPU kernel.

Example

use blstrs::{Fp, Fp2, G1Affine, G2Affine, Scalar};
use ec_gpu_gen::SourceBuilder;

let source = SourceBuilder::new()
    .add_fft::<Scalar>()
    .add_multiexp::<G1Affine, Fp>()
    .add_multiexp::<G2Affine, Fp2>()
    .build_32_bit_limbs();

Implementations

impl SourceBuilder

pub fn new() -> Self

Create a new configuration to generation a GPU kernel.

pub fn add_field<F>(self) -> Selfwhere
    F: GpuField + 'static,

Add a field to the configuration.

If it is an extension field, then the extension field and the sub-field is added.

pub fn add_fft<F>(self) -> Selfwhere
    F: GpuField + 'static,

Add an FFT kernel function to the configuration.

pub fn add_multiexp<C, F>(self) -> Selfwhere
    C: PrimeCurveAffine + GpuName,
    C::Scalar: GpuField,
    F: GpuField + 'static,

Add an Multiexp kernel function to the configuration.

The field must be given explicitly as currently it cannot derived from the curve point directly.

pub fn append_source(self, source: String) -> Self

Appends some given source at the end of the generated source.

This is useful for cases where you use this library as building block, but have your own kernel implementation. If this function is is called several times, then those sources are appended in that call order.

pub fn build_32_bit_limbs(&self) -> String

Generate the GPU kernel source code based on the current configuration with 32-bit limbs.

On CUDA 32-bit limbs are recommended.

pub fn build_64_bit_limbs(&self) -> String

Generate the GPU kernel source code based on the current configuration with 64-bit limbs.

On OpenCL 32-bit limbs are recommended.

Trait Implementations

impl Default for SourceBuilder

fn default() -> Self

Returns the “default value” for a type.