pulp::gpu_audio::detail Namespace Reference

Functions
bool	configure_gpu_convolver_trial (GpuConvolver &, const GpuConvolverTrialConfig &) noexcept
bool	drain_gpu_convolver_trial_records (GpuConvolver &, std::vector< SharedIoTraceRecord > &) noexcept
bool	drain_gpu_convolver_trial_admissions (GpuConvolver &, std::vector< SharedIoTraceAdmission > &) noexcept
std::uint64_t	gpu_convolver_trial_engine_id (const GpuConvolver &) noexcept
bool	prepare_gpu_multi_convolver_with (GpuMultiConvolver &, std::unique_ptr< render::GpuCompute >) noexcept
void	circular_box_blur (const std::vector< float > &m, std::vector< float > &out, std::uint32_t radius, float inv_kernel, std::uint32_t n)
	Circular box blur of width (2*radius+1) over m[0..n) into out[0..n), in O(n) via a sliding running sum (not O(n*radius) direct summation, and not a prefix sum — a windowed add-one/drop-one sum, which only ever holds the current window and so does not accumulate the precision loss of subtracting two large cumulative totals).

Function Documentation

circular_box_blur()

void pulp::gpu_audio::detail::circular_box_blur	(	const std::vector< float > &	m,
		std::vector< float > &	out,
		std::uint32_t	radius,
		float	inv_kernel,
		std::uint32_t	n )

Circular box blur of width (2*radius+1) over m[0..n) into out[0..n), in O(n) via a sliding running sum (not O(n*radius) direct summation, and not a prefix sum — a windowed add-one/drop-one sum, which only ever holds the current window and so does not accumulate the precision loss of subtracting two large cumulative totals).

radius must be < n/2 so the window is shorter than the circle and the add/drop bins never alias; smear_kernel bounds it at n/32. Exposed for unit testing against the direct summation.

configure_gpu_convolver_trial()

bool pulp::gpu_audio::detail::configure_gpu_convolver_trial ( GpuConvolver & , const GpuConvolverTrialConfig &  )

noexcept

drain_gpu_convolver_trial_admissions()

bool pulp::gpu_audio::detail::drain_gpu_convolver_trial_admissions ( GpuConvolver & , std::vector< SharedIoTraceAdmission > &  )

noexcept

drain_gpu_convolver_trial_records()

bool pulp::gpu_audio::detail::drain_gpu_convolver_trial_records ( GpuConvolver & , std::vector< SharedIoTraceRecord > &  )

noexcept

gpu_convolver_trial_engine_id()

std::uint64_t pulp::gpu_audio::detail::gpu_convolver_trial_engine_id ( const GpuConvolver & )

noexcept

prepare_gpu_multi_convolver_with()

bool pulp::gpu_audio::detail::prepare_gpu_multi_convolver_with ( GpuMultiConvolver & , std::unique_ptr< render::GpuCompute >  )

noexcept