tpglibs Namespace Reference

Namespaces
namespace	testapp

Classes
class	AbstractFactory
	General singleton, abstract factory. More...
class	AbstractProcessor
	Abstract signal processor. More...
class	AVXAbsRunSumProcessor
	AVX signal processor: Calculates the running sum of the signal's absolute value. More...
class	AVXFactory
	AVX typed abstract factory. More...
class	AVXFixedPedestalSubtractProcessor
	AVX signal processor: Fixes the estimated pedestal after a configured amount of time. More...
class	AVXFrugalPedestalSubtractProcessor
	AVX signal processor: Estimates the pedestal and subtracts. More...
class	AVXPipeline
	AVX typed TPG pipeline. More...
class	AVXProcessor
	AVX typed abstract signal processor. More...
class	AVXRunSumProcessor
	AVX signal processor: Calculates the running sum of the signal. More...
class	AVXThresholdProcessor
	AVX signal processor: Passes signals above a threshold. More...
class	NaiveAbsRunSumProcessor
	Naive signal processor: Calculates the running sum of the signal's absolute value. More...
class	NaiveFactory
	Naive typed abstract factory. More...
class	NaiveFrugalPedestalSubtractProcessor
	Naive signal processor: Estimates the pedestal and subtracts. More...
class	NaiveProcessor
	Naive typed abstract signal processor. More...
class	NaiveRunSumProcessor
	Naive signal processor: Calculates the running sum of the signal. More...
class	NaiveThresholdProcessor
	Naive signal processor: Passes signals above a threshold. More...
class	ProcessorInternalStateBufferManager
	Manages the internal state storage buffers for a processor. More...
class	ProcessorInternalStateNameRegistry
	Registry of internal state names. More...
struct	ProcessorMetricArray
	Dynamic array of processor metrics, templated on signal type. More...
class	TPGenerator
	TPG driving class. More...
class	TPGPipeline
	Abstract class for the TPG pipeline. More...

Functions
__m256i	_mm256_div_epi16 (const __m256i &va, const int16_t &b)
	Hack-ish AVX division function.
void	_mm256_print_epi16 (const __m256i &input)
	AVX printing function.
int16_t	_naive_div_int16 (const int16_t &a, const int16_t &b)
	Naive model of AVX division in AVXUtils.

Function Documentation

_mm256_div_epi16()

__m256i tpglibs::_mm256_div_epi16 ( const __m256i & va, const int16_t & b )

inline

Hack-ish AVX division function.

Derivation from https://stackoverflow.com/questions/42442325/how-to-divide-a-m256i-vector-by-an-integer-variable

Parameters

va	AVX2 register as the dividend.
b	Integer as the divisor.

Returns

AVX2 register whose elements were divided by b.

Definition at line 29 of file AVXUtils.hpp.

                                                                     {
  // NaiveUtils.hpp has the lay version.
  __m256i vb = _mm256_set1_epi16(32768 / b);
  return _mm256_mulhrs_epi16(va, vb);
}

_mm256_print_epi16()

void tpglibs::_mm256_print_epi16 ( const __m256i & input )

inline

AVX printing function.

Function naming follows the format of typical AVX2 functions:

• _mm256 -> 256-bit register.
• epi16 -> Vector elements are 16-bit signed integers. Does not print a new line.

Parameters

input

A 256-bit register.

Definition at line 46 of file AVXUtils.hpp.

                                                     {
  std::array<int16_t, 16> prints;
  _mm256_storeu_si256(reinterpret_cast<__m256i*>(prints.begin()), input);
 
  fmt::print("{}", prints);
}

_naive_div_int16()

int16_t tpglibs::_naive_div_int16 ( const int16_t & a, const int16_t & b )

inline

Naive model of AVX division in AVXUtils.

This is to produce consistent results with AVX division. The rounding is unorthodox, but it is systematic.

Parameters

a	The dividend.
b	The divisor.

Returns

a / b with funky rounding.

Definition at line 26 of file NaiveUtils.hpp.

                                                                    {
  int16_t vb = (1 << 15) / b;         //  1 / b * 2^15
  int32_t mulhrs = a * vb;            //  a / b * 2^15
  mulhrs = (mulhrs >> 14) + 1;        // (a / b * 2^15) * 2^-14 + 1 ~ a / b * 2 + 1
  mulhrs = mulhrs >> 1;               //~ a / b. The +1 causes unorthodox rounding.
  return (int16_t)(mulhrs);
}