openswmm::simd Namespace Reference

Functions
void	add (const double *OPENSWMM_RESTRICT a, const double *OPENSWMM_RESTRICT b, double *OPENSWMM_RESTRICT dst, std::size_t n) noexcept
	Compute dst[i] = a[i] + b[i] over n doubles.
void	multiply (const double *OPENSWMM_RESTRICT a, const double *OPENSWMM_RESTRICT b, double *OPENSWMM_RESTRICT dst, std::size_t n) noexcept
	Compute dst[i] = a[i] * b[i] over n doubles.
double	min (const double *a, std::size_t n) noexcept
	Reduce an array to its minimum value.
double	max (const double *a, std::size_t n) noexcept
	Reduce an array to its maximum value.
void	clamp (double *a, double lo, double hi, std::size_t n) noexcept
	Clamp each array element in place to an inclusive interval.
double	dot (const double *OPENSWMM_RESTRICT a, const double *OPENSWMM_RESTRICT b, std::size_t n) noexcept
	Sum corresponding products from two arrays.
constexpr std::size_t	lane_width () noexcept
	Report the compile-time number of doubles per SIMD register.
void	sqrt_array (const double *OPENSWMM_RESTRICT a, double *OPENSWMM_RESTRICT dst, std::size_t n) noexcept
	Compute dst[i] = sqrt(a[i]) over n doubles.
void	fabs_array (const double *OPENSWMM_RESTRICT a, double *OPENSWMM_RESTRICT dst, std::size_t n) noexcept
	Compute dst[i] = fabs(a[i]) over n doubles.
void	fma_array (const double *OPENSWMM_RESTRICT a, const double *OPENSWMM_RESTRICT b, const double *OPENSWMM_RESTRICT c, double *OPENSWMM_RESTRICT dst, std::size_t n) noexcept
	Compute dst[i] = a[i] * b[i] + c[i] over n doubles.

Function Documentation

add()

void openswmm::simd::add ( const double *OPENSWMM_RESTRICT a, const double *OPENSWMM_RESTRICT b, double *OPENSWMM_RESTRICT dst, std::size_t n )

inlinenoexcept

Compute dst[i] = a[i] + b[i] over n doubles.

Parameters

[in]	a	Borrowed readable array of at least n elements.
[in]	b	Borrowed readable array of at least n elements.
[out]	dst	Borrowed writable array of at least n elements.
[in]	n	Element count; any size, including zero, is supported.

Precondition

Source and destination arrays do not overlap. For n > 0 all pointers are non-null. No SIMD alignment is required.

Note

No storage is retained. A scalar tail handles partial SIMD lanes.

clamp()

void openswmm::simd::clamp ( double * a, double lo, double hi, std::size_t n )

inlinenoexcept

Clamp each array element in place to an inclusive interval.

Parameters

[in,out]	a	Borrowed writable array of at least n doubles; non-null when n > 0. No alignment requirement applies.
[in]	lo	Lower bound, in the same units as the array elements.
[in]	hi	Upper bound; must be >= lo.
[in]	n	Element count; zero leaves the array untouched.

Note

No storage is retained. NaN behavior follows backend comparisons.

dot()

double openswmm::simd::dot ( const double *OPENSWMM_RESTRICT a, const double *OPENSWMM_RESTRICT b, std::size_t n )

inlinenoexcept

Sum corresponding products from two arrays.

Parameters

[in]	a	Borrowed readable array of at least n doubles.
[in]	b	Borrowed readable array of at least n doubles.
[in]	n	Element count; zero returns zero without accessing the arrays.

Returns

Sum of a[i] * b[i], in the product of the inputs' units.

Precondition

Arrays are non-null for n > 0 and satisfy the restrict alias contract.

Note

No alignment is required. SIMD lanes may use fused multiply-add and a different summation order; scalar tails use ordinary accumulation. No Kahan compensation is performed and no storage is retained.

fabs_array()

void openswmm::simd::fabs_array ( const double *OPENSWMM_RESTRICT a, double *OPENSWMM_RESTRICT dst, std::size_t n )

inlinenoexcept

Compute dst[i] = fabs(a[i]) over n doubles.

Parameters

[in]	a	Borrowed readable array of at least n doubles.
[out]	dst	Borrowed writable array of at least n doubles.
[in]	n	Element count; zero performs no reads or writes.

Precondition

Arrays do not overlap and are non-null for n > 0.

Note

Output values are nonnegative magnitudes. No storage is retained; alignment is unrestricted and the compiler may vectorize the loop.

fma_array()

void openswmm::simd::fma_array ( const double *OPENSWMM_RESTRICT a, const double *OPENSWMM_RESTRICT b, const double *OPENSWMM_RESTRICT c, double *OPENSWMM_RESTRICT dst, std::size_t n )

inlinenoexcept

Compute dst[i] = a[i] * b[i] + c[i] over n doubles.

Parameters

[in]	a	Borrowed first-factor array with at least n elements.
[in]	b	Borrowed second-factor array with at least n elements.
[in]	c	Borrowed addend array with at least n elements, in product units.
[out]	dst	Borrowed writable result array with at least n elements.
[in]	n	Element count; zero performs no reads or writes.

Precondition

Arrays do not overlap and are non-null for n > 0.

Note

No storage is retained and no alignment is required. The expression uses ordinary multiply and add; fusion depends on compiler settings.

lane_width()

std::size_t openswmm::simd::lane_width ( )

constexprnoexcept

Report the compile-time number of doubles per SIMD register.

Returns

Four for AVX2, two for NEON, or one for the scalar backend.

max()

double openswmm::simd::max ( const double * a, std::size_t n )

inlinenoexcept

Reduce an array to its maximum value.

Parameters

[in]	a	Non-null borrowed array of at least n readable doubles.
[in]	n	Element count, at least one; checked by assert in debug builds.

Returns

Maximum element for finite inputs. NaN behavior follows the selected SIMD intrinsics or scalar comparisons and may differ between backends.

Note

No SIMD alignment is required and no storage is retained.

min()

double openswmm::simd::min ( const double * a, std::size_t n )

inlinenoexcept

Reduce an array to its minimum value.

Parameters

[in]	a	Non-null borrowed array of at least n readable doubles.
[in]	n	Element count, at least one; checked by assert in debug builds.

Returns

Minimum element for finite inputs. NaN behavior follows the selected SIMD intrinsics or scalar comparisons and may differ between backends.

Note

No SIMD alignment is required and no storage is retained.

multiply()

void openswmm::simd::multiply ( const double *OPENSWMM_RESTRICT a, const double *OPENSWMM_RESTRICT b, double *OPENSWMM_RESTRICT dst, std::size_t n )

inlinenoexcept

Compute dst[i] = a[i] * b[i] over n doubles.

Parameters

[in]	a	Borrowed readable array of at least n elements.
[in]	b	Borrowed readable array of at least n elements.
[out]	dst	Borrowed writable array of at least n elements.
[in]	n	Element count; any size, including zero, is supported.

Precondition

Source and destination arrays do not overlap. For n > 0 all pointers are non-null. No SIMD alignment is required.

Note

No storage is retained. A scalar tail handles partial SIMD lanes.

Here is the caller graph for this function:

sqrt_array()

void openswmm::simd::sqrt_array ( const double *OPENSWMM_RESTRICT a, double *OPENSWMM_RESTRICT dst, std::size_t n )

inlinenoexcept

Compute dst[i] = sqrt(a[i]) over n doubles.

Parameters

[in]	a	Borrowed readable array of at least n doubles.
[out]	dst	Borrowed writable array of at least n doubles.
[in]	n	Element count; zero performs no reads or writes.

Precondition

Arrays do not overlap and are non-null for n > 0.

Note

Negative inputs follow std::sqrt domain behavior. No storage is retained; alignment is unrestricted and the compiler may vectorize the loop.