vmul-wide ( a b -- lo hi )

Vocabulary

math.vectors

Inputs

a	a sequence
b	a sequence

Outputs

lo	a sequence
hi	a sequence

Word description

Multiplies matching 8-, 16-, or 32-bit integer SIMD vectors exactly into lanes of twice the width and the same signedness. Returns the lower input lanes first, then the upper input lanes.

Definition

IN: math.vectors

GENERIC: vmul-wide ( a b -- lo hi )

Methods

USING: accessors cpu.architecture kernel math.vectors
math.vectors.simd math.vectors.simd.intrinsics ;

M:: char-16 vmul-wide ( a b -- lo hi )
    a char-16? [
        a underlying>> b underlying>> char-16-rep
        (simd-mul-wide-head) short-8 boa
        a underlying>> b underlying>> char-16-rep
        (simd-mul-wide-tail) short-8 boa
    ] [ a bad-simd-vector ] if ; inline


USING: accessors cpu.architecture kernel math.vectors
math.vectors.simd math.vectors.simd.intrinsics ;

M:: int-4 vmul-wide ( a b -- lo hi )
    a int-4? [
        a underlying>> b underlying>> int-4-rep
        (simd-mul-wide-head) longlong-2 boa
        a underlying>> b underlying>> int-4-rep
        (simd-mul-wide-tail) longlong-2 boa
    ] [ a bad-simd-vector ] if ; inline


USING: accessors cpu.architecture kernel math.vectors
math.vectors.simd math.vectors.simd.intrinsics ;

M:: short-8 vmul-wide ( a b -- lo hi )
    a short-8? [
        a underlying>> b underlying>> short-8-rep
        (simd-mul-wide-head) int-4 boa
        a underlying>> b underlying>> short-8-rep
        (simd-mul-wide-tail) int-4 boa
    ] [ a bad-simd-vector ] if ; inline


USING: kernel math.vectors math.vectors.simd ;

M: simd-128 vmul-wide drop bad-simd-vector ;


USING: accessors cpu.architecture kernel math.vectors
math.vectors.simd math.vectors.simd.intrinsics ;

M:: uchar-16 vmul-wide ( a b -- lo hi )
    a uchar-16? [
        a underlying>> b underlying>> uchar-16-rep
        (simd-mul-wide-head) ushort-8 boa
        a underlying>> b underlying>> uchar-16-rep
        (simd-mul-wide-tail) ushort-8 boa
    ] [ a bad-simd-vector ] if ; inline


USING: accessors cpu.architecture kernel math.vectors
math.vectors.simd math.vectors.simd.intrinsics ;

M:: uint-4 vmul-wide ( a b -- lo hi )
    a uint-4? [
        a underlying>> b underlying>> uint-4-rep
        (simd-mul-wide-head) ulonglong-2 boa
        a underlying>> b underlying>> uint-4-rep
        (simd-mul-wide-tail) ulonglong-2 boa
    ] [ a bad-simd-vector ] if ; inline


USING: accessors cpu.architecture kernel math.vectors
math.vectors.simd math.vectors.simd.intrinsics ;

M:: ushort-8 vmul-wide ( a b -- lo hi )
    a ushort-8? [
        a underlying>> b underlying>> ushort-8-rep
        (simd-mul-wide-head) uint-4 boa
        a underlying>> b underlying>> ushort-8-rep
        (simd-mul-wide-tail) uint-4 boa
    ] [ a bad-simd-vector ] if ; inline