real_fjsp_2hv_template.c File Reference

#include <x86intrin.h>
#include <stdio.h>
#include <stdlib.h>

Macros
#define	__forceinline   __attribute__((always_inline)) static

Functions
	for (i=2;i< nb;i++)
	if (nq==i)
	_SSE_STORE (q, q1)

Variables
int	nb = *pnb
int	nq = *pldq
int	ldq = *pldq
int	ldh = *pldh
int	worked_on = 0
__SSE_DATATYPE	x1 = _SSE_LOAD(&q[ldq])
__SSE_DATATYPE	x2 = _SSE_LOAD(&q[ldq+offset])
__SSE_DATATYPE	x3 = _SSE_LOAD(&q[ldq+2*offset])
__SSE_DATATYPE	x4 = _SSE_LOAD(&q[ldq+3*offset])
__SSE_DATATYPE	x5 = _SSE_LOAD(&q[ldq+4*offset])
__SSE_DATATYPE	x6 = _SSE_LOAD(&q[ldq+5*offset])
__SSE_DATATYPE	h2 = _SSE_MUL(h1, vs)
__SSE_DATATYPE	q1 = _SSE_LOAD(q)
__SSE_DATATYPE	y1 = _SSE_ADD(q1, _SSE_MUL(x1, h1))
__SSE_DATATYPE	q2 = _SSE_LOAD(&q[offset])
__SSE_DATATYPE	y2 = _SSE_ADD(q2, _SSE_MUL(x2, h1))
__SSE_DATATYPE	q3 = _SSE_LOAD(&q[2*offset])
__SSE_DATATYPE	y3 = _SSE_ADD(q3, _SSE_MUL(x3, h1))
__SSE_DATATYPE	q4 = _SSE_LOAD(&q[3*offset])
__SSE_DATATYPE	y4 = _SSE_ADD(q4, _SSE_MUL(x4, h1))
__SSE_DATATYPE	q5 = _SSE_LOAD(&q[4*offset])
__SSE_DATATYPE	y5 = _SSE_ADD(q5, _SSE_MUL(x5, h1))
__SSE_DATATYPE	q6 = _SSE_LOAD(&q[5*offset])
__SSE_DATATYPE	y6 = _SSE_ADD(q6, _SSE_MUL(x6, h1))
	h1 = _SSE_XOR(tau1, sign)
_SSE_STORE &[offset]	q

Macro Definition Documentation

__forceinline

#define __forceinline   __attribute__((always_inline)) static

Function Documentation

_SSE_STORE()

_SSE_STORE	(	q	,
		q1
	)

for()

for

(

)

if()

if

(

nq

= = i

)

Variable Documentation

h1

h1 = _SSE_XOR(tau1, sign)

h2

__SSE_DATATYPE h2 = _SSE_MUL(h1, vs)

ldh

int ldh = *pldh

ldq

int ldq = *pldq

nb

int nb = *pnb

nq

int nq = *pldq

q

_SSE_STORE&[nb*ldq] q

q1

__SSE_DATATYPE q1 = _SSE_LOAD(q)

q2

__SSE_DATATYPE q2 = _SSE_LOAD(&q[offset])

q3

__SSE_DATATYPE q3 = _SSE_LOAD(&q[2*offset])

q4

__SSE_DATATYPE q4 = _SSE_LOAD(&q[3*offset])

q5

__SSE_DATATYPE q5 = _SSE_LOAD(&q[4*offset])

q6

q6 = _SSE_LOAD(&q[5*offset])

worked_on

worked_on = 0

x1

__SSE_DATATYPE x1 = _SSE_LOAD(&q[ldq])

Unrolled kernel that computes #ifdef DOUBLE_PRECISION_REAL 12 rows of Q simultaneously, a #endif #ifdef SINGLE_PRECISION_REAL 24 rows of Q simultaneously, a #endif matrix Vector product with two householder vectors + a rank 2 update is performed

Unrolled kernel that computes #ifdef DOUBLE_PRECISION_REAL 10 rows of Q simultaneously, a #endif #ifdef SINGLE_PRECISION_REAL 20 rows of Q simultaneously, a #endif matrix Vector product with two householder vectors + a rank 2 update is performed

Unrolled kernel that computes #ifdef DOUBLE_PRECISION_REAL 8 rows of Q simultaneously, a #endif #ifdef SINGLE_PRECISION_REAL 16 rows of Q simultaneously, a #endif matrix Vector product with two householder vectors + a rank 2 update is performed

Unrolled kernel that computes #ifdef DOUBLE_PRECISION_REAL 6 rows of Q simultaneously, a #endif #ifdef SINGLE_PRECISION_REAL 12 rows of Q simultaneously, a #endif matrix Vector product with two householder vectors + a rank 2 update is performed

Unrolled kernel that computes #ifdef DOUBLE_PRECISION_REAL 4 rows of Q simultaneously, a #endif #ifdef SINGLE_PRECISION_REAL 8 rows of Q simultaneously, a #endif matrix Vector product with two householder vectors + a rank 2 update is performed

Unrolled kernel that computes #ifdef DOUBLE_PRECISION_REAL 2 rows of Q simultaneously, a #endif #ifdef SINGLE_PRECISION_REAL 4 rows of Q simultaneously, a #endif matrix Vector product with two householder vectors + a rank 2 update is performed

x2

__SSE_DATATYPE x2 = _SSE_LOAD(&q[ldq+offset])

x3

__SSE_DATATYPE x3 = _SSE_LOAD(&q[ldq+2*offset])

x4

__SSE_DATATYPE x4 = _SSE_LOAD(&q[ldq+3*offset])

x5

__SSE_DATATYPE x5 = _SSE_LOAD(&q[ldq+4*offset])

x6

x6 = _SSE_LOAD(&q[ldq+5*offset])

y1

__SSE_DATATYPE y1 = _SSE_ADD(q1, _SSE_MUL(x1, h1))

y2

__SSE_DATATYPE y2 = _SSE_ADD(q2, _SSE_MUL(x2, h1))

y3

__SSE_DATATYPE y3 = _SSE_ADD(q3, _SSE_MUL(x3, h1))

y4

__SSE_DATATYPE y4 = _SSE_ADD(q4, _SSE_MUL(x4, h1))

y5

__SSE_DATATYPE y5 = _SSE_ADD(q5, _SSE_MUL(x5, h1))

y6

y6 = _SSE_ADD(q6, _SSE_MUL(x6, h1))