zrot.s

Optimized GotoBLAS libraries

/*********************************************************************/
/*                                                                   */
/*             Optimized BLAS libraries                              */
/*                     By Kazushige Goto <[e-mail]>     */
/*                                                                   */
/* Copyright (c) The University of Texas, 2005. All rights reserved. */
/* UNIVERSITY EXPRESSLY DISCLAIMS ANY AND ALL WARRANTIES CONCERNING  */
/* THIS SOFTWARE AND DOCUMENTATION, INCLUDING ANY WARRANTIES OF      */
/* MERCHANTABILITY, FITNESS FOR ANY PARTICULAR PURPOSE,              */
/* NON-INFRINGEMENT AND WARRANTIES OF PERFORMANCE, AND ANY WARRANTY  */
/* THAT MIGHT OTHERWISE ARISE FROM COURSE OF DEALING OR USAGE OF     */
/* TRADE. NO WARRANTY IS EITHER EXPRESS OR IMPLIED WITH RESPECT TO   */
/* THE USE OF THE SOFTWARE OR DOCUMENTATION.                         */
/* Under no circumstances shall University be liable for incidental, */
/* special, indirect, direct or consequential damages or loss of     */
/* profits, interruption of business, or related expenses which may  */
/* arise from use of Software or Documentation, including but not    */
/* limited to those resulting from defects in Software and/or        */
/* Documentation, or loss or inaccuracy of data of any kind.         */
/*********************************************************************/

#define ASSEMBLER
#include "common.h"

#define N	%i0
#define X	%i1
#define INCX	%i2
#define Y	%i3
#define INCY	%i4
#define I	%i5

#define XX	%l0
#define YY	%l1

#ifdef DOUBLE
#define a1	%f4
#define a2	%f6
#define a3	%f8
#define a4	%f10
#define a5	%f12
#define a6	%f14
#define a7	%f16
#define a8	%f18
#define b1	%f20
#define b2	%f22
#define b3	%f24
#define b4	%f26
#define b5	%f28
#define b6	%f30
#define b7	%f32
#define b8	%f34

#define c1	%f36
#define c2	%f38
#define c3	%f40
#define c4	%f42
#define c5	%f44
#define c6	%f46
#define c7	%f48
#define c8	%f50

#define t1	%f52
#define t2	%f54
#define t3	%f56
#define t4	%f58
#else
#define a1	%f2
#define a2	%f3
#define a3	%f4
#define a4	%f5
#define a5	%f6
#define a6	%f7
#define a7	%f8
#define a8	%f9
#define b1	%f10
#define b2	%f11
#define b3	%f12
#define b4	%f13
#define b5	%f14
#define b6	%f15
#define b7	%f16
#define b8	%f17

#define c1	%f18
#define c2	%f19
#define c3	%f20
#define c4	%f21
#define c5	%f22
#define c6	%f23
#define c7	%f24
#define c8	%f25

#define t1	%f26
#define t2	%f27
#define t3	%f28
#define t4	%f29
#endif

#ifdef DOUBLE
#define C	%f0
#define S	%f2
#else
#define C	%f0
#define S	%f1
#endif

	PROLOGUE
	SAVESP

#ifndef __64BIT__

#ifdef DOUBLE
	st	%i5, [%fp + 88]

	LDF	[%fp + 88], C
	LDF	[%fp + 96], S
#else
	st	%i5, [%fp + 88]

	LDF	[%fp + 88], C
	LDF	[%fp + 92], S
#endif
#else
#ifdef DOUBLE
	FMOV	%f10, C
	FMOV	%f12, S
#else
	FMOV	%f11, C
	FMOV	%f13, S
#endif
#endif

	cmp	N, 0
	ble	.LL19

	sll	INCX, ZBASE_SHIFT, INCX
	sll	INCY, ZBASE_SHIFT, INCY

	cmp	INCX, 2 * SIZE
	bne	.LL50
	nop

	cmp	INCY, 2 * SIZE
	bne	.LL50
	nop

	sra	N, 2, I
	cmp	I, 0
	ble,pn	%icc, .LL15
	nop

	LDF	[X +  0 * SIZE], a1
	LDF	[Y +  0 * SIZE], b1
	LDF	[X +  1 * SIZE], a2
	LDF	[Y +  1 * SIZE], b2
	LDF	[X +  2 * SIZE], a3
	LDF	[Y +  2 * SIZE], b3
	LDF	[X +  3 * SIZE], a4
	LDF	[Y +  3 * SIZE], b4

	LDF	[X +  4 * SIZE], a5
	LDF	[Y +  4 * SIZE], b5
	LDF	[X +  5 * SIZE], a6
	LDF	[Y +  5 * SIZE], b6
	LDF	[X +  6 * SIZE], a7
	LDF	[Y +  6 * SIZE], b7
	LDF	[X +  7 * SIZE], a8
	LDF	[Y +  7 * SIZE], b8

	FMUL	C, a1, c1
	FMUL	S, b1, c2
	FMUL	C, b1, c3
	LDF	[Y +  8 * SIZE], b1
	FMUL	S, a1, c4
	LDF	[X +  8 * SIZE], a1

	FMUL	C, a2, c5
	FMUL	S, b2, c6
	FADD	c1, c2, t1

	FMUL	C, b2, c7
	LDF	[Y +  9 * SIZE], b2
	FMUL	S, a2, c8
	LDF	[X +  9 * SIZE], a2
	FSUB	c3, c4, t2

	addcc	I, -1, I
	ble,pt	%icc, .LL12
	nop

#define PREFETCHSIZE 64

.LL11:
	FMUL	C, a3, c1
	nop
	prefetch [Y  + PREFETCHSIZE * SIZE], 1
	nop

	FMUL	S, b3, c2
	STF	t1, [X +  0 * SIZE]
	FADD	c5, c6, t3
	nop

	FMUL	C, b3, c3
	LDF	[Y + 10 * SIZE], b3
	nop
	nop

	FMUL	S, a3, c4
	STF	t2, [Y +  0 * SIZE]
	FSUB	c7, c8, t4
	nop

	FMUL	C, a4, c5
	LDF	[X + 10 * SIZE], a3
	nop
	nop

	FMUL	S, b4, c6
	STF	t3, [X +  1 * SIZE]
	FADD	c1, c2, t1
	nop

	FMUL	C, b4, c7
	LDF	[Y + 11 * SIZE], b4
	nop
	nop

	FMUL	S, a4, c8
	STF	t4, [Y +  1 * SIZE]
	FSUB	c3, c4, t2
	nop

	FMUL	C, a5, c1
	LDF	[X + 11 * SIZE], a4
	nop
	nop

	FMUL	S, b5, c2
	STF	t1, [X +  2 * SIZE]
	FADD	c5, c6, t3
	nop

	FMUL	C, b5, c3
	LDF	[Y + 12 * SIZE], b5
	nop
	nop

	FMUL	S, a5, c4
	STF	t2, [Y +  2 * SIZE]
	FSUB	c7, c8, t4
	nop

	FMUL	C, a6, c5
	LDF	[X + 12 * SIZE], a5
	nop
	nop

	FMUL	S, b6, c6
	STF	t3, [X +  3 * SIZE]
	FADD	c1, c2, t1
	nop

	FMUL	C, b6, c7
	LDF	[Y + 13 * SIZE], b6
	nop
	nop

	FMUL	S, a6, c8
	STF	t4, [Y +  3 * SIZE]
	FSUB	c3, c4, t2
	nop

	FMUL	C, a7, c1
	LDF	[X + 13 * SIZE], a6
	nop
	nop

	FMUL	S, b7, c2
	STF	t1, [X +  4 * SIZE]
	FADD	c5, c6, t3
	nop

	FMUL	C, b7, c3
	LDF	[Y + 14 * SIZE], b7
	nop
	nop

	FMUL	S, a7, c4
	STF	t2, [Y +  4 * SIZE]
	FSUB	c7, c8, t4
	nop

	FMUL	C, a8, c5
	LDF	[X + 14 * SIZE], a7
	nop
	nop

	FMUL	S, b8, c6
	STF	t3, [X +  5 * SIZE]
	FADD	c1, c2, t1
	nop

	FMUL	C, b8, c7
	LDF	[Y + 15 * SIZE], b8
	nop
	nop

	FMUL	S, a8, c8
	STF	t4, [Y +  5 * SIZE]
	FSUB	c3, c4, t2
	nop

	FMUL	C, a1, c1
	LDF	[X + 15 * SIZE], a8
	addcc	I, -1, I
	nop

	FMUL	S, b1, c2
	STF	t1, [X +  6 * SIZE]
	FADD	c5, c6, t3
	nop

	FMUL	C, b1, c3
	LDF	[Y + 16 * SIZE], b1
	nop
	nop

	FMUL	S, a1, c4
	STF	t2, [Y +  6 * SIZE]
	FSUB	c7, c8, t4
	nop

	FMUL	C, a2, c5
	LDF	[X + 16 * SIZE], a1
	add	Y, 8 * SIZE, Y
	nop

	FMUL	S, b2, c6
	STF	t3, [X +  7 * SIZE]
	FADD	c1, c2, t1
	nop

	FMUL	C, b2, c7
	LDF	[Y +  9 * SIZE], b2
	add	X, 8 * SIZE, X
	nop

	FMUL	S, a2, c8
	STF	t4, [Y -  1 * SIZE]
	FSUB	c3, c4, t2
	nop

	bg,pt	%icc, .LL11
	LDF	[X +  9 * SIZE], a2


.LL12:
	FMUL	C, a3, c1
	FMUL	S, b3, c2
	STF	t1, [X +  0 * SIZE]
	FADD	c5, c6, t3

	FMUL	C, b3, c3
	FMUL	S, a3, c4
	STF	t2, [Y +  0 * SIZE]
	FSUB	c7, c8, t4


	FMUL	C, a4, c5
	FMUL	S, b4, c6
	STF	t3, [X +  1 * SIZE]
	FADD	c1, c2, t1

	FMUL	C, b4, c7
	FMUL	S, a4, c8
	STF	t4, [Y +  1 * SIZE]
	FSUB	c3, c4, t2


	FMUL	C, a5, c1
	FMUL	S, b5, c2
	STF	t1, [X +  2 * SIZE]
	FADD	c5, c6, t3

	FMUL	C, b5, c3
	FMUL	S, a5, c4
	STF	t2, [Y +  2 * SIZE]
	FSUB	c7, c8, t4

	FMUL	C, a6, c5
	FMUL	S, b6, c6
	STF	t3, [X +  3 * SIZE]
	FADD	c1, c2, t1

	FMUL	C, b6, c7
	FMUL	S, a6, c8
	STF	t4, [Y +  3 * SIZE]
	FSUB	c3, c4, t2

	FMUL	C, a7, c1
	FMUL	S, b7, c2
	STF	t1, [X +  4 * SIZE]
	FADD	c5, c6, t3

	FMUL	C, b7, c3
	FMUL	S, a7, c4
	STF	t2, [Y +  4 * SIZE]
	FSUB	c7, c8, t4

	FMUL	C, a8, c5
	FMUL	S, b8, c6
	STF	t3, [X +  5 * SIZE]
	FADD	c1, c2, t1

	FMUL	C, b8, c7
	FMUL	S, a8, c8
	STF	t4, [Y +  5 * SIZE]
	FSUB	c3, c4, t2

	FADD	c5, c6, t3
	STF	t1, [X +  6 * SIZE]

	FSUB	c7, c8, t4
	STF	t2, [Y +  6 * SIZE]

	STF	t3, [X +  7 * SIZE]
	STF	t4, [Y +  7 * SIZE]

	add	X, 8 * SIZE, X
	add	Y, 8 * SIZE, Y


.LL15:
	andcc	N, 3, I
	nop
	ble,a,pn %icc, .LL19
	nop

.LL16:
	LDF	[X + 0 * SIZE], a1
	LDF	[Y + 0 * SIZE], b1
	LDF	[X + 1 * SIZE], a2
	LDF	[Y + 1 * SIZE], b2

	FMUL	C, a1, c1
	add	X, 2 * SIZE, X
	FMUL	S, b1, c2
	add	Y, 2 * SIZE, Y

	FMUL	C, b1, c3
	addcc	I, -1, I
	FMUL	S, a1, c4
	nop

	FMUL	C, a2, c5
	FMUL	S, b2, c6
	FADD	c1, c2, c2

	FMUL	C, b2, c7
	FMUL	S, a2, c8
	FSUB	c3, c4, c4

	STF	c2, [X - 2 * SIZE]
	FADD	c5, c6, c6
	STF	c4, [Y - 2 * SIZE]
	FSUB	c7, c8, c8

	STF	c6, [X - 1 * SIZE]
	bg,pt	%icc, .LL16
	STF	c8, [Y - 1 * SIZE]

.LL19:
	return	%i7 + 8
	nop
	
.LL50:
	mov	X, XX
	mov	Y, YY

	sra	N, 2, I
	cmp	I, 0
	ble,pn	%icc, .LL55
	nop

.LL51:
	LDF	[X + 0 * SIZE], a1
	LDF	[Y + 0 * SIZE], b1
	LDF	[X + 1 * SIZE], a2
	LDF	[Y + 1 * SIZE], b2

	FMUL	C, a1, c1
	FMUL	S, b1, c2
	FMUL	C, b1, c3
	FMUL	S, a1, c4

	FMUL	C, a2, c5
	nop
	FMUL	S, b2, c6
	FADD	c1, c2, c2

	FMUL	C, b2, c7
	nop
	FMUL	S, a2, c8
	FSUB	c3, c4, c4

	STF	c2, [X + 0 * SIZE]
	FADD	c5, c6, c6
	STF	c4, [Y + 0 * SIZE]
	FSUB	c7, c8, c8

	STF	c6, [X + 1 * SIZE]
	add	X, INCX, X
	STF	c8, [Y + 1 * SIZE]
	add	Y, INCY, Y

	LDF	[X + 0 * SIZE], a1
	LDF	[Y + 0 * SIZE], b1
	LDF	[X + 1 * SIZE], a2
	LDF	[Y + 1 * SIZE], b2

	FMUL	C, a1, c1
	FMUL	S, b1, c2
	FMUL	C, b1, c3
	FMUL	S, a1, c4

	FMUL	C, a2, c5
	nop
	FMUL	S, b2, c6
	FADD	c1, c2, c2

	FMUL	C, b2, c7
	nop
	FMUL	S, a2, c8
	FSUB	c3, c4, c4

	STF	c2, [X + 0 * SIZE]
	FADD	c5, c6, c6
	STF	c4, [Y + 0 * SIZE]
	FSUB	c7, c8, c8

	STF	c6, [X + 1 * SIZE]
	add	X, INCX, X
	STF	c8, [Y + 1 * SIZE]
	add	Y, INCY, Y

	LDF	[X + 0 * SIZE], a1
	LDF	[Y + 0 * SIZE], b1
	LDF	[X + 1 * SIZE], a2
	LDF	[Y + 1 * SIZE], b2

	FMUL	C, a1, c1
	FMUL	S, b1, c2
	FMUL	C, b1, c3
	FMUL	S, a1, c4

	FMUL	C, a2, c5
	nop
	FMUL	S, b2, c6
	FADD	c1, c2, c2

	FMUL	C, b2, c7
	nop
	FMUL	S, a2, c8
	FSUB	c3, c4, c4

	STF	c2, [X + 0 * SIZE]
	FADD	c5, c6, c6
	STF	c4, [Y + 0 * SIZE]
	FSUB	c7, c8, c8

	STF	c6, [X + 1 * SIZE]
	add	X, INCX, X
	STF	c8, [Y + 1 * SIZE]
	add	Y, INCY, Y

	LDF	[X + 0 * SIZE], a1
	LDF	[Y + 0 * SIZE], b1
	LDF	[X + 1 * SIZE], a2
	LDF	[Y + 1 * SIZE], b2

	FMUL	C, a1, c1
	FMUL	S, b1, c2
	FMUL	C, b1, c3
	FMUL	S, a1, c4

	FMUL	C, a2, c5
	nop
	FMUL	S, b2, c6
	FADD	c1, c2, c2

	FMUL	C, b2, c7
	nop
	FMUL	S, a2, c8
	FSUB	c3, c4, c4

	STF	c2, [X + 0 * SIZE]
	FADD	c5, c6, c6
	STF	c4, [Y + 0 * SIZE]
	FSUB	c7, c8, c8

	STF	c6, [X + 1 * SIZE]
	add	X, INCX, X
	STF	c8, [Y + 1 * SIZE]
	add	Y, INCY, Y

	addcc	I, -1, I
	bg,pt	%icc, .LL51
	nop


.LL55:
	andcc	N, 3, I
	nop
	ble	%icc, .LL59
	nop

.LL56:
	LDF	[X + 0 * SIZE], a1
	LDF	[Y + 0 * SIZE], b1
	LDF	[X + 1 * SIZE], a2
	LDF	[Y + 1 * SIZE], b2

	FMUL	C, a1, c1
	FMUL	S, b1, c2
	FMUL	C, b1, c3
	FMUL	S, a1, c4

	FMUL	C, a2, c5
	addcc	I, -1, I
	FMUL	S, b2, c6
	FADD	c1, c2, c2

	FMUL	C, b2, c7
	nop
	FMUL	S, a2, c8
	FSUB	c3, c4, c4

	STF	c2, [X + 0 * SIZE]
	FADD	c5, c6, c6
	STF	c4, [Y + 0 * SIZE]
	FSUB	c7, c8, c8

	STF	c6, [X + 1 * SIZE]
	add	X, INCX, X
	STF	c8, [Y + 1 * SIZE]

	bg	%icc, .LL56
	add	Y, INCY, Y


.LL59:
	return	%i7 + 8
	nop

	EPILOGUE