1// Copyright ©2015 The Gonum Authors. All rights reserved.
2// Use of this source code is governed by a BSD-style
3// license that can be found in the LICENSE file.
4//
5// Some of the loop unrolling code is copied from:
6// http://golang.org/src/math/big/arith_amd64.s
7// which is distributed under these terms:
8//
9// Copyright (c) 2012 The Go Authors. All rights reserved.
10//
11// Redistribution and use in source and binary forms, with or without
12// modification, are permitted provided that the following conditions are
13// met:
14//
15//    * Redistributions of source code must retain the above copyright
16// notice, this list of conditions and the following disclaimer.
17//    * Redistributions in binary form must reproduce the above
18// copyright notice, this list of conditions and the following disclaimer
19// in the documentation and/or other materials provided with the
20// distribution.
21//    * Neither the name of Google Inc. nor the names of its
22// contributors may be used to endorse or promote products derived from
23// this software without specific prior written permission.
24//
25// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
26// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
27// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
28// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
29// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
30// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
31// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
32// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
33// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
34// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
35// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
36
37// +build !noasm,!gccgo,!safe
38
39#include "textflag.h"
40
41#define X_PTR SI
42#define Y_PTR DI
43#define DST_PTR DI
44#define IDX AX
45#define LEN CX
46#define TAIL BX
47#define INC_X R8
48#define INCx3_X R11
49#define INC_Y R9
50#define INCx3_Y R12
51#define INC_DST R9
52#define INCx3_DST R12
53#define ALPHA X0
54#define ALPHA_2 X1
55
56// func AxpyInc(alpha float64, x, y []float64, n, incX, incY, ix, iy uintptr)
57TEXT ·AxpyInc(SB), NOSPLIT, $0
58	MOVQ x_base+8(FP), X_PTR  // X_PTR = &x
59	MOVQ y_base+32(FP), Y_PTR // Y_PTR = &y
60	MOVQ n+56(FP), LEN        // LEN = n
61	CMPQ LEN, $0              // if LEN == 0 { return }
62	JE   end
63
64	MOVQ ix+80(FP), INC_X
65	MOVQ iy+88(FP), INC_Y
66	LEAQ (X_PTR)(INC_X*8), X_PTR // X_PTR = &(x[ix])
67	LEAQ (Y_PTR)(INC_Y*8), Y_PTR // Y_PTR = &(y[iy])
68	MOVQ Y_PTR, DST_PTR          // DST_PTR = Y_PTR  // Write pointer
69
70	MOVQ incX+64(FP), INC_X // INC_X = incX * sizeof(float64)
71	SHLQ $3, INC_X
72	MOVQ incY+72(FP), INC_Y // INC_Y = incY * sizeof(float64)
73	SHLQ $3, INC_Y
74
75	MOVSD alpha+0(FP), ALPHA // ALPHA = alpha
76	MOVQ  LEN, TAIL
77	ANDQ  $3, TAIL           // TAIL = n % 4
78	SHRQ  $2, LEN            // LEN = floor( n / 4 )
79	JZ    tail_start         // if LEN == 0 { goto tail_start }
80
81	MOVAPS ALPHA, ALPHA_2            // ALPHA_2 = ALPHA  for pipelining
82	LEAQ   (INC_X)(INC_X*2), INCx3_X // INCx3_X = INC_X * 3
83	LEAQ   (INC_Y)(INC_Y*2), INCx3_Y // INCx3_Y = INC_Y * 3
84
85loop:  // do {  // y[i] += alpha * x[i] unrolled 4x.
86	MOVSD (X_PTR), X2            // X_i = x[i]
87	MOVSD (X_PTR)(INC_X*1), X3
88	MOVSD (X_PTR)(INC_X*2), X4
89	MOVSD (X_PTR)(INCx3_X*1), X5
90
91	MULSD ALPHA, X2   // X_i *= a
92	MULSD ALPHA_2, X3
93	MULSD ALPHA, X4
94	MULSD ALPHA_2, X5
95
96	ADDSD (Y_PTR), X2            // X_i += y[i]
97	ADDSD (Y_PTR)(INC_Y*1), X3
98	ADDSD (Y_PTR)(INC_Y*2), X4
99	ADDSD (Y_PTR)(INCx3_Y*1), X5
100
101	MOVSD X2, (DST_PTR)              // y[i] = X_i
102	MOVSD X3, (DST_PTR)(INC_DST*1)
103	MOVSD X4, (DST_PTR)(INC_DST*2)
104	MOVSD X5, (DST_PTR)(INCx3_DST*1)
105
106	LEAQ (X_PTR)(INC_X*4), X_PTR // X_PTR = &(X_PTR[incX*4])
107	LEAQ (Y_PTR)(INC_Y*4), Y_PTR // Y_PTR = &(Y_PTR[incY*4])
108	DECQ LEN
109	JNZ  loop                    // } while --LEN > 0
110	CMPQ TAIL, $0                // if TAIL == 0 { return }
111	JE   end
112
113tail_start: // Reset Loop registers
114	MOVQ TAIL, LEN // Loop counter: LEN = TAIL
115	SHRQ $1, LEN   // LEN = floor( LEN / 2 )
116	JZ   tail_one
117
118tail_two:
119	MOVSD (X_PTR), X2              // X_i = x[i]
120	MOVSD (X_PTR)(INC_X*1), X3
121	MULSD ALPHA, X2                // X_i *= a
122	MULSD ALPHA, X3
123	ADDSD (Y_PTR), X2              // X_i += y[i]
124	ADDSD (Y_PTR)(INC_Y*1), X3
125	MOVSD X2, (DST_PTR)            // y[i] = X_i
126	MOVSD X3, (DST_PTR)(INC_DST*1)
127
128	LEAQ (X_PTR)(INC_X*2), X_PTR // X_PTR = &(X_PTR[incX*2])
129	LEAQ (Y_PTR)(INC_Y*2), Y_PTR // Y_PTR = &(Y_PTR[incY*2])
130
131	ANDQ $1, TAIL
132	JZ   end      // if TAIL == 0 { goto end }
133
134tail_one:
135	// y[i] += alpha * x[i] for the last n % 4 iterations.
136	MOVSD (X_PTR), X2   // X2 = x[i]
137	MULSD ALPHA, X2     // X2 *= a
138	ADDSD (Y_PTR), X2   // X2 += y[i]
139	MOVSD X2, (DST_PTR) // y[i] = X2
140
141end:
142	RET
143