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,!appengine,!safe
38
39#include "textflag.h"
40
41#define X_PTR SI
42#define Y_PTR DX
43#define DST_PTR DI
44#define IDX AX
45#define LEN CX
46#define TAIL BX
47#define ALPHA X0
48#define ALPHA_2 X1
49
50// func AxpyUnitaryTo(dst []float64, alpha float64, x, y []float64)
51TEXT ·AxpyUnitaryTo(SB), NOSPLIT, $0
52	MOVQ    dst_base+0(FP), DST_PTR // DST_PTR := &dst
53	MOVQ    x_base+32(FP), X_PTR    // X_PTR := &x
54	MOVQ    y_base+56(FP), Y_PTR    // Y_PTR := &y
55	MOVQ    x_len+40(FP), LEN       // LEN = min( len(x), len(y), len(dst) )
56	CMPQ    y_len+64(FP), LEN
57	CMOVQLE y_len+64(FP), LEN
58	CMPQ    dst_len+8(FP), LEN
59	CMOVQLE dst_len+8(FP), LEN
60
61	CMPQ LEN, $0
62	JE   end     // if LEN == 0 { return }
63
64	XORQ   IDX, IDX            // IDX = 0
65	MOVSD  alpha+24(FP), ALPHA
66	SHUFPD $0, ALPHA, ALPHA    // ALPHA := { alpha, alpha }
67	MOVQ   Y_PTR, TAIL         // Check memory alignment
68	ANDQ   $15, TAIL           // TAIL = &y % 16
69	JZ     no_trim             // if TAIL == 0 { goto no_trim }
70
71	// Align on 16-byte boundary
72	MOVSD (X_PTR), X2   // X2 := x[0]
73	MULSD ALPHA, X2     // X2 *= a
74	ADDSD (Y_PTR), X2   // X2 += y[0]
75	MOVSD X2, (DST_PTR) // y[0] = X2
76	INCQ  IDX           // i++
77	DECQ  LEN           // LEN--
78	JZ    end           // if LEN == 0 { return }
79
80no_trim:
81	MOVQ LEN, TAIL
82	ANDQ $7, TAIL   // TAIL := n % 8
83	SHRQ $3, LEN    // LEN = floor( n / 8 )
84	JZ   tail_start // if LEN == 0 { goto tail_start }
85
86	MOVUPS ALPHA, ALPHA_2 // ALPHA_2 := ALPHA  for pipelining
87
88loop:  // do {
89	// y[i] += alpha * x[i] unrolled 8x.
90	MOVUPS (X_PTR)(IDX*8), X2   // X_i = x[i]
91	MOVUPS 16(X_PTR)(IDX*8), X3
92	MOVUPS 32(X_PTR)(IDX*8), X4
93	MOVUPS 48(X_PTR)(IDX*8), X5
94
95	MULPD ALPHA, X2   // X_i *= alpha
96	MULPD ALPHA_2, X3
97	MULPD ALPHA, X4
98	MULPD ALPHA_2, X5
99
100	ADDPD (Y_PTR)(IDX*8), X2   // X_i += y[i]
101	ADDPD 16(Y_PTR)(IDX*8), X3
102	ADDPD 32(Y_PTR)(IDX*8), X4
103	ADDPD 48(Y_PTR)(IDX*8), X5
104
105	MOVUPS X2, (DST_PTR)(IDX*8)   // y[i] = X_i
106	MOVUPS X3, 16(DST_PTR)(IDX*8)
107	MOVUPS X4, 32(DST_PTR)(IDX*8)
108	MOVUPS X5, 48(DST_PTR)(IDX*8)
109
110	ADDQ $8, IDX  // i += 8
111	DECQ LEN
112	JNZ  loop     // } while --LEN > 0
113	CMPQ TAIL, $0 // if TAIL == 0 { return }
114	JE   end
115
116tail_start: // Reset loop registers
117	MOVQ TAIL, LEN // Loop counter: LEN = TAIL
118	SHRQ $1, LEN   // LEN = floor( TAIL / 2 )
119	JZ   tail_one  // if LEN == 0 { goto tail }
120
121tail_two: // do {
122	MOVUPS (X_PTR)(IDX*8), X2   // X2 = x[i]
123	MULPD  ALPHA, X2            // X2 *= alpha
124	ADDPD  (Y_PTR)(IDX*8), X2   // X2 += y[i]
125	MOVUPS X2, (DST_PTR)(IDX*8) // y[i] = X2
126	ADDQ   $2, IDX              // i += 2
127	DECQ   LEN
128	JNZ    tail_two             // } while --LEN > 0
129
130	ANDQ $1, TAIL
131	JZ   end      // if TAIL == 0 { goto end }
132
133tail_one:
134	MOVSD (X_PTR)(IDX*8), X2   // X2 = x[i]
135	MULSD ALPHA, X2            // X2 *= a
136	ADDSD (Y_PTR)(IDX*8), X2   // X2 += y[i]
137	MOVSD X2, (DST_PTR)(IDX*8) // y[i] = X2
138
139end:
140	RET
141