1 // Copyright ©2016 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.
5 //+build !noasm,!appengine
20 #define MOVDDUP_X2_X3 LONG $0xDA120FF2 // MOVDDUP X2, X3
21 #define MOVDDUP_X4_X5 LONG $0xEC120FF2 // MOVDDUP X4, X5
22 #define MOVDDUP_X6_X7 LONG $0xFE120FF2 // MOVDDUP X6, X7
23 #define MOVDDUP_X8_X9 LONG $0x120F45F2; BYTE $0xC8 // MOVDDUP X8, X9
25 #define ADDSUBPD_X2_X3 LONG $0xDAD00F66 // ADDSUBPD X2, X3
26 #define ADDSUBPD_X4_X5 LONG $0xECD00F66 // ADDSUBPD X4, X5
27 #define ADDSUBPD_X6_X7 LONG $0xFED00F66 // ADDSUBPD X6, X7
28 #define ADDSUBPD_X8_X9 LONG $0xD00F4566; BYTE $0xC8 // ADDSUBPD X8, X9
30 // func ScalInc(alpha complex128, x []complex128, n, inc uintptr)
31 TEXT ·ScalInc(SB), NOSPLIT, $0
32 MOVQ x_base+16(FP), SRC // SRC = &x
33 MOVQ n+40(FP), LEN // LEN = len(x)
35 JE scal_end // if LEN == 0 { return }
37 MOVQ inc+48(FP), INC // INC = inc
38 SHLQ $4, INC // INC = INC * sizeof(complex128)
39 LEAQ (INC)(INC*2), INC3 // INC3 = 3 * INC
41 MOVUPS alpha+0(FP), ALPHA // ALPHA = { imag(alpha), real(alpha) }
43 SHUFPD $0x1, ALPHA_C, ALPHA_C // ALPHA_C = { real(alpha), imag(alpha) }
45 MOVAPS ALPHA, ALPHA2 // Copy ALPHA and ALPHA_C for pipelining
46 MOVAPS ALPHA_C, ALPHA_C2
48 SHRQ $2, LEN // LEN = floor( n / 4 )
49 JZ scal_tail // if BX == 0 { goto scal_tail }
52 MOVUPS (SRC), X2 // X_i = { imag(x[i]), real(x[i]) }
53 MOVUPS (SRC)(INC*1), X4
54 MOVUPS (SRC)(INC*2), X6
55 MOVUPS (SRC)(INC3*1), X8
57 // X_(i+1) = { real(x[i], real(x[i]) }
63 // X_i = { imag(x[i]), imag(x[i]) }
69 // X_i = { real(ALPHA) * imag(x[i]), imag(ALPHA) * imag(x[i]) }
70 // X_(i+1) = { imag(ALPHA) * real(x[i]), real(ALPHA) * real(x[i]) }
81 // imag(result[i]): imag(ALPHA)*real(x[i]) + real(ALPHA)*imag(x[i]),
82 // real(result[i]): real(ALPHA)*real(x[i]) - imag(ALPHA)*imag(x[i])
89 MOVUPS X3, (DST) // x[i] = X_(i+1)
90 MOVUPS X5, (DST)(INC*1)
91 MOVUPS X7, (DST)(INC*2)
92 MOVUPS X9, (DST)(INC3*1)
94 LEAQ (SRC)(INC*4), SRC // SRC = &(SRC[inc*4])
96 JNZ scal_loop // } while --BX > 0
99 ANDQ $3, TAIL // TAIL = TAIL % 4
100 JE scal_end // if TAIL == 0 { return }
102 scal_tail_loop: // do {
103 MOVUPS (SRC), X2 // X_i = { imag(x[i]), real(x[i]) }
104 MOVDDUP_X2_X3 // X_(i+1) = { real(x[i], real(x[i]) }
105 SHUFPD $0x3, X2, X2 // X_i = { imag(x[i]), imag(x[i]) }
106 MULPD ALPHA_C, X2 // X_i = { real(ALPHA) * imag(x[i]), imag(ALPHA) * imag(x[i]) }
107 MULPD ALPHA, X3 // X_(i+1) = { imag(ALPHA) * real(x[i]), real(ALPHA) * real(x[i]) }
110 // imag(result[i]): imag(ALPHA)*real(x[i]) + real(ALPHA)*imag(x[i]),
111 // real(result[i]): real(ALPHA)*real(x[i]) - imag(ALPHA)*imag(x[i])
115 MOVUPS X3, (DST) // x[i] = X_i
116 ADDQ INC, SRC // SRC = &(SRC[incX])
118 JNZ scal_tail_loop // } while --TAIL > 0