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
9 #define MOVSHDUP_X3_X2 LONG $0xD3160FF3 // MOVSHDUP X3, X2
10 #define MOVSHDUP_X5_X4 LONG $0xE5160FF3 // MOVSHDUP X5, X4
11 #define MOVSHDUP_X7_X6 LONG $0xF7160FF3 // MOVSHDUP X7, X6
12 #define MOVSHDUP_X9_X8 LONG $0x160F45F3; BYTE $0xC1 // MOVSHDUP X9, X8
14 #define MOVSLDUP_X3_X3 LONG $0xDB120FF3 // MOVSLDUP X3, X3
15 #define MOVSLDUP_X5_X5 LONG $0xED120FF3 // MOVSLDUP X5, X5
16 #define MOVSLDUP_X7_X7 LONG $0xFF120FF3 // MOVSLDUP X7, X7
17 #define MOVSLDUP_X9_X9 LONG $0x120F45F3; BYTE $0xC9 // MOVSLDUP X9, X9
19 #define ADDSUBPS_X2_X3 LONG $0xDAD00FF2 // ADDSUBPS X2, X3
20 #define ADDSUBPS_X4_X5 LONG $0xECD00FF2 // ADDSUBPS X4, X5
21 #define ADDSUBPS_X6_X7 LONG $0xFED00FF2 // ADDSUBPS X6, X7
22 #define ADDSUBPS_X8_X9 LONG $0xD00F45F2; BYTE $0xC8 // ADDSUBPS X8, X9
35 // func DotuInc(x, y []complex64, n, incX, incY, ix, iy uintptr) (sum complex64)
36 TEXT ·DotuInc(SB), NOSPLIT, $0
37 MOVQ x_base+0(FP), X_PTR // X_PTR = &x
38 MOVQ y_base+24(FP), Y_PTR // Y_PTR = &y
39 PXOR SUM, SUM // SUM = 0
40 PXOR P_SUM, P_SUM // P_SUM = 0
41 MOVQ n+48(FP), LEN // LEN = n
42 CMPQ LEN, $0 // if LEN == 0 { return }
46 LEAQ (X_PTR)(INC_X*8), X_PTR // X_PTR = &(X_PTR[ix])
47 LEAQ (Y_PTR)(INC_Y*8), Y_PTR // Y_PTR = &(Y_PTR[iy])
48 MOVQ incX+56(FP), INC_X // INC_X = incX * sizeof(complex64)
50 MOVQ incY+64(FP), INC_Y // INC_Y = incY * sizeof(complex64)
54 ANDQ $3, TAIL // TAIL = LEN % 4
55 SHRQ $2, LEN // LEN = floor( LEN / 4 )
56 JZ dotu_tail // if TAIL == 0 { goto dotu_tail }
58 LEAQ (INC_X)(INC_X*2), INCx3_X // INCx3_X = INC_X * 3
59 LEAQ (INC_Y)(INC_Y*2), INCx3_Y // INCx3_Y = INC_Y * 3
62 MOVSD (X_PTR), X3 // X_i = { imag(x[i]), real(x[i]) }
63 MOVSD (X_PTR)(INC_X*1), X5
64 MOVSD (X_PTR)(INC_X*2), X7
65 MOVSD (X_PTR)(INCx3_X*1), X9
67 // X_(i-1) = { imag(x[i]), imag(x[i]) }
73 // X_i = { real(x[i]), real(x[i]) }
79 // X_j = { imag(y[i]), real(y[i]) }
81 MOVSD (Y_PTR)(INC_Y*1), X11
82 MOVSD (Y_PTR)(INC_Y*2), X12
83 MOVSD (Y_PTR)(INCx3_Y*1), X13
85 // X_i = { imag(y[i]) * real(x[i]), real(y[i]) * real(x[i]) }
91 // X_j = { real(y[i]), imag(y[i]) }
92 SHUFPS $0xB1, X10, X10
93 SHUFPS $0xB1, X11, X11
94 SHUFPS $0xB1, X12, X12
95 SHUFPS $0xB1, X13, X13
97 // X_(i-1) = { real(y[i]) * imag(x[i]), imag(y[i]) * imag(x[i]) }
104 // imag(result[i]): imag(y[i]) * real(x[i]) + real(y[i]) * imag(x[i]),
105 // real(result[i]): real(y[i]) * real(x[i]) - imag(y[i]) * imag(x[i]) }
117 LEAQ (X_PTR)(INC_X*4), X_PTR // X_PTR = &(X_PTR[INC_X*4])
118 LEAQ (Y_PTR)(INC_Y*4), Y_PTR // Y_PTR = &(Y_PTR[INC_Y*4])
121 JNZ dotu_loop // } while --LEN > 0
123 ADDPS P_SUM, SUM // SUM = { P_SUM + SUM }
124 CMPQ TAIL, $0 // if TAIL == 0 { return }
128 MOVSD (X_PTR), X3 // X_i = { imag(x[i]), real(x[i]) }
129 MOVSHDUP_X3_X2 // X_(i-1) = { imag(x[i]), imag(x[i]) }
130 MOVSLDUP_X3_X3 // X_i = { real(x[i]), real(x[i]) }
131 MOVUPS (Y_PTR), X10 // X_j = { imag(y[i]), real(y[i]) }
132 MULPS X10, X3 // X_i = { imag(y[i]) * real(x[i]), real(y[i]) * real(x[i]) }
133 SHUFPS $0x1, X10, X10 // X_j = { real(y[i]), imag(y[i]) }
134 MULPS X10, X2 // X_(i-1) = { real(y[i]) * imag(x[i]), imag(y[i]) * imag(x[i]) }
137 // imag(result[i]): imag(y[i])*real(x[i]) + real(y[i])*imag(x[i]),
138 // real(result[i]): real(y[i])*real(x[i]) - imag(y[i])*imag(x[i]) }
140 ADDPS X3, SUM // SUM += X_i
141 ADDQ INC_X, X_PTR // X_PTR += INC_X
142 ADDQ INC_Y, Y_PTR // Y_PTR += INC_Y
144 JNZ dotu_tail // } while --TAIL > 0
147 MOVSD SUM, sum+88(FP) // return SUM