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 MOVDDUP_XPTR_IDX_8__X3 LONG $0x1C120FF2; BYTE $0xC6 // MOVDDUP (SI)(AX*8), X3
10 #define MOVDDUP_16_XPTR_IDX_8__X5 LONG $0x6C120FF2; WORD $0x10C6 // MOVDDUP 16(SI)(AX*8), X5
11 #define MOVDDUP_32_XPTR_IDX_8__X7 LONG $0x7C120FF2; WORD $0x20C6 // MOVDDUP 32(SI)(AX*8), X7
12 #define MOVDDUP_48_XPTR_IDX_8__X9 LONG $0x120F44F2; WORD $0xC64C; BYTE $0x30 // MOVDDUP 48(SI)(AX*8), X9
14 #define MOVDDUP_XPTR_IIDX_8__X2 LONG $0x14120FF2; BYTE $0xD6 // MOVDDUP (SI)(DX*8), X2
15 #define MOVDDUP_16_XPTR_IIDX_8__X4 LONG $0x64120FF2; WORD $0x10D6 // MOVDDUP 16(SI)(DX*8), X4
16 #define MOVDDUP_32_XPTR_IIDX_8__X6 LONG $0x74120FF2; WORD $0x20D6 // MOVDDUP 32(SI)(DX*8), X6
17 #define MOVDDUP_48_XPTR_IIDX_8__X8 LONG $0x120F44F2; WORD $0xD644; BYTE $0x30 // MOVDDUP 48(SI)(DX*8), X8
19 #define ADDSUBPD_X2_X3 LONG $0xDAD00F66 // ADDSUBPD X2, X3
20 #define ADDSUBPD_X4_X5 LONG $0xECD00F66 // ADDSUBPD X4, X5
21 #define ADDSUBPD_X6_X7 LONG $0xFED00F66 // ADDSUBPD X6, X7
22 #define ADDSUBPD_X8_X9 LONG $0xD00F4566; BYTE $0xC8 // ADDSUBPD X8, X9
33 // func DotuUnitary(x, y []complex128) (sum complex128)
34 TEXT ·DotuUnitary(SB), NOSPLIT, $0
35 MOVQ x_base+0(FP), X_PTR // X_PTR = &x
36 MOVQ y_base+24(FP), Y_PTR // Y_PTR = &y
37 MOVQ x_len+8(FP), LEN // LEN = min( len(x), len(y) )
38 CMPQ y_len+32(FP), LEN
39 CMOVQLE y_len+32(FP), LEN
40 PXOR SUM, SUM // SUM = 0
41 CMPQ LEN, $0 // if LEN == 0 { return }
43 PXOR P_SUM, P_SUM // P_SUM = 0
44 XORQ IDX, IDX // IDX = 0
47 ANDQ $3, TAIL // TAIL = floor( LEN / 4 )
48 SHRQ $2, LEN // LEN = LEN % 4
49 JZ dot_tail // if LEN == 0 { goto dot_tail }
52 MOVDDUP_XPTR_IDX_8__X3 // X_(i+1) = { real(x[i], real(x[i]) }
53 MOVDDUP_16_XPTR_IDX_8__X5
54 MOVDDUP_32_XPTR_IDX_8__X7
55 MOVDDUP_48_XPTR_IDX_8__X9
57 MOVDDUP_XPTR_IIDX_8__X2 // X_i = { imag(x[i]), imag(x[i]) }
58 MOVDDUP_16_XPTR_IIDX_8__X4
59 MOVDDUP_32_XPTR_IIDX_8__X6
60 MOVDDUP_48_XPTR_IIDX_8__X8
62 // X_j = { imag(y[i]), real(y[i]) }
63 MOVUPS (Y_PTR)(IDX*8), X10
64 MOVUPS 16(Y_PTR)(IDX*8), X11
65 MOVUPS 32(Y_PTR)(IDX*8), X12
66 MOVUPS 48(Y_PTR)(IDX*8), X13
68 // X_(i+1) = { imag(a) * real(x[i]), real(a) * real(x[i]) }
74 // X_j = { real(y[i]), imag(y[i]) }
80 // X_i = { real(a) * imag(x[i]), imag(a) * imag(x[i]) }
87 // imag(result[i]): imag(a)*real(x[i]) + real(a)*imag(x[i]),
88 // real(result[i]): real(a)*real(x[i]) - imag(a)*imag(x[i])
101 ADDQ $8, IDX // IDX += 8
102 ADDQ $8, I_IDX // I_IDX += 8
104 JNZ dot_loop // } while --LEN > 0
105 ADDPD P_SUM, SUM // SUM += P_SUM
106 CMPQ TAIL, $0 // if TAIL == 0 { return }
110 MOVDDUP_XPTR_IDX_8__X3 // X_(i+1) = { real(x[i] , real(x[i]) }
111 MOVDDUP_XPTR_IIDX_8__X2 // X_i = { imag(x[i]) , imag(x[i]) }
112 MOVUPS (Y_PTR)(IDX*8), X10 // X_j = { imag(y[i]) , real(y[i]) }
113 MULPD X10, X3 // X_(i+1) = { imag(a) * real(x[i]), real(a) * real(x[i]) }
114 SHUFPD $0x1, X10, X10 // X_j = { real(y[i]) , imag(y[i]) }
115 MULPD X10, X2 // X_i = { real(a) * imag(x[i]), imag(a) * imag(x[i]) }
118 // imag(result[i]): imag(a)*real(x[i]) + real(a)*imag(x[i]),
119 // real(result[i]): real(a)*real(x[i]) - imag(a)*imag(x[i])
122 ADDPD X3, SUM // psum += result[i]
123 ADDQ $2, IDX // IDX += 2
124 ADDQ $2, I_IDX // I_IDX += 2
126 JNZ dot_tail // } while TAIL > 0
129 MOVUPS SUM, sum+48(FP)