github.com/gopherd/gonum@v0.0.4/internal/asm/c128/scalUnitary_amd64.s (about)

     1  // Copyright ©2017 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  // +build !noasm,!gccgo,!safe
     6  
     7  #include "textflag.h"
     8  
     9  #define SRC SI
    10  #define DST SI
    11  #define LEN CX
    12  #define IDX AX
    13  #define TAIL BX
    14  #define ALPHA X0
    15  #define ALPHA_C X1
    16  #define ALPHA2 X10
    17  #define ALPHA_C2 X11
    18  
    19  #define MOVDDUP_X2_X3    LONG $0xDA120FF2 // MOVDDUP X2, X3
    20  #define MOVDDUP_X4_X5    LONG $0xEC120FF2 // MOVDDUP X4, X5
    21  #define MOVDDUP_X6_X7    LONG $0xFE120FF2 // MOVDDUP X6, X7
    22  #define MOVDDUP_X8_X9    LONG $0x120F45F2; BYTE $0xC8 // MOVDDUP X8, X9
    23  
    24  #define ADDSUBPD_X2_X3    LONG $0xDAD00F66 // ADDSUBPD X2, X3
    25  #define ADDSUBPD_X4_X5    LONG $0xECD00F66 // ADDSUBPD X4, X5
    26  #define ADDSUBPD_X6_X7    LONG $0xFED00F66 // ADDSUBPD X6, X7
    27  #define ADDSUBPD_X8_X9    LONG $0xD00F4566; BYTE $0xC8 // ADDSUBPD X8, X9
    28  
    29  // func ScalUnitary(alpha complex128, x []complex128)
    30  TEXT ·ScalUnitary(SB), NOSPLIT, $0
    31  	MOVQ x_base+16(FP), SRC // SRC = &x
    32  	MOVQ x_len+24(FP), LEN  // LEN = len(x)
    33  	CMPQ LEN, $0            // if LEN == 0 { return }
    34  	JE   scal_end
    35  
    36  	MOVUPS alpha+0(FP), ALPHA     // ALPHA = { imag(alpha), real(alpha) }
    37  	MOVAPS ALPHA, ALPHA_C
    38  	SHUFPD $0x1, ALPHA_C, ALPHA_C // ALPHA_C = { real(alpha), imag(alpha) }
    39  
    40  	XORQ   IDX, IDX          // IDX = 0
    41  	MOVAPS ALPHA, ALPHA2     // Copy ALPHA and ALPHA_C for pipelining
    42  	MOVAPS ALPHA_C, ALPHA_C2
    43  	MOVQ   LEN, TAIL
    44  	SHRQ   $2, LEN           // LEN = floor( n / 4 )
    45  	JZ     scal_tail         // if BX == 0 { goto scal_tail }
    46  
    47  scal_loop: // do {
    48  	MOVUPS (SRC)(IDX*8), X2   // X_i = { imag(x[i]), real(x[i]) }
    49  	MOVUPS 16(SRC)(IDX*8), X4
    50  	MOVUPS 32(SRC)(IDX*8), X6
    51  	MOVUPS 48(SRC)(IDX*8), X8
    52  
    53  	// X_(i+1) = { real(x[i], real(x[i]) }
    54  	MOVDDUP_X2_X3
    55  	MOVDDUP_X4_X5
    56  	MOVDDUP_X6_X7
    57  	MOVDDUP_X8_X9
    58  
    59  	// X_i = { imag(x[i]), imag(x[i]) }
    60  	SHUFPD $0x3, X2, X2
    61  	SHUFPD $0x3, X4, X4
    62  	SHUFPD $0x3, X6, X6
    63  	SHUFPD $0x3, X8, X8
    64  
    65  	// X_i     = { real(ALPHA) * imag(x[i]), imag(ALPHA) * imag(x[i])  }
    66  	// X_(i+1) = { imag(ALPHA) * real(x[i]), real(ALPHA) * real(x[i])  }
    67  	MULPD ALPHA_C, X2
    68  	MULPD ALPHA, X3
    69  	MULPD ALPHA_C2, X4
    70  	MULPD ALPHA2, X5
    71  	MULPD ALPHA_C, X6
    72  	MULPD ALPHA, X7
    73  	MULPD ALPHA_C2, X8
    74  	MULPD ALPHA2, X9
    75  
    76  	// X_(i+1) = {
    77  	//	imag(result[i]):  imag(ALPHA)*real(x[i]) + real(ALPHA)*imag(x[i]),
    78  	//	real(result[i]):  real(ALPHA)*real(x[i]) - imag(ALPHA)*imag(x[i])
    79  	//  }
    80  	ADDSUBPD_X2_X3
    81  	ADDSUBPD_X4_X5
    82  	ADDSUBPD_X6_X7
    83  	ADDSUBPD_X8_X9
    84  
    85  	MOVUPS X3, (DST)(IDX*8)   // x[i] = X_(i+1)
    86  	MOVUPS X5, 16(DST)(IDX*8)
    87  	MOVUPS X7, 32(DST)(IDX*8)
    88  	MOVUPS X9, 48(DST)(IDX*8)
    89  	ADDQ   $8, IDX            // IDX += 8
    90  	DECQ   LEN
    91  	JNZ    scal_loop          // } while --LEN > 0
    92  
    93  scal_tail:
    94  	ANDQ $3, TAIL // TAIL = TAIL % 4
    95  	JZ   scal_end // if TAIL == 0 { return }
    96  
    97  scal_tail_loop: // do {
    98  	MOVUPS (SRC)(IDX*8), X2 // X_i = { imag(x[i]), real(x[i]) }
    99  	MOVDDUP_X2_X3           // X_(i+1) = { real(x[i], real(x[i]) }
   100  	SHUFPD $0x3, X2, X2     // X_i = { imag(x[i]), imag(x[i]) }
   101  	MULPD  ALPHA_C, X2      // X_i     = { real(ALPHA) * imag(x[i]), imag(ALPHA) * imag(x[i])  }
   102  	MULPD  ALPHA, X3        // X_(i+1) = { imag(ALPHA) * real(x[i]), real(ALPHA) * real(x[i])  }
   103  
   104  	// X_(i+1) = {
   105  	//	imag(result[i]):  imag(ALPHA)*real(x[i]) + real(ALPHA)*imag(x[i]),
   106  	//	real(result[i]):  real(ALPHA)*real(x[i]) - imag(ALPHA)*imag(x[i])
   107  	//  }
   108  	ADDSUBPD_X2_X3
   109  
   110  	MOVUPS X3, (DST)(IDX*8) // x[i] = X_(i+1)
   111  	ADDQ   $2, IDX          // IDX += 2
   112  	DECQ   TAIL
   113  	JNZ    scal_tail_loop   // }  while --LEN > 0
   114  
   115  scal_end:
   116  	RET