github.com/gopherd/gonum@v0.0.4/internal/asm/c64/dotuinc_amd64.s (about) 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. 4 5 // +build !noasm,!gccgo,!safe 6 7 #include "textflag.h" 8 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 13 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 18 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 23 24 #define X_PTR SI 25 #define Y_PTR DI 26 #define LEN CX 27 #define TAIL BX 28 #define SUM X0 29 #define P_SUM X1 30 #define INC_X R8 31 #define INCx3_X R9 32 #define INC_Y R10 33 #define INCx3_Y R11 34 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 } 43 JE dotu_end 44 MOVQ ix+72(FP), INC_X 45 MOVQ iy+80(FP), INC_Y 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) 49 SHLQ $3, INC_X 50 MOVQ incY+64(FP), INC_Y // INC_Y = incY * sizeof(complex64) 51 SHLQ $3, INC_Y 52 53 MOVQ LEN, TAIL 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 } 57 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 60 61 dotu_loop: // do { 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 66 67 // X_(i-1) = { imag(x[i]), imag(x[i]) } 68 MOVSHDUP_X3_X2 69 MOVSHDUP_X5_X4 70 MOVSHDUP_X7_X6 71 MOVSHDUP_X9_X8 72 73 // X_i = { real(x[i]), real(x[i]) } 74 MOVSLDUP_X3_X3 75 MOVSLDUP_X5_X5 76 MOVSLDUP_X7_X7 77 MOVSLDUP_X9_X9 78 79 // X_j = { imag(y[i]), real(y[i]) } 80 MOVSD (Y_PTR), X10 81 MOVSD (Y_PTR)(INC_Y*1), X11 82 MOVSD (Y_PTR)(INC_Y*2), X12 83 MOVSD (Y_PTR)(INCx3_Y*1), X13 84 85 // X_i = { imag(y[i]) * real(x[i]), real(y[i]) * real(x[i]) } 86 MULPS X10, X3 87 MULPS X11, X5 88 MULPS X12, X7 89 MULPS X13, X9 90 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 96 97 // X_(i-1) = { real(y[i]) * imag(x[i]), imag(y[i]) * imag(x[i]) } 98 MULPS X10, X2 99 MULPS X11, X4 100 MULPS X12, X6 101 MULPS X13, X8 102 103 // 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]) } 106 ADDSUBPS_X2_X3 107 ADDSUBPS_X4_X5 108 ADDSUBPS_X6_X7 109 ADDSUBPS_X8_X9 110 111 // SUM += X_i 112 ADDPS X3, SUM 113 ADDPS X5, P_SUM 114 ADDPS X7, SUM 115 ADDPS X9, P_SUM 116 117 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]) 119 120 DECQ LEN 121 JNZ dotu_loop // } while --LEN > 0 122 123 ADDPS P_SUM, SUM // SUM = { P_SUM + SUM } 124 CMPQ TAIL, $0 // if TAIL == 0 { return } 125 JE dotu_end 126 127 dotu_tail: // do { 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]) } 135 136 // 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]) } 139 ADDSUBPS_X2_X3 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 143 DECQ TAIL 144 JNZ dotu_tail // } while --TAIL > 0 145 146 dotu_end: 147 MOVSD SUM, sum+88(FP) // return SUM 148 RET