2 # Copyright 2002,2005 Free Software Foundation, Inc.
4 # This file is part of GNU Radio
6 # GNU Radio is free software; you can redistribute it and/or modify
7 # it under the terms of the GNU General Public License as published by
8 # the Free Software Foundation; either version 3, or (at your option)
11 # GNU Radio is distributed in the hope that it will be useful,
12 # but WITHOUT ANY WARRANTY; without even the implied warranty of
13 # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14 # GNU General Public License for more details.
16 # You should have received a copy of the GNU General Public License
17 # along with GNU Radio; see the file COPYING. If not, write to
18 # the Free Software Foundation, Inc., 51 Franklin Street,
19 # Boston, MA 02110-1301, USA.
22 # input and taps are guarenteed to be 16 byte aligned.
23 # n_2_complex_blocks is != 0
26 # fcomplex_dotprod_generic (const float *input,
27 # const float *taps, unsigned n_2_complex_blocks, float *result)
36 # sum0 += input[0] * taps[0];
37 # sum1 += input[0] * taps[1];
38 # sum2 += input[1] * taps[2];
39 # sum3 += input[1] * taps[3];
44 # } while (--n_2_complex_blocks != 0);
47 # result[0] = sum0 + sum2;
48 # result[1] = sum1 + sum3;
55 .file "fcomplex_dotprod_3dnow64.S"
59 .globl GLOB_SYMB(fcomplex_dotprod_3dnow)
60 DEF_FUNC_HEAD(fcomplex_dotprod_3dnow)
61 GLOB_SYMB(fcomplex_dotprod_3dnow):
63 # intput: rdi, taps: rsi, n_2_ccomplex_blocks: rdx, result: rcx
69 pxor %mm4, %mm4 # mm4 = 0 0
70 pxor %mm5, %mm5 # mm5 = 0 0
71 pxor %mm6, %mm6 # mm6 = 0 0
72 pxor %mm7, %mm7 # mm7 = 0 0
74 shr $1, %rax # rax = n_2_complex_blocks / 2
90 # something like ?? cycles / loop
133 # We've handled the bulk of multiplies up to here.
134 # Now accumulate the final two additions and see if original
135 # n_2_complex_blocks was odd. If so, we've got 2 more
143 # The count was odd, do 2 more taps.
144 # Note that we've already got mm0 and mm1 preloaded
145 # from the main loop.
154 # at this point mm4, mm5, mm6 and mm7 contain partial sums
160 movq %mm4, (%rcx) # result
165 FUNC_TAIL(fcomplex_dotprod_3dnow)
166 .ident "Hand coded x86_64 3DNow! assembly"
168 #if defined(__linux__) && defined(__ELF__)
169 .section .note.GNU-stack,"",%progbits