|
| 1 | +/*************************************************************************** |
| 2 | +Copyright (c) 2020, The OpenBLAS Project |
| 3 | +All rights reserved. |
| 4 | +Redistribution and use in source and binary forms, with or without |
| 5 | +modification, are permitted provided that the following conditions are |
| 6 | +met: |
| 7 | +1. Redistributions of source code must retain the above copyright |
| 8 | +notice, this list of conditions and the following disclaimer. |
| 9 | +2. Redistributions in binary form must reproduce the above copyright |
| 10 | +notice, this list of conditions and the following disclaimer in |
| 11 | +the documentation and/or other materials provided with the |
| 12 | +distribution. |
| 13 | +3. Neither the name of the OpenBLAS project nor the names of |
| 14 | +its contributors may be used to endorse or promote products |
| 15 | +derived from this software without specific prior written permission. |
| 16 | +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" |
| 17 | +AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE |
| 18 | +IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE |
| 19 | +ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE |
| 20 | +LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL |
| 21 | +DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR |
| 22 | +SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER |
| 23 | +CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, |
| 24 | +OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE |
| 25 | +USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. |
| 26 | +*****************************************************************************/ |
| 27 | + |
| 28 | +#define HAVE_KERNEL_8 1 |
| 29 | + |
| 30 | +static double ddot_kernel_8 (long n, double *x, double *y) |
| 31 | +{ |
| 32 | + double dot; |
| 33 | + |
| 34 | + __asm__ |
| 35 | + ( |
| 36 | + "dcbt 0, %2 \n\t" |
| 37 | + "dcbt 0, %3 \n\t" |
| 38 | + |
| 39 | + "xxlxor 32, 32, 32 \n\t" |
| 40 | + "xxlxor 33, 33, 33 \n\t" |
| 41 | + "xxlxor 34, 34, 34 \n\t" |
| 42 | + "xxlxor 35, 35, 35 \n\t" |
| 43 | + "xxlxor 36, 36, 36 \n\t" |
| 44 | + "xxlxor 37, 37, 37 \n\t" |
| 45 | + "xxlxor 38, 38, 38 \n\t" |
| 46 | + "xxlxor 39, 39, 39 \n\t" |
| 47 | + |
| 48 | + "lxvp 40, 0(%2) \n\t" |
| 49 | + "lxvp 42, 32(%2) \n\t" |
| 50 | + "lxvp 44, 64(%2) \n\t" |
| 51 | + "lxvp 46, 96(%2) \n\t" |
| 52 | + "lxvp 48, 0(%3) \n\t" |
| 53 | + "lxvp 50, 32(%3) \n\t" |
| 54 | + "lxvp 52, 64(%3) \n\t" |
| 55 | + "lxvp 54, 96(%3) \n\t" |
| 56 | + |
| 57 | + "addi %2, %2, 128 \n\t" |
| 58 | + "addi %3, %3, 128 \n\t" |
| 59 | + |
| 60 | + "addic. %1, %1, -16 \n\t" |
| 61 | + "ble two%= \n\t" |
| 62 | + |
| 63 | + ".align 5 \n" |
| 64 | + "one%=: \n\t" |
| 65 | + |
| 66 | + "xvmaddadp 32, 40, 48 \n\t" |
| 67 | + "xvmaddadp 33, 41, 49 \n\t" |
| 68 | + "lxvp 40, 0(%2) \n\t" |
| 69 | + "lxvp 48, 0(%3) \n\t" |
| 70 | + "xvmaddadp 34, 42, 50 \n\t" |
| 71 | + "xvmaddadp 35, 43, 51 \n\t" |
| 72 | + "lxvp 42, 32(%2) \n\t" |
| 73 | + "lxvp 50, 32(%3) \n\t" |
| 74 | + "xvmaddadp 36, 44, 52 \n\t" |
| 75 | + "xvmaddadp 37, 45, 53 \n\t" |
| 76 | + "lxvp 44, 64(%2) \n\t" |
| 77 | + "lxvp 52, 64(%3) \n\t" |
| 78 | + "xvmaddadp 38, 46, 54 \n\t" |
| 79 | + "xvmaddadp 39, 47, 55 \n\t" |
| 80 | + "lxvp 46, 96(%2) \n\t" |
| 81 | + "lxvp 54, 96(%3) \n\t" |
| 82 | + |
| 83 | + "addi %2, %2, 128 \n\t" |
| 84 | + "addi %3, %3, 128 \n\t" |
| 85 | + |
| 86 | + "addic. %1, %1, -16 \n\t" |
| 87 | + "bgt one%= \n" |
| 88 | + |
| 89 | + "two%=: \n\t" |
| 90 | + |
| 91 | + "xvmaddadp 32, 40, 48 \n\t" |
| 92 | + "xvmaddadp 33, 41, 49 \n\t" |
| 93 | + "xvmaddadp 34, 42, 50 \n\t" |
| 94 | + "xvmaddadp 35, 43, 51 \n\t" |
| 95 | + "xvmaddadp 36, 44, 52 \n\t" |
| 96 | + "xvmaddadp 37, 45, 53 \n\t" |
| 97 | + "xvmaddadp 38, 46, 54 \n\t" |
| 98 | + "xvmaddadp 39, 47, 55 \n\t" |
| 99 | + |
| 100 | + "xvadddp 32, 32, 33 \n\t" |
| 101 | + "xvadddp 34, 34, 35 \n\t" |
| 102 | + "xvadddp 36, 36, 37 \n\t" |
| 103 | + "xvadddp 38, 38, 39 \n\t" |
| 104 | + |
| 105 | + "xvadddp 32, 32, 34 \n\t" |
| 106 | + "xvadddp 36, 36, 38 \n\t" |
| 107 | + |
| 108 | + "xvadddp 32, 32, 36 \n\t" |
| 109 | + |
| 110 | + XXSWAPD_S(33,32) |
| 111 | + |
| 112 | + "xsadddp %x0, 32, 33 \n" |
| 113 | + |
| 114 | + "#dot=%0 n=%1 x=%4=%2 y=%5=%3\n" |
| 115 | + : |
| 116 | + "=d" (dot), // 0 |
| 117 | + "+r" (n), // 1 |
| 118 | + "+b" (x), // 2 |
| 119 | + "+b" (y) // 3 |
| 120 | + : |
| 121 | + "m" (*x), |
| 122 | + "m" (*y) |
| 123 | + : |
| 124 | + "cr0", |
| 125 | + "vs32","vs33","vs34","vs35","vs36","vs37","vs38","vs39", |
| 126 | + "vs40","vs41","vs42","vs43","vs44","vs45","vs46","vs47", |
| 127 | + "vs48","vs49","vs50","vs51","vs52","vs53","vs54","vs55" |
| 128 | + ); |
| 129 | + |
| 130 | + return dot; |
| 131 | +} |
0 commit comments