95 lines · plain
1/*2 * Copyright 2014 Martin Peres <martin.peres@free.fr>3 *4 * Permission is hereby granted, free of charge, to any person obtaining a5 * copy of this software and associated documentation files (the "Software"),6 * to deal in the Software without restriction, including without limitation7 * the rights to use, copy, modify, merge, publish, distribute, sublicense,8 * and/or sell copies of the Software, and to permit persons to whom the9 * Software is furnished to do so, subject to the folloing conditions:10 *11 * The above copyright notice and this permission notice shall be included in12 * all copies or substantial portions of the Software.13 *14 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR15 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,16 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL17 * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR18 * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,19 * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR20 * OTHER DEALINGS IN THE SOFTWARE.21 *22 * Authors: Martin Peres23 */24 25/******************************************************************************26 * arith data segment27 *****************************************************************************/28#ifdef INCLUDE_PROC29#endif30 31#ifdef INCLUDE_DATA32#endif33 34/******************************************************************************35 * arith code segment36 *****************************************************************************/37#ifdef INCLUDE_CODE38 39// does a 32x32 -> 64 multiplication40//41// A * B = A_lo * B_lo42// + ( A_hi * B_lo ) << 1643// + ( A_lo * B_hi ) << 1644// + ( A_hi * B_hi ) << 3245//46// $r15 - current47// $r14 - A48// $r13 - B49// $r12 - mul_lo (return)50// $r11 - mul_hi (return)51// $r0 - zero52mulu32_32_64:53 push $r1 // A_hi54 push $r2 // B_hi55 push $r3 // tmp056 push $r4 // tmp157 58 shr b32 $r1 $r14 1659 shr b32 $r2 $r13 1660 61 clear b32 $r1262 clear b32 $r1163 64 // A_lo * B_lo65 mulu $r12 $r14 $r1366 67 // ( A_hi * B_lo ) << 1668 mulu $r3 $r1 $r13 // tmp0 = A_hi * B_lo69 mov b32 $r4 $r370 and $r3 0xffff // tmp0 = tmp0_lo71 shl b32 $r3 1672 shr b32 $r4 16 // tmp1 = tmp0_hi73 add b32 $r12 $r374 adc b32 $r11 $r475 76 // ( A_lo * B_hi ) << 1677 mulu $r3 $r14 $r2 // tmp0 = A_lo * B_hi78 mov b32 $r4 $r379 and $r3 0xffff // tmp0 = tmp0_lo80 shl b32 $r3 1681 shr b32 $r4 16 // tmp1 = tmp0_hi82 add b32 $r12 $r383 adc b32 $r11 $r484 85 // ( A_hi * B_hi ) << 3286 mulu $r3 $r1 $r2 // tmp0 = A_hi * B_hi87 add b32 $r11 $r388 89 pop $r490 pop $r391 pop $r292 pop $r193 ret94#endif95