David Brazdil | 0f672f6 | 2019-12-10 10:32:29 +0000 | [diff] [blame^] | 1 | ;; SPDX-License-Identifier: GPL-2.0-or-later |
Andrew Scull | b4b6d4a | 2019-01-02 15:54:55 +0000 | [diff] [blame] | 2 | ;; Copyright (C) 2010 Texas Instruments Incorporated |
| 3 | ;; Contributed by Mark Salter <msalter@redhat.com>. |
| 4 | ;; |
Andrew Scull | b4b6d4a | 2019-01-02 15:54:55 +0000 | [diff] [blame] | 5 | |
| 6 | #include <linux/linkage.h> |
| 7 | |
| 8 | ;; uint64_t __c6xabi_mpyll(uint64_t x, uint64_t y) |
| 9 | ;; |
| 10 | ;; 64x64 multiply |
| 11 | ;; First compute partial results using 32-bit parts of x and y: |
| 12 | ;; |
| 13 | ;; b63 b32 b31 b0 |
| 14 | ;; ----------------------------- |
| 15 | ;; | 1 | 0 | |
| 16 | ;; ----------------------------- |
| 17 | ;; |
| 18 | ;; P0 = X0*Y0 |
| 19 | ;; P1 = X0*Y1 + X1*Y0 |
| 20 | ;; P2 = X1*Y1 |
| 21 | ;; |
| 22 | ;; result = (P2 << 64) + (P1 << 32) + P0 |
| 23 | ;; |
| 24 | ;; Since the result is also 64-bit, we can skip the P2 term. |
| 25 | |
| 26 | .text |
| 27 | ENTRY(__c6xabi_mpyll) |
| 28 | mpy32u .m1x A4,B4,A1:A0 ; X0*Y0 |
| 29 | b .s2 B3 |
| 30 | || mpy32u .m2x B5,A4,B1:B0 ; X0*Y1 (don't need upper 32-bits) |
| 31 | || mpy32u .m1x A5,B4,A3:A2 ; X1*Y0 (don't need upper 32-bits) |
| 32 | nop |
| 33 | nop |
| 34 | mv .s1 A0,A4 |
| 35 | add .l1x A2,B0,A5 |
| 36 | add .s1 A1,A5,A5 |
| 37 | ENDPROC(__c6xabi_mpyll) |