Daniel Dunbar | b3a6901 | 2009-06-26 16:47:03 +0000 | [diff] [blame^] | 1 | //===-- popcountti2.c - Implement __popcountti2 ----------------------------===// |
| 2 | // |
| 3 | // The LLVM Compiler Infrastructure |
| 4 | // |
| 5 | // This file is distributed under the University of Illinois Open Source |
| 6 | // License. See LICENSE.TXT for details. |
| 7 | // |
| 8 | //===----------------------------------------------------------------------===// |
| 9 | // |
| 10 | // This file implements __popcountti2 for the compiler_rt library. |
| 11 | // |
| 12 | //===----------------------------------------------------------------------===// |
| 13 | |
| 14 | #if __x86_64 |
| 15 | |
| 16 | #include "int_lib.h" |
| 17 | |
| 18 | // Returns: count of 1 bits |
| 19 | |
| 20 | si_int |
| 21 | __popcountti2(ti_int a) |
| 22 | { |
| 23 | tu_int x3 = (tu_int)a; |
| 24 | x3 = x3 - ((x3 >> 1) & (((tu_int)0x5555555555555555uLL << 64) | |
| 25 | 0x5555555555555555uLL)); |
| 26 | // Every 2 bits holds the sum of every pair of bits (64) |
| 27 | x3 = ((x3 >> 2) & (((tu_int)0x3333333333333333uLL << 64) | 0x3333333333333333uLL)) |
| 28 | + (x3 & (((tu_int)0x3333333333333333uLL << 64) | 0x3333333333333333uLL)); |
| 29 | // Every 4 bits holds the sum of every 4-set of bits (3 significant bits) (32) |
| 30 | x3 = (x3 + (x3 >> 4)) |
| 31 | & (((tu_int)0x0F0F0F0F0F0F0F0FuLL << 64) | 0x0F0F0F0F0F0F0F0FuLL); |
| 32 | // Every 8 bits holds the sum of every 8-set of bits (4 significant bits) (16) |
| 33 | du_int x2 = (du_int)(x3 + (x3 >> 64)); |
| 34 | // Every 8 bits holds the sum of every 8-set of bits (5 significant bits) (8) |
| 35 | su_int x = (su_int)(x2 + (x2 >> 32)); |
| 36 | // Every 8 bits holds the sum of every 8-set of bits (6 significant bits) (4) |
| 37 | x = x + (x >> 16); |
| 38 | // Every 8 bits holds the sum of every 8-set of bits (7 significant bits) (2) |
| 39 | // Upper 16 bits are garbage |
| 40 | return (x + (x >> 8)) & 0xFF; // (8 significant bits) |
| 41 | } |
| 42 | |
| 43 | #endif |