clzdi2.S revision 2d1fdb26e458c4ddc04155c1d421bced3ba90cd0
1/* ===-- clzdi2.c - Implement __clzdi2 -------------------------------------=== 2 * 3 * The LLVM Compiler Infrastructure 4 * 5 * This file is dual licensed under the MIT and the University of Illinois Open 6 * Source Licenses. See LICENSE.TXT for details. 7 * 8 * ===----------------------------------------------------------------------=== 9 * 10 * This file implements count leading zeros for 64bit arguments. 11 * 12 * ===----------------------------------------------------------------------=== 13 */ 14#include "../assembly.h" 15 16 .syntax unified 17 18 .text 19 .p2align 2 20DEFINE_COMPILERRT_FUNCTION(__clzdi2) 21#ifdef __ARM_FEATURE_CLZ 22#ifdef __ARMEB__ 23 cmp r0, 0 24 itee ne 25 clzne r0, r0 26 clzeq r0, r1 27 addeq r0, r0, 32 28#else 29 cmp r1, 0 30 itee ne 31 clzne r0, r1 32 clzeq r0, r0 33 addeq r0, r0, 32 34#endif 35 JMP(lr) 36#else 37 /* Assumption: n != 0 */ 38 39 /* 40 * r0: n 41 * r1: upper half of n, overwritten after check 42 * r1: count of leading zeros in n + 1 43 * r2: scratch register for shifted r0 44 */ 45#ifdef __ARMEB__ 46 cmp r0, 0 47 moveq r0, r1 48#else 49 cmp r1, 0 50 movne r0, r1 51#endif 52 movne r1, 1 53 moveq r1, 33 54 55 /* 56 * Basic block: 57 * if ((r0 >> SHIFT) == 0) 58 * r1 += SHIFT; 59 * else 60 * r0 >>= SHIFT; 61 * for descending powers of two as SHIFT. 62 */ 63#define BLOCK(shift) \ 64 lsrs r2, r0, shift; \ 65 movne r0, r2; \ 66 addeq r1, shift \ 67 68 BLOCK(16) 69 BLOCK(8) 70 BLOCK(4) 71 BLOCK(2) 72 73 /* 74 * The basic block invariants at this point are (r0 >> 2) == 0 and 75 * r0 != 0. This means 1 <= r0 <= 3 and 0 <= (r0 >> 1) <= 1. 76 * 77 * r0 | (r0 >> 1) == 0 | (r0 >> 1) == 1 | -(r0 >> 1) | 1 - (r0 >> 1) 78 * ---+----------------+----------------+------------+-------------- 79 * 1 | 1 | 0 | 0 | 1 80 * 2 | 0 | 1 | -1 | 0 81 * 3 | 0 | 1 | -1 | 0 82 * 83 * The r1's initial value of 1 compensates for the 1 here. 84 */ 85 sub r0, r1, r0, lsr #1 86 87 JMP(lr) 88#endif // __ARM_FEATURE_CLZ 89END_COMPILERRT_FUNCTION(__clzdi2) 90