| 1 | /* # 1 "libgcc1.S" */ |
| 2 | @ libgcc1 routines for ARM cpu. |
| 3 | @ Division routines, written by Richard Earnshaw, (rearnsha@armltd.co.uk) |
| 4 | /* # 145 "libgcc1.S" */ |
| 5 | dividend .req r0 |
| 6 | divisor .req r1 |
| 7 | overdone .req r2 |
| 8 | curbit .req r3 |
| 9 | /* ip .req r12 */ |
| 10 | /* sp .req r13 */ |
| 11 | /* lr .req r14 */ |
| 12 | /* pc .req r15 */ |
| 13 | .text |
| 14 | .globl __umodsi3 |
| 15 | .type __umodsi3 ,function |
| 16 | .align 0 |
| 17 | __umodsi3 : |
| 18 | cmp divisor, #0 |
| 19 | beq Ldiv0 |
| 20 | mov curbit, #1 |
| 21 | cmp dividend, divisor |
| 22 | movcc pc, lr |
| 23 | Loop1: |
| 24 | @ Unless the divisor is very big, shift it up in multiples of |
| 25 | @ four bits, since this is the amount of unwinding in the main |
| 26 | @ division loop. Continue shifting until the divisor is |
| 27 | @ larger than the dividend. |
| 28 | cmp divisor, #0x10000000 |
| 29 | cmpcc divisor, dividend |
| 30 | movcc divisor, divisor, lsl #4 |
| 31 | movcc curbit, curbit, lsl #4 |
| 32 | bcc Loop1 |
| 33 | Lbignum: |
| 34 | @ For very big divisors, we must shift it a bit at a time, or |
| 35 | @ we will be in danger of overflowing. |
| 36 | cmp divisor, #0x80000000 |
| 37 | cmpcc divisor, dividend |
| 38 | movcc divisor, divisor, lsl #1 |
| 39 | movcc curbit, curbit, lsl #1 |
| 40 | bcc Lbignum |
| 41 | Loop3: |
| 42 | @ Test for possible subtractions. On the final pass, this may |
| 43 | @ subtract too much from the dividend, so keep track of which |
| 44 | @ subtractions are done, we can fix them up afterwards... |
| 45 | mov overdone, #0 |
| 46 | cmp dividend, divisor |
| 47 | subcs dividend, dividend, divisor |
| 48 | cmp dividend, divisor, lsr #1 |
| 49 | subcs dividend, dividend, divisor, lsr #1 |
| 50 | orrcs overdone, overdone, curbit, ror #1 |
| 51 | cmp dividend, divisor, lsr #2 |
| 52 | subcs dividend, dividend, divisor, lsr #2 |
| 53 | orrcs overdone, overdone, curbit, ror #2 |
| 54 | cmp dividend, divisor, lsr #3 |
| 55 | subcs dividend, dividend, divisor, lsr #3 |
| 56 | orrcs overdone, overdone, curbit, ror #3 |
| 57 | mov ip, curbit |
| 58 | cmp dividend, #0 @ Early termination? |
| 59 | movnes curbit, curbit, lsr #4 @ No, any more bits to do? |
| 60 | movne divisor, divisor, lsr #4 |
| 61 | bne Loop3 |
| 62 | @ Any subtractions that we should not have done will be recorded in |
| 63 | @ the top three bits of "overdone". Exactly which were not needed |
| 64 | @ are governed by the position of the bit, stored in ip. |
| 65 | @ If we terminated early, because dividend became zero, |
| 66 | @ then none of the below will match, since the bit in ip will not be |
| 67 | @ in the bottom nibble. |
| 68 | ands overdone, overdone, #0xe0000000 |
| 69 | moveq pc, lr @ No fixups needed |
| 70 | tst overdone, ip, ror #3 |
| 71 | addne dividend, dividend, divisor, lsr #3 |
| 72 | tst overdone, ip, ror #2 |
| 73 | addne dividend, dividend, divisor, lsr #2 |
| 74 | tst overdone, ip, ror #1 |
| 75 | addne dividend, dividend, divisor, lsr #1 |
| 76 | mov pc, lr |
| 77 | Ldiv0: |
| 78 | str lr, [sp, #-4]! |
| 79 | bl __div0 (PLT) |
| 80 | mov r0, #0 @ about as wrong as it could be |
| 81 | ldmia sp!, {pc} |
| 82 | .size __umodsi3 , . - __umodsi3 |
| 83 | /* # 320 "libgcc1.S" */ |
| 84 | /* # 421 "libgcc1.S" */ |
| 85 | /* # 433 "libgcc1.S" */ |
| 86 | /* # 456 "libgcc1.S" */ |
| 87 | /* # 500 "libgcc1.S" */ |
| 88 | /* # 580 "libgcc1.S" */ |
| 89 | |
