Wolfgang Denk | 4646d2a | 2006-05-30 15:56:48 +0200 | [diff] [blame] | 1 | |
| 2 | .macro ARM_DIV_BODY dividend, divisor, result, curbit |
| 3 | |
| 4 | #if __LINUX_ARM_ARCH__ >= 5 |
| 5 | |
| 6 | clz \curbit, \divisor |
| 7 | clz \result, \dividend |
| 8 | sub \result, \curbit, \result |
| 9 | mov \curbit, #1 |
| 10 | mov \divisor, \divisor, lsl \result |
| 11 | mov \curbit, \curbit, lsl \result |
| 12 | mov \result, #0 |
| 13 | |
| 14 | #else |
| 15 | |
| 16 | @ Initially shift the divisor left 3 bits if possible, |
| 17 | @ set curbit accordingly. This allows for curbit to be located |
| 18 | @ at the left end of each 4 bit nibbles in the division loop |
| 19 | @ to save one loop in most cases. |
| 20 | tst \divisor, #0xe0000000 |
| 21 | moveq \divisor, \divisor, lsl #3 |
| 22 | moveq \curbit, #8 |
| 23 | movne \curbit, #1 |
| 24 | |
| 25 | @ Unless the divisor is very big, shift it up in multiples of |
| 26 | @ four bits, since this is the amount of unwinding in the main |
| 27 | @ division loop. Continue shifting until the divisor is |
| 28 | @ larger than the dividend. |
| 29 | 1: cmp \divisor, #0x10000000 |
| 30 | cmplo \divisor, \dividend |
| 31 | movlo \divisor, \divisor, lsl #4 |
| 32 | movlo \curbit, \curbit, lsl #4 |
| 33 | blo 1b |
| 34 | |
| 35 | @ For very big divisors, we must shift it a bit at a time, or |
| 36 | @ we will be in danger of overflowing. |
| 37 | 1: cmp \divisor, #0x80000000 |
| 38 | cmplo \divisor, \dividend |
| 39 | movlo \divisor, \divisor, lsl #1 |
| 40 | movlo \curbit, \curbit, lsl #1 |
| 41 | blo 1b |
| 42 | |
| 43 | mov \result, #0 |
| 44 | |
| 45 | #endif |
| 46 | |
| 47 | @ Division loop |
| 48 | 1: cmp \dividend, \divisor |
| 49 | subhs \dividend, \dividend, \divisor |
| 50 | orrhs \result, \result, \curbit |
| 51 | cmp \dividend, \divisor, lsr #1 |
| 52 | subhs \dividend, \dividend, \divisor, lsr #1 |
| 53 | orrhs \result, \result, \curbit, lsr #1 |
| 54 | cmp \dividend, \divisor, lsr #2 |
| 55 | subhs \dividend, \dividend, \divisor, lsr #2 |
| 56 | orrhs \result, \result, \curbit, lsr #2 |
| 57 | cmp \dividend, \divisor, lsr #3 |
| 58 | subhs \dividend, \dividend, \divisor, lsr #3 |
| 59 | orrhs \result, \result, \curbit, lsr #3 |
| 60 | cmp \dividend, #0 @ Early termination? |
| 61 | movnes \curbit, \curbit, lsr #4 @ No, any more bits to do? |
| 62 | movne \divisor, \divisor, lsr #4 |
| 63 | bne 1b |
| 64 | |
| 65 | .endm |
| 66 | |
| 67 | .macro ARM_DIV2_ORDER divisor, order |
| 68 | |
| 69 | #if __LINUX_ARM_ARCH__ >= 5 |
| 70 | |
| 71 | clz \order, \divisor |
| 72 | rsb \order, \order, #31 |
| 73 | |
| 74 | #else |
| 75 | |
| 76 | cmp \divisor, #(1 << 16) |
| 77 | movhs \divisor, \divisor, lsr #16 |
| 78 | movhs \order, #16 |
| 79 | movlo \order, #0 |
| 80 | |
| 81 | cmp \divisor, #(1 << 8) |
| 82 | movhs \divisor, \divisor, lsr #8 |
| 83 | addhs \order, \order, #8 |
| 84 | |
| 85 | cmp \divisor, #(1 << 4) |
| 86 | movhs \divisor, \divisor, lsr #4 |
| 87 | addhs \order, \order, #4 |
| 88 | |
| 89 | cmp \divisor, #(1 << 2) |
| 90 | addhi \order, \order, #3 |
| 91 | addls \order, \order, \divisor, lsr #1 |
| 92 | |
| 93 | #endif |
| 94 | |
| 95 | .endm |
| 96 | |
| 97 | .align 5 |
| 98 | .globl __divsi3 |
| 99 | __divsi3: |
| 100 | cmp r1, #0 |
| 101 | eor ip, r0, r1 @ save the sign of the result. |
| 102 | beq Ldiv0 |
| 103 | rsbmi r1, r1, #0 @ loops below use unsigned. |
| 104 | subs r2, r1, #1 @ division by 1 or -1 ? |
| 105 | beq 10f |
| 106 | movs r3, r0 |
| 107 | rsbmi r3, r0, #0 @ positive dividend value |
| 108 | cmp r3, r1 |
| 109 | bls 11f |
| 110 | tst r1, r2 @ divisor is power of 2 ? |
| 111 | beq 12f |
| 112 | |
| 113 | ARM_DIV_BODY r3, r1, r0, r2 |
| 114 | |
| 115 | cmp ip, #0 |
| 116 | rsbmi r0, r0, #0 |
| 117 | mov pc, lr |
| 118 | |
| 119 | 10: teq ip, r0 @ same sign ? |
| 120 | rsbmi r0, r0, #0 |
| 121 | mov pc, lr |
| 122 | |
| 123 | 11: movlo r0, #0 |
| 124 | moveq r0, ip, asr #31 |
| 125 | orreq r0, r0, #1 |
| 126 | mov pc, lr |
| 127 | |
| 128 | 12: ARM_DIV2_ORDER r1, r2 |
| 129 | |
| 130 | cmp ip, #0 |
| 131 | mov r0, r3, lsr r2 |
| 132 | rsbmi r0, r0, #0 |
| 133 | mov pc, lr |
| 134 | |
| 135 | Ldiv0: |
| 136 | |
| 137 | str lr, [sp, #-4]! |
| 138 | bl __div0 |
| 139 | mov r0, #0 @ About as wrong as it could be. |
| 140 | ldr pc, [sp], #4 |