diff options
| -rw-r--r-- | arch/arm/lib/memset.S | 85 |
1 files changed, 44 insertions, 41 deletions
diff --git a/arch/arm/lib/memset.S b/arch/arm/lib/memset.S index 650d5923ab83..d912e7397ecc 100644 --- a/arch/arm/lib/memset.S +++ b/arch/arm/lib/memset.S | |||
| @@ -19,9 +19,9 @@ | |||
| 19 | 1: subs r2, r2, #4 @ 1 do we have enough | 19 | 1: subs r2, r2, #4 @ 1 do we have enough |
| 20 | blt 5f @ 1 bytes to align with? | 20 | blt 5f @ 1 bytes to align with? |
| 21 | cmp r3, #2 @ 1 | 21 | cmp r3, #2 @ 1 |
| 22 | strltb r1, [r0], #1 @ 1 | 22 | strltb r1, [ip], #1 @ 1 |
| 23 | strleb r1, [r0], #1 @ 1 | 23 | strleb r1, [ip], #1 @ 1 |
| 24 | strb r1, [r0], #1 @ 1 | 24 | strb r1, [ip], #1 @ 1 |
| 25 | add r2, r2, r3 @ 1 (r2 = r2 - (4 - r3)) | 25 | add r2, r2, r3 @ 1 (r2 = r2 - (4 - r3)) |
| 26 | /* | 26 | /* |
| 27 | * The pointer is now aligned and the length is adjusted. Try doing the | 27 | * The pointer is now aligned and the length is adjusted. Try doing the |
| @@ -29,10 +29,14 @@ | |||
| 29 | */ | 29 | */ |
| 30 | 30 | ||
| 31 | ENTRY(memset) | 31 | ENTRY(memset) |
| 32 | ands r3, r0, #3 @ 1 unaligned? | 32 | /* |
| 33 | * Preserve the contents of r0 for the return value. | ||
| 34 | */ | ||
| 35 | mov ip, r0 | ||
| 36 | ands r3, ip, #3 @ 1 unaligned? | ||
| 33 | bne 1b @ 1 | 37 | bne 1b @ 1 |
| 34 | /* | 38 | /* |
| 35 | * we know that the pointer in r0 is aligned to a word boundary. | 39 | * we know that the pointer in ip is aligned to a word boundary. |
| 36 | */ | 40 | */ |
| 37 | orr r1, r1, r1, lsl #8 | 41 | orr r1, r1, r1, lsl #8 |
| 38 | orr r1, r1, r1, lsl #16 | 42 | orr r1, r1, r1, lsl #16 |
| @@ -43,29 +47,28 @@ ENTRY(memset) | |||
| 43 | #if ! CALGN(1)+0 | 47 | #if ! CALGN(1)+0 |
| 44 | 48 | ||
| 45 | /* | 49 | /* |
| 46 | * We need an extra register for this loop - save the return address and | 50 | * We need 2 extra registers for this loop - use r8 and the LR |
| 47 | * use the LR | ||
| 48 | */ | 51 | */ |
| 49 | str lr, [sp, #-4]! | 52 | stmfd sp!, {r8, lr} |
| 50 | mov ip, r1 | 53 | mov r8, r1 |
| 51 | mov lr, r1 | 54 | mov lr, r1 |
| 52 | 55 | ||
| 53 | 2: subs r2, r2, #64 | 56 | 2: subs r2, r2, #64 |
| 54 | stmgeia r0!, {r1, r3, ip, lr} @ 64 bytes at a time. | 57 | stmgeia ip!, {r1, r3, r8, lr} @ 64 bytes at a time. |
| 55 | stmgeia r0!, {r1, r3, ip, lr} | 58 | stmgeia ip!, {r1, r3, r8, lr} |
| 56 | stmgeia r0!, {r1, r3, ip, lr} | 59 | stmgeia ip!, {r1, r3, r8, lr} |
| 57 | stmgeia r0!, {r1, r3, ip, lr} | 60 | stmgeia ip!, {r1, r3, r8, lr} |
| 58 | bgt 2b | 61 | bgt 2b |
| 59 | ldmeqfd sp!, {pc} @ Now <64 bytes to go. | 62 | ldmeqfd sp!, {r8, pc} @ Now <64 bytes to go. |
| 60 | /* | 63 | /* |
| 61 | * No need to correct the count; we're only testing bits from now on | 64 | * No need to correct the count; we're only testing bits from now on |
| 62 | */ | 65 | */ |
| 63 | tst r2, #32 | 66 | tst r2, #32 |
| 64 | stmneia r0!, {r1, r3, ip, lr} | 67 | stmneia ip!, {r1, r3, r8, lr} |
| 65 | stmneia r0!, {r1, r3, ip, lr} | 68 | stmneia ip!, {r1, r3, r8, lr} |
| 66 | tst r2, #16 | 69 | tst r2, #16 |
| 67 | stmneia r0!, {r1, r3, ip, lr} | 70 | stmneia ip!, {r1, r3, r8, lr} |
| 68 | ldr lr, [sp], #4 | 71 | ldmfd sp!, {r8, lr} |
| 69 | 72 | ||
| 70 | #else | 73 | #else |
| 71 | 74 | ||
| @@ -74,54 +77,54 @@ ENTRY(memset) | |||
| 74 | * whole cache lines at once. | 77 | * whole cache lines at once. |
| 75 | */ | 78 | */ |
| 76 | 79 | ||
| 77 | stmfd sp!, {r4-r7, lr} | 80 | stmfd sp!, {r4-r8, lr} |
| 78 | mov r4, r1 | 81 | mov r4, r1 |
| 79 | mov r5, r1 | 82 | mov r5, r1 |
| 80 | mov r6, r1 | 83 | mov r6, r1 |
| 81 | mov r7, r1 | 84 | mov r7, r1 |
| 82 | mov ip, r1 | 85 | mov r8, r1 |
| 83 | mov lr, r1 | 86 | mov lr, r1 |
| 84 | 87 | ||
| 85 | cmp r2, #96 | 88 | cmp r2, #96 |
| 86 | tstgt r0, #31 | 89 | tstgt ip, #31 |
| 87 | ble 3f | 90 | ble 3f |
| 88 | 91 | ||
| 89 | and ip, r0, #31 | 92 | and r8, ip, #31 |
| 90 | rsb ip, ip, #32 | 93 | rsb r8, r8, #32 |
| 91 | sub r2, r2, ip | 94 | sub r2, r2, r8 |
| 92 | movs ip, ip, lsl #(32 - 4) | 95 | movs r8, r8, lsl #(32 - 4) |
| 93 | stmcsia r0!, {r4, r5, r6, r7} | 96 | stmcsia ip!, {r4, r5, r6, r7} |
| 94 | stmmiia r0!, {r4, r5} | 97 | stmmiia ip!, {r4, r5} |
| 95 | tst ip, #(1 << 30) | 98 | tst r8, #(1 << 30) |
| 96 | mov ip, r1 | 99 | mov r8, r1 |
| 97 | strne r1, [r0], #4 | 100 | strne r1, [ip], #4 |
| 98 | 101 | ||
| 99 | 3: subs r2, r2, #64 | 102 | 3: subs r2, r2, #64 |
| 100 | stmgeia r0!, {r1, r3-r7, ip, lr} | 103 | stmgeia ip!, {r1, r3-r8, lr} |
| 101 | stmgeia r0!, {r1, r3-r7, ip, lr} | 104 | stmgeia ip!, {r1, r3-r8, lr} |
| 102 | bgt 3b | 105 | bgt 3b |
| 103 | ldmeqfd sp!, {r4-r7, pc} | 106 | ldmeqfd sp!, {r4-r8, pc} |
| 104 | 107 | ||
| 105 | tst r2, #32 | 108 | tst r2, #32 |
| 106 | stmneia r0!, {r1, r3-r7, ip, lr} | 109 | stmneia ip!, {r1, r3-r8, lr} |
| 107 | tst r2, #16 | 110 | tst r2, #16 |
| 108 | stmneia r0!, {r4-r7} | 111 | stmneia ip!, {r4-r7} |
| 109 | ldmfd sp!, {r4-r7, lr} | 112 | ldmfd sp!, {r4-r8, lr} |
| 110 | 113 | ||
| 111 | #endif | 114 | #endif |
| 112 | 115 | ||
| 113 | 4: tst r2, #8 | 116 | 4: tst r2, #8 |
| 114 | stmneia r0!, {r1, r3} | 117 | stmneia ip!, {r1, r3} |
| 115 | tst r2, #4 | 118 | tst r2, #4 |
| 116 | strne r1, [r0], #4 | 119 | strne r1, [ip], #4 |
| 117 | /* | 120 | /* |
| 118 | * When we get here, we've got less than 4 bytes to zero. We | 121 | * When we get here, we've got less than 4 bytes to zero. We |
| 119 | * may have an unaligned pointer as well. | 122 | * may have an unaligned pointer as well. |
| 120 | */ | 123 | */ |
| 121 | 5: tst r2, #2 | 124 | 5: tst r2, #2 |
| 122 | strneb r1, [r0], #1 | 125 | strneb r1, [ip], #1 |
| 123 | strneb r1, [r0], #1 | 126 | strneb r1, [ip], #1 |
| 124 | tst r2, #1 | 127 | tst r2, #1 |
| 125 | strneb r1, [r0], #1 | 128 | strneb r1, [ip], #1 |
| 126 | mov pc, lr | 129 | mov pc, lr |
| 127 | ENDPROC(memset) | 130 | ENDPROC(memset) |