projects / mirror_edk2.git / blame
| Commit | Line | Data |
|---|---|---|
| a37f6605 AB |
1 | //\r |
| 2 | // Copyright (c) 2013 - 2016, Linaro Limited\r | |
| 3 | // All rights reserved.\r | |
| 4 | //\r | |
| 5 | // Redistribution and use in source and binary forms, with or without\r | |
| 6 | // modification, are permitted provided that the following conditions are met:\r | |
| 7 | // * Redistributions of source code must retain the above copyright\r | |
| 8 | // notice, this list of conditions and the following disclaimer.\r | |
| 9 | // * Redistributions in binary form must reproduce the above copyright\r | |
| 10 | // notice, this list of conditions and the following disclaimer in the\r | |
| 11 | // documentation and/or other materials provided with the distribution.\r | |
| 12 | // * Neither the name of the Linaro nor the\r | |
| 13 | // names of its contributors may be used to endorse or promote products\r | |
| 14 | // derived from this software without specific prior written permission.\r | |
| 15 | //\r | |
| 16 | // THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS\r | |
| 17 | // "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT\r | |
| 18 | // LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR\r | |
| 19 | // A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT\r | |
| 20 | // HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,\r | |
| 21 | // SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT\r | |
| 22 | // LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,\r | |
| 23 | // DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY\r | |
| 24 | // THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT\r | |
| 25 | // (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE\r | |
| 26 | // OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.\r | |
| 27 | //\r | |
| 28 | \r | |
| 29 | // Parameters and result.\r | |
| 30 | #define src1 r0\r | |
| 31 | #define src2 r1\r | |
| 32 | #define limit r2\r | |
| 33 | #define result r0\r | |
| 34 | \r | |
| 35 | // Internal variables.\r | |
| 36 | #define data1 r3\r | |
| 37 | #define data2 r4\r | |
| 38 | #define limit_wd r5\r | |
| 39 | #define diff r6\r | |
| 40 | #define tmp1 r7\r | |
| 41 | #define tmp2 r12\r | |
| 42 | #define pos r8\r | |
| 43 | #define mask r14\r | |
| 44 | \r | |
| 45 | .text\r | |
| 46 | .thumb\r | |
| 47 | .syntax unified\r | |
| 48 | .align 5\r | |
| decaac5d | 49 | .type ASM_PFX(InternalMemCompareMem), %function\r |
| a37f6605 AB |
50 | ASM_GLOBAL ASM_PFX(InternalMemCompareMem)\r |
| 51 | ASM_PFX(InternalMemCompareMem):\r | |
| 52 | push {r4-r8, lr}\r | |
| 53 | eor tmp1, src1, src2\r | |
| 54 | tst tmp1, #3\r | |
| 55 | bne .Lmisaligned4\r | |
| 56 | ands tmp1, src1, #3\r | |
| 57 | bne .Lmutual_align\r | |
| 58 | add limit_wd, limit, #3\r | |
| 59 | nop.w\r | |
| 60 | lsr limit_wd, limit_wd, #2\r | |
| 61 | \r | |
| 62 | // Start of performance-critical section -- one 32B cache line.\r | |
| 63 | .Lloop_aligned:\r | |
| 64 | ldr data1, [src1], #4\r | |
| 65 | ldr data2, [src2], #4\r | |
| 66 | .Lstart_realigned:\r | |
| 67 | subs limit_wd, limit_wd, #1\r | |
| 68 | eor diff, data1, data2 // Non-zero if differences found.\r | |
| 69 | cbnz diff, 0f\r | |
| 70 | bne .Lloop_aligned\r | |
| 71 | // End of performance-critical section -- one 32B cache line.\r | |
| 72 | \r | |
| 73 | // Not reached the limit, must have found a diff.\r | |
| 74 | 0: cbnz limit_wd, .Lnot_limit\r | |
| 75 | \r | |
| 76 | // Limit % 4 == 0 => all bytes significant.\r | |
| 77 | ands limit, limit, #3\r | |
| 78 | beq .Lnot_limit\r | |
| 79 | \r | |
| 80 | lsl limit, limit, #3 // Bits -> bytes.\r | |
| 81 | mov mask, #~0\r | |
| 82 | lsl mask, mask, limit\r | |
| 83 | bic data1, data1, mask\r | |
| 84 | bic data2, data2, mask\r | |
| 85 | \r | |
| 86 | orr diff, diff, mask\r | |
| 87 | \r | |
| 88 | .Lnot_limit:\r | |
| 89 | rev diff, diff\r | |
| 90 | rev data1, data1\r | |
| 91 | rev data2, data2\r | |
| 92 | \r | |
| 93 | // The MS-non-zero bit of DIFF marks either the first bit\r | |
| 94 | // that is different, or the end of the significant data.\r | |
| 95 | // Shifting left now will bring the critical information into the\r | |
| 96 | // top bits.\r | |
| 97 | clz pos, diff\r | |
| 98 | lsl data1, data1, pos\r | |
| 99 | lsl data2, data2, pos\r | |
| 100 | \r | |
| 101 | // But we need to zero-extend (char is unsigned) the value and then\r | |
| 102 | // perform a signed 32-bit subtraction.\r | |
| 103 | lsr data1, data1, #28\r | |
| 104 | sub result, data1, data2, lsr #28\r | |
| 105 | pop {r4-r8, pc}\r | |
| 106 | \r | |
| 107 | .Lmutual_align:\r | |
| 108 | // Sources are mutually aligned, but are not currently at an\r | |
| 109 | // alignment boundary. Round down the addresses and then mask off\r | |
| 110 | // the bytes that precede the start point.\r | |
| 111 | bic src1, src1, #3\r | |
| 112 | bic src2, src2, #3\r | |
| 113 | add limit, limit, tmp1 // Adjust the limit for the extra.\r | |
| 60fe5e8a | 114 | lsl tmp1, tmp1, #3 // Bytes beyond alignment -> bits.\r |
| a37f6605 | 115 | ldr data1, [src1], #4\r |
| 60fe5e8a | 116 | rsb tmp1, tmp1, #32 // Bits to alignment -32.\r |
| a37f6605 AB |
117 | ldr data2, [src2], #4\r |
| 118 | mov tmp2, #~0\r | |
| 119 | \r | |
| 120 | // Little-endian. Early bytes are at LSB.\r | |
| 121 | lsr tmp2, tmp2, tmp1 // Shift (tmp1 & 31).\r | |
| 122 | add limit_wd, limit, #3\r | |
| 123 | orr data1, data1, tmp2\r | |
| 124 | orr data2, data2, tmp2\r | |
| 125 | lsr limit_wd, limit_wd, #2\r | |
| 126 | b .Lstart_realigned\r | |
| 127 | \r | |
| 128 | .Lmisaligned4:\r | |
| 129 | sub limit, limit, #1\r | |
| 130 | 1:\r | |
| 131 | // Perhaps we can do better than this.\r | |
| 132 | ldrb data1, [src1], #1\r | |
| 133 | ldrb data2, [src2], #1\r | |
| 134 | subs limit, limit, #1\r | |
| 135 | it cs\r | |
| eab26788 | 136 | cmpcs.n data1, data2\r |
| a37f6605 AB |
137 | beq 1b\r |
| 138 | sub result, data1, data2\r | |
| 139 | pop {r4-r8, pc}\r |