Sam Walsh / F7_Ethernet

Forked from STM32F7 internet for nucleo F746ZG

Dependents:   Nucleo_F746ZG_Ethernet_MQTT_Ultrasound

Fork of F7_Ethernet by Dieter Graef

lwip-sys/arch/checksum.c@0:d26c1b55cfca, 2016-06-19 (annotated)

Committer:
DieterGraef
Date:
Sun Jun 19 16:23:40 2016 +0000
Revision:
0:d26c1b55cfca
Ethernet Library for Nucleo stm32f746ZG and Disco stm32f746NG  works under arm and gcc environment

Who changed what in which revision?

UserRevisionLine numberNew contents of line
DieterGraef 0:d26c1b55cfca 1 /* Copyright (C) 2013 - Adam Green (https://github.com/adamgreen)
DieterGraef 0:d26c1b55cfca 2
DieterGraef 0:d26c1b55cfca 3 Licensed under the Apache License, Version 2.0 (the "License");
DieterGraef 0:d26c1b55cfca 4 you may not use this file except in compliance with the License.
DieterGraef 0:d26c1b55cfca 5 You may obtain a copy of the License at
DieterGraef 0:d26c1b55cfca 6
DieterGraef 0:d26c1b55cfca 7 http://www.apache.org/licenses/LICENSE-2.0
DieterGraef 0:d26c1b55cfca 8
DieterGraef 0:d26c1b55cfca 9 Unless required by applicable law or agreed to in writing, software
DieterGraef 0:d26c1b55cfca 10 distributed under the License is distributed on an "AS IS" BASIS,
DieterGraef 0:d26c1b55cfca 11 WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
DieterGraef 0:d26c1b55cfca 12 See the License for the specific language governing permissions and
DieterGraef 0:d26c1b55cfca 13 limitations under the License.
DieterGraef 0:d26c1b55cfca 14 */
DieterGraef 0:d26c1b55cfca 15 #if defined(TOOLCHAIN_GCC) && defined(__thumb2__)
DieterGraef 0:d26c1b55cfca 16
DieterGraef 0:d26c1b55cfca 17
DieterGraef 0:d26c1b55cfca 18 /* This is a hand written Thumb-2 assembly language version of the
DieterGraef 0:d26c1b55cfca 19 algorithm 3 version of lwip_standard_chksum in lwIP's inet_chksum.c. It
DieterGraef 0:d26c1b55cfca 20 performs the checksumming 32-bits at a time and even unrolls the loop to
DieterGraef 0:d26c1b55cfca 21 perform two of these 32-bit adds per loop iteration.
DieterGraef 0:d26c1b55cfca 22
DieterGraef 0:d26c1b55cfca 23 Returns:
DieterGraef 0:d26c1b55cfca 24 16-bit 1's complement summation (not inversed).
DieterGraef 0:d26c1b55cfca 25
DieterGraef 0:d26c1b55cfca 26 NOTE: This function does return a uint16_t from the assembly language code
DieterGraef 0:d26c1b55cfca 27 but is marked as void so that GCC doesn't issue warning because it
DieterGraef 0:d26c1b55cfca 28 doesn't know about this low level return.
DieterGraef 0:d26c1b55cfca 29 */
DieterGraef 0:d26c1b55cfca 30 __attribute__((naked)) void /*uint16_t*/ thumb2_checksum(const void* pData, int length)
DieterGraef 0:d26c1b55cfca 31 {
DieterGraef 0:d26c1b55cfca 32 __asm (
DieterGraef 0:d26c1b55cfca 33 ".syntax unified\n"
DieterGraef 0:d26c1b55cfca 34 ".thumb\n"
DieterGraef 0:d26c1b55cfca 35
DieterGraef 0:d26c1b55cfca 36 // Push non-volatile registers we use on stack. Push link register too to
DieterGraef 0:d26c1b55cfca 37 // keep stack 8-byte aligned and allow single pop to restore and return.
DieterGraef 0:d26c1b55cfca 38 " push {r4, lr}\n"
DieterGraef 0:d26c1b55cfca 39 // Initialize sum, r2, to 0.
DieterGraef 0:d26c1b55cfca 40 " movs r2, #0\n"
DieterGraef 0:d26c1b55cfca 41 // Remember whether pData was at odd address in r3. This is used later to
DieterGraef 0:d26c1b55cfca 42 // know if it needs to swap the result since the summation will be done at
DieterGraef 0:d26c1b55cfca 43 // an offset of 1, rather than 0.
DieterGraef 0:d26c1b55cfca 44 " ands r3, r0, #1\n"
DieterGraef 0:d26c1b55cfca 45 // Need to 2-byte align? If not skip ahead.
DieterGraef 0:d26c1b55cfca 46 " beq 1$\n"
DieterGraef 0:d26c1b55cfca 47 // We can return if there are no bytes to sum.
DieterGraef 0:d26c1b55cfca 48 " cbz r1, 9$\n"
DieterGraef 0:d26c1b55cfca 49
DieterGraef 0:d26c1b55cfca 50 // 2-byte align.
DieterGraef 0:d26c1b55cfca 51 // Place the first data byte in odd summation location since it needs to be
DieterGraef 0:d26c1b55cfca 52 // swapped later. It's ok to overwrite r2 here as it only had a value of 0
DieterGraef 0:d26c1b55cfca 53 // up until now. Advance r0 pointer and decrement r1 length as we go.
DieterGraef 0:d26c1b55cfca 54 " ldrb r2, [r0], #1\n"
DieterGraef 0:d26c1b55cfca 55 " lsls r2, r2, #8\n"
DieterGraef 0:d26c1b55cfca 56 " subs r1, r1, #1\n"
DieterGraef 0:d26c1b55cfca 57
DieterGraef 0:d26c1b55cfca 58 // Need to 4-byte align? If not skip ahead.
DieterGraef 0:d26c1b55cfca 59 "1$:\n"
DieterGraef 0:d26c1b55cfca 60 " ands r4, r0, #3\n"
DieterGraef 0:d26c1b55cfca 61 " beq 2$\n"
DieterGraef 0:d26c1b55cfca 62 // Have more than 1 byte left to align? If not skip ahead to take care of
DieterGraef 0:d26c1b55cfca 63 // trailing byte.
DieterGraef 0:d26c1b55cfca 64 " cmp r1, #2\n"
DieterGraef 0:d26c1b55cfca 65 " blt 7$\n"
DieterGraef 0:d26c1b55cfca 66
DieterGraef 0:d26c1b55cfca 67 // 4-byte align.
DieterGraef 0:d26c1b55cfca 68 " ldrh r4, [r0], #2\n"
DieterGraef 0:d26c1b55cfca 69 " adds r2, r2, r4\n"
DieterGraef 0:d26c1b55cfca 70 " subs r1, r1, #2\n"
DieterGraef 0:d26c1b55cfca 71
DieterGraef 0:d26c1b55cfca 72 // Main summing loop which sums up data 2 words at a time.
DieterGraef 0:d26c1b55cfca 73 // Make sure that we have more than 7 bytes left to sum.
DieterGraef 0:d26c1b55cfca 74 "2$:\n"
DieterGraef 0:d26c1b55cfca 75 " cmp r1, #8\n"
DieterGraef 0:d26c1b55cfca 76 " blt 3$\n"
DieterGraef 0:d26c1b55cfca 77 // Sum next two words. Applying previous upper 16-bit carry to
DieterGraef 0:d26c1b55cfca 78 // lower 16-bits.
DieterGraef 0:d26c1b55cfca 79 " ldr r4, [r0], #4\n"
DieterGraef 0:d26c1b55cfca 80 " adds r2, r4\n"
DieterGraef 0:d26c1b55cfca 81 " adc r2, r2, #0\n"
DieterGraef 0:d26c1b55cfca 82 " ldr r4, [r0], #4\n"
DieterGraef 0:d26c1b55cfca 83 " adds r2, r4\n"
DieterGraef 0:d26c1b55cfca 84 " adc r2, r2, #0\n"
DieterGraef 0:d26c1b55cfca 85 " subs r1, r1, #8\n"
DieterGraef 0:d26c1b55cfca 86 " b 2$\n"
DieterGraef 0:d26c1b55cfca 87
DieterGraef 0:d26c1b55cfca 88 // Sum up any remaining half-words.
DieterGraef 0:d26c1b55cfca 89 "3$:\n"
DieterGraef 0:d26c1b55cfca 90 // Make sure that we have more than 1 byte left to sum.
DieterGraef 0:d26c1b55cfca 91 " cmp r1, #2\n"
DieterGraef 0:d26c1b55cfca 92 " blt 7$\n"
DieterGraef 0:d26c1b55cfca 93 // Sum up next half word, continue to apply carry.
DieterGraef 0:d26c1b55cfca 94 " ldrh r4, [r0], #2\n"
DieterGraef 0:d26c1b55cfca 95 " adds r2, r4\n"
DieterGraef 0:d26c1b55cfca 96 " adc r2, r2, #0\n"
DieterGraef 0:d26c1b55cfca 97 " subs r1, r1, #2\n"
DieterGraef 0:d26c1b55cfca 98 " b 3$\n"
DieterGraef 0:d26c1b55cfca 99
DieterGraef 0:d26c1b55cfca 100 // Handle trailing byte, if it exists
DieterGraef 0:d26c1b55cfca 101 "7$:\n"
DieterGraef 0:d26c1b55cfca 102 " cbz r1, 8$\n"
DieterGraef 0:d26c1b55cfca 103 " ldrb r4, [r0]\n"
DieterGraef 0:d26c1b55cfca 104 " adds r2, r4\n"
DieterGraef 0:d26c1b55cfca 105 " adc r2, r2, #0\n"
DieterGraef 0:d26c1b55cfca 106
DieterGraef 0:d26c1b55cfca 107 // Fold 32-bit checksum into 16-bit checksum.
DieterGraef 0:d26c1b55cfca 108 "8$:\n"
DieterGraef 0:d26c1b55cfca 109 " ubfx r4, r2, #16, #16\n"
DieterGraef 0:d26c1b55cfca 110 " ubfx r2, r2, #0, #16\n"
DieterGraef 0:d26c1b55cfca 111 " adds r2, r4\n"
DieterGraef 0:d26c1b55cfca 112 " ubfx r4, r2, #16, #16\n"
DieterGraef 0:d26c1b55cfca 113 " ubfx r2, r2, #0, #16\n"
DieterGraef 0:d26c1b55cfca 114 " adds r2, r4\n"
DieterGraef 0:d26c1b55cfca 115
DieterGraef 0:d26c1b55cfca 116 // Swap bytes if started at odd address
DieterGraef 0:d26c1b55cfca 117 " cbz r3, 9$\n"
DieterGraef 0:d26c1b55cfca 118 " rev16 r2, r2\n"
DieterGraef 0:d26c1b55cfca 119
DieterGraef 0:d26c1b55cfca 120 // Return final sum.
DieterGraef 0:d26c1b55cfca 121 "9$: mov r0, r2\n"
DieterGraef 0:d26c1b55cfca 122 " pop {r4, pc}\n"
DieterGraef 0:d26c1b55cfca 123 );
DieterGraef 0:d26c1b55cfca 124 }
DieterGraef 0:d26c1b55cfca 125
DieterGraef 0:d26c1b55cfca 126 #endif