From 5c952cf0245421feb4644f2e71487c0b2e1dbd13 Mon Sep 17 00:00:00 2001 From: wdenk Date: Sun, 10 Oct 2004 21:27:30 +0000 Subject: Patches by Scott McNutt, 24 Aug 2004: - Add support for Altera Nios-II processors. - Add support for Psyent PCI-5441 board. - Add support for Psyent PK1C20 board. --- cpu/nios2/start.S | 211 ++++++++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 211 insertions(+) create mode 100644 cpu/nios2/start.S (limited to 'cpu/nios2/start.S') diff --git a/cpu/nios2/start.S b/cpu/nios2/start.S new file mode 100644 index 0000000000..281d42c212 --- /dev/null +++ b/cpu/nios2/start.S @@ -0,0 +1,211 @@ +/* + * (C) Copyright 2004, Psyent Corporation + * Scott McNutt + * + * See file CREDITS for list of people who contributed to this + * project. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License as + * published by the Free Software Foundation; either version 2 of + * the License, or (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 59 Temple Place, Suite 330, Boston, + * MA 02111-1307 USA + */ + + +#include +#include + +/************************************************************************* + * RESTART + ************************************************************************/ + + .text + .global _start + +_start: + /* ICACHE INIT -- only the icache line at the reset address + * is invalidated at reset. So the init must stay within + * the cache line size (8 words). If GERMS is used, we'll + * just be invalidating the cache a second time. If cache + * is not implemented initi behaves as nop. + */ + movhi r4, %hi(CFG_ICACHELINE_SIZE) + ori r4, r4, %lo(CFG_ICACHELINE_SIZE) + movhi r5, %hi(CFG_ICACHE_SIZE) + ori r5, r5, %lo(CFG_ICACHE_SIZE) + mov r6, r0 +0: initi r6 + add r6, r6, r4 + bltu r6, r5, 0b + + /* INTERRUPTS -- for now, all interrupts masked and globally + * disabled. + */ + wrctl status, r0 /* Disable interrupts */ + wrctl ienable, r0 /* All disabled */ + + /* DCACHE INIT -- if dcache not implemented, initd behaves as + * nop. + */ + movhi r4, %hi(CFG_DCACHELINE_SIZE) + ori r4, r4, %lo(CFG_DCACHELINE_SIZE) + movhi r5, %hi(CFG_DCACHE_SIZE) + ori r5, r5, %lo(CFG_DCACHE_SIZE) + mov r6, r0 +1: initd 0(r6) + add r6, r6, r4 + bltu r6, r5, 1b + + /* RELOCATE CODE, DATA & COMMAND TABLE -- the following code + * assumes code, data and the command table are all + * contiguous. This lets us relocate everything as a single + * block. Make sure the linker script matches this ;-) + */ + nextpc r4 +_cur: movhi r5, %hi(_cur - _start) + ori r5, r5, %lo(_cur - _start) + sub r4, r4, r5 /* r4 <- cur _start */ + mov r8, r4 + movhi r5, %hi(_start) + ori r5, r5, %lo(_start) /* r5 <- linked _start */ + beq r4, r5, 3f + + movhi r6, %hi(_edata) + ori r6, r6, %lo(_edata) +2: ldwio r7, 0(r4) + addi r4, r4, 4 + stwio r7, 0(r5) + addi r5, r5, 4 + bne r5, r6, 2b +3: + + /* ZERO BSS/SBSS -- bss and sbss are assumed to be adjacent + * and between __bss_start and _end. + */ + movhi r5, %hi(__bss_start) + ori r5, r5, %lo(__bss_start) + movhi r6, %hi(_end) + ori r6, r6, %lo(_end) + beq r5, r6, 5f + +4: stwio r0, 0(r5) + addi r5, r5, 4 + bne r5, r6, 4b +5: + + /* GLOBAL POINTER -- the global pointer is used to reference + * "small data" (see -G switch). The linker script must + * provide the gp address. + */ + movhi gp, %hi(_gp) + ori gp, gp, %lo(_gp) + + /* JUMP TO RELOC ADDR */ + movhi r4, %hi(_reloc) + ori r4, r4, %lo(_reloc) + jmp r4 +_reloc: + + /* COPY EXCEPTION TRAMPOLINE -- copy the tramp to the + * exception address. + */ +#if !defined(CONFIG_ROM_STUBS) + movhi r4, %hi(_except_start) + ori r4, r4, %lo(_except_start) + movhi r5, %hi(_except_end) + ori r5, r5, %lo(_except_end) + movhi r6, %hi(CFG_EXCEPTION_ADDR) + ori r6, r6, %lo(CFG_EXCEPTION_ADDR) + +6: ldwio r7, 0(r4) + stwio r7, 0(r6) + addi r4, r4, 4 + addi r6, r6, 4 + bne r4, r5, 6b +#endif + + /* STACK INIT -- zero top two words for call back chain. + */ + movhi sp, %hi(CFG_INIT_SP) + ori sp, sp, %lo(CFG_INIT_SP) + addi sp, sp, -8 + stw r0, 0(sp) + stw r0, 4(sp) + mov fp, sp + + /* + * Call board_init -- never returns + */ + movhi r4, %hi(board_init@h) + ori r4, r4, %lo(board_init@h) + callr r4 + + /* NEVER RETURNS -- but branch to the _start just + * in case ;-) + */ + br _start + + /* EXCEPTION TRAMPOLINE -- the following gets copied + * to the exception address. + */ +_except_start: + movhi et, %hi(_exception) + ori et, et, %lo(_exception) + jmp et +_except_end: + + +/* + * dly_clks -- Nios2 (like Nios1) doesn't have a timebase in + * the core. For simple delay loops, we do our best by counting + * instruction cycles. + * + * Instruction performance varies based on the core. For cores + * with icache and static/dynamic branch prediction (II/f, II/s): + * + * Normal ALU (e.g. add, cmp, etc): 1 cycle + * Branch (correctly predicted, taken): 2 cycles + * Negative offset is predicted (II/s). + * + * For cores without icache and no branch prediction (II/e): + * + * Normal ALU (e.g. add, cmp, etc): 6 cycles + * Branch (no prediction): 6 cycles + * + * For simplicity, if an instruction cache is implemented we + * assume II/f or II/s. Otherwise, we use the II/e. + * + */ + .globl dly_clks + +dly_clks: + +#if (CFG_ICACHE_SIZE > 0) + subi r4, r4, 3 /* 3 clocks/loop */ +#else + subi r4, r4, 12 /* 12 clocks/loop */ +#endif + bge r4, r0, dly_clks + ret + + +#if !defined(CONFIG_IDENT_STRING) +#define CONFIG_IDENT_STRING "" +#endif + .data + .globl version_string + +version_string: + .ascii U_BOOT_VERSION + .ascii " (", __DATE__, " - ", __TIME__, ")" + .ascii CONFIG_IDENT_STRING, "\0" -- cgit