#if __riscv_xlen == 64
# define STORE    sd
# define LOAD     ld
# define LOG_REGBYTES 3
#else
# define STORE    sw
# define LOAD     lw
# define LOG_REGBYTES 2
#endif
#define REGBYTES (1 << LOG_REGBYTES)

/*
    Entry point of all programs (_start).

    It initializes DWARF call frame information, the stack pointer, the
    frame pointer (needed for closures to work in start_rust) and the global
    pointer. Then it calls _start_rust.
*/

.section .init, "ax"
.global _start

_start:
    /* Jump to the absolute address defined by the linker script. */
    // for 32bit
    .if __riscv_xlen == 32
    lui ra, %hi(_abs_start)
    jr %lo(_abs_start)(ra)
    .endif

    // for 64bit
    .if __riscv_xlen == 64
.option push
.option norelax // to prevent an unsupported R_RISCV_ALIGN relocation from being generated
1:
    auipc ra, %pcrel_hi(1f)
    ld ra, %pcrel_lo(1b)(ra)
    jr ra
    .align  3
1:
    .dword _abs_start
.option pop
    .endif

_abs_start:
    .cfi_startproc
    .cfi_undefined ra

    #ifdef SMODE
    csrw sie, 0     // interrupt disable 
    csrw sip, 0     // no pending interrupts
    #else
    csrw mie, 0
    csrw mip, 0
    #endif
    

    li  x1, 0
    li  x2, 0
    li  x3, 0
    li  x4, 0
    li  x5, 0
    li  x6, 0
    li  x7, 0
    li  x8, 0
    li  x9, 0
    // a0..a2 (x10..x12) skipped
    li  x13,0
    li  x14,0
    li  x15,0
    li  x16,0
    li  x17,0
    li  x18,0
    li  x19,0
    li  x20,0
    li  x21,0
    li  x22,0
    li  x23,0
    li  x24,0
    li  x25,0
    li  x26,0
    li  x27,0
    li  x28,0
    li  x29,0
    li  x30,0
    li  x31,0

    .option push
    .option norelax
    la gp, __global_pointer$
    .option pop

    #ifdef SMODE
    // there is no equivalent of mhartid in supervisor mode.
    // instead, the hartid is passed as paramter by SMODE
    mv t2, a0   
    #else 
    csrr t2, mhartid
    #endif
    lui t0, %hi(_max_hart_id)
    add t0, t0, %lo(_max_hart_id)
    bgtu t2, t0, abort

    // Allocate stacks
    la sp, _stack_start
    lui t0, %hi(_hart_stack_size)
    add t0, t0, %lo(_hart_stack_size)
#ifdef __riscv_mul
    mul t0, t2, t0
#else
    beqz t2, 2f  // Jump if single-hart
    mv t1, t2
    mv t3, t0
1:
    add t0, t0, t3
    addi t1, t1, -1
    bnez t1, 1b
2:
#endif
    sub sp, sp, t0

    // Set frame pointer
    add s0, sp, zero

    jal zero, _start_rust

    .cfi_endproc

/*
    Trap entry point (_start_trap)

    Saves caller saved registers ra, t0..6, a0..7, calls _start_trap_rust,
    restores caller saved registers and then returns.
*/
.section .trap, "ax"
.global default_start_trap

default_start_trap:
    addi sp, sp, -16*REGBYTES

    STORE ra, 0*REGBYTES(sp)
    STORE t0, 1*REGBYTES(sp)
    STORE t1, 2*REGBYTES(sp)
    STORE t2, 3*REGBYTES(sp)
    STORE t3, 4*REGBYTES(sp)
    STORE t4, 5*REGBYTES(sp)
    STORE t5, 6*REGBYTES(sp)
    STORE t6, 7*REGBYTES(sp)
    STORE a0, 8*REGBYTES(sp)
    STORE a1, 9*REGBYTES(sp)
    STORE a2, 10*REGBYTES(sp)
    STORE a3, 11*REGBYTES(sp)
    STORE a4, 12*REGBYTES(sp)
    STORE a5, 13*REGBYTES(sp)
    STORE a6, 14*REGBYTES(sp)
    STORE a7, 15*REGBYTES(sp)

    add a0, sp, zero
    jal ra, _start_trap_rust

    LOAD ra, 0*REGBYTES(sp)
    LOAD t0, 1*REGBYTES(sp)
    LOAD t1, 2*REGBYTES(sp)
    LOAD t2, 3*REGBYTES(sp)
    LOAD t3, 4*REGBYTES(sp)
    LOAD t4, 5*REGBYTES(sp)
    LOAD t5, 6*REGBYTES(sp)
    LOAD t6, 7*REGBYTES(sp)
    LOAD a0, 8*REGBYTES(sp)
    LOAD a1, 9*REGBYTES(sp)
    LOAD a2, 10*REGBYTES(sp)
    LOAD a3, 11*REGBYTES(sp)
    LOAD a4, 12*REGBYTES(sp)
    LOAD a5, 13*REGBYTES(sp)
    LOAD a6, 14*REGBYTES(sp)
    LOAD a7, 15*REGBYTES(sp)

    addi sp, sp, 16*REGBYTES
    #ifdef SMODE
    sret
    #else
    mret
    #endif

/* Make sure there is an abort when linking */
.section .text.abort
.globl abort
abort:
    j abort