in.c

#include <stdio.h>
#include <stdint.h>
#include <stddef.h>
#include <malloc.h>

typedef struct gc_root gc_root_t;
typedef struct gc_block_meta gc_block_meta_t;
typedef struct gc_block gc_block_t;

#define BLOCK_ALLOC_SIZE 4096
#define BLOCK_POOL_SIZE (BLOCK_ALLOC_SIZE - sizeof(gc_block_meta_t))

static gc_block_t volatile *volatile gc_block_head = NULL;
static gc_root_t volatile *volatile gc_large_root_head = NULL;

struct gc_root {
    gc_root_t *next;
    unsigned refs;
};

struct gc_block_meta {
    gc_block_t *next;
    gc_root_t *root;
    unsigned lock;
    size_t upto;
};

#pragma pack(push, 0)

struct gc_block {
    gc_block_meta_t meta;
    uint8_t pool[BLOCK_POOL_SIZE];
};

#pragma pack(pop)

void *gc_malloc_large(size_t size) {
    return (void*)0;
}

void *gc_malloc(size_t size) {
    if (size > BLOCK_POOL_SIZE) {
        return gc_malloc_large(size);
    }

    size_t upto_n, upto_o;
    gc_block_t volatile *block = gc_block_head;
    do {
    l_retry:
        upto_o = block->meta.upto;
        upto_n = upto_o + size;
        if (upto_n > BLOCK_POOL_SIZE) {
            if (__sync_bool_compare_and_swap(&block->meta.lock, 0, 1)) {
                //gc_block_t *block_n = 
                // We have the lock
                return (void*)0;
            }
            while (!block->meta.lock);
            block = gc_block_head;
            goto l_retry;
        }
    }
    while (!__sync_bool_compare_and_swap(&block->meta.upto, upto_o, upto_n));

    gc_root_t *root_n = (gc_root_t*)(&block->pool[upto_o]), *root_o;
    root_n->refs = 0;
    do {
        root_o = block->meta.root;
        root_n->next = root_o;
    }
    while (!__sync_bool_compare_and_swap(&block->meta.root, root_o, root_n));

    return root_n;
}

int main() {
    printf("%d", gc_malloc(300));
    return 0;
}

out.s

   0x00000001004010e0 <+0>:     cmp    rcx,0xfe0
   0x00000001004010e7 <+7>:     ja     0x10040113f <gc_malloc+95>
   0x00000001004010e9 <+9>:     mov    rdx,QWORD PTR [rip+0x4f28]        # 0x406018
   0x00000001004010f0 <+16>:    xor    r11d,r11d
   0x00000001004010f3 <+19>:    mov    r10d,0x1
   0x00000001004010f9 <+25>:    mov    r8,QWORD PTR [rdx+0x18]
   0x00000001004010fd <+29>:    lea    r9,[rcx+r8*1]
   0x0000000100401101 <+33>:    cmp    r9,0xfe0
   0x0000000100401108 <+40>:    ja     0x100401134 <gc_malloc+84>
   0x000000010040110a <+42>:    mov    rax,r8
   0x000000010040110d <+45>:    lock cmpxchg QWORD PTR [rdx+0x18],r9
   0x0000000100401113 <+51>:    jne    0x1004010f9 <gc_malloc+25>
   0x0000000100401115 <+53>:    lea    rcx,[rdx+r8*1+0x20]
   0x000000010040111a <+58>:    mov    DWORD PTR [rcx+0x8],0x0
   0x0000000100401121 <+65>:    mov    rax,QWORD PTR [rdx+0x8]
   0x0000000100401125 <+69>:    mov    QWORD PTR [rcx],rax
   0x0000000100401128 <+72>:    lock cmpxchg QWORD PTR [rdx+0x8],rcx
   0x000000010040112e <+78>:    jne    0x100401121 <gc_malloc+65>
   0x0000000100401130 <+80>:    mov    rax,rcx
   0x0000000100401133 <+83>:    ret
   0x0000000100401134 <+84>:    mov    eax,r11d
   0x0000000100401137 <+87>:    lock cmpxchg DWORD PTR [rdx+0x10],r10d
   0x000000010040113d <+93>:    jne    0x100401142 <gc_malloc+98>
   0x000000010040113f <+95>:    xor    eax,eax
   0x0000000100401141 <+97>:    ret
   0x0000000100401142 <+98>:    mov    eax,DWORD PTR [rdx+0x10]
   0x0000000100401145 <+101>:   test   eax,eax
   0x0000000100401147 <+103>:   je     0x100401142 <gc_malloc+98>
   0x0000000100401149 <+105>:   mov    rdx,QWORD PTR [rip+0x4ec8]        # 0x406018
   0x0000000100401150 <+112>:   jmp    0x1004010f9 <gc_malloc+25>