Skip to content

Preevaluation #2

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
Bleb1k wants to merge 2 commits into
base: main
Choose a base branch
from
Open

Preevaluation #2

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
4c269e4
evaluation mostly works
Bleb1k Feb 1, 2026
0749a60
replaced write with printf
Bleb1k Feb 1, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
202 changes: 159 additions & 43 deletions bq.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -32,6 +32,7 @@
#include "cvector.h"

//#define PRINT_BUT_DONT_EXEC
#define EVAL_LIMIT 5000

#define code_trap() code_append("\xcc")

Expand Down Expand Up @@ -164,27 +165,10 @@ handler(int signum, siginfo_t *si, void *ucontext)
die("could not protect tape memory underflow guard page:");
}

int
main(int argc, char *argv[])
void
parse(cvector(Instr) instrs, const char *source)
{
if unlikely (argc != 2)
usage(argv[0]);

int fd = open(argv[1], O_RDONLY);
if unlikely (fd < 0)
die("cannot access '%s':", argv[1]);

struct stat st;
fstat(fd, &st);

char *txt = emalloc(st.st_size);
read(fd, txt, st.st_size);
close(fd);

cvector(Instr) instrs = NULL;
cvector_reserve(instrs, (size_t)st.st_size);

for (char *s = txt; likely(*s); s++) {
for (const char *s = source; likely(*s); s++) {
Instr instr;
switch (*s) {
case '>':
Expand Down Expand Up @@ -267,9 +251,83 @@ main(int argc, char *argv[])
default: break;
}
}
}

free(txt);
typedef struct {
size_t cur_instruction;
size_t instr_skip_count;
bool finished;
} CodegenOptions;

void
evaluate(cvector(Instr) instrs, CodegenOptions *cg_opts)
{
cvector(char) output_acc = NULL;
cvector(size_t) jump_stack = NULL;
size_t instr_skip_count = 0;
bool finished = false;

size_t i, evaluated_count = 0;
for (i = 0; likely(i < cvector_size(instrs)); i++) {
if (evaluated_count++ >= EVAL_LIMIT)
goto end_eval;
Instr instr = instrs[i];

switch (instr.op) {
case OP_MOVE: tape += instr.arg; break;
case OP_ADD: *tape += instr.arg % 256; break;
case OP_OUTPUT: cvector_push_back(output_acc, *tape); break;
case OP_INPUT: goto end_eval;
case OP_JUMP_RIGHT:
if likely (*tape) {
cvector_push_back(jump_stack, i);
} else {
int depth = 1;
while (depth) {
switch (instrs[++i].op) {
case OP_JUMP_RIGHT: depth += 1; break;
case OP_JUMP_LEFT: depth -= 1; break;
default:;
}
}
}
break;
case OP_JUMP_LEFT:
if unlikely (!cvector_size(jump_stack))
die("jump stack underflow");
if likely (*tape)
i = *cvector_back(jump_stack);
else
cvector_pop_back(jump_stack);
break;
case OP_ADD_TO: *(tape + instr.arg) += *tape; // falls through
case OP_CLEAR: *tape = 0; break;
case OP_MOVE_UNTIL:
while (*tape)
tape += instr.arg;
break;
default: __builtin_unreachable();
}
if (!cvector_size(jump_stack))
instr_skip_count = i;
}
finished = true;

end_eval:;
cvector_push_back(output_acc, '\0');
printf("%s", output_acc);
cvector_free(output_acc);

*cg_opts = (CodegenOptions){
.cur_instruction = i,
.instr_skip_count = instr_skip_count,
.finished = finished,
};
}

void
codegen(void **fn, cvector(Instr) instrs, CodegenOptions options)
{
cvector(unsigned char) code = NULL;
cvector_reserve(code, max(cvector_size(instrs) * 5, 128));

Expand All @@ -282,6 +340,9 @@ main(int argc, char *argv[])
cvector(uintptr_t) uflowpatches = NULL;
cvector_reserve(uflowpatches, max(cvector_size(instrs) / 200, 32));

if (options.finished)
goto skip_jit;

#define code_append(snip_) \
do { \
size_t snip_size_ = sizeof snip_ / sizeof *snip_ - 1; \
Expand All @@ -308,18 +369,27 @@ main(int argc, char *argv[])
cvector_set_size(code, off + i); \
} while (0)

const char snip[] = "\x49\xbd\x00\x00\x00\x00\x00\x00\x00\x00" // movabs r13, imm64
"\x49\xbe\x00\x00\x00\x00\x00\x00\x00\x00" // movabs r14, imm64
"\x48\x89\xfb"; // mov rbx, rdi

code_append(snip);
*(void **)(code + cvector_size(code) - 21) = stdin;
*(void **)(code + cvector_size(code) - 11) = stdout;
unsigned char *entry = NULL;
{
const char snip[] = "\x49\xbd\x00\x00\x00\x00\x00\x00\x00\x00" // movabs r13, imm64
"\x49\xbe\x00\x00\x00\x00\x00\x00\x00\x00" // movabs r14, imm64
"\x48\x89\xfb"; // mov rbx, rdi

size_t icacheline = sysconf(_SC_LEVEL1_ICACHE_LINESIZE);
code_append(snip);
*(void **)(code + cvector_size(code) - 21) = stdin;
*(void **)(code + cvector_size(code) - 11) = stdout;
}
{
const char snip[] = "\xe9\x00\x00\x00\x00"; // jmp rel32
code_append(snip);
entry = code + cvector_size(code);
}
size_t icacheline = sysconf(_SC_LEVEL1_ICACHE_LINESIZE);
code_align(icacheline);

for (size_t i = 0; likely(i < cvector_size(instrs)); i++) {
for (size_t i = options.instr_skip_count; likely(i < cvector_size(instrs)); i++) {
if unlikely (i == options.cur_instruction)
*(uint32_t *)(entry - 4) = code + cvector_size(code) - entry;
Instr instr = instrs[i];

switch (instr.op) {
Expand Down Expand Up @@ -546,36 +616,68 @@ main(int argc, char *argv[])
if unlikely (cvector_size(jmps) != 0)
die("unterminated [");

skip_jit:;
cvector_free(instrs);
cvector_free(jmps);

code_append("\xc3"); // ret

void *fn = mmap(NULL, cvector_size(code), PROT_WRITE, MAP_PRIVATE | MAP_ANONYMOUS, -1, 0);
if unlikely (!fn)
#ifdef PRINT_BUT_DONT_EXEC
fwrite(code, cvector_size(code), 1, stdout);
#else
*fn = mmap(NULL, cvector_size(code), PROT_WRITE, MAP_PRIVATE | MAP_ANONYMOUS, -1, 0);
if unlikely (!*fn)
die("could not allocate executable memory:");

extern int __overflow(FILE *, int);
for (size_t i = 0; i < cvector_size(overflowpatches); i++)
*(int *)&code[overflowpatches[i]] = (uintptr_t)__overflow - ((uintptr_t)fn + overflowpatches[i] + 4);

cvector_free(overflowpatches);
*(int *)&code[overflowpatches[i]] = (uintptr_t)__overflow - ((uintptr_t)*fn + overflowpatches[i] + 4);

extern int __uflow(FILE *);
for (size_t i = 0; i < cvector_size(uflowpatches); i++)
*(int *)&code[uflowpatches[i]] = (uintptr_t)__uflow - ((uintptr_t)fn + uflowpatches[i] + 4);
*(int *)&code[uflowpatches[i]] = (uintptr_t)__uflow - ((uintptr_t)*fn + uflowpatches[i] + 4);

memcpy(*fn, code, cvector_size(code));
mprotect(*fn, cvector_size(code), PROT_EXEC);
#endif

cvector_free(overflowpatches);
cvector_free(uflowpatches);
cvector_free(code);

#ifdef PRINT_BUT_DONT_EXEC
fwrite(code, cvector_size(code), 1, stdout);
#undef code_append
#undef code_align
}

return 0;
#endif
int
main(int argc, char *argv[])
{
if unlikely (argc != 2)
usage(argv[0]);

memcpy(fn, code, cvector_size(code));
mprotect(fn, cvector_size(code), PROT_EXEC);
cvector_free(code);
int fd = open(argv[1], O_RDONLY);
if unlikely (fd < 0)
die("cannot access '%s':", argv[1]);

struct stat st;
fstat(fd, &st);

// ----- Read Source -----

char *txt = emalloc(st.st_size);
read(fd, txt, st.st_size);
close(fd);

// ----- Parse -----

cvector(Instr) instrs = NULL;
cvector_reserve(instrs, (size_t)st.st_size);

parse(instrs, txt);

free(txt);

// ----- Prepare tape -----

size_t tapesize = 30000;
size_t pagesize = getpagesize();
Expand All @@ -602,8 +704,22 @@ main(int argc, char *argv[])
if unlikely (sigaction(SIGSEGV, &sa, NULL) < 0)
die("could not prepare tape memory guard page:");

// ----- Preevaluate -----

CodegenOptions cg_opts = { 0 };
evaluate(instrs, &cg_opts);

// ----- Codegen -----

void *fn = NULL;
codegen(&fn, instrs, cg_opts);

// ----- Run -----

#ifndef PRINT_BUT_DONT_EXEC
(*(void (**)(void *))&fn)(tape);
// leak tape on purpose
#endif

return 0;
}