From 8c0c0b0dad6a7ddf2b3075db2285dce8130ef2d0 Mon Sep 17 00:00:00 2001 From: zinc Date: Thu, 15 Oct 2026 14:25:01 +0330 Subject: actors: micro workers map a copy-on-write snapshot too; inherited input is only the unread bytes; worker stacks 4 MB --- src/core/core.c | 2 +- src/micro/micro.c | 32 ++++++++++++++--- src/micro/port_linux.c | 94 +++++++++++++++++++++++++++++++++++++++++++++----- 3 files changed, 113 insertions(+), 15 deletions(-) diff --git a/src/core/core.c b/src/core/core.c index 77c4a99..648f54a 100644 --- a/src/core/core.c +++ b/src/core/core.c @@ -7688,7 +7688,7 @@ static Cell context_start(Cell entry, Cell channel) { memcpy(context->callback_snapshot, ffi_cb, sizeof ffi_cb); memcpy(context->symbol_snapshot, ffi_sym, sizeof ffi_sym); memcpy(context->library_snapshot, ffi_lib, sizeof ffi_lib); context->ffi_symbols = ffi_nsym; context->ffi_libraries = ffi_nlib; context->ffi_libm = ffi_tried_libm; #endif - pthread_attr_init(&attributes); pthread_attr_setstacksize(&attributes, 16u << 20); + pthread_attr_init(&attributes); pthread_attr_setstacksize(&attributes, 4u << 20); /* the vm keeps its own stacks in vm memory; 4 MB of address space covers the deepest native recursion (jit compile, o3) */ i = pthread_create(&context->thread, &attributes, context_worker, context); pthread_attr_destroy(&attributes); if (i) goto failed; context->parent_channel.next = context_channels; context_channels = &context->parent_channel; diff --git a/src/micro/micro.c b/src/micro/micro.c index 39a22b2..9924d5e 100644 --- a/src/micro/micro.c +++ b/src/micro/micro.c @@ -647,6 +647,11 @@ extern size_t micro_context_port_size(void); extern void micro_context_port_save(void *); extern void micro_context_port_restore(const void *); extern void micro_context_port_free(void); +#ifndef _WIN32 +extern Cell *micro_mem_clone(void); +extern void micro_mem_adopt(Cell *mem, UCell words); +extern void micro_mem_release(Cell *mem); +#endif #define CONTEXT_HANDLE_LO 0x40000000 typedef struct ContextChannel { Cell handle; int fd; struct ContextChannel *next; } ContextChannel; typedef struct VmContext { @@ -658,6 +663,7 @@ typedef struct VmContext { UCell words, limit, initial; int sp, rp, exit_code, requested_exit, enhancement_count, input_flag, input_known; Cell *memory; + int memory_mapped; void *port; Xop operations[XPOOL]; Enh enhancements[NENH]; @@ -683,15 +689,26 @@ static int context_close(Cell handle) { } static void context_snapshot_free(VmContext *context) { free(context->port); - if (context->memory) { free(context->memory); atomic_fetch_sub(&context_memory_live, 1); } + if (context->memory) { +#ifndef _WIN32 + if (context->memory_mapped) micro_mem_release(context->memory); else +#endif + free(context->memory); + atomic_fetch_sub(&context_memory_live, 1); + } } static void *context_worker(void *opaque) { VmContext *context = opaque; context_current = context; context->worker_channel.next = context_channels; context_channels = &context->worker_channel; micro_context_port_restore(context->port); free(context->port); context->port = 0; memory_limit = context->limit; memory_initial = context->initial; + if (context->memory_mapped) { +#ifndef _WIN32 + micro_mem_adopt(context->memory, context->words); context->memory = 0; +#endif + } if (!mem_grow_words(context->words)) { context->exit_code = -1; context_close(context->worker_channel.handle); return 0; } - memcpy(m, context->memory, (size_t)context->words * sizeof(Cell)); free(context->memory); context->memory = 0; + if (context->memory) { memcpy(m, context->memory, (size_t)context->words * sizeof(Cell)); free(context->memory); context->memory = 0; } stack_control = context->stack; jflag = context->input_flag; jknown = context->input_known; xuse = context->enhancement_count; memcpy(xops, context->operations, sizeof xops); memcpy(enh, context->enhancements, sizeof enh); int sp = context->sp, rp = context->rp; Cell entry = context->entry; @@ -714,12 +731,17 @@ static Cell context_start(Cell entry, Cell channel, int sp, int rp) { context->parent_channel.handle = CONTEXT_HANDLE_LO + atomic_fetch_add(&context_sequence, 1) + 1; context->worker_channel.handle = CONTEXT_HANDLE_LO + atomic_fetch_add(&context_sequence, 1) + 1; context->parent_channel.fd = fd[0]; context->worker_channel.fd = fd[1]; atomic_init(&context->stop, 0); - context->port = malloc(micro_context_port_size()); context->memory = malloc((size_t)mw * sizeof(Cell)); + context->port = malloc(micro_context_port_size()); +#ifndef _WIN32 + context->memory = micro_mem_clone(); context->memory_mapped = context->memory != 0; + if (!context->memory) +#endif + context->memory = malloc((size_t)mw * sizeof(Cell)); if (context->memory) atomic_fetch_add(&context_memory_live, 1); if (!context->port || !context->memory) goto failed; - micro_context_port_save(context->port); memcpy(context->memory, m, (size_t)mw * sizeof(Cell)); context->memory[channel] = context->worker_channel.handle; + micro_context_port_save(context->port); if (!context->memory_mapped) memcpy(context->memory, m, (size_t)mw * sizeof(Cell)); context->memory[channel] = context->worker_channel.handle; memcpy(context->operations, xops, sizeof xops); memcpy(context->enhancements, enh, sizeof enh); - pthread_attr_t attributes; pthread_attr_init(&attributes); pthread_attr_setstacksize(&attributes, 16u << 20); + pthread_attr_t attributes; pthread_attr_init(&attributes); pthread_attr_setstacksize(&attributes, 4u << 20); /* the vm keeps its own stacks in vm memory; 4 MB of address space covers the deepest native recursion (jit compile, o3) */ int result = pthread_create(&context->thread, &attributes, context_worker, context); pthread_attr_destroy(&attributes); if (result) goto failed; context->parent_channel.next = context_channels; context_channels = &context->parent_channel; diff --git a/src/micro/port_linux.c b/src/micro/port_linux.c index 6245f4f..a39e3e8 100644 --- a/src/micro/port_linux.c +++ b/src/micro/port_linux.c @@ -38,16 +38,38 @@ #define VM_TLS #endif -/* the machine's memory: a block that grows by realloc, zero in the new part. the - base can move on a grow, which micro.c takes into account (it re-reads m). */ +/* the machine's memory. with threads (actors) on a posix host it is one large + reservation that grows in place, so that a worker's memory can be a + copy-on-write map of a snapshot of its parent (micro_mem_clone); a machine that + outgrows the reservation moves to a block that grows by realloc. without threads it + is a block that grows by realloc, zero in the new part. the base can move on a + grow, which micro.c takes into account (it re-reads m). */ static VM_TLS Cell *lmem; static VM_TLS UCell lcap; +#if defined(ZY_THREADS) && !defined(_WIN32) +#include +#ifdef __linux__ +#include +#endif +#define RESW ((UCell)1 << 28) /* words reserved per machine: 2 GB of address space */ +static VM_TLS int lres; /* lmem is a reservation of RESW words */ +static Cell *res_map(void) { void *p = mmap(0, RESW * sizeof(Cell), PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANON | MAP_NORESERVE, -1, 0); return p == MAP_FAILED ? 0 : p; } +#endif Cell *micro_mem_alloc(UCell words) { Cell *n; if (words <= lcap && lmem) return lmem; if (words > ((UCell)-1) / sizeof(Cell)) return NULL; +#if defined(ZY_THREADS) && !defined(_WIN32) + if (!lmem && words <= RESW && (lmem = res_map())) { lres = 1; lcap = words; return lmem; } + if (lres && words <= RESW) { lcap = words; return lmem; } /* the reservation reads zero past lcap */ + if (lres) { + n = malloc(words * sizeof(Cell)); if (!n) return NULL; + memcpy(n, lmem, lcap * sizeof(Cell)); memset(n + lcap, 0, (words - lcap) * sizeof(Cell)); + munmap(lmem, RESW * sizeof(Cell)); lres = 0; lmem = n; lcap = words; return n; + } +#endif n = realloc(lmem, words * sizeof(Cell)); if (!n) return NULL; memset(n + lcap, 0, (words - lcap) * sizeof(Cell)); @@ -56,15 +78,62 @@ Cell *micro_mem_alloc(UCell words) return n; } +#if defined(ZY_THREADS) && !defined(_WIN32) +/* a worker's memory is a private map of an unchanging snapshot of its parent's, + plus the pages the parent changed since; see the same scheme in src/core/core.c */ +#define SNAP_PAGE 512 +static VM_TLS int snap_fd = -1; static VM_TLS Cell *snap_view; static VM_TLS UCell snap_words; +static void snap_drop(void) { if (snap_view) munmap(snap_view, snap_words * sizeof(Cell)); if (snap_fd >= 0) close(snap_fd); snap_fd = -1; snap_view = 0; snap_words = 0; } +static int snap_take(void) +{ + UCell words = (lcap + SNAP_PAGE - 1) / SNAP_PAGE * SNAP_PAGE; size_t bytes = words * sizeof(Cell), done = 0; int fd = -1; void *view; +#ifdef __linux__ +#ifdef SYS_memfd_create + fd = (int)syscall(SYS_memfd_create, "zym-snapshot", 1u); +#endif +#else + { char name[64]; snprintf(name, sizeof name, "/zym-%d-%p", (int)getpid(), (void *)&name); fd = shm_open(name, O_RDWR | O_CREAT | O_EXCL, 0600); if (fd >= 0) shm_unlink(name); } +#endif + if (fd < 0) return -1; + if (ftruncate(fd, (off_t)bytes) < 0) { close(fd); return -1; } + while (done < bytes) { ssize_t k = pwrite(fd, (char *)lmem + done, bytes - done, (off_t)done); if (k <= 0) { close(fd); return -1; } done += (size_t)k; } + view = mmap(0, bytes, PROT_READ, MAP_SHARED, fd, 0); + if (view == MAP_FAILED) { close(fd); return -1; } + if (mmap(lmem, bytes, PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_FIXED | MAP_NORESERVE, fd, 0) == MAP_FAILED) { munmap(view, bytes); close(fd); return -1; } + snap_drop(); snap_fd = fd; snap_view = view; snap_words = words; return 0; +} +Cell *micro_mem_clone(void) +{ + UCell p, dirty = 0; Cell *c; + if (!lres || !lmem) return 0; + if (snap_fd >= 0) { + for (p = 0; p < lcap; p += SNAP_PAGE) if (p >= snap_words || memcmp(lmem + p, snap_view + p, SNAP_PAGE * sizeof(Cell))) dirty++; + if (dirty * 4 > snap_words / SNAP_PAGE) snap_drop(); + } + if (snap_fd < 0 && snap_take() < 0) return 0; + c = res_map(); if (!c) return 0; + if (mmap(c, snap_words * sizeof(Cell), PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_FIXED | MAP_NORESERVE, snap_fd, 0) == MAP_FAILED) { munmap(c, RESW * sizeof(Cell)); return 0; } + for (p = 0; p < lcap; p += SNAP_PAGE) { + UCell k = lcap - p < SNAP_PAGE ? lcap - p : SNAP_PAGE; + if (p >= snap_words || memcmp(lmem + p, snap_view + p, k * sizeof(Cell))) memcpy(c + p, lmem + p, k * sizeof(Cell)); + } + return c; +} +void micro_mem_adopt(Cell *mem, UCell words) { lmem = mem; lcap = words; lres = 1; } +void micro_mem_release(Cell *mem) { if (mem) munmap(mem, RESW * sizeof(Cell)); } +#endif + #if defined(ZY_THREADS) -static VM_TLS unsigned char input_bytes[65536]; +/* allocated when the machine first reads, so a worker that never reads has none */ +static VM_TLS unsigned char *input_buffer; +#define input_bytes (input_buffer ? input_buffer : (input_buffer = malloc(65536)) ? input_buffer : (fputs("zym: out of memory\n", stderr), exit(2), input_buffer)) static VM_TLS int input_position, input_count, input_ended; static int input_fill(int wait) { if (input_position < input_count) return 1; if (input_ended) return 0; if (!wait) { struct pollfd descriptor = {0, POLLIN, 0}; if (poll(&descriptor, 1, 0) <= 0) return 0; } else fflush(stdout); - int n = (int)read(0, input_bytes, sizeof input_bytes); + int n = (int)read(0, input_bytes, 65536); if (n <= 0) { input_ended = 1; return 0; } input_position = 0; input_count = n; return 1; } @@ -392,22 +461,29 @@ int main(int argc, char **argv) { #ifdef ZY_THREADS struct PortSnapshot { - unsigned char bytes[65536]; int position, count, ended, argc, envc; char **argv, **envp; + unsigned char bytes[]; /* the input read ahead and not consumed yet */ }; -size_t micro_context_port_size(void) { return sizeof(struct PortSnapshot); } +size_t micro_context_port_size(void) { return sizeof(struct PortSnapshot) + (size_t)(input_position < input_count ? input_count - input_position : 0); } void micro_context_port_save(void *destination) { struct PortSnapshot *snapshot = destination; - memcpy(snapshot->bytes, input_bytes, sizeof input_bytes); + if (input_position < input_count) memcpy(snapshot->bytes, input_bytes + input_position, (size_t)(input_count - input_position)); snapshot->position = input_position; snapshot->count = input_count; snapshot->ended = input_ended; snapshot->argc = zy_argc; snapshot->argv = zy_argv; snapshot->envc = zy_envc; snapshot->envp = zy_envp; } void micro_context_port_restore(const void *source) { const struct PortSnapshot *snapshot = source; - memcpy(input_bytes, snapshot->bytes, sizeof input_bytes); input_position = snapshot->position; input_count = snapshot->count; input_ended = snapshot->ended; + if (input_position < input_count) memcpy(input_bytes + input_position, snapshot->bytes, (size_t)(input_count - input_position)); zy_argc = snapshot->argc; zy_argv = snapshot->argv; zy_envc = snapshot->envc; zy_envp = snapshot->envp; } -void micro_context_port_free(void) { free(lmem); lmem = 0; lcap = 0; } +void micro_context_port_free(void) { + free(input_buffer); input_buffer = 0; +#ifndef _WIN32 + if (lres) { munmap(lmem, RESW * sizeof(Cell)); lres = 0; snap_drop(); } else +#endif + free(lmem); + lmem = 0; lcap = 0; +} #endif -- cgit v1.2.3