From: Oleg Babin via Tarantool-patches <tarantool-patches@dev.tarantool.org> To: Igor Munkin <imun@tarantool.org> Cc: tarantool-patches@dev.tarantool.org, v.shpilevoy@tarantool.org Subject: Re: [Tarantool-patches] [PATCH v2 3/3] fiber: keep reference to userdata if fiber created once Date: Fri, 3 Sep 2021 23:48:58 +0300 [thread overview] Message-ID: <6e5f684a-bf7b-3767-1e31-25ed59416c61@tarantool.org> (raw) In-Reply-To: <20210903122314.GV5743@tarantool.org> Thanks for your review. See my answers below. On 03.09.2021 15:23, Igor Munkin wrote: > Oleg, > > Thanks for the fixes! I still have some comments regarding the new > version of this patch. > > On 02.09.21, olegrok@tarantool.org wrote: >> From: Oleg Babin <babinoleg@mail.ru> >> >> This patch reworks approach to fiber management in Lua. Before >> this patch each action that should return fiber led to new >> userdata creation that was quite slow and made GC suffer. This >> patch introduces new field in struct fiber to store a reference to >> userdata that was created once for a fiber. It allows speedup >> operations as fiber.self() and fiber.id(). >> Simple benchmark shows that access to fiber storage is faster in >> two times, fiber.find() - 2-3 times and fiber.new/create functions >> don't have any changes. >> --- >> src/lib/core/fiber.c | 5 ++ >> src/lib/core/fiber.h | 5 ++ >> src/lua/fiber.c | 113 +++++++++++++------------------------------ >> 3 files changed, 44 insertions(+), 79 deletions(-) >> >> diff --git a/src/lib/core/fiber.c b/src/lib/core/fiber.c >> index 588b78504..39b67f940 100644 >> --- a/src/lib/core/fiber.c >> +++ b/src/lib/core/fiber.c >> @@ -32,6 +32,7 @@ >> >> #include <trivia/config.h> >> #include <trivia/util.h> >> +#include <lauxlib.h> > Totally agree with Vlad here. I guess you have two options: > * Use the approach similar to <cord_on_yield>. > * Introduce a constant in src/lib/core/fiber.h and add a static assert > in src/lua/utils.h whether this value is the same as in LuaJIT. > > Maybe there are other options Vlad can suggest. I've found that there is lua_nil_ref value in our code. For my purpose I've added lua_no_ref. Incremental diff will be below. Also I will send new series since this change was done in separate patch. > >> #include <errno.h> >> #include <stdio.h> >> #include <stdlib.h> > <snipped> > >> diff --git a/src/lua/fiber.c b/src/lua/fiber.c >> index 5575f2079..ab0e895eb 100644 >> --- a/src/lua/fiber.c > +++ b/src/lua/fiber.c >> @@ -87,27 +87,28 @@ luaL_testcancel(struct lua_State *L) > <snipped> > >> + struct fiber *f = event; >> + int storage_ref = f->storage.lua.storage_ref; > Minor: Why do you need both <storage_ref> and <ref> variables? After > your change they can be freely dropped. Removed > >> + luaL_unref(tarantool_L, LUA_REGISTRYINDEX, storage_ref); >> + f->storage.lua.storage_ref = LUA_NOREF; >> + int ref = f->storage.lua.fid_ref; >> + luaL_unref(tarantool_L, LUA_REGISTRYINDEX, ref); >> + f->storage.lua.fid_ref = LUA_NOREF; >> + trigger_clear(trigger); >> + free(trigger); >> + return 0; >> } >> >> /** >> @@ -116,42 +117,26 @@ lbox_create_weak_table(struct lua_State *L, const char *name) > <snipped> > >> + int ref = f->storage.lua.fid_ref; > Minor: It would be nice also to rename <ref> to <fid_ref>. Fixed. >> + if (ref == LUA_NOREF) { > <snipped> > >> -- >> 2.32.0 >> diff --git a/src/lib/core/fiber.c b/src/lib/core/fiber.c index 39b67f940..1e5b2fb80 100644 --- a/src/lib/core/fiber.c +++ b/src/lib/core/fiber.c @@ -32,7 +32,6 @@ #include <trivia/config.h> #include <trivia/util.h> -#include <lauxlib.h> #include <errno.h> #include <stdio.h> #include <stdlib.h> @@ -45,6 +44,7 @@ #include "errinj.h" extern void cord_on_yield(void); +extern int luaL_no_ref; #if ENABLE_FIBER_TOP #include <x86intrin.h> /* __rdtscp() */ @@ -896,8 +896,8 @@ fiber_recycle(struct fiber *fiber) fiber->f = NULL; fiber->wait_pad = NULL; memset(&fiber->storage, 0, sizeof(fiber->storage)); - fiber->storage.lua.storage_ref = LUA_NOREF; - fiber->storage.lua.fid_ref = LUA_NOREF; + fiber->storage.lua.storage_ref = luaL_no_ref; + fiber->storage.lua.fid_ref = luaL_no_ref; unregister_fid(fiber); fiber->fid = 0; region_free(&fiber->gc); @@ -1239,8 +1239,8 @@ fiber_new_ex(const char *name, const struct fiber_attr *fiber_attr, return NULL; } memset(fiber, 0, sizeof(struct fiber)); - fiber->storage.lua.storage_ref = LUA_NOREF; - fiber->storage.lua.fid_ref = LUA_NOREF; + fiber->storage.lua.storage_ref = luaL_no_ref; + fiber->storage.lua.fid_ref = luaL_no_ref; if (fiber_stack_create(fiber, &cord()->slabc, fiber_attr->stack_size)) { diff --git a/src/lua/fiber.c b/src/lua/fiber.c index f82f4032f..70addd8f6 100644 --- a/src/lua/fiber.c +++ b/src/lua/fiber.c @@ -100,11 +100,9 @@ static int lbox_fiber_on_stop(struct trigger *trigger, void *event) { struct fiber *f = event; - int storage_ref = f->storage.lua.storage_ref; - luaL_unref(tarantool_L, LUA_REGISTRYINDEX, storage_ref); + luaL_unref(tarantool_L, LUA_REGISTRYINDEX, f->storage.lua.storage_ref); f->storage.lua.storage_ref = LUA_NOREF; - int ref = f->storage.lua.fid_ref; - luaL_unref(tarantool_L, LUA_REGISTRYINDEX, ref); + luaL_unref(tarantool_L, LUA_REGISTRYINDEX, f->storage.lua.fid_ref); f->storage.lua.fid_ref = LUA_NOREF; trigger_clear(trigger); free(trigger); @@ -117,8 +115,8 @@ lbox_fiber_on_stop(struct trigger *trigger, void *event) static void lbox_pushfiber(struct lua_State *L, struct fiber *f) { - int ref = f->storage.lua.fid_ref; - if (ref == LUA_NOREF) { + int fid_ref = f->storage.lua.fid_ref; + if (fid_ref == LUA_NOREF) { struct trigger *t = malloc(sizeof(*t)); if (t == NULL) { diag_set(OutOfMemory, sizeof(*t), "malloc", "t"); @@ -133,10 +131,10 @@ lbox_pushfiber(struct lua_State *L, struct fiber *f) *ptr = fid; luaL_getmetatable(L, fiberlib_name); lua_setmetatable(L, -2); - ref = luaL_ref(L, LUA_REGISTRYINDEX); - f->storage.lua.fid_ref = ref; + fid_ref = luaL_ref(L, LUA_REGISTRYINDEX); + f->storage.lua.fid_ref = fid_ref; } - lua_rawgeti(L, LUA_REGISTRYINDEX, ref); + lua_rawgeti(L, LUA_REGISTRYINDEX, fid_ref); } static struct fiber * diff --git a/src/lua/utils.c b/src/lua/utils.c index c71cd4857..c252d7cdf 100644 --- a/src/lua/utils.c +++ b/src/lua/utils.c @@ -40,6 +40,7 @@ #include "uuid/tt_uuid.h" int luaL_nil_ref = LUA_REFNIL; +int luaL_no_ref = LUA_NOREF; static int luaT_newthread_ref = LUA_NOREF; diff --git a/test/unit/core_test_utils.c b/test/unit/core_test_utils.c index 23452bbfd..9b1789e75 100644 --- a/test/unit/core_test_utils.c +++ b/test/unit/core_test_utils.c @@ -35,3 +35,5 @@ void cord_on_yield(void) { } + +int luaL_no_ref = -2;
next prev parent reply other threads:[~2021-09-03 20:49 UTC|newest] Thread overview: 8+ messages / expand[flat|nested] mbox.gz Atom feed top 2021-09-02 18:13 [Tarantool-patches] [PATCH v2 0/3] " Oleg Babin via Tarantool-patches 2021-09-02 18:13 ` [Tarantool-patches] [PATCH v2 1/3] fiber: rename ref to storage_ref Oleg Babin via Tarantool-patches 2021-09-02 18:13 ` [Tarantool-patches] [PATCH v2 2/3] fiber: pass struct fiber into lbox_pushfiber instead of id Oleg Babin via Tarantool-patches 2021-09-02 18:13 ` [Tarantool-patches] [PATCH v2 3/3] fiber: keep reference to userdata if fiber created once Oleg Babin via Tarantool-patches 2021-09-02 19:47 ` Oleg Babin via Tarantool-patches 2021-09-03 12:23 ` Igor Munkin via Tarantool-patches 2021-09-03 20:48 ` Oleg Babin via Tarantool-patches [this message] 2021-09-05 15:07 ` Vladislav Shpilevoy via Tarantool-patches
Reply instructions: You may reply publicly to this message via plain-text email using any one of the following methods: * Save the following mbox file, import it into your mail client, and reply-to-all from there: mbox Avoid top-posting and favor interleaved quoting: https://en.wikipedia.org/wiki/Posting_style#Interleaved_style * Reply using the --to, --cc, and --in-reply-to switches of git-send-email(1): git send-email \ --in-reply-to=6e5f684a-bf7b-3767-1e31-25ed59416c61@tarantool.org \ --to=tarantool-patches@dev.tarantool.org \ --cc=imun@tarantool.org \ --cc=olegrok@tarantool.org \ --cc=v.shpilevoy@tarantool.org \ --subject='Re: [Tarantool-patches] [PATCH v2 3/3] fiber: keep reference to userdata if fiber created once' \ /path/to/YOUR_REPLY https://kernel.org/pub/software/scm/git/docs/git-send-email.html * If your mail client supports setting the In-Reply-To header via mailto: links, try the mailto: link
This is a public inbox, see mirroring instructions for how to clone and mirror all data and code used for this inbox