io.c: recycle garbage on write

* string.c (STR_IS_SHARED_M): new flag to mark shared mulitple times
  (STR_SET_SHARED): set STR_IS_SHARED_M
  (rb_str_tmp_frozen_acquire, rb_str_tmp_frozen_release): new functions
  (str_new_frozen): set/unset STR_IS_SHARED_M as appropriate
* internal.h: declare new functions
* io.c (fwrite_arg, fwrite_do, fwrite_end): new
  (io_fwrite): use new functions

Introduce rb_str_tmp_frozen_acquire and rb_str_tmp_frozen_release
to manage a hidden, frozen string.  Reuse one bit of the embed
length for shared strings as STR_IS_SHARED_M to indicate a string
has been shared multiple times.  In the common case, the string
is only shared once so the object slot can be reclaimed immediately.

minimum results in each 3 measurements. (time and size)

Execution time (sec)
name                            trunk   built
io_copy_stream_write            0.682   0.254
io_copy_stream_write_socket     1.225   0.751

Speedup ratio: compare with the result of `trunk' (greater is better)
name    built
io_copy_stream_write            2.680
io_copy_stream_write_socket     1.630

Memory usage (last size) (B)
name                            trunk           built
io_copy_stream_write            95436800.000    6512640.000
io_copy_stream_write_socket     117628928.000   7127040.000

Memory consuming ratio (size) with the result of `trunk' (greater is better)
name    built
io_copy_stream_write            14.654
io_copy_stream_write_socket     16.505

git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@57469 b2dd03c8-39d4-4d8f-98ff-823fe69b080e

internal.h

@@ -1475,6 +1475,8 @@ VALUE rb_id_quote_unprintable(ID);
 char *rb_str_fill_terminator(VALUE str, const int termlen);
 void rb_str_change_terminator_length(VALUE str, const int oldtermlen, const int termlen);
 VALUE rb_str_locktmp_ensure(VALUE str, VALUE (*func)(VALUE), VALUE arg);
+VALUE rb_str_tmp_frozen_acquire(VALUE str);
+void rb_str_tmp_frozen_release(VALUE str, VALUE tmp);
 VALUE rb_str_chomp_string(VALUE str, VALUE chomp);
 #ifdef RUBY_ENCODING_H
 VALUE rb_external_str_with_enc(VALUE str, rb_encoding *eenc);

io.c

@@ -1419,10 +1419,40 @@ do_writeconv(VALUE str, rb_io_t *fptr, int *converted)
     return str;
 }
 
+struct fwrite_arg {
+    VALUE orig;
+    VALUE tmp;
+    rb_io_t *fptr;
+    int nosync;
+};
+
+static VALUE
+fwrite_do(VALUE arg)
+{
+    struct fwrite_arg *fa = (struct fwrite_arg *)arg;
+    const char *ptr;
+    long len;
+
+    RSTRING_GETMEM(fa->tmp, ptr, len);
+
+    return (VALUE)io_binwrite(fa->tmp, ptr, len, fa->fptr, fa->nosync);
+}
+
+static VALUE
+fwrite_end(VALUE arg)
+{
+    struct fwrite_arg *fa = (struct fwrite_arg *)arg;
+
+    rb_str_tmp_frozen_release(fa->orig, fa->tmp);
+
+    return Qfalse;
+}
+
 static long
 io_fwrite(VALUE str, rb_io_t *fptr, int nosync)
 {
     int converted = 0;
+    struct fwrite_arg fa;
 #ifdef _WIN32
     if (fptr->mode & FMODE_TTY) {
 	long len = rb_w32_write_console(str, fptr->fd);
@@ -1432,11 +1462,13 @@ io_fwrite(VALUE str, rb_io_t *fptr, int nosync)
     str = do_writeconv(str, fptr, &converted);
     if (converted)
 	OBJ_FREEZE(str);
-    else
-	str = rb_str_new_frozen(str);
 
-    return io_binwrite(str, RSTRING_PTR(str), RSTRING_LEN(str),
-		       fptr, nosync);
+    fa.orig = str;
+    fa.tmp = rb_str_tmp_frozen_acquire(str);
+    fa.fptr = fptr;
+    fa.nosync = nosync;
+
+    return (long)rb_ensure(fwrite_do, (VALUE)&fa, fwrite_end, (VALUE)&fa);
 }
 
 ssize_t

string.c

@@ -70,6 +70,7 @@ VALUE rb_cSymbol;
  * 1:     RSTRING_NOEMBED
  * 2:     STR_SHARED (== ELTS_SHARED)
  * 2-6:   RSTRING_EMBED_LEN (5 bits == 32)
+ * 6:     STR_IS_SHARED_M (shared, when RSTRING_NOEMBED==1 && klass==0)
  * 7:     STR_TMPLOCK
  * 8-9:   ENC_CODERANGE (2 bits)
  * 10-16: ENCODING (7 bits == 128)
@@ -79,6 +80,7 @@ VALUE rb_cSymbol;
  */
 
 #define RUBY_MAX_CHAR_LEN 16
+#define STR_IS_SHARED_M FL_USER6
 #define STR_TMPLOCK FL_USER7
 #define STR_NOFREE FL_USER18
 #define STR_FAKESTR FL_USER19
@@ -150,6 +152,8 @@ VALUE rb_cSymbol;
     if (!FL_TEST(str, STR_FAKESTR)) { \
 	RB_OBJ_WRITE((str), &RSTRING(str)->as.heap.aux.shared, (shared_str)); \
 	FL_SET((str), STR_SHARED); \
+	if (RBASIC_CLASS((shared_str)) == 0) /* for CoW-friendliness */ \
+	    FL_SET_RAW((shared_str), STR_IS_SHARED_M); \
     } \
 } while (0)
 
@@ -1127,6 +1131,45 @@ rb_str_new_frozen(VALUE orig)
     return str;
 }
 
+VALUE
+rb_str_tmp_frozen_acquire(VALUE orig)
+{
+    VALUE tmp;
+
+    if (OBJ_FROZEN_RAW(orig)) return orig;
+
+    tmp = str_new_frozen(0, orig);
+    OBJ_INFECT(tmp, orig);
+
+    return tmp;
+}
+
+void
+rb_str_tmp_frozen_release(VALUE orig, VALUE tmp)
+{
+    if (RBASIC_CLASS(tmp) != 0)
+	return;
+
+    if (FL_TEST_RAW(orig, STR_SHARED) &&
+	    !FL_TEST_RAW(orig, STR_TMPLOCK|RUBY_FL_FREEZE)) {
+	VALUE shared = RSTRING(orig)->as.heap.aux.shared;
+
+	if (shared == tmp && !FL_TEST_RAW(tmp, STR_IS_SHARED_M)) {
+	    FL_UNSET_RAW(orig, STR_SHARED);
+	    assert(RSTRING(orig)->as.heap.ptr == RSTRING(tmp)->as.heap.ptr);
+	    assert(RSTRING(orig)->as.heap.len == RSTRING(tmp)->as.heap.len);
+	    RSTRING(orig)->as.heap.aux.capa = RSTRING(tmp)->as.heap.aux.capa;
+	    RBASIC(orig)->flags |= RBASIC(tmp)->flags & STR_NOFREE;
+	    assert(OBJ_FROZEN_RAW(tmp));
+	    rb_gc_force_recycle(tmp);
+	}
+    }
+    else if (STR_EMBED_P(tmp)) {
+	assert(OBJ_FROZEN_RAW(tmp));
+	rb_gc_force_recycle(tmp);
+    }
+}
+
 static VALUE
 str_new_frozen(VALUE klass, VALUE orig)
 {
@@ -1152,6 +1195,8 @@ str_new_frozen(VALUE klass, VALUE orig)
 		RSTRING(str)->as.heap.len -= ofs + rest;
 	    }
 	    else {
+		if (RBASIC_CLASS(shared) == 0)
+		    FL_SET_RAW(shared, STR_IS_SHARED_M);
 		return shared;
 	    }
 	}
@@ -1171,6 +1216,8 @@ str_new_frozen(VALUE klass, VALUE orig)
 	    RBASIC(str)->flags |= RBASIC(orig)->flags & STR_NOFREE;
 	    RBASIC(orig)->flags &= ~STR_NOFREE;
 	    STR_SET_SHARED(orig, str);
+	    if (klass == 0)
+		FL_UNSET_RAW(str, STR_IS_SHARED_M);
 	}
     }