[GR-18163] Store the conversion from native to managed string inside …

…the RubyString when matching regexps PullRequest: truffleruby/4315
oracle · Jul 22, 2024 · 16fdb0a · 16fdb0a
2 parents 938589f + 8b3e47b
commit 16fdb0a
Show file tree

Hide file tree

Showing 3 changed files with 27 additions and 1 deletion.
diff --git a/CHANGELOG.md b/CHANGELOG.md
@@ -47,6 +47,7 @@ Performance:
 
 * Fix inline caching for Regexp creation from Strings (#3492, @andrykonchin, @eregon).
 * Optimize `Integer#pow` method for small modulus values (#3544, @andrykonchin).
+* Avoid repeated copies from native to managed string when matching Regexps on a native string (#2193, @eregon).
 
 Changes:
 

diff --git a/src/main/java/org/truffleruby/core/regexp/TruffleRegexpNodes.java b/src/main/java/org/truffleruby/core/regexp/TruffleRegexpNodes.java
@@ -64,6 +64,7 @@
 import org.truffleruby.core.kernel.KernelNodes.SameOrEqualNode;
 import org.truffleruby.core.regexp.RegexpNodes.ToSNode;
 import org.truffleruby.core.string.ATStringWithEncoding;
+import org.truffleruby.core.string.StringHelperNodes.StringToTruffleStringInplaceNode;
 import org.truffleruby.core.string.TStringBuilder;
 import org.truffleruby.core.string.TStringWithEncoding;
 import org.truffleruby.core.string.RubyString;
@@ -897,6 +898,7 @@ static Object matchInRegionTRegex(
                 boolean atStart,
                 int startPos,
                 boolean createMatchData,
+                @Cached StringToTruffleStringInplaceNode stringToTruffleStringInplaceNode,
                 @Cached TruffleString.SwitchEncodingNode switchEncodingNode,
                 @Cached InlinedConditionProfile createMatchDataProfile,
                 @Cached InlinedConditionProfile matchFoundProfile,
@@ -916,11 +918,12 @@ static Object matchInRegionTRegex(
                 @Cached LazyMatchInRegionNode fallbackMatchInRegionNode,
                 @Cached LazyTruffleStringSubstringByteIndexNode substringByteIndexNode,
                 @Bind("this") Node node) {
-            final Object tRegex;
+            stringToTruffleStringInplaceNode.execute(node, string);
             final RubyEncoding negotiatedEncoding = prepareRegexpEncodingNode.executePrepare(node, regexp, string);
             var tstring = switchEncodingNode.execute(libString.getTString(string), negotiatedEncoding.tencoding);
             final int byteLength = tstring.byteLength(negotiatedEncoding.tencoding);
 
+            final Object tRegex;
             if (tRegexIncompatibleProfile
                     .profile(node, toPos < fromPos || toPos != byteLength || fromPos < 0) ||
                     tRegexCouldNotCompileProfile.profile(node, (tRegex = tRegexCompileNode.executeTRegexCompile(

diff --git a/src/main/java/org/truffleruby/core/string/StringHelperNodes.java b/src/main/java/org/truffleruby/core/string/StringHelperNodes.java
@@ -621,4 +621,26 @@ static RubyString stringAppend(Node node, Object string, Object other,
         }
     }
 
+    /** Stores the TruffleString back in the Ruby String to avoid repeated copies from native to managed */
+    @GenerateInline
+    @GenerateCached(false)
+    public abstract static class StringToTruffleStringInplaceNode extends RubyBaseNode {
+
+        public abstract TruffleString execute(Node node, Object string);
+
+        @Specialization
+        static TruffleString immutable(ImmutableRubyString string) {
+            return string.tstring;
+        }
+
+        @Specialization
+        static TruffleString mutable(RubyString string,
+                @Cached RubyStringLibrary libString,
+                @Cached(inline = false) TruffleString.AsTruffleStringNode asTruffleStringNode) {
+            TruffleString tstring = asTruffleStringNode.execute(string.tstring, libString.getTEncoding(string));
+            string.setTString(tstring);
+            return tstring;
+        }
+    }
+
 }