[builtins,arm64] Use direct calls for embedded builtins.

This is a port of e5630ea9:
"[builtins,x64] pc-relative builtin-to-builtin calls".

Bug: v8:6666
Change-Id: Id53bf45c6a4fedb1c972ced72984b6673808ee96
Reviewed-on: https://chromium-review.googlesource.com/1126249Reviewed-by: Jakob Gruber <jgruber@chromium.org>
Commit-Queue: Georgia Kouveli <georgia.kouveli@arm.com>
Cr-Commit-Position: refs/heads/master@{#54243}

src/arm64/macro-assembler-arm64.cc

@@ -1931,7 +1931,13 @@ void TurboAssembler::Jump(Handle<Code> code, RelocInfo::Mode rmode,
                           Condition cond) {
   DCHECK(RelocInfo::IsCodeTarget(rmode));
   if (FLAG_embedded_builtins) {
-    if (root_array_available_ && options().isolate_independent_code) {
+    if (root_array_available_ && options().isolate_independent_code &&
+        !Builtins::IsIsolateIndependentBuiltin(*code)) {
+      // Calls to embedded targets are initially generated as standard
+      // pc-relative calls below. When creating the embedded blob, call offsets
+      // are patched up to point directly to the off-heap instruction start.
+      // Note: It is safe to dereference {code} above since code generation
+      // for builtins and code stubs happens on the main thread.
       UseScratchRegisterScope temps(this);
       Register scratch = temps.AcquireX();
       IndirectLoadConstant(scratch, code);
@@ -2004,7 +2010,13 @@ void TurboAssembler::Call(Handle<Code> code, RelocInfo::Mode rmode) {
 #endif
 
   if (FLAG_embedded_builtins) {
-    if (root_array_available_ && options().isolate_independent_code) {
+    if (root_array_available_ && options().isolate_independent_code &&
+        !Builtins::IsIsolateIndependentBuiltin(*code)) {
+      // Calls to embedded targets are initially generated as standard
+      // pc-relative calls below. When creating the embedded blob, call offsets
+      // are patched up to point directly to the off-heap instruction start.
+      // Note: It is safe to dereference {code} above since code generation
+      // for builtins and code stubs happens on the main thread.
       UseScratchRegisterScope temps(this);
       Register scratch = temps.AcquireX();
       IndirectLoadConstant(scratch, code);

src/snapshot/snapshot-common.cc

@@ -340,7 +340,7 @@ void FinalizeEmbeddedCodeTargets(Isolate* isolate, EmbeddedData* blob) {
     RelocIterator on_heap_it(code, kRelocMask);
     RelocIterator off_heap_it(blob, code, kRelocMask);
 
-#ifdef V8_TARGET_ARCH_X64
+#if defined(V8_TARGET_ARCH_X64) || defined(V8_TARGET_ARCH_ARM64)
     while (!on_heap_it.done()) {
       DCHECK(!off_heap_it.done());
 
@@ -357,12 +357,12 @@ void FinalizeEmbeddedCodeTargets(Isolate* isolate, EmbeddedData* blob) {
     }
     DCHECK(off_heap_it.done());
 #else
-    // Architectures other than x64 do not use pc-relative calls and thus must
-    // not contain embedded code targets. Instead, we use an indirection through
-    // the root register.
+    // Architectures other than x64 and arm64 do not use pc-relative calls and
+    // thus must not contain embedded code targets. Instead, we use an
+    // indirection through the root register.
     CHECK(on_heap_it.done());
     CHECK(off_heap_it.done());
-#endif  // V8_TARGET_ARCH_X64
+#endif  // defined(V8_TARGET_ARCH_X64) || defined(V8_TARGET_ARCH_ARM64)
   }
 }
 }  // namespace