When entering nterp, take a fast path for instance calls with 1 argument.
Such methods only take 'this' as an argument and we don't need to fetch
the shorty.
We can make this optimization when doing nterp->compiled as a follow-up,
by checking that the next instruction after the invoke is not
move-result(-wide).
Test: test.py
Bug: 112676029
Change-Id: Ibc7b4d4ca1c636f4ad6572484e0990ccdbd63293
diff --git a/tools/cpp-define-generator/art_method.def b/tools/cpp-define-generator/art_method.def
index 7b5606f..c2e18b1 100644
--- a/tools/cpp-define-generator/art_method.def
+++ b/tools/cpp-define-generator/art_method.def
@@ -25,6 +25,8 @@
art::kAccStatic)
ASM_DEFINE(ART_METHOD_IMT_MASK,
art::ImTable::kSizeTruncToPowerOfTwo - 1)
+ASM_DEFINE(ART_METHOD_IS_STATIC_FLAG_BIT,
+ art::MostSignificantBit(art::kAccStatic))
ASM_DEFINE(ART_METHOD_DECLARING_CLASS_OFFSET,
art::ArtMethod::DeclaringClassOffset().Int32Value())
ASM_DEFINE(ART_METHOD_JNI_OFFSET_32,