configurable inlining to help with opts

3 files changed, 14 insertions, 4 deletions

diff --git a/Cargo.toml b/Cargo.toml
index e8734cc..34f8af7 100644
--- a/Cargo.toml
+++ b/Cargo.toml
@@ -52,3 +52,7 @@ colors = ["yaxpeax-arch/colors"]
 # This enables some capstone benchmarks over the same
 # instruction bytes used to bench this code.
 capstone_bench = []
+
+# this disables a lot of inlining to make it easier for me to measure
+# likelihood of codepaths for typical instruction sequences
+profiling = []
diff --git a/src/long_mode/display.rs b/src/long_mode/display.rs
index 77111b8..d0aef26 100644
--- a/src/long_mode/display.rs
+++ b/src/long_mode/display.rs
@@ -384,7 +384,6 @@ pub trait DisplaySink: fmt::Write {
     /// utf-8 string. this may corrupt Rust strings.
     unsafe fn write_lt_32(&mut self, s: &str) -> Result<(), core::fmt::Error> {
         self.write_str(s)
-
     }
     /// write a string to this sink that is less than 16 bytes. this is provided for optimization
     /// opportunities when writing a variable-length string with known max size.
@@ -5769,12 +5768,14 @@ impl Instruction {
         Ok(())
     }
 
+    #[cfg_attr(feature="profiling", inline(never))]
     pub fn write_to<T: DisplaySink>(&self, out: &mut T) -> fmt::Result {
         contextualize_intel(self, out)
 //        self.display_with(DisplayStyle::Intel).contextualize(&NoColors, 0, Some(&NoContext), out)
     }
 }
 
+#[cfg_attr(feature="profiling", inline(never))]
 fn contextualize_intel<T: DisplaySink>(instr: &Instruction, out: &mut T) -> fmt::Result {
     if instr.xacquire() {
         out.write_fixed_size("xacquire ")?;
diff --git a/src/long_mode/mod.rs b/src/long_mode/mod.rs
index e6d0a02..5a5b89a 100644
--- a/src/long_mode/mod.rs
+++ b/src/long_mode/mod.rs
@@ -4383,6 +4383,7 @@ impl Instruction {
     /// if the work you expect to do per-operand is very small, constructing an `Operand` and
     /// dispatching on tags may be a substantial factor of overall runtime. `visit_operand` can
     /// reduce total overhead in such cases.
+    #[cfg_attr(features="profiling", inline(never))]
     fn visit_operand<T: OperandVisitor>(&self, i: u8, visitor: &mut T) -> Result<T::Ok, T::Error> {
         assert!(i < 4);
         let spec = self.operands[i as usize];
@@ -6009,7 +6010,8 @@ const OPCODES: [OpcodeRecord; 256] = [
 ];
 
 #[allow(non_snake_case)]
-#[inline(always)]
+#[cfg_attr(feature="profiling", inline(never))]
+#[cfg_attr(not(feature="profiling"), inline(always))]
 pub(self) fn read_E<
     T: Reader<<Arch as yaxpeax_arch::Arch>::Address, <Arch as yaxpeax_arch::Arch>::Word>,
     S: DescriptionSink<FieldDescription>,
@@ -6638,6 +6640,7 @@ impl DecodeCtx {
         self.rb_size
     }
 
+#[cfg_attr(feature="profiling", inline(never))]
 fn read_opc_hotpath<
     T: Reader<<Arch as yaxpeax_arch::Arch>::Address, <Arch as yaxpeax_arch::Arch>::Word>,
     S: DescriptionSink<FieldDescription>,
@@ -6694,7 +6697,8 @@ fn read_opc_hotpath<
     }
 }
 
-#[inline(always)]
+#[cfg_attr(feature="profiling", inline(never))]
+#[cfg_attr(not(feature="profiling"), inline(always))]
 fn read_with_annotations<
     T: Reader<<Arch as yaxpeax_arch::Arch>::Address, <Arch as yaxpeax_arch::Arch>::Word>,
     S: DescriptionSink<FieldDescription>,
@@ -6912,7 +6916,8 @@ fn read_avx_prefixed<
     return Ok(());
 }
 
-#[inline(always)]
+#[cfg_attr(feature="profiling", inline(never))]
+#[cfg_attr(not(feature="profiling"), inline(always))]
 fn read_operands<
     T: Reader<<Arch as yaxpeax_arch::Arch>::Address, <Arch as yaxpeax_arch::Arch>::Word>,
     S: DescriptionSink<FieldDescription>