From e1b3633e9595abdd4714a18bae9d5b0aeaed8a80 Mon Sep 17 00:00:00 2001
From: iximeow <me@iximeow.net>
Date: Thu, 21 Apr 2022 02:35:38 -0700
Subject: support a fast path through the decoder for [rex-prefixed]opcode
 insts

the overwhelming majority of x86 instructions are either a single-byte
opcode or a single-byte opcode with a rex prefix. supporting these
specially means that we don't have to length-check on every byte or
go through the full decode loop while reading the most likely
instructions. this is a significant improvement on typical x86 streams,
but comes at a moderate penalty for crafted x86 instructions.

the penalty is still not very bad, as the fast path is exited in favor
of the full decode loop as soon as we see a non-rex prefix byte; this
adds maybe a dozen instructions to the slow path.
---
 src/long_mode/mod.rs | 49 ++++++++++++++++++++++++++++++++++++++++++++++++-
 1 file changed, 48 insertions(+), 1 deletion(-)

(limited to 'src')

diff --git a/src/long_mode/mod.rs b/src/long_mode/mod.rs
index bc14a50..afa6642 100644
--- a/src/long_mode/mod.rs
+++ b/src/long_mode/mod.rs
@@ -6414,6 +6414,51 @@ impl fmt::Display for FieldDescription {
     }
 }
 
+fn read_opc_hotpath<
+    T: Reader<<Arch as yaxpeax_arch::Arch>::Address, <Arch as yaxpeax_arch::Arch>::Word>,
+    S: DescriptionSink<FieldDescription>,
+>(mut b: u8, mut record: OpcodeRecord, nextb: &mut u8, next_rec: &mut OpcodeRecord, words: &mut T, instruction: &mut Instruction, sink: &mut S) -> Result<Option<OperandCode>, DecodeError> {
+    if b >= 0x40 && b < 0x50 {
+        sink.record((words.offset() - 1) as u32 * 8, (words.offset() - 1) as u32 * 8 + 7, FieldDescription {
+            desc: InnerDescription::RexPrefix(b),
+            id: words.offset() as u32 * 8 - 8,
+        });
+        b = words.next().ok().ok_or(DecodeError::ExhaustedInput)?;
+        record = unsafe {
+            core::ptr::read_volatile(&OPCODES[b as usize])
+        };
+        instruction.prefixes.rex_from(b);
+    }
+
+    if let Interpretation::Instruction(opc) = record.0 {
+        if words.offset() > 1 {
+            sink.record(
+                words.offset() as u32 * 8 - 8 - 1, words.offset() as u32 * 8 - 8 - 1,
+                InnerDescription::Boundary("prefixes end")
+                    .with_id(words.offset() as u32 * 8 - 9)
+            );
+        }
+        if opc != Opcode::Invalid {
+            sink.record((words.offset() - 1) as u32 * 8, (words.offset() - 1) as u32 * 8 + 7, FieldDescription {
+                desc: InnerDescription::Opcode(opc),
+                id: words.offset() as u32 * 8 - 8,
+            });
+        }
+        sink.record((words.offset() - 1) as u32 * 8, (words.offset() - 1) as u32 * 8 + 7, FieldDescription {
+            desc: InnerDescription::OperandCode(OperandCodeWrapper { code: record.1 }),
+            id: words.offset() as u32 * 8 - 8 + 1,
+        });
+        instruction.mem_size = 0;
+        instruction.operand_count = 2;
+        instruction.opcode = opc;
+        return Ok(Some(record.1));
+    } else {
+        *nextb = b;
+        *next_rec = record;
+        return Ok(None);
+    }
+}
+
 #[inline(always)]
 fn read_with_annotations<
     T: Reader<<Arch as yaxpeax_arch::Arch>::Address, <Arch as yaxpeax_arch::Arch>::Word>,
@@ -6429,7 +6474,9 @@ fn read_with_annotations<
     // default operands to [RegRRR, Nothing, Nothing, Nothing]
     instruction.operands = unsafe { core::mem::transmute(0x00_00_00_01) };
 
-    let record: OperandCode = {
+    let record: OperandCode = if let Some(rec) = read_opc_hotpath(nextb, next_rec, &mut nextb, &mut next_rec, words, instruction, sink)? {
+        rec
+    } else {
         let prefixes = &mut instruction.prefixes;
         let record = loop {
             let record = next_rec;
-- 
cgit v1.1