From da09fa13a86c59e45005a0e6ae5bfb802fd58f0d Mon Sep 17 00:00:00 2001
From: Luke Street <luke@street.dev>
Date: Thu, 26 Sep 2024 23:21:28 -0600
Subject: [PATCH] Update README.md & version 0.3.2

---
 Cargo.lock       |  8 ++++----
 Cargo.toml       |  4 ++--
 README.md        | 17 ++++++++---------
 fuzz/src/main.rs | 16 ++++++++--------
 4 files changed, 22 insertions(+), 23 deletions(-)

diff --git a/Cargo.lock b/Cargo.lock
index ebd0798..ed4435b 100644
--- a/Cargo.lock
+++ b/Cargo.lock
@@ -253,18 +253,18 @@ checksum = "439ee305def115ba05938db6eb1644ff94165c5ab5e9420d1c1bcedbba909391"
 
 [[package]]
 name = "ppc750cl"
-version = "0.3.1"
+version = "0.3.2"
 
 [[package]]
 name = "ppc750cl-asm"
-version = "0.3.1"
+version = "0.3.2"
 dependencies = [
  "phf",
 ]
 
 [[package]]
 name = "ppc750cl-fuzz"
-version = "0.3.1"
+version = "0.3.2"
 dependencies = [
  "clap",
  "num_cpus",
@@ -273,7 +273,7 @@ dependencies = [
 
 [[package]]
 name = "ppc750cl-genisa"
-version = "0.3.1"
+version = "0.3.2"
 dependencies = [
  "anyhow",
  "log",
diff --git a/Cargo.toml b/Cargo.toml
index 665502f..8ad9ede 100644
--- a/Cargo.toml
+++ b/Cargo.toml
@@ -7,10 +7,10 @@ panic = "abort"
 
 [profile.release-lto]
 inherits = "release"
-lto = "thin"
+lto = true
 
 [workspace.package]
-version = "0.3.1"
+version = "0.3.2"
 edition = "2021"
 authors = ["Luke Street <luke@street.dev>"]
 license = "MIT OR Apache-2.0"
diff --git a/README.md b/README.md
index da3e7ea..13076e6 100644
--- a/README.md
+++ b/README.md
@@ -8,7 +8,7 @@
 [rustdoc]: https://docs.rs/ppc750cl
 [Rust Version]: https://img.shields.io/badge/rust-1.74+-blue.svg?maxAge=3600
 
-Rust tools for working with the PowerPC 750CL family of processors.
+Rust tools for working with the PowerPC 750CL / 750CXe family of processors.
 
 ### Building
 
@@ -22,14 +22,14 @@ cargo build --release
 For those unfamiliar with PowerPC, here are some basics.
 - PowerPC 7xx is a family of RISC CPUs produced from 1997 to 2012.
   - They operate with 32-bit words and every instruction is 32-bits wide.
-- This project focuses (only) on compatibility with the PowerPC 750CL.
-  - This chip is famously packaged as codename "Broadway" for the Nintendo Wii.
-  - Its predecessor PowerPC 750CXe is used in the Nintendo GameCube.
-  - It adds a "paired-singles" SIMD unit and a bunch of other instructions.
+- This project focuses (only) on compatibility with the PowerPC 750CL and 750CXe.
+  - PowerPC 750CL ("Broadway") is used in the Nintendo Wii.
+  - PowerPC 750CXe ("Gekko") is used in the Nintendo GameCube.
+  - They add a "paired-singles" SIMD unit and a bunch of other instructions.
 
 ### isa.yaml
 
-The file [isa.yaml](./isa.yaml) contains a full definition of the PowerPC 750CL instruction set.
+The file [isa.yaml](./isa.yaml) contains a full definition of the PowerPC 750CL / 750CXe instruction set.
 
 It powers the disassembler and assembler.
 
@@ -37,7 +37,6 @@ Similarly to LLVM TableGen, the program `ppc750cl-genisa` generates a Rust file
 
 ### Safety & Correctness
 
-- This project does not use `unsafe` Rust code outside of testing utils.
 - The disassembler has been fuzzed over all ~4.29 billion possible instructions (via `ppc750cl-fuzz`).
 - It is safe to run the disassembler over untrusted byte arrays.
 - However, no guarantees on correctness are made (yet). Expect bugs.
@@ -46,5 +45,5 @@ Similarly to LLVM TableGen, the program `ppc750cl-genisa` generates a Rust file
 
 With a single thread on Ryzen 9 3900X:
 
-- Disassembling & printing: ~5M insn/s (~20 MB/s)
-- Disassembling only: ~50M insn/s (~200 MB/s)
+- Disassembling & printing: ~12M insn/s (~50 MB/s)
+- Disassembling only: ~275M insn/s (~1 GB/s)
diff --git a/fuzz/src/main.rs b/fuzz/src/main.rs
index f8ba0ce..986c9d3 100644
--- a/fuzz/src/main.rs
+++ b/fuzz/src/main.rs
@@ -11,7 +11,7 @@ fn main() {
         .arg(
             clap::Arg::new("threads")
                 .short('t')
-                .long("--threads")
+                .long("threads")
                 .help("Number of threads to use (default num CPUs)"),
         )
         .get_matches();
@@ -22,8 +22,8 @@ fn main() {
     };
     let start = Instant::now();
     let fuzzer = MultiFuzzer::new(threads);
-    fuzzer.run();
-    println!("Finished in {:.2}s", start.elapsed().as_secs_f32());
+    let avg = fuzzer.run();
+    println!("Finished in {:.2}s (avg {}/s)", start.elapsed().as_secs_f32(), avg);
 }
 
 #[derive(Clone)]
@@ -70,14 +70,17 @@ impl MultiFuzzer {
         });
     }
 
-    fn run(&self) {
+    fn run(&self) -> f32 {
         self.dispatch_progress_monitor();
+        let start = Instant::now();
         let handles: Vec<_> = self.threads.iter().map(|t| t.dispatch()).collect();
         for handle in handles {
             // TODO This doesn't panic immediately, since we'll block on thread zero
             //      for most of the time.
             handle.join().expect("thread panicked");
         }
+        let end = start.elapsed();
+        u32::MAX as f32 / end.as_secs_f32() / self.threads.len() as f32
     }
 }
 
@@ -113,11 +116,8 @@ impl Fuzzer {
 
 struct DevNull;
 
-impl std::io::Write for DevNull {
+impl Write for DevNull {
     fn write(&mut self, buf: &[u8]) -> std::io::Result<usize> {
-        buf.iter().for_each(|b| unsafe {
-            std::ptr::read_volatile(b);
-        });
         Ok(buf.len())
     }
     fn flush(&mut self) -> std::io::Result<()> {