From 9b538419de240f8b06326173e0b697d789865ddb Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Jo=C3=A3o=20Magalh=C3=A3es?= <joamag@gmail.com>
Date: Sat, 4 Mar 2023 17:57:08 +0000
Subject: [PATCH] feat: efficient multi-cycle per clock APU Improves
performance of APU by an order of maginute at the cost of less accurate
emulation.
---
src/apu.rs | 150 +++++++++++++++++++++++++----------------------------
1 file changed, 71 insertions(+), 79 deletions(-)
diff --git a/src/apu.rs b/src/apu.rs
index deb08753..03893a90 100644
--- a/src/apu.rs
+++ b/src/apu.rs
@@ -17,7 +17,7 @@ pub enum Channel {
}
pub struct Apu {
- ch1_timer: u16,
+ ch1_timer: i16,
ch1_sequence: u8,
ch1_envelope_sequence: u8,
ch1_envelope_enabled: bool,
@@ -35,7 +35,7 @@ pub struct Apu {
ch1_length_stop: bool,
ch1_enabled: bool,
- ch2_timer: u16,
+ ch2_timer: i16,
ch2_sequence: u8,
ch2_envelope_sequence: u8,
ch2_envelope_enabled: bool,
@@ -49,7 +49,7 @@ pub struct Apu {
ch2_length_stop: bool,
ch2_enabled: bool,
- ch3_timer: u16,
+ ch3_timer: i16,
ch3_position: u8,
ch3_output: u8,
ch3_dac: bool,
@@ -67,7 +67,7 @@ pub struct Apu {
sampling_rate: u16,
sequencer: u16,
sequencer_step: u8,
- output_timer: u16,
+ output_timer: i16,
audio_buffer: VecDeque<u8>,
audio_buffer_max: usize,
}
@@ -190,9 +190,60 @@ impl Apu {
}
pub fn clock(&mut self, cycles: u8) {
- // @TODO the performance here requires improvement
- for _ in 0..cycles {
- self.tick();
+ self.sequencer += cycles as u16;
+ if self.sequencer >= 8192 {
+ // each of these steps runs at 512/8 Hz = 64Hz,
+ // meaning a complete loop runs at 512 Hz
+ match self.sequencer_step {
+ 0 => {
+ self.tick_length_all();
+ }
+ 1 => (),
+ 2 => {
+ self.tick_ch1_sweep();
+ self.tick_length_all();
+ }
+ 3 => (),
+ 4 => {
+ self.tick_length_all();
+ }
+ 5 => (),
+ 6 => {
+ self.tick_ch1_sweep();
+ self.tick_length_all();
+ }
+ 7 => {
+ self.tick_envelope_all();
+ }
+ _ => (),
+ }
+
+ self.sequencer -= 8192;
+ self.sequencer_step = (self.sequencer_step + 1) & 7;
+ }
+
+ self.tick_ch_all(cycles);
+
+ self.output_timer = self.output_timer.saturating_sub(cycles as i16);
+ if self.output_timer <= 0 {
+ // verifies if we've reached the maximum allowed size for the
+ // audio buffer and if that's the case an item is removed from
+ // the buffer (avoiding overflow) and then then the new audio
+ // volume item is added to the queue
+ if self.audio_buffer.len() >= self.audio_buffer_max {
+ self.audio_buffer.pop_front();
+ self.audio_buffer.pop_front();
+ }
+ if self.left_enabled {
+ self.audio_buffer.push_back(self.output());
+ }
+ if self.right_enabled {
+ self.audio_buffer.push_back(self.output());
+ }
+
+ // @TODO the CPU clock is hardcoded here, we must handle situations
+ // where there's some kind of overclock
+ self.output_timer += (4194304.0 / self.sampling_rate as f32) as i16;
}
}
@@ -309,65 +360,6 @@ impl Apu {
self.audio_buffer.clear();
}
- #[inline(always)]
- fn tick(&mut self) {
- self.sequencer += 1;
- if self.sequencer >= 8192 {
- // each of these steps runs at 512/8 Hz = 64Hz,
- // meaning a complete loop runs at 512 Hz
- match self.sequencer_step {
- 0 => {
- self.tick_length_all();
- }
- 1 => (),
- 2 => {
- self.tick_ch1_sweep();
- self.tick_length_all();
- }
- 3 => (),
- 4 => {
- self.tick_length_all();
- }
- 5 => (),
- 6 => {
- self.tick_ch1_sweep();
- self.tick_length_all();
- }
- 7 => {
- self.tick_envelope_all();
- }
- _ => (),
- }
-
- self.sequencer = 0;
- self.sequencer_step = (self.sequencer_step + 1) & 7;
- }
-
- self.tick_ch_all();
-
- self.output_timer = self.output_timer.saturating_sub(1);
- if self.output_timer == 0 {
- // verifies if we've reached the maximum allowed size for the
- // audio buffer and if that's the case an item is removed from
- // the buffer (avoiding overflow) and then then the new audio
- // volume item is added to the queue
- if self.audio_buffer.len() >= self.audio_buffer_max {
- self.audio_buffer.pop_front();
- self.audio_buffer.pop_front();
- }
- if self.left_enabled {
- self.audio_buffer.push_back(self.output());
- }
- if self.right_enabled {
- self.audio_buffer.push_back(self.output());
- }
-
- // @TODO the CPU clock is hardcoded here, we must handle situations
- // where there's some kind of overclock
- self.output_timer = (4194304.0 / self.sampling_rate as f32) as u16;
- }
- }
-
#[inline(always)]
fn tick_length_all(&mut self) {
self.tick_length(Channel::Ch1);
@@ -477,15 +469,15 @@ impl Apu {
}
#[inline(always)]
- fn tick_ch_all(&mut self) {
- self.tick_ch1();
- self.tick_ch2();
- self.tick_ch3();
+ fn tick_ch_all(&mut self, cycles: u8) {
+ self.tick_ch1(cycles);
+ self.tick_ch2(cycles);
+ self.tick_ch3(cycles);
}
#[inline(always)]
- fn tick_ch1(&mut self) {
- self.ch1_timer = self.ch1_timer.saturating_sub(1);
+ fn tick_ch1(&mut self, cycles: u8) {
+ self.ch1_timer = self.ch1_timer.saturating_sub(cycles as i16);
if self.ch1_timer > 0 {
return;
}
@@ -501,13 +493,13 @@ impl Apu {
self.ch1_output = 0;
}
- self.ch1_timer = (2048 - self.ch1_wave_length) << 2;
+ self.ch1_timer += ((2048 - self.ch1_wave_length) << 2) as i16;
self.ch1_sequence = (self.ch1_sequence + 1) & 7;
}
#[inline(always)]
- fn tick_ch2(&mut self) {
- self.ch2_timer = self.ch2_timer.saturating_sub(1);
+ fn tick_ch2(&mut self, cycles: u8) {
+ self.ch2_timer = self.ch2_timer.saturating_sub(cycles as i16);
if self.ch2_timer > 0 {
return;
}
@@ -523,13 +515,13 @@ impl Apu {
self.ch2_output = 0;
}
- self.ch2_timer = (2048 - self.ch2_wave_length) << 2;
+ self.ch2_timer += ((2048 - self.ch2_wave_length) << 2) as i16;
self.ch2_sequence = (self.ch2_sequence + 1) & 7;
}
#[inline(always)]
- fn tick_ch3(&mut self) {
- self.ch3_timer = self.ch3_timer.saturating_sub(1);
+ fn tick_ch3(&mut self, cycles: u8) {
+ self.ch3_timer = self.ch3_timer.saturating_sub(cycles as i16);
if self.ch3_timer > 0 {
return;
}
@@ -552,7 +544,7 @@ impl Apu {
self.ch3_output = 0;
}
- self.ch3_timer = (2048 - self.ch3_wave_length) << 1;
+ self.ch3_timer += ((2048 - self.ch3_wave_length) << 1) as i16;
self.ch3_position = (self.ch3_position + 1) & 31;
}
}
--
GitLab