pbwraith/gopher64-netplay
Archived
1
0
Fork 0
Code Issues Pull Requests 3 Actions Packages Projects Releases Wiki Activity

optimize some reads/writes (#571)

Browse Source
This commit is contained in:
Logan McNaughton 2025-09-15 18:26:45 +02:00 committed by GitHub
parent daee53a626
commit 100238f198
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 138 additions and 43 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

181
src/device/rsp_su_instructions.rs
View File

@ -304,25 +304,44 @@ pub fn lh(device: &mut device::Device, opcode: u32) {
let address =
device.rsp.cpu.gpr[rs(opcode) as usize].wrapping_add(imm(opcode) as i16 as i32 as u32);
let mut w = [0; 2];
w[0] = device.rsp.mem[address as usize & 0xFFF];
w[1] = device.rsp.mem[(address as usize + 1) & 0xFFF];
// Optimize for aligned access when possible
let addr_masked = address as usize & 0xFFF;
let value = if addr_masked <= 0xFFE {
u16::from_be_bytes(
device.rsp.mem[addr_masked..addr_masked + 2]
.try_into()
.unwrap(),
)
} else {
let w0 = device.rsp.mem[addr_masked];
let w1 = device.rsp.mem[(addr_masked + 1) & 0xFFF];
((w0 as u16) << 8) | w1 as u16
};
device.rsp.cpu.gpr[rt(opcode) as usize] =
(((w[0] as u16) << 8) | w[1] as u16) as i16 as i32 as u32
device.rsp.cpu.gpr[rt(opcode) as usize] = value as i16 as i32 as u32;
}
pub fn lw(device: &mut device::Device, opcode: u32) {
let address =
device.rsp.cpu.gpr[rs(opcode) as usize].wrapping_add(imm(opcode) as i16 as i32 as u32);
let mut w = [0; 4];
for (i, item) in w.iter_mut().enumerate() {
*item = device.rsp.mem[(address as usize + i) & 0xFFF];
}
// Optimize for aligned access when possible
let addr_masked = address as usize & 0xFFF;
let value = if addr_masked <= 0xFFC {
u32::from_be_bytes(
device.rsp.mem[addr_masked..addr_masked + 4]
.try_into()
.unwrap(),
)
} else {
let mut value = 0;
for i in 0..4 {
value |= (device.rsp.mem[(address as usize + i) & 0xFFF] as u32) << (8 * (3 - i));
}
value
};
device.rsp.cpu.gpr[rt(opcode) as usize] =
((w[0] as u32) << 24) | ((w[1] as u32) << 16) | ((w[2] as u32) << 8) | (w[3] as u32)
device.rsp.cpu.gpr[rt(opcode) as usize] = value;
}
pub fn lbu(device: &mut device::Device, opcode: u32) {
@ -336,24 +355,44 @@ pub fn lhu(device: &mut device::Device, opcode: u32) {
let address =
device.rsp.cpu.gpr[rs(opcode) as usize].wrapping_add(imm(opcode) as i16 as i32 as u32);
let mut w = [0; 2];
w[0] = device.rsp.mem[address as usize & 0xFFF];
w[1] = device.rsp.mem[(address as usize + 1) & 0xFFF];
// Optimize for aligned access when possible
let addr_masked = address as usize & 0xFFF;
let value = if addr_masked <= 0xFFE {
u16::from_be_bytes(
device.rsp.mem[addr_masked..addr_masked + 2]
.try_into()
.unwrap(),
)
} else {
let w0 = device.rsp.mem[addr_masked];
let w1 = device.rsp.mem[(addr_masked + 1) & 0xFFF];
((w0 as u16) << 8) | w1 as u16
};
device.rsp.cpu.gpr[rt(opcode) as usize] = (((w[0] as u16) << 8) | w[1] as u16) as u32
device.rsp.cpu.gpr[rt(opcode) as usize] = value as u32;
}
pub fn lwu(device: &mut device::Device, opcode: u32) {
let address =
device.rsp.cpu.gpr[rs(opcode) as usize].wrapping_add(imm(opcode) as i16 as i32 as u32);
let mut w = [0; 4];
for (i, item) in w.iter_mut().enumerate() {
*item = device.rsp.mem[(address as usize + i) & 0xFFF];
}
// Optimize for aligned access when possible
let addr_masked = address as usize & 0xFFF;
let value = if addr_masked <= 0xFFC {
u32::from_be_bytes(
device.rsp.mem[addr_masked..addr_masked + 4]
.try_into()
.unwrap(),
)
} else {
let mut value = 0;
for i in 0..4 {
value |= (device.rsp.mem[(address as usize + i) & 0xFFF] as u32) << (8 * (3 - i));
}
value
};
device.rsp.cpu.gpr[rt(opcode) as usize] =
((w[0] as u32) << 24) | ((w[1] as u32) << 16) | ((w[2] as u32) << 8) | (w[3] as u32)
device.rsp.cpu.gpr[rt(opcode) as usize] = value;
}
pub fn sb(device: &mut device::Device, opcode: u32) {
@ -367,18 +406,30 @@ pub fn sh(device: &mut device::Device, opcode: u32) {
let address =
device.rsp.cpu.gpr[rs(opcode) as usize].wrapping_add(imm(opcode) as i16 as i32 as u32);
device.rsp.mem[address as usize & 0xFFF] = (device.rsp.cpu.gpr[rt(opcode) as usize] >> 8) as u8;
device.rsp.mem[(address as usize + 1) & 0xFFF] =
(device.rsp.cpu.gpr[rt(opcode) as usize]) as u8;
let addr_masked = address as usize & 0xFFF;
let value = device.rsp.cpu.gpr[rt(opcode) as usize] as u16;
// Optimize for aligned access when possible
if addr_masked <= 0xFFE {
device.rsp.mem[addr_masked..addr_masked + 2].copy_from_slice(&value.to_be_bytes());
} else {
device.rsp.mem[addr_masked] = (value >> 8) as u8;
device.rsp.mem[(addr_masked + 1) & 0xFFF] = value as u8;
}
}
pub fn sw(device: &mut device::Device, opcode: u32) {
let address =
device.rsp.cpu.gpr[rs(opcode) as usize].wrapping_add(imm(opcode) as i16 as i32 as u32);
for i in 0..4 {
device.rsp.mem[(address as usize + i) & 0xFFF] =
(device.rsp.cpu.gpr[rt(opcode) as usize] >> ((3 - i) * 8)) as u8;
let addr_masked = address as usize & 0xFFF;
let value = device.rsp.cpu.gpr[rt(opcode) as usize];
// Optimize for aligned access when possible
if addr_masked <= 0xFFC {
device.rsp.mem[addr_masked..addr_masked + 4].copy_from_slice(&value.to_be_bytes());
} else {
for i in 0..4 {
device.rsp.mem[(address as usize + i) & 0xFFF] = (value >> ((3 - i) * 8)) as u8;
}
}
}
@ -737,10 +788,21 @@ pub fn llv(device: &mut device::Device, opcode: u32) {
let mut element = velement(opcode);
if element.is_multiple_of(4) {
let mut value = 0;
for i in 0..4 {
value |= (device.rsp.mem[((address + i) & 0xFFF) as usize] as u32) << (8 * (3 - i));
}
// Optimize for aligned access when possible
let addr_masked = address as usize & 0xFFF;
let value = if addr_masked <= 0xFFC {
u32::from_be_bytes(
device.rsp.mem[addr_masked..addr_masked + 4]
.try_into()
.unwrap(),
)
} else {
let mut value = 0;
for i in 0..4 {
value |= (device.rsp.mem[(address as usize + i) & 0xFFF] as u32) << (8 * (3 - i));
}
value
};
modify_vpr32(
&mut device.rsp.cpu.vpr[rt(opcode) as usize],
element / 4,
@ -767,10 +829,21 @@ pub fn ldv(device: &mut device::Device, opcode: u32) {
let mut element = velement(opcode);
if element.is_multiple_of(8) {
let mut value = 0;
for i in 0..8 {
value |= (device.rsp.mem[((address + i) & 0xFFF) as usize] as u64) << (8 * (7 - i));
}
// Optimize for aligned access when possible
let addr_masked = address as usize & 0xFFF;
let value = if addr_masked <= 0xFF8 {
u64::from_be_bytes(
device.rsp.mem[addr_masked..addr_masked + 8]
.try_into()
.unwrap(),
)
} else {
let mut value = 0;
for i in 0..8 {
value |= (device.rsp.mem[(address as usize + i) & 0xFFF] as u64) << (8 * (7 - i));
}
value
};
modify_vpr64(
&mut device.rsp.cpu.vpr[rt(opcode) as usize],
element / 8,
@ -797,10 +870,21 @@ pub fn lqv(device: &mut device::Device, opcode: u32) {
let mut element = velement(opcode);
if element == 0 && address.is_multiple_of(16) {
let mut value = 0;
for i in 0..16 {
value |= (device.rsp.mem[((address + i) & 0xFFF) as usize] as u128) << (8 * (15 - i));
}
// Optimize for aligned 16-byte access
let addr_masked = address as usize & 0xFFF;
let value = if addr_masked <= 0xFF0 {
u128::from_be_bytes(
device.rsp.mem[addr_masked..addr_masked + 16]
.try_into()
.unwrap(),
)
} else {
let mut value = 0;
for i in 0..16 {
value |= (device.rsp.mem[(address as usize + i) & 0xFFF] as u128) << (8 * (15 - i));
}
value
};
modify_vpr128(&mut device.rsp.cpu.vpr[rt(opcode) as usize], value);
} else {
let end = std::cmp::min(16 + element - ((address & 15) as u8), 16);
@ -823,10 +907,21 @@ pub fn lrv(device: &mut device::Device, opcode: u32) {
let mut element = 16u8.wrapping_sub(((address & 15) as u8).wrapping_sub(velement(opcode)));
address &= !15;
if element == 0 {
let mut value = 0;
for i in 0..16 {
value |= (device.rsp.mem[((address + i) & 0xFFF) as usize] as u128) << (8 * (15 - i));
}
// Optimize for aligned 16-byte access
let addr_masked = address as usize & 0xFFF;
let value = if addr_masked <= 0xFF0 {
u128::from_be_bytes(
device.rsp.mem[addr_masked..addr_masked + 16]
.try_into()
.unwrap(),
)
} else {
let mut value = 0;
for i in 0..16 {
value |= (device.rsp.mem[(address as usize + i) & 0xFFF] as u128) << (8 * (15 - i));
}
value
};
modify_vpr128(&mut device.rsp.cpu.vpr[rt(opcode) as usize], value);
} else {
while element < 16 {