From 080b1aa04c9f039fbcd087088adc61e1aee8acd8 Mon Sep 17 00:00:00 2001
From: Chris Duncan <chris@zoso.dev>
Date: Tue, 14 Jan 2025 08:16:14 -0800
Subject: [PATCH] Replace more scalar addition with vector addition.

---
 src/shaders/compute.wgsl | 336 +++++++++++++++++++++++++--------------
 1 file changed, 216 insertions(+), 120 deletions(-)

diff --git a/src/shaders/compute.wgsl b/src/shaders/compute.wgsl
index ea92d9a..ec65fff 100644
--- a/src/shaders/compute.wgsl
+++ b/src/shaders/compute.wgsl
@@ -331,11 +331,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v29 = xor0;
 
 	// c = c + d
-	o0 = v20 + v28;
-	o1 = v21 + v29;
-	o1 = o1 + select(0u, 1u, o0 < v20);
-	v20 = o0;
-	v21 = o1;
+	v_2021.x = v20;
+	v_2021.y = v21;
+	v_2829.x = v28;
+	v_2829.y = v29;
+	v_2021 = v_2021 + v_2829 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_2021.x + v_2829.x < v_2021.x);
+	v20 = v_2021.x;
+	v21 = v_2021.y;
+	v28 = v_2829.x;
+	v29 = v_2829.y;
 
 	// b = rotr64(b ^ c, 24)
 	xor0 = v12 ^ v20;
@@ -371,11 +375,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v29 = (xor1 >> 16u) ^ (xor0 << 16u);
 
 	// c = c + d
-	o0 = v20 + v28;
-	o1 = v21 + v29;
-	o1 = o1 + select(0u, 1u, o0 < v20);
-	v20 = o0;
-	v21 = o1;
+	v_2021.x = v20;
+	v_2021.y = v21;
+	v_2829.x = v28;
+	v_2829.y = v29;
+	v_2021 = v_2021 + v_2829 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_2021.x + v_2829.x < v_2021.x);
+	v20 = v_2021.x;
+	v21 = v_2021.y;
+	v28 = v_2829.x;
+	v29 = v_2829.y;
 
 	// b = rotr64(b ^ c, 63)
 	xor0 = v12 ^ v20;
@@ -1012,11 +1020,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v29 = xor0;
 
 	// c = c + d
-	o0 = v20 + v28;
-	o1 = v21 + v29;
-	o1 = o1 + select(0u, 1u, o0 < v20);
-	v20 = o0;
-	v21 = o1;
+	v_2021.x = v20;
+	v_2021.y = v21;
+	v_2829.x = v28;
+	v_2829.y = v29;
+	v_2021 = v_2021 + v_2829 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_2021.x + v_2829.x < v_2021.x);
+	v20 = v_2021.x;
+	v21 = v_2021.y;
+	v28 = v_2829.x;
+	v29 = v_2829.y;
 
 	// b = rotr64(b ^ c, 24)
 	xor0 = v12 ^ v20;
@@ -1052,11 +1064,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v29 = (xor1 >> 16u) ^ (xor0 << 16u);
 
 	// c = c + d
-	o0 = v20 + v28;
-	o1 = v21 + v29;
-	o1 = o1 + select(0u, 1u, o0 < v20);
-	v20 = o0;
-	v21 = o1;
+	v_2021.x = v20;
+	v_2021.y = v21;
+	v_2829.x = v28;
+	v_2829.y = v29;
+	v_2021 = v_2021 + v_2829 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_2021.x + v_2829.x < v_2021.x);
+	v20 = v_2021.x;
+	v21 = v_2021.y;
+	v28 = v_2829.x;
+	v29 = v_2829.y;
 
 	// b = rotr64(b ^ c, 63)
 	xor0 = v12 ^ v20;
@@ -1681,11 +1697,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v29 = xor0;
 
 	// c = c + d
-	o0 = v20 + v28;
-	o1 = v21 + v29;
-	o1 = o1 + select(0u, 1u, o0 < v20);
-	v20 = o0;
-	v21 = o1;
+	v_2021.x = v20;
+	v_2021.y = v21;
+	v_2829.x = v28;
+	v_2829.y = v29;
+	v_2021 = v_2021 + v_2829 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_2021.x + v_2829.x < v_2021.x);
+	v20 = v_2021.x;
+	v21 = v_2021.y;
+	v28 = v_2829.x;
+	v29 = v_2829.y;
 
 	// b = rotr64(b ^ c, 24)
 	xor0 = v12 ^ v20;
@@ -1718,11 +1738,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v29 = (xor1 >> 16u) ^ (xor0 << 16u);
 
 	// c = c + d
-	o0 = v20 + v28;
-	o1 = v21 + v29;
-	o1 = o1 + select(0u, 1u, o0 < v20);
-	v20 = o0;
-	v21 = o1;
+	v_2021.x = v20;
+	v_2021.y = v21;
+	v_2829.x = v28;
+	v_2829.y = v29;
+	v_2021 = v_2021 + v_2829 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_2021.x + v_2829.x < v_2021.x);
+	v20 = v_2021.x;
+	v21 = v_2021.y;
+	v28 = v_2829.x;
+	v29 = v_2829.y;
 
 	// b = rotr64(b ^ c, 63)
 	xor0 = v12 ^ v20;
@@ -2347,11 +2371,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v29 = xor0;
 
 	// c = c + d
-	o0 = v20 + v28;
-	o1 = v21 + v29;
-	o1 = o1 + select(0u, 1u, o0 < v20);
-	v20 = o0;
-	v21 = o1;
+	v_2021.x = v20;
+	v_2021.y = v21;
+	v_2829.x = v28;
+	v_2829.y = v29;
+	v_2021 = v_2021 + v_2829 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_2021.x + v_2829.x < v_2021.x);
+	v20 = v_2021.x;
+	v21 = v_2021.y;
+	v28 = v_2829.x;
+	v29 = v_2829.y;
 
 	// b = rotr64(b ^ c, 24)
 	xor0 = v12 ^ v20;
@@ -2387,11 +2415,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v29 = (xor1 >> 16u) ^ (xor0 << 16u);
 
 	// c = c + d
-	o0 = v20 + v28;
-	o1 = v21 + v29;
-	o1 = o1 + select(0u, 1u, o0 < v20);
-	v20 = o0;
-	v21 = o1;
+	v_2021.x = v20;
+	v_2021.y = v21;
+	v_2829.x = v28;
+	v_2829.y = v29;
+	v_2021 = v_2021 + v_2829 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_2021.x + v_2829.x < v_2021.x);
+	v20 = v_2021.x;
+	v21 = v_2021.y;
+	v28 = v_2829.x;
+	v29 = v_2829.y;
 
 	// b = rotr64(b ^ c, 63)
 	xor0 = v12 ^ v20;
@@ -3016,11 +3048,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v29 = xor0;
 
 	// c = c + d
-	o0 = v20 + v28;
-	o1 = v21 + v29;
-	o1 = o1 + select(0u, 1u, o0 < v20);
-	v20 = o0;
-	v21 = o1;
+	v_2021.x = v20;
+	v_2021.y = v21;
+	v_2829.x = v28;
+	v_2829.y = v29;
+	v_2021 = v_2021 + v_2829 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_2021.x + v_2829.x < v_2021.x);
+	v20 = v_2021.x;
+	v21 = v_2021.y;
+	v28 = v_2829.x;
+	v29 = v_2829.y;
 
 	// b = rotr64(b ^ c, 24)
 	xor0 = v12 ^ v20;
@@ -3053,11 +3089,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v29 = (xor1 >> 16u) ^ (xor0 << 16u);
 
 	// c = c + d
-	o0 = v20 + v28;
-	o1 = v21 + v29;
-	o1 = o1 + select(0u, 1u, o0 < v20);
-	v20 = o0;
-	v21 = o1;
+	v_2021.x = v20;
+	v_2021.y = v21;
+	v_2829.x = v28;
+	v_2829.y = v29;
+	v_2021 = v_2021 + v_2829 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_2021.x + v_2829.x < v_2021.x);
+	v20 = v_2021.x;
+	v21 = v_2021.y;
+	v28 = v_2829.x;
+	v29 = v_2829.y;
 
 	// b = rotr64(b ^ c, 63)
 	xor0 = v12 ^ v20;
@@ -3685,11 +3725,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v29 = xor0;
 
 	// c = c + d
-	o0 = v20 + v28;
-	o1 = v21 + v29;
-	o1 = o1 + select(0u, 1u, o0 < v20);
-	v20 = o0;
-	v21 = o1;
+	v_2021.x = v20;
+	v_2021.y = v21;
+	v_2829.x = v28;
+	v_2829.y = v29;
+	v_2021 = v_2021 + v_2829 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_2021.x + v_2829.x < v_2021.x);
+	v20 = v_2021.x;
+	v21 = v_2021.y;
+	v28 = v_2829.x;
+	v29 = v_2829.y;
 
 	// b = rotr64(b ^ c, 24)
 	xor0 = v12 ^ v20;
@@ -3725,11 +3769,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v29 = (xor1 >> 16u) ^ (xor0 << 16u);
 
 	// c = c + d
-	o0 = v20 + v28;
-	o1 = v21 + v29;
-	o1 = o1 + select(0u, 1u, o0 < v20);
-	v20 = o0;
-	v21 = o1;
+	v_2021.x = v20;
+	v_2021.y = v21;
+	v_2829.x = v28;
+	v_2829.y = v29;
+	v_2021 = v_2021 + v_2829 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_2021.x + v_2829.x < v_2021.x);
+	v20 = v_2021.x;
+	v21 = v_2021.y;
+	v28 = v_2829.x;
+	v29 = v_2829.y;
 
 	// b = rotr64(b ^ c, 63)
 	xor0 = v12 ^ v20;
@@ -4357,11 +4405,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v29 = xor0;
 
 	// c = c + d
-	o0 = v20 + v28;
-	o1 = v21 + v29;
-	o1 = o1 + select(0u, 1u, o0 < v20);
-	v20 = o0;
-	v21 = o1;
+	v_2021.x = v20;
+	v_2021.y = v21;
+	v_2829.x = v28;
+	v_2829.y = v29;
+	v_2021 = v_2021 + v_2829 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_2021.x + v_2829.x < v_2021.x);
+	v20 = v_2021.x;
+	v21 = v_2021.y;
+	v28 = v_2829.x;
+	v29 = v_2829.y;
 
 	// b = rotr64(b ^ c, 24)
 	xor0 = v12 ^ v20;
@@ -4397,11 +4449,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v29 = (xor1 >> 16u) ^ (xor0 << 16u);
 
 	// c = c + d
-	o0 = v20 + v28;
-	o1 = v21 + v29;
-	o1 = o1 + select(0u, 1u, o0 < v20);
-	v20 = o0;
-	v21 = o1;
+	v_2021.x = v20;
+	v_2021.y = v21;
+	v_2829.x = v28;
+	v_2829.y = v29;
+	v_2021 = v_2021 + v_2829 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_2021.x + v_2829.x < v_2021.x);
+	v20 = v_2021.x;
+	v21 = v_2021.y;
+	v28 = v_2829.x;
+	v29 = v_2829.y;
 
 	// b = rotr64(b ^ c, 63)
 	xor0 = v12 ^ v20;
@@ -5029,11 +5085,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v29 = xor0;
 
 	// c = c + d
-	o0 = v20 + v28;
-	o1 = v21 + v29;
-	o1 = o1 + select(0u, 1u, o0 < v20);
-	v20 = o0;
-	v21 = o1;
+	v_2021.x = v20;
+	v_2021.y = v21;
+	v_2829.x = v28;
+	v_2829.y = v29;
+	v_2021 = v_2021 + v_2829 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_2021.x + v_2829.x < v_2021.x);
+	v20 = v_2021.x;
+	v21 = v_2021.y;
+	v28 = v_2829.x;
+	v29 = v_2829.y;
 
 	// b = rotr64(b ^ c, 24)
 	xor0 = v12 ^ v20;
@@ -5066,11 +5126,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v29 = (xor1 >> 16u) ^ (xor0 << 16u);
 
 	// c = c + d
-	o0 = v20 + v28;
-	o1 = v21 + v29;
-	o1 = o1 + select(0u, 1u, o0 < v20);
-	v20 = o0;
-	v21 = o1;
+	v_2021.x = v20;
+	v_2021.y = v21;
+	v_2829.x = v28;
+	v_2829.y = v29;
+	v_2021 = v_2021 + v_2829 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_2021.x + v_2829.x < v_2021.x);
+	v20 = v_2021.x;
+	v21 = v_2021.y;
+	v28 = v_2829.x;
+	v29 = v_2829.y;
 
 	// b = rotr64(b ^ c, 63)
 	xor0 = v12 ^ v20;
@@ -5698,11 +5762,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v29 = xor0;
 
 	// c = c + d
-	o0 = v20 + v28;
-	o1 = v21 + v29;
-	o1 = o1 + select(0u, 1u, o0 < v20);
-	v20 = o0;
-	v21 = o1;
+	v_2021.x = v20;
+	v_2021.y = v21;
+	v_2829.x = v28;
+	v_2829.y = v29;
+	v_2021 = v_2021 + v_2829 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_2021.x + v_2829.x < v_2021.x);
+	v20 = v_2021.x;
+	v21 = v_2021.y;
+	v28 = v_2829.x;
+	v29 = v_2829.y;
 
 	// b = rotr64(b ^ c, 24)
 	xor0 = v12 ^ v20;
@@ -5735,11 +5803,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v29 = (xor1 >> 16u) ^ (xor0 << 16u);
 
 	// c = c + d
-	o0 = v20 + v28;
-	o1 = v21 + v29;
-	o1 = o1 + select(0u, 1u, o0 < v20);
-	v20 = o0;
-	v21 = o1;
+	v_2021.x = v20;
+	v_2021.y = v21;
+	v_2829.x = v28;
+	v_2829.y = v29;
+	v_2021 = v_2021 + v_2829 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_2021.x + v_2829.x < v_2021.x);
+	v20 = v_2021.x;
+	v21 = v_2021.y;
+	v28 = v_2829.x;
+	v29 = v_2829.y;
 
 	// b = rotr64(b ^ c, 63)
 	xor0 = v12 ^ v20;
@@ -6361,11 +6433,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v29 = xor0;
 
 	// c = c + d
-	o0 = v20 + v28;
-	o1 = v21 + v29;
-	o1 = o1 + select(0u, 1u, o0 < v20);
-	v20 = o0;
-	v21 = o1;
+	v_2021.x = v20;
+	v_2021.y = v21;
+	v_2829.x = v28;
+	v_2829.y = v29;
+	v_2021 = v_2021 + v_2829 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_2021.x + v_2829.x < v_2021.x);
+	v20 = v_2021.x;
+	v21 = v_2021.y;
+	v28 = v_2829.x;
+	v29 = v_2829.y;
 
 	// b = rotr64(b ^ c, 24)
 	xor0 = v12 ^ v20;
@@ -6401,11 +6477,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v29 = (xor1 >> 16u) ^ (xor0 << 16u);
 
 	// c = c + d
-	o0 = v20 + v28;
-	o1 = v21 + v29;
-	o1 = o1 + select(0u, 1u, o0 < v20);
-	v20 = o0;
-	v21 = o1;
+	v_2021.x = v20;
+	v_2021.y = v21;
+	v_2829.x = v28;
+	v_2829.y = v29;
+	v_2021 = v_2021 + v_2829 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_2021.x + v_2829.x < v_2021.x);
+	v20 = v_2021.x;
+	v21 = v_2021.y;
+	v28 = v_2829.x;
+	v29 = v_2829.y;
 
 	// b = rotr64(b ^ c, 63)
 	xor0 = v12 ^ v20;
@@ -7021,11 +7101,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v29 = xor0;
 
 	// c = c + d
-	o0 = v20 + v28;
-	o1 = v21 + v29;
-	o1 = o1 + select(0u, 1u, o0 < v20);
-	v20 = o0;
-	v21 = o1;
+	v_2021.x = v20;
+	v_2021.y = v21;
+	v_2829.x = v28;
+	v_2829.y = v29;
+	v_2021 = v_2021 + v_2829 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_2021.x + v_2829.x < v_2021.x);
+	v20 = v_2021.x;
+	v21 = v_2021.y;
+	v28 = v_2829.x;
+	v29 = v_2829.y;
 
 	// b = rotr64(b ^ c, 24)
 	xor0 = v12 ^ v20;
@@ -7061,11 +7145,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v29 = (xor1 >> 16u) ^ (xor0 << 16u);
 
 	// c = c + d
-	o0 = v20 + v28;
-	o1 = v21 + v29;
-	o1 = o1 + select(0u, 1u, o0 < v20);
-	v20 = o0;
-	v21 = o1;
+	v_2021.x = v20;
+	v_2021.y = v21;
+	v_2829.x = v28;
+	v_2829.y = v29;
+	v_2021 = v_2021 + v_2829 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_2021.x + v_2829.x < v_2021.x);
+	v20 = v_2021.x;
+	v21 = v_2021.y;
+	v28 = v_2829.x;
+	v29 = v_2829.y;
 
 	// b = rotr64(b ^ c, 63)
 	xor0 = v12 ^ v20;
@@ -7702,11 +7790,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v29 = xor0;
 
 	// c = c + d
-	o0 = v20 + v28;
-	o1 = v21 + v29;
-	o1 = o1 + select(0u, 1u, o0 < v20);
-	v20 = o0;
-	v21 = o1;
+	v_2021.x = v20;
+	v_2021.y = v21;
+	v_2829.x = v28;
+	v_2829.y = v29;
+	v_2021 = v_2021 + v_2829 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_2021.x + v_2829.x < v_2021.x);
+	v20 = v_2021.x;
+	v21 = v_2021.y;
+	v28 = v_2829.x;
+	v29 = v_2829.y;
 
 	// b = rotr64(b ^ c, 24)
 	xor0 = v12 ^ v20;
@@ -7742,11 +7834,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v29 = (xor1 >> 16u) ^ (xor0 << 16u);
 
 	// c = c + d
-	o0 = v20 + v28;
-	o1 = v21 + v29;
-	o1 = o1 + select(0u, 1u, o0 < v20);
-	v20 = o0;
-	v21 = o1;
+	v_2021.x = v20;
+	v_2021.y = v21;
+	v_2829.x = v28;
+	v_2829.y = v29;
+	v_2021 = v_2021 + v_2829 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_2021.x + v_2829.x < v_2021.x);
+	v20 = v_2021.x;
+	v21 = v_2021.y;
+	v28 = v_2829.x;
+	v29 = v_2829.y;
 
 	// b = rotr64(b ^ c, 63)
 	xor0 = v12 ^ v20;
-- 
2.34.1