From: Chris Duncan <chris@zoso.dev>
Date: Tue, 14 Jan 2025 16:21:51 +0000 (-0800)
Subject: Replace more scalar addition with vector addition.
X-Git-Url: https://zoso.dev/?a=commitdiff_plain;h=b299692f121ca1075f8a3a7f18d5e19aabd145aa;p=nano-pow.git

Replace more scalar addition with vector addition.
---

diff --git a/src/shaders/compute.wgsl b/src/shaders/compute.wgsl
index 3b9c769..53610b6 100644
--- a/src/shaders/compute.wgsl
+++ b/src/shaders/compute.wgsl
@@ -427,11 +427,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v31 = xor0;
 
 	// c = c + d
-	o0 = v22 + v30;
-	o1 = v23 + v31;
-	o1 = o1 + select(0u, 1u, o0 < v22);
-	v22 = o0;
-	v23 = o1;
+	v_2223.x = v22;
+	v_2223.y = v23;
+	v_3031.x = v30;
+	v_3031.y = v31;
+	v_2223 = v_2223 + v_3031 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_2223.x + v_3031.x < v_2223.x);
+	v22 = v_2223.x;
+	v23 = v_2223.y;
+	v30 = v_3031.x;
+	v31 = v_3031.y;
 
 	// b = rotr64(b ^ c, 24)
 	xor0 = v14 ^ v22;
@@ -467,11 +471,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v31 = (xor1 >> 16u) ^ (xor0 << 16u);
 
 	// c = c + d
-	o0 = v22 + v30;
-	o1 = v23 + v31;
-	o1 = o1 + select(0u, 1u, o0 < v22);
-	v22 = o0;
-	v23 = o1;
+	v_2223.x = v22;
+	v_2223.y = v23;
+	v_3031.x = v30;
+	v_3031.y = v31;
+	v_2223 = v_2223 + v_3031 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_2223.x + v_3031.x < v_2223.x);
+	v22 = v_2223.x;
+	v23 = v_2223.y;
+	v30 = v_3031.x;
+	v31 = v_3031.y;
 
 	// b = rotr64(b ^ c, 63)
 	xor0 = v14 ^ v22;
@@ -1124,11 +1132,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v31 = xor0;
 
 	// c = c + d
-	o0 = v22 + v30;
-	o1 = v23 + v31;
-	o1 = o1 + select(0u, 1u, o0 < v22);
-	v22 = o0;
-	v23 = o1;
+	v_2223.x = v22;
+	v_2223.y = v23;
+	v_3031.x = v30;
+	v_3031.y = v31;
+	v_2223 = v_2223 + v_3031 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_2223.x + v_3031.x < v_2223.x);
+	v22 = v_2223.x;
+	v23 = v_2223.y;
+	v30 = v_3031.x;
+	v31 = v_3031.y;
 
 	// b = rotr64(b ^ c, 24)
 	xor0 = v14 ^ v22;
@@ -1164,11 +1176,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v31 = (xor1 >> 16u) ^ (xor0 << 16u);
 
 	// c = c + d
-	o0 = v22 + v30;
-	o1 = v23 + v31;
-	o1 = o1 + select(0u, 1u, o0 < v22);
-	v22 = o0;
-	v23 = o1;
+	v_2223.x = v22;
+	v_2223.y = v23;
+	v_3031.x = v30;
+	v_3031.y = v31;
+	v_2223 = v_2223 + v_3031 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_2223.x + v_3031.x < v_2223.x);
+	v22 = v_2223.x;
+	v23 = v_2223.y;
+	v30 = v_3031.x;
+	v31 = v_3031.y;
 
 	// b = rotr64(b ^ c, 63)
 	xor0 = v14 ^ v22;
@@ -1806,11 +1822,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v31 = xor0;
 
 	// c = c + d
-	o0 = v22 + v30;
-	o1 = v23 + v31;
-	o1 = o1 + select(0u, 1u, o0 < v22);
-	v22 = o0;
-	v23 = o1;
+	v_2223.x = v22;
+	v_2223.y = v23;
+	v_3031.x = v30;
+	v_3031.y = v31;
+	v_2223 = v_2223 + v_3031 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_2223.x + v_3031.x < v_2223.x);
+	v22 = v_2223.x;
+	v23 = v_2223.y;
+	v30 = v_3031.x;
+	v31 = v_3031.y;
 
 	// b = rotr64(b ^ c, 24)
 	xor0 = v14 ^ v22;
@@ -1846,11 +1866,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v31 = (xor1 >> 16u) ^ (xor0 << 16u);
 
 	// c = c + d
-	o0 = v22 + v30;
-	o1 = v23 + v31;
-	o1 = o1 + select(0u, 1u, o0 < v22);
-	v22 = o0;
-	v23 = o1;
+	v_2223.x = v22;
+	v_2223.y = v23;
+	v_3031.x = v30;
+	v_3031.y = v31;
+	v_2223 = v_2223 + v_3031 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_2223.x + v_3031.x < v_2223.x);
+	v22 = v_2223.x;
+	v23 = v_2223.y;
+	v30 = v_3031.x;
+	v31 = v_3031.y;
 
 	// b = rotr64(b ^ c, 63)
 	xor0 = v14 ^ v22;
@@ -2491,11 +2515,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v31 = xor0;
 
 	// c = c + d
-	o0 = v22 + v30;
-	o1 = v23 + v31;
-	o1 = o1 + select(0u, 1u, o0 < v22);
-	v22 = o0;
-	v23 = o1;
+	v_2223.x = v22;
+	v_2223.y = v23;
+	v_3031.x = v30;
+	v_3031.y = v31;
+	v_2223 = v_2223 + v_3031 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_2223.x + v_3031.x < v_2223.x);
+	v22 = v_2223.x;
+	v23 = v_2223.y;
+	v30 = v_3031.x;
+	v31 = v_3031.y;
 
 	// b = rotr64(b ^ c, 24)
 	xor0 = v14 ^ v22;
@@ -2531,11 +2559,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v31 = (xor1 >> 16u) ^ (xor0 << 16u);
 
 	// c = c + d
-	o0 = v22 + v30;
-	o1 = v23 + v31;
-	o1 = o1 + select(0u, 1u, o0 < v22);
-	v22 = o0;
-	v23 = o1;
+	v_2223.x = v22;
+	v_2223.y = v23;
+	v_3031.x = v30;
+	v_3031.y = v31;
+	v_2223 = v_2223 + v_3031 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_2223.x + v_3031.x < v_2223.x);
+	v22 = v_2223.x;
+	v23 = v_2223.y;
+	v30 = v_3031.x;
+	v31 = v_3031.y;
 
 	// b = rotr64(b ^ c, 63)
 	xor0 = v14 ^ v22;
@@ -3173,11 +3205,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v31 = xor0;
 
 	// c = c + d
-	o0 = v22 + v30;
-	o1 = v23 + v31;
-	o1 = o1 + select(0u, 1u, o0 < v22);
-	v22 = o0;
-	v23 = o1;
+	v_2223.x = v22;
+	v_2223.y = v23;
+	v_3031.x = v30;
+	v_3031.y = v31;
+	v_2223 = v_2223 + v_3031 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_2223.x + v_3031.x < v_2223.x);
+	v22 = v_2223.x;
+	v23 = v_2223.y;
+	v30 = v_3031.x;
+	v31 = v_3031.y;
 
 	// b = rotr64(b ^ c, 24)
 	xor0 = v14 ^ v22;
@@ -3213,11 +3249,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v31 = (xor1 >> 16u) ^ (xor0 << 16u);
 
 	// c = c + d
-	o0 = v22 + v30;
-	o1 = v23 + v31;
-	o1 = o1 + select(0u, 1u, o0 < v22);
-	v22 = o0;
-	v23 = o1;
+	v_2223.x = v22;
+	v_2223.y = v23;
+	v_3031.x = v30;
+	v_3031.y = v31;
+	v_2223 = v_2223 + v_3031 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_2223.x + v_3031.x < v_2223.x);
+	v22 = v_2223.x;
+	v23 = v_2223.y;
+	v30 = v_3031.x;
+	v31 = v_3031.y;
 
 	// b = rotr64(b ^ c, 63)
 	xor0 = v14 ^ v22;
@@ -3861,11 +3901,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v31 = xor0;
 
 	// c = c + d
-	o0 = v22 + v30;
-	o1 = v23 + v31;
-	o1 = o1 + select(0u, 1u, o0 < v22);
-	v22 = o0;
-	v23 = o1;
+	v_2223.x = v22;
+	v_2223.y = v23;
+	v_3031.x = v30;
+	v_3031.y = v31;
+	v_2223 = v_2223 + v_3031 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_2223.x + v_3031.x < v_2223.x);
+	v22 = v_2223.x;
+	v23 = v_2223.y;
+	v30 = v_3031.x;
+	v31 = v_3031.y;
 
 	// b = rotr64(b ^ c, 24)
 	xor0 = v14 ^ v22;
@@ -3898,11 +3942,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v31 = (xor1 >> 16u) ^ (xor0 << 16u);
 
 	// c = c + d
-	o0 = v22 + v30;
-	o1 = v23 + v31;
-	o1 = o1 + select(0u, 1u, o0 < v22);
-	v22 = o0;
-	v23 = o1;
+	v_2223.x = v22;
+	v_2223.y = v23;
+	v_3031.x = v30;
+	v_3031.y = v31;
+	v_2223 = v_2223 + v_3031 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_2223.x + v_3031.x < v_2223.x);
+	v22 = v_2223.x;
+	v23 = v_2223.y;
+	v30 = v_3031.x;
+	v31 = v_3031.y;
 
 	// b = rotr64(b ^ c, 63)
 	xor0 = v14 ^ v22;
@@ -4546,11 +4594,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v31 = xor0;
 
 	// c = c + d
-	o0 = v22 + v30;
-	o1 = v23 + v31;
-	o1 = o1 + select(0u, 1u, o0 < v22);
-	v22 = o0;
-	v23 = o1;
+	v_2223.x = v22;
+	v_2223.y = v23;
+	v_3031.x = v30;
+	v_3031.y = v31;
+	v_2223 = v_2223 + v_3031 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_2223.x + v_3031.x < v_2223.x);
+	v22 = v_2223.x;
+	v23 = v_2223.y;
+	v30 = v_3031.x;
+	v31 = v_3031.y;
 
 	// b = rotr64(b ^ c, 24)
 	xor0 = v14 ^ v22;
@@ -4586,11 +4638,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v31 = (xor1 >> 16u) ^ (xor0 << 16u);
 
 	// c = c + d
-	o0 = v22 + v30;
-	o1 = v23 + v31;
-	o1 = o1 + select(0u, 1u, o0 < v22);
-	v22 = o0;
-	v23 = o1;
+	v_2223.x = v22;
+	v_2223.y = v23;
+	v_3031.x = v30;
+	v_3031.y = v31;
+	v_2223 = v_2223 + v_3031 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_2223.x + v_3031.x < v_2223.x);
+	v22 = v_2223.x;
+	v23 = v_2223.y;
+	v30 = v_3031.x;
+	v31 = v_3031.y;
 
 	// b = rotr64(b ^ c, 63)
 	xor0 = v14 ^ v22;
@@ -5231,11 +5287,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v31 = xor0;
 
 	// c = c + d
-	o0 = v22 + v30;
-	o1 = v23 + v31;
-	o1 = o1 + select(0u, 1u, o0 < v22);
-	v22 = o0;
-	v23 = o1;
+	v_2223.x = v22;
+	v_2223.y = v23;
+	v_3031.x = v30;
+	v_3031.y = v31;
+	v_2223 = v_2223 + v_3031 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_2223.x + v_3031.x < v_2223.x);
+	v22 = v_2223.x;
+	v23 = v_2223.y;
+	v30 = v_3031.x;
+	v31 = v_3031.y;
 
 	// b = rotr64(b ^ c, 24)
 	xor0 = v14 ^ v22;
@@ -5271,11 +5331,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v31 = (xor1 >> 16u) ^ (xor0 << 16u);
 
 	// c = c + d
-	o0 = v22 + v30;
-	o1 = v23 + v31;
-	o1 = o1 + select(0u, 1u, o0 < v22);
-	v22 = o0;
-	v23 = o1;
+	v_2223.x = v22;
+	v_2223.y = v23;
+	v_3031.x = v30;
+	v_3031.y = v31;
+	v_2223 = v_2223 + v_3031 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_2223.x + v_3031.x < v_2223.x);
+	v22 = v_2223.x;
+	v23 = v_2223.y;
+	v30 = v_3031.x;
+	v31 = v_3031.y;
 
 	// b = rotr64(b ^ c, 63)
 	xor0 = v14 ^ v22;
@@ -5916,11 +5980,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v31 = xor0;
 
 	// c = c + d
-	o0 = v22 + v30;
-	o1 = v23 + v31;
-	o1 = o1 + select(0u, 1u, o0 < v22);
-	v22 = o0;
-	v23 = o1;
+	v_2223.x = v22;
+	v_2223.y = v23;
+	v_3031.x = v30;
+	v_3031.y = v31;
+	v_2223 = v_2223 + v_3031 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_2223.x + v_3031.x < v_2223.x);
+	v22 = v_2223.x;
+	v23 = v_2223.y;
+	v30 = v_3031.x;
+	v31 = v_3031.y;
 
 	// b = rotr64(b ^ c, 24)
 	xor0 = v14 ^ v22;
@@ -5956,11 +6024,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v31 = (xor1 >> 16u) ^ (xor0 << 16u);
 
 	// c = c + d
-	o0 = v22 + v30;
-	o1 = v23 + v31;
-	o1 = o1 + select(0u, 1u, o0 < v22);
-	v22 = o0;
-	v23 = o1;
+	v_2223.x = v22;
+	v_2223.y = v23;
+	v_3031.x = v30;
+	v_3031.y = v31;
+	v_2223 = v_2223 + v_3031 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_2223.x + v_3031.x < v_2223.x);
+	v22 = v_2223.x;
+	v23 = v_2223.y;
+	v30 = v_3031.x;
+	v31 = v_3031.y;
 
 	// b = rotr64(b ^ c, 63)
 	xor0 = v14 ^ v22;
@@ -6598,11 +6670,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v31 = xor0;
 
 	// c = c + d
-	o0 = v22 + v30;
-	o1 = v23 + v31;
-	o1 = o1 + select(0u, 1u, o0 < v22);
-	v22 = o0;
-	v23 = o1;
+	v_2223.x = v22;
+	v_2223.y = v23;
+	v_3031.x = v30;
+	v_3031.y = v31;
+	v_2223 = v_2223 + v_3031 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_2223.x + v_3031.x < v_2223.x);
+	v22 = v_2223.x;
+	v23 = v_2223.y;
+	v30 = v_3031.x;
+	v31 = v_3031.y;
 
 	// b = rotr64(b ^ c, 24)
 	xor0 = v14 ^ v22;
@@ -6638,11 +6714,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v31 = (xor1 >> 16u) ^ (xor0 << 16u);
 
 	// c = c + d
-	o0 = v22 + v30;
-	o1 = v23 + v31;
-	o1 = o1 + select(0u, 1u, o0 < v22);
-	v22 = o0;
-	v23 = o1;
+	v_2223.x = v22;
+	v_2223.y = v23;
+	v_3031.x = v30;
+	v_3031.y = v31;
+	v_2223 = v_2223 + v_3031 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_2223.x + v_3031.x < v_2223.x);
+	v22 = v_2223.x;
+	v23 = v_2223.y;
+	v30 = v_3031.x;
+	v31 = v_3031.y;
 
 	// b = rotr64(b ^ c, 63)
 	xor0 = v14 ^ v22;
@@ -7277,11 +7357,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v31 = xor0;
 
 	// c = c + d
-	o0 = v22 + v30;
-	o1 = v23 + v31;
-	o1 = o1 + select(0u, 1u, o0 < v22);
-	v22 = o0;
-	v23 = o1;
+	v_2223.x = v22;
+	v_2223.y = v23;
+	v_3031.x = v30;
+	v_3031.y = v31;
+	v_2223 = v_2223 + v_3031 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_2223.x + v_3031.x < v_2223.x);
+	v22 = v_2223.x;
+	v23 = v_2223.y;
+	v30 = v_3031.x;
+	v31 = v_3031.y;
 
 	// b = rotr64(b ^ c, 24)
 	xor0 = v14 ^ v22;
@@ -7317,11 +7401,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v31 = (xor1 >> 16u) ^ (xor0 << 16u);
 
 	// c = c + d
-	o0 = v22 + v30;
-	o1 = v23 + v31;
-	o1 = o1 + select(0u, 1u, o0 < v22);
-	v22 = o0;
-	v23 = o1;
+	v_2223.x = v22;
+	v_2223.y = v23;
+	v_3031.x = v30;
+	v_3031.y = v31;
+	v_2223 = v_2223 + v_3031 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_2223.x + v_3031.x < v_2223.x);
+	v22 = v_2223.x;
+	v23 = v_2223.y;
+	v30 = v_3031.x;
+	v31 = v_3031.y;
 
 	// b = rotr64(b ^ c, 63)
 	xor0 = v14 ^ v22;
@@ -7974,11 +8062,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v31 = xor0;
 
 	// c = c + d
-	o0 = v22 + v30;
-	o1 = v23 + v31;
-	o1 = o1 + select(0u, 1u, o0 < v22);
-	v22 = o0;
-	v23 = o1;
+	v_2223.x = v22;
+	v_2223.y = v23;
+	v_3031.x = v30;
+	v_3031.y = v31;
+	v_2223 = v_2223 + v_3031 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_2223.x + v_3031.x < v_2223.x);
+	v22 = v_2223.x;
+	v23 = v_2223.y;
+	v30 = v_3031.x;
+	v31 = v_3031.y;
 
 	// b = rotr64(b ^ c, 24)
 	xor0 = v14 ^ v22;
@@ -8014,11 +8106,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v31 = (xor1 >> 16u) ^ (xor0 << 16u);
 
 	// c = c + d
-	o0 = v22 + v30;
-	o1 = v23 + v31;
-	o1 = o1 + select(0u, 1u, o0 < v22);
-	v22 = o0;
-	v23 = o1;
+	v_2223.x = v22;
+	v_2223.y = v23;
+	v_3031.x = v30;
+	v_3031.y = v31;
+	v_2223 = v_2223 + v_3031 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_2223.x + v_3031.x < v_2223.x);
+	v22 = v_2223.x;
+	v23 = v_2223.y;
+	v30 = v_3031.x;
+	v31 = v_3031.y;
 
 	// b = rotr64(b ^ c, 63)
 	xor0 = v14 ^ v22;