From: Chris Duncan <chris@zoso.dev>
Date: Tue, 14 Jan 2025 16:28:14 +0000 (-0800)
Subject: Replace more scalar addition with vector addition.
X-Git-Tag: v2.0.0~133
X-Git-Url: https://zoso.dev/?a=commitdiff_plain;h=9410edf379d49716fe2b2be4712d7c2a106b015a;p=nano-pow.git

Replace more scalar addition with vector addition.
---

diff --git a/src/shaders/compute.wgsl b/src/shaders/compute.wgsl
index ff73ed2..c3fd9de 100644
--- a/src/shaders/compute.wgsl
+++ b/src/shaders/compute.wgsl
@@ -592,11 +592,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	*/
 
 	// a = a + b
-	o0 = v2 + v12;
-	o1 = v3 + v13;
-	o1 = o1 + select(0u, 1u, o0 < v2);
-	v2 = o0;
-	v3 = o1;
+	v_23.x = v2;
+	v_23.y = v3;
+	v_1213.x = v12;
+	v_1213.y = v13;
+	v_23 = v_23 + v_1213 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_23.x + v_1213.x < v_23.x);
+	v2 = v_23.x;
+	v3 = v_23.y;
+	v12 = v_1213.x;
+	v13 = v_1213.y;
 
 	// // a = a + m[sigma[r][2*i+0]]
 	// // skip since adding 0u does nothing
@@ -628,11 +632,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v13 = (xor1 >> 24u) ^ (xor0 << 8u);
 
 	// a = a + b
-	o0 = v2 + v12;
-	o1 = v3 + v13;
-	o1 = o1 + select(0u, 1u, o0 < v2);
-	v2 = o0;
-	v3 = o1;
+	v_23.x = v2;
+	v_23.y = v3;
+	v_1213.x = v12;
+	v_1213.y = v13;
+	v_23 = v_23 + v_1213 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_23.x + v_1213.x < v_23.x);
+	v2 = v_23.x;
+	v3 = v_23.y;
+	v12 = v_1213.x;
+	v13 = v_1213.y;
 
 	// // a = a + m[sigma[r][2*i+1]]
 	// // skip since adding 0u does nothing
@@ -1310,11 +1318,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	*/
 
 	// a = a + b
-	o0 = v2 + v12;
-	o1 = v3 + v13;
-	o1 = o1 + select(0u, 1u, o0 < v2);
-	v2 = o0;
-	v3 = o1;
+	v_23.x = v2;
+	v_23.y = v3;
+	v_1213.x = v12;
+	v_1213.y = v13;
+	v_23 = v_23 + v_1213 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_23.x + v_1213.x < v_23.x);
+	v2 = v_23.x;
+	v3 = v_23.y;
+	v12 = v_1213.x;
+	v13 = v_1213.y;
 
 	// a = a + m[sigma[r][2*i+0]]
 	o0 = v2 + m0;
@@ -1343,11 +1355,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v13 = (xor1 >> 24u) ^ (xor0 << 8u);
 
 	// a = a + b
-	o0 = v2 + v12;
-	o1 = v3 + v13;
-	o1 = o1 + select(0u, 1u, o0 < v2);
-	v2 = o0;
-	v3 = o1;
+	v_23.x = v2;
+	v_23.y = v3;
+	v_1213.x = v12;
+	v_1213.y = v13;
+	v_23 = v_23 + v_1213 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_23.x + v_1213.x < v_23.x);
+	v2 = v_23.x;
+	v3 = v_23.y;
+	v12 = v_1213.x;
+	v13 = v_1213.y;
 
 	// a = a + m[sigma[r][2*i+1]]
 	o0 = v2 + m4;
@@ -2019,11 +2035,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	*/
 
 	// a = a + b
-	o0 = v2 + v12;
-	o1 = v3 + v13;
-	o1 = o1 + select(0u, 1u, o0 < v2);
-	v2 = o0;
-	v3 = o1;
+	v_23.x = v2;
+	v_23.y = v3;
+	v_1213.x = v12;
+	v_1213.y = v13;
+	v_23 = v_23 + v_1213 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_23.x + v_1213.x < v_23.x);
+	v2 = v_23.x;
+	v3 = v_23.y;
+	v12 = v_1213.x;
+	v13 = v_1213.y;
 
 	// a = a + m[sigma[r][2*i+0]]
 	o0 = v2 + m6;
@@ -2052,11 +2072,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v13 = (xor1 >> 24u) ^ (xor0 << 8u);
 
 	// a = a + b
-	o0 = v2 + v12;
-	o1 = v3 + v13;
-	o1 = o1 + select(0u, 1u, o0 < v2);
-	v2 = o0;
-	v3 = o1;
+	v_23.x = v2;
+	v_23.y = v3;
+	v_1213.x = v12;
+	v_1213.y = v13;
+	v_23 = v_23 + v_1213 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_23.x + v_1213.x < v_23.x);
+	v2 = v_23.x;
+	v3 = v_23.y;
+	v12 = v_1213.x;
+	v13 = v_1213.y;
 
 	// // a = a + m[sigma[r][2*i+1]]
 	// // skip since adding 0u does nothing
@@ -2725,11 +2749,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	*/
 
 	// a = a + b
-	o0 = v2 + v12;
-	o1 = v3 + v13;
-	o1 = o1 + select(0u, 1u, o0 < v2);
-	v2 = o0;
-	v3 = o1;
+	v_23.x = v2;
+	v_23.y = v3;
+	v_1213.x = v12;
+	v_1213.y = v13;
+	v_23 = v_23 + v_1213 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_23.x + v_1213.x < v_23.x);
+	v2 = v_23.x;
+	v3 = v_23.y;
+	v12 = v_1213.x;
+	v13 = v_1213.y;
 
 	// // a = a + m[sigma[r][2*i+0]]
 	// // skip since adding 0u does nothing
@@ -2761,11 +2789,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v13 = (xor1 >> 24u) ^ (xor0 << 8u);
 
 	// a = a + b
-	o0 = v2 + v12;
-	o1 = v3 + v13;
-	o1 = o1 + select(0u, 1u, o0 < v2);
-	v2 = o0;
-	v3 = o1;
+	v_23.x = v2;
+	v_23.y = v3;
+	v_1213.x = v12;
+	v_1213.y = v13;
+	v_23 = v_23 + v_1213 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_23.x + v_1213.x < v_23.x);
+	v2 = v_23.x;
+	v3 = v_23.y;
+	v12 = v_1213.x;
+	v13 = v_1213.y;
 
 	// // a = a + m[sigma[r][2*i+1]]
 	// // skip since adding 0u does nothing
@@ -3431,11 +3463,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	*/
 
 	// a = a + b
-	o0 = v2 + v12;
-	o1 = v3 + v13;
-	o1 = o1 + select(0u, 1u, o0 < v2);
-	v2 = o0;
-	v3 = o1;
+	v_23.x = v2;
+	v_23.y = v3;
+	v_1213.x = v12;
+	v_1213.y = v13;
+	v_23 = v_23 + v_1213 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_23.x + v_1213.x < v_23.x);
+	v2 = v_23.x;
+	v3 = v_23.y;
+	v12 = v_1213.x;
+	v13 = v_1213.y;
 
 	// // a = a + m[sigma[r][2*i+0]]
 	// // skip since adding 0u does nothing
@@ -3467,11 +3503,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v13 = (xor1 >> 24u) ^ (xor0 << 8u);
 
 	// a = a + b
-	o0 = v2 + v12;
-	o1 = v3 + v13;
-	o1 = o1 + select(0u, 1u, o0 < v2);
-	v2 = o0;
-	v3 = o1;
+	v_23.x = v2;
+	v_23.y = v3;
+	v_1213.x = v12;
+	v_1213.y = v13;
+	v_23 = v_23 + v_1213 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_23.x + v_1213.x < v_23.x);
+	v2 = v_23.x;
+	v3 = v_23.y;
+	v12 = v_1213.x;
+	v13 = v_1213.y;
 
 	// // a = a + m[sigma[r][2*i+1]]
 	// // skip since adding 0u does nothing
@@ -4140,11 +4180,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	*/
 
 	// a = a + b
-	o0 = v2 + v12;
-	o1 = v3 + v13;
-	o1 = o1 + select(0u, 1u, o0 < v2);
-	v2 = o0;
-	v3 = o1;
+	v_23.x = v2;
+	v_23.y = v3;
+	v_1213.x = v12;
+	v_1213.y = v13;
+	v_23 = v_23 + v_1213 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_23.x + v_1213.x < v_23.x);
+	v2 = v_23.x;
+	v3 = v_23.y;
+	v12 = v_1213.x;
+	v13 = v_1213.y;
 
 	// // a = a + m[sigma[r][2*i+0]]
 	// // skip since adding 0u does nothing
@@ -4176,11 +4220,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v13 = (xor1 >> 24u) ^ (xor0 << 8u);
 
 	// a = a + b
-	o0 = v2 + v12;
-	o1 = v3 + v13;
-	o1 = o1 + select(0u, 1u, o0 < v2);
-	v2 = o0;
-	v3 = o1;
+	v_23.x = v2;
+	v_23.y = v3;
+	v_1213.x = v12;
+	v_1213.y = v13;
+	v_23 = v_23 + v_1213 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_23.x + v_1213.x < v_23.x);
+	v2 = v_23.x;
+	v3 = v_23.y;
+	v12 = v_1213.x;
+	v13 = v_1213.y;
 
 	// // a = a + m[sigma[r][2*i+1]]
 	// // skip since adding 0u does nothing
@@ -4852,11 +4900,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	*/
 
 	// a = a + b
-	o0 = v2 + v12;
-	o1 = v3 + v13;
-	o1 = o1 + select(0u, 1u, o0 < v2);
-	v2 = o0;
-	v3 = o1;
+	v_23.x = v2;
+	v_23.y = v3;
+	v_1213.x = v12;
+	v_1213.y = v13;
+	v_23 = v_23 + v_1213 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_23.x + v_1213.x < v_23.x);
+	v2 = v_23.x;
+	v3 = v_23.y;
+	v12 = v_1213.x;
+	v13 = v_1213.y;
 
 	// // a = a + m[sigma[r][2*i+0]]
 	// // skip since adding 0u does nothing
@@ -4888,11 +4940,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v13 = (xor1 >> 24u) ^ (xor0 << 8u);
 
 	// a = a + b
-	o0 = v2 + v12;
-	o1 = v3 + v13;
-	o1 = o1 + select(0u, 1u, o0 < v2);
-	v2 = o0;
-	v3 = o1;
+	v_23.x = v2;
+	v_23.y = v3;
+	v_1213.x = v12;
+	v_1213.y = v13;
+	v_23 = v_23 + v_1213 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_23.x + v_1213.x < v_23.x);
+	v2 = v_23.x;
+	v3 = v_23.y;
+	v12 = v_1213.x;
+	v13 = v_1213.y;
 
 	// a = a + m[sigma[r][2*i+1]]
 	o0 = v2 + m6;
@@ -5561,11 +5617,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	*/
 
 	// a = a + b
-	o0 = v2 + v12;
-	o1 = v3 + v13;
-	o1 = o1 + select(0u, 1u, o0 < v2);
-	v2 = o0;
-	v3 = o1;
+	v_23.x = v2;
+	v_23.y = v3;
+	v_1213.x = v12;
+	v_1213.y = v13;
+	v_23 = v_23 + v_1213 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_23.x + v_1213.x < v_23.x);
+	v2 = v_23.x;
+	v3 = v_23.y;
+	v12 = v_1213.x;
+	v13 = v_1213.y;
 
 	// // a = a + m[sigma[r][2*i+0]]
 	// // skip since adding 0u does nothing
@@ -5597,11 +5657,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v13 = (xor1 >> 24u) ^ (xor0 << 8u);
 
 	// a = a + b
-	o0 = v2 + v12;
-	o1 = v3 + v13;
-	o1 = o1 + select(0u, 1u, o0 < v2);
-	v2 = o0;
-	v3 = o1;
+	v_23.x = v2;
+	v_23.y = v3;
+	v_1213.x = v12;
+	v_1213.y = v13;
+	v_23 = v_23 + v_1213 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_23.x + v_1213.x < v_23.x);
+	v2 = v_23.x;
+	v3 = v_23.y;
+	v12 = v_1213.x;
+	v13 = v_1213.y;
 
 	// a = a + m[sigma[r][2*i+1]]
 	o0 = v2 + m8;
@@ -6270,11 +6334,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	*/
 
 	// a = a + b
-	o0 = v2 + v12;
-	o1 = v3 + v13;
-	o1 = o1 + select(0u, 1u, o0 < v2);
-	v2 = o0;
-	v3 = o1;
+	v_23.x = v2;
+	v_23.y = v3;
+	v_1213.x = v12;
+	v_1213.y = v13;
+	v_23 = v_23 + v_1213 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_23.x + v_1213.x < v_23.x);
+	v2 = v_23.x;
+	v3 = v_23.y;
+	v12 = v_1213.x;
+	v13 = v_1213.y;
 
 	// // a = a + m[sigma[r][2*i+0]]
 	// // skip since adding 0u does nothing
@@ -6306,11 +6374,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v13 = (xor1 >> 24u) ^ (xor0 << 8u);
 
 	// a = a + b
-	o0 = v2 + v12;
-	o1 = v3 + v13;
-	o1 = o1 + select(0u, 1u, o0 < v2);
-	v2 = o0;
-	v3 = o1;
+	v_23.x = v2;
+	v_23.y = v3;
+	v_1213.x = v12;
+	v_1213.y = v13;
+	v_23 = v_23 + v_1213 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_23.x + v_1213.x < v_23.x);
+	v2 = v_23.x;
+	v3 = v_23.y;
+	v12 = v_1213.x;
+	v13 = v_1213.y;
 
 	// // a = a + m[sigma[r][2*i+1]]
 	// // skip since adding 0u does nothing
@@ -6979,11 +7051,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	*/
 
 	// a = a + b
-	o0 = v2 + v12;
-	o1 = v3 + v13;
-	o1 = o1 + select(0u, 1u, o0 < v2);
-	v2 = o0;
-	v3 = o1;
+	v_23.x = v2;
+	v_23.y = v3;
+	v_1213.x = v12;
+	v_1213.y = v13;
+	v_23 = v_23 + v_1213 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_23.x + v_1213.x < v_23.x);
+	v2 = v_23.x;
+	v3 = v_23.y;
+	v12 = v_1213.x;
+	v13 = v_1213.y;
 
 	// // a = a + m[sigma[r][2*i+0]]
 	// // skip since adding 0u does nothing
@@ -7015,11 +7091,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v13 = (xor1 >> 24u) ^ (xor0 << 8u);
 
 	// a = a + b
-	o0 = v2 + v12;
-	o1 = v3 + v13;
-	o1 = o1 + select(0u, 1u, o0 < v2);
-	v2 = o0;
-	v3 = o1;
+	v_23.x = v2;
+	v_23.y = v3;
+	v_1213.x = v12;
+	v_1213.y = v13;
+	v_23 = v_23 + v_1213 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_23.x + v_1213.x < v_23.x);
+	v2 = v_23.x;
+	v3 = v_23.y;
+	v12 = v_1213.x;
+	v13 = v_1213.y;
 
 	// // a = a + m[sigma[r][2*i+1]]
 	// // skip since adding 0u does nothing
@@ -7682,11 +7762,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	*/
 
 	// a = a + b
-	o0 = v2 + v12;
-	o1 = v3 + v13;
-	o1 = o1 + select(0u, 1u, o0 < v2);
-	v2 = o0;
-	v3 = o1;
+	v_23.x = v2;
+	v_23.y = v3;
+	v_1213.x = v12;
+	v_1213.y = v13;
+	v_23 = v_23 + v_1213 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_23.x + v_1213.x < v_23.x);
+	v2 = v_23.x;
+	v3 = v_23.y;
+	v12 = v_1213.x;
+	v13 = v_1213.y;
 
 	// // a = a + m[sigma[r][2*i+0]]
 	// // skip since adding 0u does nothing
@@ -7718,11 +7802,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v13 = (xor1 >> 24u) ^ (xor0 << 8u);
 
 	// a = a + b
-	o0 = v2 + v12;
-	o1 = v3 + v13;
-	o1 = o1 + select(0u, 1u, o0 < v2);
-	v2 = o0;
-	v3 = o1;
+	v_23.x = v2;
+	v_23.y = v3;
+	v_1213.x = v12;
+	v_1213.y = v13;
+	v_23 = v_23 + v_1213 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_23.x + v_1213.x < v_23.x);
+	v2 = v_23.x;
+	v3 = v_23.y;
+	v12 = v_1213.x;
+	v13 = v_1213.y;
 
 	// // a = a + m[sigma[r][2*i+1]]
 	// // skip since adding 0u does nothing
@@ -8400,11 +8488,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	*/
 
 	// a = a + b
-	o0 = v2 + v12;
-	o1 = v3 + v13;
-	o1 = o1 + select(0u, 1u, o0 < v2);
-	v2 = o0;
-	v3 = o1;
+	v_23.x = v2;
+	v_23.y = v3;
+	v_1213.x = v12;
+	v_1213.y = v13;
+	v_23 = v_23 + v_1213 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_23.x + v_1213.x < v_23.x);
+	v2 = v_23.x;
+	v3 = v_23.y;
+	v12 = v_1213.x;
+	v13 = v_1213.y;
 
 	// a = a + m[sigma[r][2*i+0]]
 	o0 = v2 + m0;
@@ -8433,11 +8525,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v13 = (xor1 >> 24u) ^ (xor0 << 8u);
 
 	// a = a + b
-	o0 = v2 + v12;
-	o1 = v3 + v13;
-	o1 = o1 + select(0u, 1u, o0 < v2);
-	v2 = o0;
-	v3 = o1;
+	v_23.x = v2;
+	v_23.y = v3;
+	v_1213.x = v12;
+	v_1213.y = v13;
+	v_23 = v_23 + v_1213 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_23.x + v_1213.x < v_23.x);
+	v2 = v_23.x;
+	v3 = v_23.y;
+	v12 = v_1213.x;
+	v13 = v_1213.y;
 
 	// a = a + m[sigma[r][2*i+1]]
 	o0 = v2 + m4;