From: Chris Duncan <chris@zoso.dev>
Date: Tue, 14 Jan 2025 14:54:11 +0000 (-0800)
Subject: Replace more scalar addition with vector addition. Interestingly, found that commenti... 
X-Git-Url: https://zoso.dev/?a=commitdiff_plain;h=00da999f8bb23486dc37519287bc09b802ae0772;p=nano-pow.git

Replace more scalar addition with vector addition. Interestingly, found that commenting out `	v24 = v_2425.x;` or `v25 = v_2425.y;` still results in a correct nonce. This presents an opportunity to prune even more instructions, so this will be pursued in a separate branch.
---

diff --git a/src/shaders/compute.wgsl b/src/shaders/compute.wgsl
index 4dcc796..e59c9b7 100644
--- a/src/shaders/compute.wgsl
+++ b/src/shaders/compute.wgsl
@@ -127,8 +127,10 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	*/
 
 	// a = a + b
-	// v1 = v1 + v9 + select(0u, 1u, v0 + v8 < v0);
-	// v0 = v0 + v8;
+	v_01.x = v0;
+	v_01.y = v1;
+	v_89.x = v8;
+	v_89.y = v9;
 	v_01 = v_01 + v_89 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_01.x + v_89.x < v_01.x);
 	v0 = v_01.x;
 	v1 = v_01.y;
@@ -150,11 +152,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v25 = xor0;
 
 	// c = c + d
-	o0 = v16 + v24;
-	o1 = v17 + v25;
-	o1 = o1 + select(0u, 1u, o0 < v16);
-	v16 = o0;
-	v17 = o1;
+	v_1617.x = v16;
+	v_1617.y = v17;
+	v_2425.x = v24;
+	v_2425.y = v25;
+	v_1617 = v_1617 + v_2425 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_1617.x + v_2425.x < v_1617.x);
+	v16 = v_1617.x;
+	v17 = v_1617.y;
+	v24 = v_2425.x;
+	v25 = v_2425.y;
 
 	// b = rotr64(b ^ c, 24)
 	xor0 = v8 ^ v16;
@@ -163,9 +169,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v9 = (xor1 >> 24u) ^ (xor0 << 8u);
 
 	// a = a + b
-	v1 = v1 + v9;
-	v1 = v1 + select(0u, 1u, v0 + v8 < v0);
-	v0 = v0 + v8;
+	v_01.x = v0;
+	v_01.y = v1;
+	v_89.x = v8;
+	v_89.y = v9;
+	v_01 = v_01 + v_89 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_01.x + v_89.x < v_01.x);
+	v0 = v_01.x;
+	v1 = v_01.y;
+	v8 = v_89.x;
+	v9 = v_89.y;
 
 	// a = a + m[sigma[r][2*i+1]]
 	o0 = v0 + m2;
@@ -181,11 +193,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v25 = (xor1 >> 16u) ^ (xor0 << 16u);
 
 	// c = c + d
-	o0 = v16 + v24;
-	o1 = v17 + v25;
-	o1 = o1 + select(0u, 1u, o0 < v16);
-	v16 = o0;
-	v17 = o1;
+	v_1617.x = v16;
+	v_1617.y = v17;
+	v_2425.x = v24;
+	v_2425.y = v25;
+	v_1617 = v_1617 + v_2425 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_1617.x + v_2425.x < v_1617.x);
+	v16 = v_1617.x;
+	v17 = v_1617.y;
+	v24 = v_2425.x;
+	v25 = v_2425.y;
 
 	// b = rotr64(b ^ c, 63)
 	xor0 = v8 ^ v16;
@@ -202,19 +218,16 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	*/
 
 	// a = a + b
+	v_23.x = v2;
+	v_23.y = v3;
+	v_1011.x = v10;
+	v_1011.y = v11;
 	v_23 = v_23 + v_1011 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_23.x + v_1011.x < v_23.x);
 	v2 = v_23.x;
 	v3 = v_23.y;
 	v10 = v_1011.x;
 	v11 = v_1011.y;
 
-	// o0 = v2 + v10;
-	// o1 = v3 + v11;
-	// o1 = o1 + select(0u, 1u, o0 < v2);
-	// v2 = o0;
-	// v3 = o1;
-
-
 	// a = a + m[sigma[r][2*i+0]]
 	o0 = v2 + m4;
 	o1 = v3 + m5;
@@ -242,11 +255,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v11 = (xor1 >> 24u) ^ (xor0 << 8u);
 
 	// a = a + b
-	o0 = v2 + v10;
-	o1 = v3 + v11;
-	o1 = o1 + select(0u, 1u, o0 < v2);
-	v2 = o0;
-	v3 = o1;
+	v_23.x = v2;
+	v_23.y = v3;
+	v_1011.x = v10;
+	v_1011.y = v11;
+	v_23 = v_23 + v_1011 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_23.x + v_1011.x < v_23.x);
+	v2 = v_23.x;
+	v3 = v_23.y;
+	v10 = v_1011.x;
+	v11 = v_1011.y;
 
 	// a = a + m[sigma[r][2*i+1]]
 	o0 = v2 + m6;
@@ -764,9 +781,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	*/
 
 	// a = a + b
-	v1 = v1 + v9;
-	v1 = v1 + select(0u, 1u, v0 + v8 < v0);
-	v0 = v0 + v8;
+	v_01.x = v0;
+	v_01.y = v1;
+	v_89.x = v8;
+	v_89.y = v9;
+	v_01 = v_01 + v_89 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_01.x + v_89.x < v_01.x);
+	v0 = v_01.x;
+	v1 = v_01.y;
+	v8 = v_89.x;
+	v9 = v_89.y;
 
 	// // a = a + m[sigma[r][2*i+0]]
 	// // skip since adding 0u does nothing
@@ -785,11 +808,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v25 = xor0;
 
 	// c = c + d
-	o0 = v16 + v24;
-	o1 = v17 + v25;
-	o1 = o1 + select(0u, 1u, o0 < v16);
-	v16 = o0;
-	v17 = o1;
+	v_1617.x = v16;
+	v_1617.y = v17;
+	v_2425.x = v24;
+	v_2425.y = v25;
+	v_1617 = v_1617 + v_2425 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_1617.x + v_2425.x < v_1617.x);
+	v16 = v_1617.x;
+	v17 = v_1617.y;
+	v24 = v_2425.x;
+	v25 = v_2425.y;
 
 	// b = rotr64(b ^ c, 24)
 	xor0 = v8 ^ v16;
@@ -798,9 +825,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v9 = (xor1 >> 24u) ^ (xor0 << 8u);
 
 	// a = a + b
-	v1 = v1 + v9;
-	v1 = v1 + select(0u, 1u, v0 + v8 < v0);
-	v0 = v0 + v8;
+	v_01.x = v0;
+	v_01.y = v1;
+	v_89.x = v8;
+	v_89.y = v9;
+	v_01 = v_01 + v_89 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_01.x + v_89.x < v_01.x);
+	v0 = v_01.x;
+	v1 = v_01.y;
+	v8 = v_89.x;
+	v9 = v_89.y;
 
 	// // a = a + m[sigma[r][2*i+1]]
 	// // skip since adding 0u does nothing
@@ -819,11 +852,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v25 = (xor1 >> 16u) ^ (xor0 << 16u);
 
 	// c = c + d
-	o0 = v16 + v24;
-	o1 = v17 + v25;
-	o1 = o1 + select(0u, 1u, o0 < v16);
-	v16 = o0;
-	v17 = o1;
+	v_1617.x = v16;
+	v_1617.y = v17;
+	v_2425.x = v24;
+	v_2425.y = v25;
+	v_1617 = v_1617 + v_2425 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_1617.x + v_2425.x < v_1617.x);
+	v16 = v_1617.x;
+	v17 = v_1617.y;
+	v24 = v_2425.x;
+	v25 = v_2425.y;
 
 	// b = rotr64(b ^ c, 63)
 	xor0 = v8 ^ v16;
@@ -840,11 +877,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	*/
 
 	// a = a + b
-	o0 = v2 + v10;
-	o1 = v3 + v11;
-	o1 = o1 + select(0u, 1u, o0 < v2);
-	v2 = o0;
-	v3 = o1;
+	v_23.x = v2;
+	v_23.y = v3;
+	v_1011.x = v10;
+	v_1011.y = v11;
+	v_23 = v_23 + v_1011 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_23.x + v_1011.x < v_23.x);
+	v2 = v_23.x;
+	v3 = v_23.y;
+	v10 = v_1011.x;
+	v11 = v_1011.y;
 
 	// a = a + m[sigma[r][2*i+0]]
 	o0 = v2 + m8;
@@ -873,11 +914,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v11 = (xor1 >> 24u) ^ (xor0 << 8u);
 
 	// a = a + b
-	o0 = v2 + v10;
-	o1 = v3 + v11;
-	o1 = o1 + select(0u, 1u, o0 < v2);
-	v2 = o0;
-	v3 = o1;
+	v_23.x = v2;
+	v_23.y = v3;
+	v_1011.x = v10;
+	v_1011.y = v11;
+	v_23 = v_23 + v_1011 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_23.x + v_1011.x < v_23.x);
+	v2 = v_23.x;
+	v3 = v_23.y;
+	v10 = v_1011.x;
+	v11 = v_1011.y;
 
 	// // a = a + m[sigma[r][2*i+1]]
 	// // skip since adding 0u does nothing
@@ -1389,9 +1434,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	*/
 
 	// a = a + b
-	v1 = v1 + v9;
-	v1 = v1 + select(0u, 1u, v0 + v8 < v0);
-	v0 = v0 + v8;
+	v_01.x = v0;
+	v_01.y = v1;
+	v_89.x = v8;
+	v_89.y = v9;
+	v_01 = v_01 + v_89 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_01.x + v_89.x < v_01.x);
+	v0 = v_01.x;
+	v1 = v_01.y;
+	v8 = v_89.x;
+	v9 = v_89.y;
 
 	// // a = a + m[sigma[r][2*i+0]]
 	// // skip since adding 0u does nothing
@@ -1410,11 +1461,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v25 = xor0;
 
 	// c = c + d
-	o0 = v16 + v24;
-	o1 = v17 + v25;
-	o1 = o1 + select(0u, 1u, o0 < v16);
-	v16 = o0;
-	v17 = o1;
+	v_1617.x = v16;
+	v_1617.y = v17;
+	v_2425.x = v24;
+	v_2425.y = v25;
+	v_1617 = v_1617 + v_2425 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_1617.x + v_2425.x < v_1617.x);
+	v16 = v_1617.x;
+	v17 = v_1617.y;
+	v24 = v_2425.x;
+	v25 = v_2425.y;
 
 	// b = rotr64(b ^ c, 24)
 	xor0 = v8 ^ v16;
@@ -1423,9 +1478,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v9 = (xor1 >> 24u) ^ (xor0 << 8u);
 
 	// a = a + b
-	v1 = v1 + v9;
-	v1 = v1 + select(0u, 1u, v0 + v8 < v0);
-	v0 = v0 + v8;
+	v_01.x = v0;
+	v_01.y = v1;
+	v_89.x = v8;
+	v_89.y = v9;
+	v_01 = v_01 + v_89 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_01.x + v_89.x < v_01.x);
+	v0 = v_01.x;
+	v1 = v_01.y;
+	v8 = v_89.x;
+	v9 = v_89.y;
 
 	// // a = a + m[sigma[r][2*i+1]]
 	// // skip since adding 0u does nothing
@@ -1444,11 +1505,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v25 = (xor1 >> 16u) ^ (xor0 << 16u);
 
 	// c = c + d
-	o0 = v16 + v24;
-	o1 = v17 + v25;
-	o1 = o1 + select(0u, 1u, o0 < v16);
-	v16 = o0;
-	v17 = o1;
+	v_1617.x = v16;
+	v_1617.y = v17;
+	v_2425.x = v24;
+	v_2425.y = v25;
+	v_1617 = v_1617 + v_2425 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_1617.x + v_2425.x < v_1617.x);
+	v16 = v_1617.x;
+	v17 = v_1617.y;
+	v24 = v_2425.x;
+	v25 = v_2425.y;
 
 	// b = rotr64(b ^ c, 63)
 	xor0 = v8 ^ v16;
@@ -1465,11 +1530,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	*/
 
 	// a = a + b
-	o0 = v2 + v10;
-	o1 = v3 + v11;
-	o1 = o1 + select(0u, 1u, o0 < v2);
-	v2 = o0;
-	v3 = o1;
+	v_23.x = v2;
+	v_23.y = v3;
+	v_1011.x = v10;
+	v_1011.y = v11;
+	v_23 = v_23 + v_1011 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_23.x + v_1011.x < v_23.x);
+	v2 = v_23.x;
+	v3 = v_23.y;
+	v10 = v_1011.x;
+	v11 = v_1011.y;
 
 	// // a = a + m[sigma[r][2*i+0]]
 	// // skip since adding 0u does nothing
@@ -1501,11 +1570,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v11 = (xor1 >> 24u) ^ (xor0 << 8u);
 
 	// a = a + b
-	o0 = v2 + v10;
-	o1 = v3 + v11;
-	o1 = o1 + select(0u, 1u, o0 < v2);
-	v2 = o0;
-	v3 = o1;
+	v_23.x = v2;
+	v_23.y = v3;
+	v_1011.x = v10;
+	v_1011.y = v11;
+	v_23 = v_23 + v_1011 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_23.x + v_1011.x < v_23.x);
+	v2 = v_23.x;
+	v3 = v_23.y;
+	v10 = v_1011.x;
+	v11 = v_1011.y;
 
 	// a = a + m[sigma[r][2*i+1]]
 	o0 = v2 + m0;
@@ -2014,9 +2087,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	*/
 
 	// a = a + b
-	v1 = v1 + v9;
-	v1 = v1 + select(0u, 1u, v0 + v8 < v0);
-	v0 = v0 + v8;
+	v_01.x = v0;
+	v_01.y = v1;
+	v_89.x = v8;
+	v_89.y = v9;
+	v_01 = v_01 + v_89 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_01.x + v_89.x < v_01.x);
+	v0 = v_01.x;
+	v1 = v_01.y;
+	v8 = v_89.x;
+	v9 = v_89.y;
 
 	// // a = a + m[sigma[r][2*i+0]]
 	// // skip since adding 0u does nothing
@@ -2035,11 +2114,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v25 = xor0;
 
 	// c = c + d
-	o0 = v16 + v24;
-	o1 = v17 + v25;
-	o1 = o1 + select(0u, 1u, o0 < v16);
-	v16 = o0;
-	v17 = o1;
+	v_1617.x = v16;
+	v_1617.y = v17;
+	v_2425.x = v24;
+	v_2425.y = v25;
+	v_1617 = v_1617 + v_2425 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_1617.x + v_2425.x < v_1617.x);
+	v16 = v_1617.x;
+	v17 = v_1617.y;
+	v24 = v_2425.x;
+	v25 = v_2425.y;
 
 	// b = rotr64(b ^ c, 24)
 	xor0 = v8 ^ v16;
@@ -2048,9 +2131,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v9 = (xor1 >> 24u) ^ (xor0 << 8u);
 
 	// a = a + b
-	v1 = v1 + v9;
-	v1 = v1 + select(0u, 1u, v0 + v8 < v0);
-	v0 = v0 + v8;
+	v_01.x = v0;
+	v_01.y = v1;
+	v_89.x = v8;
+	v_89.y = v9;
+	v_01 = v_01 + v_89 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_01.x + v_89.x < v_01.x);
+	v0 = v_01.x;
+	v1 = v_01.y;
+	v8 = v_89.x;
+	v9 = v_89.y;
 
 	// // a = a + m[sigma[r][2*i+1]]
 	// // skip since adding 0u does nothing
@@ -2069,11 +2158,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v25 = (xor1 >> 16u) ^ (xor0 << 16u);
 
 	// c = c + d
-	o0 = v16 + v24;
-	o1 = v17 + v25;
-	o1 = o1 + select(0u, 1u, o0 < v16);
-	v16 = o0;
-	v17 = o1;
+	v_1617.x = v16;
+	v_1617.y = v17;
+	v_2425.x = v24;
+	v_2425.y = v25;
+	v_1617 = v_1617 + v_2425 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_1617.x + v_2425.x < v_1617.x);
+	v16 = v_1617.x;
+	v17 = v_1617.y;
+	v24 = v_2425.x;
+	v25 = v_2425.y;
 
 	// b = rotr64(b ^ c, 63)
 	xor0 = v8 ^ v16;
@@ -2090,11 +2183,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	*/
 
 	// a = a + b
-	o0 = v2 + v10;
-	o1 = v3 + v11;
-	o1 = o1 + select(0u, 1u, o0 < v2);
-	v2 = o0;
-	v3 = o1;
+	v_23.x = v2;
+	v_23.y = v3;
+	v_1011.x = v10;
+	v_1011.y = v11;
+	v_23 = v_23 + v_1011 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_23.x + v_1011.x < v_23.x);
+	v2 = v_23.x;
+	v3 = v_23.y;
+	v10 = v_1011.x;
+	v11 = v_1011.y;
 
 	// a = a + m[sigma[r][2*i+0]]
 	o0 = v2 + m6;
@@ -2123,11 +2220,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v11 = (xor1 >> 24u) ^ (xor0 << 8u);
 
 	// a = a + b
-	o0 = v2 + v10;
-	o1 = v3 + v11;
-	o1 = o1 + select(0u, 1u, o0 < v2);
-	v2 = o0;
-	v3 = o1;
+	v_23.x = v2;
+	v_23.y = v3;
+	v_1011.x = v10;
+	v_1011.y = v11;
+	v_23 = v_23 + v_1011 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_23.x + v_1011.x < v_23.x);
+	v2 = v_23.x;
+	v3 = v_23.y;
+	v10 = v_1011.x;
+	v11 = v_1011.y;
 
 	// a = a + m[sigma[r][2*i+1]]
 	o0 = v2 + m2;
@@ -2639,9 +2740,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	*/
 
 	// a = a + b
-	v1 = v1 + v9;
-	v1 = v1 + select(0u, 1u, v0 + v8 < v0);
-	v0 = v0 + v8;
+	v_01.x = v0;
+	v_01.y = v1;
+	v_89.x = v8;
+	v_89.y = v9;
+	v_01 = v_01 + v_89 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_01.x + v_89.x < v_01.x);
+	v0 = v_01.x;
+	v1 = v_01.y;
+	v8 = v_89.x;
+	v9 = v_89.y;
 
 	// // a = a + m[sigma[r][2*i+0]]
 	// // skip since adding 0u does nothing
@@ -2660,11 +2767,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v25 = xor0;
 
 	// c = c + d
-	o0 = v16 + v24;
-	o1 = v17 + v25;
-	o1 = o1 + select(0u, 1u, o0 < v16);
-	v16 = o0;
-	v17 = o1;
+	v_1617.x = v16;
+	v_1617.y = v17;
+	v_2425.x = v24;
+	v_2425.y = v25;
+	v_1617 = v_1617 + v_2425 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_1617.x + v_2425.x < v_1617.x);
+	v16 = v_1617.x;
+	v17 = v_1617.y;
+	v24 = v_2425.x;
+	v25 = v_2425.y;
 
 	// b = rotr64(b ^ c, 24)
 	xor0 = v8 ^ v16;
@@ -2673,9 +2784,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v9 = (xor1 >> 24u) ^ (xor0 << 8u);
 
 	// a = a + b
-	v1 = v1 + v9;
-	v1 = v1 + select(0u, 1u, v0 + v8 < v0);
-	v0 = v0 + v8;
+	v_01.x = v0;
+	v_01.y = v1;
+	v_89.x = v8;
+	v_89.y = v9;
+	v_01 = v_01 + v_89 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_01.x + v_89.x < v_01.x);
+	v0 = v_01.x;
+	v1 = v_01.y;
+	v8 = v_89.x;
+	v9 = v_89.y;
 
 	// a = a + m[sigma[r][2*i+1]]
 	o0 = v0 + m0;
@@ -2691,11 +2808,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v25 = (xor1 >> 16u) ^ (xor0 << 16u);
 
 	// c = c + d
-	o0 = v16 + v24;
-	o1 = v17 + v25;
-	o1 = o1 + select(0u, 1u, o0 < v16);
-	v16 = o0;
-	v17 = o1;
+	v_1617.x = v16;
+	v_1617.y = v17;
+	v_2425.x = v24;
+	v_2425.y = v25;
+	v_1617 = v_1617 + v_2425 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_1617.x + v_2425.x < v_1617.x);
+	v16 = v_1617.x;
+	v17 = v_1617.y;
+	v24 = v_2425.x;
+	v25 = v_2425.y;
 
 	// b = rotr64(b ^ c, 63)
 	xor0 = v8 ^ v16;
@@ -2712,11 +2833,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	*/
 
 	// a = a + b
-	o0 = v2 + v10;
-	o1 = v3 + v11;
-	o1 = o1 + select(0u, 1u, o0 < v2);
-	v2 = o0;
-	v3 = o1;
+	v_23.x = v2;
+	v_23.y = v3;
+	v_1011.x = v10;
+	v_1011.y = v11;
+	v_23 = v_23 + v_1011 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_23.x + v_1011.x < v_23.x);
+	v2 = v_23.x;
+	v3 = v_23.y;
+	v10 = v_1011.x;
+	v11 = v_1011.y;
 
 	// // a = a + m[sigma[r][2*i+0]]
 	// // skip since adding 0u does nothing
@@ -2748,11 +2873,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v11 = (xor1 >> 24u) ^ (xor0 << 8u);
 
 	// a = a + b
-	o0 = v2 + v10;
-	o1 = v3 + v11;
-	o1 = o1 + select(0u, 1u, o0 < v2);
-	v2 = o0;
-	v3 = o1;
+	v_23.x = v2;
+	v_23.y = v3;
+	v_1011.x = v10;
+	v_1011.y = v11;
+	v_23 = v_23 + v_1011 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_23.x + v_1011.x < v_23.x);
+	v2 = v_23.x;
+	v3 = v_23.y;
+	v10 = v_1011.x;
+	v11 = v_1011.y;
 
 	// // a = a + m[sigma[r][2*i+1]]
 	// // skip since adding 0u does nothing
@@ -3264,9 +3393,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	*/
 
 	// a = a + b
-	v1 = v1 + v9;
-	v1 = v1 + select(0u, 1u, v0 + v8 < v0);
-	v0 = v0 + v8;
+	v_01.x = v0;
+	v_01.y = v1;
+	v_89.x = v8;
+	v_89.y = v9;
+	v_01 = v_01 + v_89 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_01.x + v_89.x < v_01.x);
+	v0 = v_01.x;
+	v1 = v_01.y;
+	v8 = v_89.x;
+	v9 = v_89.y;
 
 	// a = a + m[sigma[r][2*i+0]]
 	o0 = v0 + m4;
@@ -3282,11 +3417,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v25 = xor0;
 
 	// c = c + d
-	o0 = v16 + v24;
-	o1 = v17 + v25;
-	o1 = o1 + select(0u, 1u, o0 < v16);
-	v16 = o0;
-	v17 = o1;
+	v_1617.x = v16;
+	v_1617.y = v17;
+	v_2425.x = v24;
+	v_2425.y = v25;
+	v_1617 = v_1617 + v_2425 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_1617.x + v_2425.x < v_1617.x);
+	v16 = v_1617.x;
+	v17 = v_1617.y;
+	v24 = v_2425.x;
+	v25 = v_2425.y;
 
 	// b = rotr64(b ^ c, 24)
 	xor0 = v8 ^ v16;
@@ -3295,9 +3434,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v9 = (xor1 >> 24u) ^ (xor0 << 8u);
 
 	// a = a + b
-	v1 = v1 + v9;
-	v1 = v1 + select(0u, 1u, v0 + v8 < v0);
-	v0 = v0 + v8;
+	v_01.x = v0;
+	v_01.y = v1;
+	v_89.x = v8;
+	v_89.y = v9;
+	v_01 = v_01 + v_89 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_01.x + v_89.x < v_01.x);
+	v0 = v_01.x;
+	v1 = v_01.y;
+	v8 = v_89.x;
+	v9 = v_89.y;
 
 	// // a = a + m[sigma[r][2*i+1]]
 	// // skip since adding 0u does nothing
@@ -3316,11 +3461,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v25 = (xor1 >> 16u) ^ (xor0 << 16u);
 
 	// c = c + d
-	o0 = v16 + v24;
-	o1 = v17 + v25;
-	o1 = o1 + select(0u, 1u, o0 < v16);
-	v16 = o0;
-	v17 = o1;
+	v_1617.x = v16;
+	v_1617.y = v17;
+	v_2425.x = v24;
+	v_2425.y = v25;
+	v_1617 = v_1617 + v_2425 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_1617.x + v_2425.x < v_1617.x);
+	v16 = v_1617.x;
+	v17 = v_1617.y;
+	v24 = v_2425.x;
+	v25 = v_2425.y;
 
 	// b = rotr64(b ^ c, 63)
 	xor0 = v8 ^ v16;
@@ -3337,11 +3486,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	*/
 
 	// a = a + b
-	o0 = v2 + v10;
-	o1 = v3 + v11;
-	o1 = o1 + select(0u, 1u, o0 < v2);
-	v2 = o0;
-	v3 = o1;
+	v_23.x = v2;
+	v_23.y = v3;
+	v_1011.x = v10;
+	v_1011.y = v11;
+	v_23 = v_23 + v_1011 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_23.x + v_1011.x < v_23.x);
+	v2 = v_23.x;
+	v3 = v_23.y;
+	v10 = v_1011.x;
+	v11 = v_1011.y;
 
 	// // a = a + m[sigma[r][2*i+0]]
 	// // skip since adding 0u does nothing
@@ -3373,11 +3526,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v11 = (xor1 >> 24u) ^ (xor0 << 8u);
 
 	// a = a + b
-	o0 = v2 + v10;
-	o1 = v3 + v11;
-	o1 = o1 + select(0u, 1u, o0 < v2);
-	v2 = o0;
-	v3 = o1;
+	v_23.x = v2;
+	v_23.y = v3;
+	v_1011.x = v10;
+	v_1011.y = v11;
+	v_23 = v_23 + v_1011 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_23.x + v_1011.x < v_23.x);
+	v2 = v_23.x;
+	v3 = v_23.y;
+	v10 = v_1011.x;
+	v11 = v_1011.y;
 
 	// // a = a + m[sigma[r][2*i+1]]
 	// // skip since adding 0u does nothing
@@ -3889,9 +4046,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	*/
 
 	// a = a + b
-	v1 = v1 + v9;
-	v1 = v1 + select(0u, 1u, v0 + v8 < v0);
-	v0 = v0 + v8;
+	v_01.x = v0;
+	v_01.y = v1;
+	v_89.x = v8;
+	v_89.y = v9;
+	v_01 = v_01 + v_89 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_01.x + v_89.x < v_01.x);
+	v0 = v_01.x;
+	v1 = v_01.y;
+	v8 = v_89.x;
+	v9 = v_89.y;
 
 	// // a = a + m[sigma[r][2*i+0]]
 	// // skip since adding 0u does nothing
@@ -3910,11 +4073,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v25 = xor0;
 
 	// c = c + d
-	o0 = v16 + v24;
-	o1 = v17 + v25;
-	o1 = o1 + select(0u, 1u, o0 < v16);
-	v16 = o0;
-	v17 = o1;
+	v_1617.x = v16;
+	v_1617.y = v17;
+	v_2425.x = v24;
+	v_2425.y = v25;
+	v_1617 = v_1617 + v_2425 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_1617.x + v_2425.x < v_1617.x);
+	v16 = v_1617.x;
+	v17 = v_1617.y;
+	v24 = v_2425.x;
+	v25 = v_2425.y;
 
 	// b = rotr64(b ^ c, 24)
 	xor0 = v8 ^ v16;
@@ -3923,9 +4090,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v9 = (xor1 >> 24u) ^ (xor0 << 8u);
 
 	// a = a + b
-	v1 = v1 + v9;
-	v1 = v1 + select(0u, 1u, v0 + v8 < v0);
-	v0 = v0 + v8;
+	v_01.x = v0;
+	v_01.y = v1;
+	v_89.x = v8;
+	v_89.y = v9;
+	v_01 = v_01 + v_89 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_01.x + v_89.x < v_01.x);
+	v0 = v_01.x;
+	v1 = v_01.y;
+	v8 = v_89.x;
+	v9 = v_89.y;
 
 	// // a = a + m[sigma[r][2*i+1]]
 	// // skip since adding 0u does nothing
@@ -3944,11 +4117,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v25 = (xor1 >> 16u) ^ (xor0 << 16u);
 
 	// c = c + d
-	o0 = v16 + v24;
-	o1 = v17 + v25;
-	o1 = o1 + select(0u, 1u, o0 < v16);
-	v16 = o0;
-	v17 = o1;
+	v_1617.x = v16;
+	v_1617.y = v17;
+	v_2425.x = v24;
+	v_2425.y = v25;
+	v_1617 = v_1617 + v_2425 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_1617.x + v_2425.x < v_1617.x);
+	v16 = v_1617.x;
+	v17 = v_1617.y;
+	v24 = v_2425.x;
+	v25 = v_2425.y;
 
 	// b = rotr64(b ^ c, 63)
 	xor0 = v8 ^ v16;
@@ -3965,11 +4142,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	*/
 
 	// a = a + b
-	o0 = v2 + v10;
-	o1 = v3 + v11;
-	o1 = o1 + select(0u, 1u, o0 < v2);
-	v2 = o0;
-	v3 = o1;
+	v_23.x = v2;
+	v_23.y = v3;
+	v_1011.x = v10;
+	v_1011.y = v11;
+	v_23 = v_23 + v_1011 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_23.x + v_1011.x < v_23.x);
+	v2 = v_23.x;
+	v3 = v_23.y;
+	v10 = v_1011.x;
+	v11 = v_1011.y;
 
 	// a = a + m[sigma[r][2*i+0]]
 	o0 = v2 + m2;
@@ -3998,11 +4179,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v11 = (xor1 >> 24u) ^ (xor0 << 8u);
 
 	// a = a + b
-	o0 = v2 + v10;
-	o1 = v3 + v11;
-	o1 = o1 + select(0u, 1u, o0 < v2);
-	v2 = o0;
-	v3 = o1;
+	v_23.x = v2;
+	v_23.y = v3;
+	v_1011.x = v10;
+	v_1011.y = v11;
+	v_23 = v_23 + v_1011 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_23.x + v_1011.x < v_23.x);
+	v2 = v_23.x;
+	v3 = v_23.y;
+	v10 = v_1011.x;
+	v11 = v_1011.y;
 
 	// // a = a + m[sigma[r][2*i+1]]
 	// // skip since adding 0u does nothing
@@ -4514,9 +4699,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	*/
 
 	// a = a + b
-	v1 = v1 + v9;
-	v1 = v1 + select(0u, 1u, v0 + v8 < v0);
-	v0 = v0 + v8;
+	v_01.x = v0;
+	v_01.y = v1;
+	v_89.x = v8;
+	v_89.y = v9;
+	v_01 = v_01 + v_89 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_01.x + v_89.x < v_01.x);
+	v0 = v_01.x;
+	v1 = v_01.y;
+	v8 = v_89.x;
+	v9 = v_89.y;
 
 	// // a = a + m[sigma[r][2*i+0]]
 	// // skip since adding 0u does nothing
@@ -4535,11 +4726,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v25 = xor0;
 
 	// c = c + d
-	o0 = v16 + v24;
-	o1 = v17 + v25;
-	o1 = o1 + select(0u, 1u, o0 < v16);
-	v16 = o0;
-	v17 = o1;
+	v_1617.x = v16;
+	v_1617.y = v17;
+	v_2425.x = v24;
+	v_2425.y = v25;
+	v_1617 = v_1617 + v_2425 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_1617.x + v_2425.x < v_1617.x);
+	v16 = v_1617.x;
+	v17 = v_1617.y;
+	v24 = v_2425.x;
+	v25 = v_2425.y;
 
 	// b = rotr64(b ^ c, 24)
 	xor0 = v8 ^ v16;
@@ -4548,9 +4743,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v9 = (xor1 >> 24u) ^ (xor0 << 8u);
 
 	// a = a + b
-	v1 = v1 + v9;
-	v1 = v1 + select(0u, 1u, v0 + v8 < v0);
-	v0 = v0 + v8;
+	v_01.x = v0;
+	v_01.y = v1;
+	v_89.x = v8;
+	v_89.y = v9;
+	v_01 = v_01 + v_89 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_01.x + v_89.x < v_01.x);
+	v0 = v_01.x;
+	v1 = v_01.y;
+	v8 = v_89.x;
+	v9 = v_89.y;
 
 	// // a = a + m[sigma[r][2*i+1]]
 	// // skip since adding 0u does nothing
@@ -4569,11 +4770,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v25 = (xor1 >> 16u) ^ (xor0 << 16u);
 
 	// c = c + d
-	o0 = v16 + v24;
-	o1 = v17 + v25;
-	o1 = o1 + select(0u, 1u, o0 < v16);
-	v16 = o0;
-	v17 = o1;
+	v_1617.x = v16;
+	v_1617.y = v17;
+	v_2425.x = v24;
+	v_2425.y = v25;
+	v_1617 = v_1617 + v_2425 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_1617.x + v_2425.x < v_1617.x);
+	v16 = v_1617.x;
+	v17 = v_1617.y;
+	v24 = v_2425.x;
+	v25 = v_2425.y;
 
 	// b = rotr64(b ^ c, 63)
 	xor0 = v8 ^ v16;
@@ -4590,11 +4795,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	*/
 
 	// a = a + b
-	o0 = v2 + v10;
-	o1 = v3 + v11;
-	o1 = o1 + select(0u, 1u, o0 < v2);
-	v2 = o0;
-	v3 = o1;
+	v_23.x = v2;
+	v_23.y = v3;
+	v_1011.x = v10;
+	v_1011.y = v11;
+	v_23 = v_23 + v_1011 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_23.x + v_1011.x < v_23.x);
+	v2 = v_23.x;
+	v3 = v_23.y;
+	v10 = v_1011.x;
+	v11 = v_1011.y;
 
 	// // a = a + m[sigma[r][2*i+0]]
 	// // skip since adding 0u does nothing
@@ -4626,11 +4835,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v11 = (xor1 >> 24u) ^ (xor0 << 8u);
 
 	// a = a + b
-	o0 = v2 + v10;
-	o1 = v3 + v11;
-	o1 = o1 + select(0u, 1u, o0 < v2);
-	v2 = o0;
-	v3 = o1;
+	v_23.x = v2;
+	v_23.y = v3;
+	v_1011.x = v10;
+	v_1011.y = v11;
+	v_23 = v_23 + v_1011 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_23.x + v_1011.x < v_23.x);
+	v2 = v_23.x;
+	v3 = v_23.y;
+	v10 = v_1011.x;
+	v11 = v_1011.y;
 
 	// // a = a + m[sigma[r][2*i+1]]
 	// // skip since adding 0u does nothing
@@ -5139,9 +5352,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	*/
 
 	// a = a + b
-	v1 = v1 + v9;
-	v1 = v1 + select(0u, 1u, v0 + v8 < v0);
-	v0 = v0 + v8;
+	v_01.x = v0;
+	v_01.y = v1;
+	v_89.x = v8;
+	v_89.y = v9;
+	v_01 = v_01 + v_89 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_01.x + v_89.x < v_01.x);
+	v0 = v_01.x;
+	v1 = v_01.y;
+	v8 = v_89.x;
+	v9 = v_89.y;
 
 	// // a = a + m[sigma[r][2*i+0]]
 	// // skip since adding 0u does nothing
@@ -5160,11 +5379,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v25 = xor0;
 
 	// c = c + d
-	o0 = v16 + v24;
-	o1 = v17 + v25;
-	o1 = o1 + select(0u, 1u, o0 < v16);
-	v16 = o0;
-	v17 = o1;
+	v_1617.x = v16;
+	v_1617.y = v17;
+	v_2425.x = v24;
+	v_2425.y = v25;
+	v_1617 = v_1617 + v_2425 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_1617.x + v_2425.x < v_1617.x);
+	v16 = v_1617.x;
+	v17 = v_1617.y;
+	v24 = v_2425.x;
+	v25 = v_2425.y;
 
 	// b = rotr64(b ^ c, 24)
 	xor0 = v8 ^ v16;
@@ -5173,9 +5396,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v9 = (xor1 >> 24u) ^ (xor0 << 8u);
 
 	// a = a + b
-	v1 = v1 + v9;
-	v1 = v1 + select(0u, 1u, v0 + v8 < v0);
-	v0 = v0 + v8;
+	v_01.x = v0;
+	v_01.y = v1;
+	v_89.x = v8;
+	v_89.y = v9;
+	v_01 = v_01 + v_89 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_01.x + v_89.x < v_01.x);
+	v0 = v_01.x;
+	v1 = v_01.y;
+	v8 = v_89.x;
+	v9 = v_89.y;
 
 	// // a = a + m[sigma[r][2*i+1]]
 	// // skip since adding 0u does nothing
@@ -5194,11 +5423,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v25 = (xor1 >> 16u) ^ (xor0 << 16u);
 
 	// c = c + d
-	o0 = v16 + v24;
-	o1 = v17 + v25;
-	o1 = o1 + select(0u, 1u, o0 < v16);
-	v16 = o0;
-	v17 = o1;
+	v_1617.x = v16;
+	v_1617.y = v17;
+	v_2425.x = v24;
+	v_2425.y = v25;
+	v_1617 = v_1617 + v_2425 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_1617.x + v_2425.x < v_1617.x);
+	v16 = v_1617.x;
+	v17 = v_1617.y;
+	v24 = v_2425.x;
+	v25 = v_2425.y;
 
 	// b = rotr64(b ^ c, 63)
 	xor0 = v8 ^ v16;
@@ -5215,11 +5448,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	*/
 
 	// a = a + b
-	o0 = v2 + v10;
-	o1 = v3 + v11;
-	o1 = o1 + select(0u, 1u, o0 < v2);
-	v2 = o0;
-	v3 = o1;
+	v_23.x = v2;
+	v_23.y = v3;
+	v_1011.x = v10;
+	v_1011.y = v11;
+	v_23 = v_23 + v_1011 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_23.x + v_1011.x < v_23.x);
+	v2 = v_23.x;
+	v3 = v_23.y;
+	v10 = v_1011.x;
+	v11 = v_1011.y;
 
 	// // a = a + m[sigma[r][2*i+0]]
 	// // skip since adding 0u does nothing
@@ -5251,11 +5488,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v11 = (xor1 >> 24u) ^ (xor0 << 8u);
 
 	// a = a + b
-	o0 = v2 + v10;
-	o1 = v3 + v11;
-	o1 = o1 + select(0u, 1u, o0 < v2);
-	v2 = o0;
-	v3 = o1;
+	v_23.x = v2;
+	v_23.y = v3;
+	v_1011.x = v10;
+	v_1011.y = v11;
+	v_23 = v_23 + v_1011 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_23.x + v_1011.x < v_23.x);
+	v2 = v_23.x;
+	v3 = v_23.y;
+	v10 = v_1011.x;
+	v11 = v_1011.y;
 
 	// // a = a + m[sigma[r][2*i+1]]
 	// // skip since adding 0u does nothing
@@ -5764,9 +6005,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	*/
 
 	// a = a + b
-	v1 = v1 + v9;
-	v1 = v1 + select(0u, 1u, v0 + v8 < v0);
-	v0 = v0 + v8;
+	v_01.x = v0;
+	v_01.y = v1;
+	v_89.x = v8;
+	v_89.y = v9;
+	v_01 = v_01 + v_89 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_01.x + v_89.x < v_01.x);
+	v0 = v_01.x;
+	v1 = v_01.y;
+	v8 = v_89.x;
+	v9 = v_89.y;
 
 	// // a = a + m[sigma[r][2*i+0]]
 	// // skip since adding 0u does nothing
@@ -5785,11 +6032,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v25 = xor0;
 
 	// c = c + d
-	o0 = v16 + v24;
-	o1 = v17 + v25;
-	o1 = o1 + select(0u, 1u, o0 < v16);
-	v16 = o0;
-	v17 = o1;
+	v_1617.x = v16;
+	v_1617.y = v17;
+	v_2425.x = v24;
+	v_2425.y = v25;
+	v_1617 = v_1617 + v_2425 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_1617.x + v_2425.x < v_1617.x);
+	v16 = v_1617.x;
+	v17 = v_1617.y;
+	v24 = v_2425.x;
+	v25 = v_2425.y;
 
 	// b = rotr64(b ^ c, 24)
 	xor0 = v8 ^ v16;
@@ -5798,9 +6049,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v9 = (xor1 >> 24u) ^ (xor0 << 8u);
 
 	// a = a + b
-	v1 = v1 + v9;
-	v1 = v1 + select(0u, 1u, v0 + v8 < v0);
-	v0 = v0 + v8;
+	v_01.x = v0;
+	v_01.y = v1;
+	v_89.x = v8;
+	v_89.y = v9;
+	v_01 = v_01 + v_89 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_01.x + v_89.x < v_01.x);
+	v0 = v_01.x;
+	v1 = v_01.y;
+	v8 = v_89.x;
+	v9 = v_89.y;
 
 	// a = a + m[sigma[r][2*i+1]]
 	o0 = v0 + m4;
@@ -5816,11 +6073,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v25 = (xor1 >> 16u) ^ (xor0 << 16u);
 
 	// c = c + d
-	o0 = v16 + v24;
-	o1 = v17 + v25;
-	o1 = o1 + select(0u, 1u, o0 < v16);
-	v16 = o0;
-	v17 = o1;
+	v_1617.x = v16;
+	v_1617.y = v17;
+	v_2425.x = v24;
+	v_2425.y = v25;
+	v_1617 = v_1617 + v_2425 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_1617.x + v_2425.x < v_1617.x);
+	v16 = v_1617.x;
+	v17 = v_1617.y;
+	v24 = v_2425.x;
+	v25 = v_2425.y;
 
 	// b = rotr64(b ^ c, 63)
 	xor0 = v8 ^ v16;
@@ -5837,11 +6098,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	*/
 
 	// a = a + b
-	o0 = v2 + v10;
-	o1 = v3 + v11;
-	o1 = o1 + select(0u, 1u, o0 < v2);
-	v2 = o0;
-	v3 = o1;
+	v_23.x = v2;
+	v_23.y = v3;
+	v_1011.x = v10;
+	v_1011.y = v11;
+	v_23 = v_23 + v_1011 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_23.x + v_1011.x < v_23.x);
+	v2 = v_23.x;
+	v3 = v_23.y;
+	v10 = v_1011.x;
+	v11 = v_1011.y;
 
 	// // a = a + m[sigma[r][2*i+0]]
 	// // skip since adding 0u does nothing
@@ -5873,11 +6138,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v11 = (xor1 >> 24u) ^ (xor0 << 8u);
 
 	// a = a + b
-	o0 = v2 + v10;
-	o1 = v3 + v11;
-	o1 = o1 + select(0u, 1u, o0 < v2);
-	v2 = o0;
-	v3 = o1;
+	v_23.x = v2;
+	v_23.y = v3;
+	v_1011.x = v10;
+	v_1011.y = v11;
+	v_23 = v_23 + v_1011 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_23.x + v_1011.x < v_23.x);
+	v2 = v_23.x;
+	v3 = v_23.y;
+	v10 = v_1011.x;
+	v11 = v_1011.y;
 
 	// a = a + m[sigma[r][2*i+1]]
 	o0 = v2 + m8;
@@ -6389,9 +6658,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	*/
 
 	// a = a + b
-	v1 = v1 + v9;
-	v1 = v1 + select(0u, 1u, v0 + v8 < v0);
-	v0 = v0 + v8;
+	v_01.x = v0;
+	v_01.y = v1;
+	v_89.x = v8;
+	v_89.y = v9;
+	v_01 = v_01 + v_89 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_01.x + v_89.x < v_01.x);
+	v0 = v_01.x;
+	v1 = v_01.y;
+	v8 = v_89.x;
+	v9 = v_89.y;
 
 	// a = a + m[sigma[r][2*i+0]]
 	o0 = v0 + m0;
@@ -6407,11 +6682,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v25 = xor0;
 
 	// c = c + d
-	o0 = v16 + v24;
-	o1 = v17 + v25;
-	o1 = o1 + select(0u, 1u, o0 < v16);
-	v16 = o0;
-	v17 = o1;
+	v_1617.x = v16;
+	v_1617.y = v17;
+	v_2425.x = v24;
+	v_2425.y = v25;
+	v_1617 = v_1617 + v_2425 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_1617.x + v_2425.x < v_1617.x);
+	v16 = v_1617.x;
+	v17 = v_1617.y;
+	v24 = v_2425.x;
+	v25 = v_2425.y;
 
 	// b = rotr64(b ^ c, 24)
 	xor0 = v8 ^ v16;
@@ -6420,9 +6699,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v9 = (xor1 >> 24u) ^ (xor0 << 8u);
 
 	// a = a + b
-	v1 = v1 + v9;
-	v1 = v1 + select(0u, 1u, v0 + v8 < v0);
-	v0 = v0 + v8;
+	v_01.x = v0;
+	v_01.y = v1;
+	v_89.x = v8;
+	v_89.y = v9;
+	v_01 = v_01 + v_89 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_01.x + v_89.x < v_01.x);
+	v0 = v_01.x;
+	v1 = v_01.y;
+	v8 = v_89.x;
+	v9 = v_89.y;
 
 	// a = a + m[sigma[r][2*i+1]]
 	o0 = v0 + m2;
@@ -6438,11 +6723,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v25 = (xor1 >> 16u) ^ (xor0 << 16u);
 
 	// c = c + d
-	o0 = v16 + v24;
-	o1 = v17 + v25;
-	o1 = o1 + select(0u, 1u, o0 < v16);
-	v16 = o0;
-	v17 = o1;
+	v_1617.x = v16;
+	v_1617.y = v17;
+	v_2425.x = v24;
+	v_2425.y = v25;
+	v_1617 = v_1617 + v_2425 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_1617.x + v_2425.x < v_1617.x);
+	v16 = v_1617.x;
+	v17 = v_1617.y;
+	v24 = v_2425.x;
+	v25 = v_2425.y;
 
 	// b = rotr64(b ^ c, 63)
 	xor0 = v8 ^ v16;
@@ -6459,11 +6748,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	*/
 
 	// a = a + b
-	o0 = v2 + v10;
-	o1 = v3 + v11;
-	o1 = o1 + select(0u, 1u, o0 < v2);
-	v2 = o0;
-	v3 = o1;
+	v_23.x = v2;
+	v_23.y = v3;
+	v_1011.x = v10;
+	v_1011.y = v11;
+	v_23 = v_23 + v_1011 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_23.x + v_1011.x < v_23.x);
+	v2 = v_23.x;
+	v3 = v_23.y;
+	v10 = v_1011.x;
+	v11 = v_1011.y;
 
 	// a = a + m[sigma[r][2*i+0]]
 	o0 = v2 + m4;
@@ -6492,11 +6785,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v11 = (xor1 >> 24u) ^ (xor0 << 8u);
 
 	// a = a + b
-	o0 = v2 + v10;
-	o1 = v3 + v11;
-	o1 = o1 + select(0u, 1u, o0 < v2);
-	v2 = o0;
-	v3 = o1;
+	v_23.x = v2;
+	v_23.y = v3;
+	v_1011.x = v10;
+	v_1011.y = v11;
+	v_23 = v_23 + v_1011 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_23.x + v_1011.x < v_23.x);
+	v2 = v_23.x;
+	v3 = v_23.y;
+	v10 = v_1011.x;
+	v11 = v_1011.y;
 
 	// a = a + m[sigma[r][2*i+1]]
 	o0 = v2 + m6;
@@ -7014,9 +7311,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	*/
 
 	// a = a + b
-	v1 = v1 + v9;
-	v1 = v1 + select(0u, 1u, v0 + v8 < v0);
-	v0 = v0 + v8;
+	v_01.x = v0;
+	v_01.y = v1;
+	v_89.x = v8;
+	v_89.y = v9;
+	v_01 = v_01 + v_89 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_01.x + v_89.x < v_01.x);
+	v0 = v_01.x;
+	v1 = v_01.y;
+	v8 = v_89.x;
+	v9 = v_89.y;
 
 	// // a = a + m[sigma[r][2*i+0]]
 	// // skip since adding 0u does nothing
@@ -7035,11 +7338,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v25 = xor0;
 
 	// c = c + d
-	o0 = v16 + v24;
-	o1 = v17 + v25;
-	o1 = o1 + select(0u, 1u, o0 < v16);
-	v16 = o0;
-	v17 = o1;
+	v_1617.x = v16;
+	v_1617.y = v17;
+	v_2425.x = v24;
+	v_2425.y = v25;
+	v_1617 = v_1617 + v_2425 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_1617.x + v_2425.x < v_1617.x);
+	v16 = v_1617.x;
+	v17 = v_1617.y;
+	v24 = v_2425.x;
+	v25 = v_2425.y;
 
 	// b = rotr64(b ^ c, 24)
 	xor0 = v8 ^ v16;
@@ -7048,9 +7355,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v9 = (xor1 >> 24u) ^ (xor0 << 8u);
 
 	// a = a + b
-	v1 = v1 + v9;
-	v1 = v1 + select(0u, 1u, v0 + v8 < v0);
-	v0 = v0 + v8;
+	v_01.x = v0;
+	v_01.y = v1;
+	v_89.x = v8;
+	v_89.y = v9;
+	v_01 = v_01 + v_89 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_01.x + v_89.x < v_01.x);
+	v0 = v_01.x;
+	v1 = v_01.y;
+	v8 = v_89.x;
+	v9 = v_89.y;
 
 	// // a = a + m[sigma[r][2*i+1]]
 	// // skip since adding 0u does nothing
@@ -7069,11 +7382,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v25 = (xor1 >> 16u) ^ (xor0 << 16u);
 
 	// c = c + d
-	o0 = v16 + v24;
-	o1 = v17 + v25;
-	o1 = o1 + select(0u, 1u, o0 < v16);
-	v16 = o0;
-	v17 = o1;
+	v_1617.x = v16;
+	v_1617.y = v17;
+	v_2425.x = v24;
+	v_2425.y = v25;
+	v_1617 = v_1617 + v_2425 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_1617.x + v_2425.x < v_1617.x);
+	v16 = v_1617.x;
+	v17 = v_1617.y;
+	v24 = v_2425.x;
+	v25 = v_2425.y;
 
 	// b = rotr64(b ^ c, 63)
 	xor0 = v8 ^ v16;
@@ -7090,11 +7407,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	*/
 
 	// a = a + b
-	o0 = v2 + v10;
-	o1 = v3 + v11;
-	o1 = o1 + select(0u, 1u, o0 < v2);
-	v2 = o0;
-	v3 = o1;
+	v_23.x = v2;
+	v_23.y = v3;
+	v_1011.x = v10;
+	v_1011.y = v11;
+	v_23 = v_23 + v_1011 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_23.x + v_1011.x < v_23.x);
+	v2 = v_23.x;
+	v3 = v_23.y;
+	v10 = v_1011.x;
+	v11 = v_1011.y;
 
 	// a = a + m[sigma[r][2*i+0]]
 	o0 = v2 + m8;
@@ -7123,11 +7444,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v11 = (xor1 >> 24u) ^ (xor0 << 8u);
 
 	// a = a + b
-	o0 = v2 + v10;
-	o1 = v3 + v11;
-	o1 = o1 + select(0u, 1u, o0 < v2);
-	v2 = o0;
-	v3 = o1;
+	v_23.x = v2;
+	v_23.y = v3;
+	v_1011.x = v10;
+	v_1011.y = v11;
+	v_23 = v_23 + v_1011 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_23.x + v_1011.x < v_23.x);
+	v2 = v_23.x;
+	v3 = v_23.y;
+	v10 = v_1011.x;
+	v11 = v_1011.y;
 
 	// // a = a + m[sigma[r][2*i+1]]
 	// // skip since adding 0u does nothing