Replace more scalar addition with vector addition.

diff --git a/src/shaders/compute.wgsl b/src/shaders/compute.wgsl
index c3fd9dee094b9dd32d16c10d50cc0a1aeff392a2..d3b1628bfa1db69e77831c75c5523b62ec3f34c3 100644 (file)
--- a/src/shaders/compute.wgsl
+++ b/src/shaders/compute.wgsl
@@ -619,11 +619,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        v25 = xor0;
 
        // c = c + d
-       o0 = v22 + v24;
-       o1 = v23 + v25;
-       o1 = o1 + select(0u, 1u, o0 < v22);
-       v22 = o0;
-       v23 = o1;
+       v_2223.x = v22;
+       v_2223.y = v23;
+       v_2425.x = v24;
+       v_2425.y = v25;
+       v_2223 = v_2223 + v_2425 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_2223.x + v_2425.x < v_2223.x);
+       v22 = v_2223.x;
+       v23 = v_2223.y;
+       v24 = v_2425.x;
+       v25 = v_2425.y;
 
        // b = rotr64(b ^ c, 24)
        xor0 = v12 ^ v22;
@@ -659,11 +663,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        v25 = (xor1 >> 16u) ^ (xor0 << 16u);
 
        // c = c + d
-       o0 = v22 + v24;
-       o1 = v23 + v25;
-       o1 = o1 + select(0u, 1u, o0 < v22);
-       v22 = o0;
-       v23 = o1;
+       v_2223.x = v22;
+       v_2223.y = v23;
+       v_2425.x = v24;
+       v_2425.y = v25;
+       v_2223 = v_2223 + v_2425 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_2223.x + v_2425.x < v_2223.x);
+       v22 = v_2223.x;
+       v23 = v_2223.y;
+       v24 = v_2425.x;
+       v25 = v_2425.y;
 
        // b = rotr64(b ^ c, 63)
        xor0 = v12 ^ v22;
@@ -680,11 +688,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        */
 
        // a = a + b
-       o0 = v4 + v14;
-       o1 = v5 + v15;
-       o1 = o1 + select(0u, 1u, o0 < v4);
-       v4 = o0;
-       v5 = o1;
+       v_45.x = v4;
+       v_45.y = v5;
+       v_1415.x = v14;
+       v_1415.y = v15;
+       v_45 = v_45 + v_1415 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_45.x + v_1415.x < v_45.x);
+       v4 = v_45.x;
+       v5 = v_45.y;
+       v14 = v_1415.x;
+       v15 = v_1415.y;
 
        // // a = a + m[sigma[r][2*i+0]]
        // // skip since adding 0u does nothing
@@ -716,11 +728,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        v15 = (xor1 >> 24u) ^ (xor0 << 8u);
 
        // a = a + b
-       o0 = v4 + v14;
-       o1 = v5 + v15;
-       o1 = o1 + select(0u, 1u, o0 < v4);
-       v4 = o0;
-       v5 = o1;
+       v_45.x = v4;
+       v_45.y = v5;
+       v_1415.x = v14;
+       v_1415.y = v15;
+       v_45 = v_45 + v_1415 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_45.x + v_1415.x < v_45.x);
+       v4 = v_45.x;
+       v5 = v_45.y;
+       v14 = v_1415.x;
+       v15 = v_1415.y;
 
        // // a = a + m[sigma[r][2*i+1]]
        // // skip since adding 0u does nothing
@@ -1342,11 +1358,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        v25 = xor0;
 
        // c = c + d
-       o0 = v22 + v24;
-       o1 = v23 + v25;
-       o1 = o1 + select(0u, 1u, o0 < v22);
-       v22 = o0;
-       v23 = o1;
+       v_2223.x = v22;
+       v_2223.y = v23;
+       v_2425.x = v24;
+       v_2425.y = v25;
+       v_2223 = v_2223 + v_2425 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_2223.x + v_2425.x < v_2223.x);
+       v22 = v_2223.x;
+       v23 = v_2223.y;
+       v24 = v_2425.x;
+       v25 = v_2425.y;
 
        // b = rotr64(b ^ c, 24)
        xor0 = v12 ^ v22;
@@ -1379,11 +1399,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        v25 = (xor1 >> 16u) ^ (xor0 << 16u);
 
        // c = c + d
-       o0 = v22 + v24;
-       o1 = v23 + v25;
-       o1 = o1 + select(0u, 1u, o0 < v22);
-       v22 = o0;
-       v23 = o1;
+       v_2223.x = v22;
+       v_2223.y = v23;
+       v_2425.x = v24;
+       v_2425.y = v25;
+       v_2223 = v_2223 + v_2425 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_2223.x + v_2425.x < v_2223.x);
+       v22 = v_2223.x;
+       v23 = v_2223.y;
+       v24 = v_2425.x;
+       v25 = v_2425.y;
 
        // b = rotr64(b ^ c, 63)
        xor0 = v12 ^ v22;
@@ -1400,11 +1424,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        */
 
        // a = a + b
-       o0 = v4 + v14;
-       o1 = v5 + v15;
-       o1 = o1 + select(0u, 1u, o0 < v4);
-       v4 = o0;
-       v5 = o1;
+       v_45.x = v4;
+       v_45.y = v5;
+       v_1415.x = v14;
+       v_1415.y = v15;
+       v_45 = v_45 + v_1415 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_45.x + v_1415.x < v_45.x);
+       v4 = v_45.x;
+       v5 = v_45.y;
+       v14 = v_1415.x;
+       v15 = v_1415.y;
 
        // // a = a + m[sigma[r][2*i+0]]
        // // skip since adding 0u does nothing
@@ -1436,11 +1464,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        v15 = (xor1 >> 24u) ^ (xor0 << 8u);
 
        // a = a + b
-       o0 = v4 + v14;
-       o1 = v5 + v15;
-       o1 = o1 + select(0u, 1u, o0 < v4);
-       v4 = o0;
-       v5 = o1;
+       v_45.x = v4;
+       v_45.y = v5;
+       v_1415.x = v14;
+       v_1415.y = v15;
+       v_45 = v_45 + v_1415 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_45.x + v_1415.x < v_45.x);
+       v4 = v_45.x;
+       v5 = v_45.y;
+       v14 = v_1415.x;
+       v15 = v_1415.y;
 
        // // a = a + m[sigma[r][2*i+1]]
        // // skip since adding 0u does nothing
@@ -2059,11 +2091,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        v25 = xor0;
 
        // c = c + d
-       o0 = v22 + v24;
-       o1 = v23 + v25;
-       o1 = o1 + select(0u, 1u, o0 < v22);
-       v22 = o0;
-       v23 = o1;
+       v_2223.x = v22;
+       v_2223.y = v23;
+       v_2425.x = v24;
+       v_2425.y = v25;
+       v_2223 = v_2223 + v_2425 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_2223.x + v_2425.x < v_2223.x);
+       v22 = v_2223.x;
+       v23 = v_2223.y;
+       v24 = v_2425.x;
+       v25 = v_2425.y;
 
        // b = rotr64(b ^ c, 24)
        xor0 = v12 ^ v22;
@@ -2099,11 +2135,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        v25 = (xor1 >> 16u) ^ (xor0 << 16u);
 
        // c = c + d
-       o0 = v22 + v24;
-       o1 = v23 + v25;
-       o1 = o1 + select(0u, 1u, o0 < v22);
-       v22 = o0;
-       v23 = o1;
+       v_2223.x = v22;
+       v_2223.y = v23;
+       v_2425.x = v24;
+       v_2425.y = v25;
+       v_2223 = v_2223 + v_2425 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_2223.x + v_2425.x < v_2223.x);
+       v22 = v_2223.x;
+       v23 = v_2223.y;
+       v24 = v_2425.x;
+       v25 = v_2425.y;
 
        // b = rotr64(b ^ c, 63)
        xor0 = v12 ^ v22;
@@ -2120,11 +2160,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        */
 
        // a = a + b
-       o0 = v4 + v14;
-       o1 = v5 + v15;
-       o1 = o1 + select(0u, 1u, o0 < v4);
-       v4 = o0;
-       v5 = o1;
+       v_45.x = v4;
+       v_45.y = v5;
+       v_1415.x = v14;
+       v_1415.y = v15;
+       v_45 = v_45 + v_1415 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_45.x + v_1415.x < v_45.x);
+       v4 = v_45.x;
+       v5 = v_45.y;
+       v14 = v_1415.x;
+       v15 = v_1415.y;
 
        // // a = a + m[sigma[r][2*i+0]]
        // // skip since adding 0u does nothing
@@ -2156,11 +2200,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        v15 = (xor1 >> 24u) ^ (xor0 << 8u);
 
        // a = a + b
-       o0 = v4 + v14;
-       o1 = v5 + v15;
-       o1 = o1 + select(0u, 1u, o0 < v4);
-       v4 = o0;
-       v5 = o1;
+       v_45.x = v4;
+       v_45.y = v5;
+       v_1415.x = v14;
+       v_1415.y = v15;
+       v_45 = v_45 + v_1415 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_45.x + v_1415.x < v_45.x);
+       v4 = v_45.x;
+       v5 = v_45.y;
+       v14 = v_1415.x;
+       v15 = v_1415.y;
 
        // a = a + m[sigma[r][2*i+1]]
        o0 = v4 + m2;
@@ -2776,11 +2824,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        v25 = xor0;
 
        // c = c + d
-       o0 = v22 + v24;
-       o1 = v23 + v25;
-       o1 = o1 + select(0u, 1u, o0 < v22);
-       v22 = o0;
-       v23 = o1;
+       v_2223.x = v22;
+       v_2223.y = v23;
+       v_2425.x = v24;
+       v_2425.y = v25;
+       v_2223 = v_2223 + v_2425 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_2223.x + v_2425.x < v_2223.x);
+       v22 = v_2223.x;
+       v23 = v_2223.y;
+       v24 = v_2425.x;
+       v25 = v_2425.y;
 
        // b = rotr64(b ^ c, 24)
        xor0 = v12 ^ v22;
@@ -2816,11 +2868,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        v25 = (xor1 >> 16u) ^ (xor0 << 16u);
 
        // c = c + d
-       o0 = v22 + v24;
-       o1 = v23 + v25;
-       o1 = o1 + select(0u, 1u, o0 < v22);
-       v22 = o0;
-       v23 = o1;
+       v_2223.x = v22;
+       v_2223.y = v23;
+       v_2425.x = v24;
+       v_2425.y = v25;
+       v_2223 = v_2223 + v_2425 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_2223.x + v_2425.x < v_2223.x);
+       v22 = v_2223.x;
+       v23 = v_2223.y;
+       v24 = v_2425.x;
+       v25 = v_2425.y;
 
        // b = rotr64(b ^ c, 63)
        xor0 = v12 ^ v22;
@@ -2837,11 +2893,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        */
 
        // a = a + b
-       o0 = v4 + v14;
-       o1 = v5 + v15;
-       o1 = o1 + select(0u, 1u, o0 < v4);
-       v4 = o0;
-       v5 = o1;
+       v_45.x = v4;
+       v_45.y = v5;
+       v_1415.x = v14;
+       v_1415.y = v15;
+       v_45 = v_45 + v_1415 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_45.x + v_1415.x < v_45.x);
+       v4 = v_45.x;
+       v5 = v_45.y;
+       v14 = v_1415.x;
+       v15 = v_1415.y;
 
        // a = a + m[sigma[r][2*i+0]]
        o0 = v4 + m8;
@@ -2870,11 +2930,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        v15 = (xor1 >> 24u) ^ (xor0 << 8u);
 
        // a = a + b
-       o0 = v4 + v14;
-       o1 = v5 + v15;
-       o1 = o1 + select(0u, 1u, o0 < v4);
-       v4 = o0;
-       v5 = o1;
+       v_45.x = v4;
+       v_45.y = v5;
+       v_1415.x = v14;
+       v_1415.y = v15;
+       v_45 = v_45 + v_1415 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_45.x + v_1415.x < v_45.x);
+       v4 = v_45.x;
+       v5 = v_45.y;
+       v14 = v_1415.x;
+       v15 = v_1415.y;
 
        // a = a + m[sigma[r][2*i+1]]
        o0 = v4 + m0;
@@ -3490,11 +3554,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        v25 = xor0;
 
        // c = c + d
-       o0 = v22 + v24;
-       o1 = v23 + v25;
-       o1 = o1 + select(0u, 1u, o0 < v22);
-       v22 = o0;
-       v23 = o1;
+       v_2223.x = v22;
+       v_2223.y = v23;
+       v_2425.x = v24;
+       v_2425.y = v25;
+       v_2223 = v_2223 + v_2425 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_2223.x + v_2425.x < v_2223.x);
+       v22 = v_2223.x;
+       v23 = v_2223.y;
+       v24 = v_2425.x;
+       v25 = v_2425.y;
 
        // b = rotr64(b ^ c, 24)
        xor0 = v12 ^ v22;
@@ -3530,11 +3598,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        v25 = (xor1 >> 16u) ^ (xor0 << 16u);
 
        // c = c + d
-       o0 = v22 + v24;
-       o1 = v23 + v25;
-       o1 = o1 + select(0u, 1u, o0 < v22);
-       v22 = o0;
-       v23 = o1;
+       v_2223.x = v22;
+       v_2223.y = v23;
+       v_2425.x = v24;
+       v_2425.y = v25;
+       v_2223 = v_2223 + v_2425 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_2223.x + v_2425.x < v_2223.x);
+       v22 = v_2223.x;
+       v23 = v_2223.y;
+       v24 = v_2425.x;
+       v25 = v_2425.y;
 
        // b = rotr64(b ^ c, 63)
        xor0 = v12 ^ v22;
@@ -3551,11 +3623,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        */
 
        // a = a + b
-       o0 = v4 + v14;
-       o1 = v5 + v15;
-       o1 = o1 + select(0u, 1u, o0 < v4);
-       v4 = o0;
-       v5 = o1;
+       v_45.x = v4;
+       v_45.y = v5;
+       v_1415.x = v14;
+       v_1415.y = v15;
+       v_45 = v_45 + v_1415 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_45.x + v_1415.x < v_45.x);
+       v4 = v_45.x;
+       v5 = v_45.y;
+       v14 = v_1415.x;
+       v15 = v_1415.y;
 
        // // a = a + m[sigma[r][2*i+0]]
        // // skip since adding 0u does nothing
@@ -3587,11 +3663,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        v15 = (xor1 >> 24u) ^ (xor0 << 8u);
 
        // a = a + b
-       o0 = v4 + v14;
-       o1 = v5 + v15;
-       o1 = o1 + select(0u, 1u, o0 < v4);
-       v4 = o0;
-       v5 = o1;
+       v_45.x = v4;
+       v_45.y = v5;
+       v_1415.x = v14;
+       v_1415.y = v15;
+       v_45 = v_45 + v_1415 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_45.x + v_1415.x < v_45.x);
+       v4 = v_45.x;
+       v5 = v_45.y;
+       v14 = v_1415.x;
+       v15 = v_1415.y;
 
        // // a = a + m[sigma[r][2*i+1]]
        // // skip since adding 0u does nothing
@@ -4207,11 +4287,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        v25 = xor0;
 
        // c = c + d
-       o0 = v22 + v24;
-       o1 = v23 + v25;
-       o1 = o1 + select(0u, 1u, o0 < v22);
-       v22 = o0;
-       v23 = o1;
+       v_2223.x = v22;
+       v_2223.y = v23;
+       v_2425.x = v24;
+       v_2425.y = v25;
+       v_2223 = v_2223 + v_2425 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_2223.x + v_2425.x < v_2223.x);
+       v22 = v_2223.x;
+       v23 = v_2223.y;
+       v24 = v_2425.x;
+       v25 = v_2425.y;
 
        // b = rotr64(b ^ c, 24)
        xor0 = v12 ^ v22;
@@ -4247,11 +4331,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        v25 = (xor1 >> 16u) ^ (xor0 << 16u);
 
        // c = c + d
-       o0 = v22 + v24;
-       o1 = v23 + v25;
-       o1 = o1 + select(0u, 1u, o0 < v22);
-       v22 = o0;
-       v23 = o1;
+       v_2223.x = v22;
+       v_2223.y = v23;
+       v_2425.x = v24;
+       v_2425.y = v25;
+       v_2223 = v_2223 + v_2425 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_2223.x + v_2425.x < v_2223.x);
+       v22 = v_2223.x;
+       v23 = v_2223.y;
+       v24 = v_2425.x;
+       v25 = v_2425.y;
 
        // b = rotr64(b ^ c, 63)
        xor0 = v12 ^ v22;
@@ -4268,11 +4356,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        */
 
        // a = a + b
-       o0 = v4 + v14;
-       o1 = v5 + v15;
-       o1 = o1 + select(0u, 1u, o0 < v4);
-       v4 = o0;
-       v5 = o1;
+       v_45.x = v4;
+       v_45.y = v5;
+       v_1415.x = v14;
+       v_1415.y = v15;
+       v_45 = v_45 + v_1415 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_45.x + v_1415.x < v_45.x);
+       v4 = v_45.x;
+       v5 = v_45.y;
+       v14 = v_1415.x;
+       v15 = v_1415.y;
 
        // // a = a + m[sigma[r][2*i+0]]
        // // skip since adding 0u does nothing
@@ -4304,11 +4396,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        v15 = (xor1 >> 24u) ^ (xor0 << 8u);
 
        // a = a + b
-       o0 = v4 + v14;
-       o1 = v5 + v15;
-       o1 = o1 + select(0u, 1u, o0 < v4);
-       v4 = o0;
-       v5 = o1;
+       v_45.x = v4;
+       v_45.y = v5;
+       v_1415.x = v14;
+       v_1415.y = v15;
+       v_45 = v_45 + v_1415 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_45.x + v_1415.x < v_45.x);
+       v4 = v_45.x;
+       v5 = v_45.y;
+       v14 = v_1415.x;
+       v15 = v_1415.y;
 
        // // a = a + m[sigma[r][2*i+1]]
        // // skip since adding 0u does nothing
@@ -4927,11 +5023,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        v25 = xor0;
 
        // c = c + d
-       o0 = v22 + v24;
-       o1 = v23 + v25;
-       o1 = o1 + select(0u, 1u, o0 < v22);
-       v22 = o0;
-       v23 = o1;
+       v_2223.x = v22;
+       v_2223.y = v23;
+       v_2425.x = v24;
+       v_2425.y = v25;
+       v_2223 = v_2223 + v_2425 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_2223.x + v_2425.x < v_2223.x);
+       v22 = v_2223.x;
+       v23 = v_2223.y;
+       v24 = v_2425.x;
+       v25 = v_2425.y;
 
        // b = rotr64(b ^ c, 24)
        xor0 = v12 ^ v22;
@@ -4964,11 +5064,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        v25 = (xor1 >> 16u) ^ (xor0 << 16u);
 
        // c = c + d
-       o0 = v22 + v24;
-       o1 = v23 + v25;
-       o1 = o1 + select(0u, 1u, o0 < v22);
-       v22 = o0;
-       v23 = o1;
+       v_2223.x = v22;
+       v_2223.y = v23;
+       v_2425.x = v24;
+       v_2425.y = v25;
+       v_2223 = v_2223 + v_2425 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_2223.x + v_2425.x < v_2223.x);
+       v22 = v_2223.x;
+       v23 = v_2223.y;
+       v24 = v_2425.x;
+       v25 = v_2425.y;
 
        // b = rotr64(b ^ c, 63)
        xor0 = v12 ^ v22;
@@ -4985,11 +5089,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        */
 
        // a = a + b
-       o0 = v4 + v14;
-       o1 = v5 + v15;
-       o1 = o1 + select(0u, 1u, o0 < v4);
-       v4 = o0;
-       v5 = o1;
+       v_45.x = v4;
+       v_45.y = v5;
+       v_1415.x = v14;
+       v_1415.y = v15;
+       v_45 = v_45 + v_1415 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_45.x + v_1415.x < v_45.x);
+       v4 = v_45.x;
+       v5 = v_45.y;
+       v14 = v_1415.x;
+       v15 = v_1415.y;
 
        // // a = a + m[sigma[r][2*i+0]]
        // // skip since adding 0u does nothing
@@ -5021,11 +5129,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        v15 = (xor1 >> 24u) ^ (xor0 << 8u);
 
        // a = a + b
-       o0 = v4 + v14;
-       o1 = v5 + v15;
-       o1 = o1 + select(0u, 1u, o0 < v4);
-       v4 = o0;
-       v5 = o1;
+       v_45.x = v4;
+       v_45.y = v5;
+       v_1415.x = v14;
+       v_1415.y = v15;
+       v_45 = v_45 + v_1415 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_45.x + v_1415.x < v_45.x);
+       v4 = v_45.x;
+       v5 = v_45.y;
+       v14 = v_1415.x;
+       v15 = v_1415.y;
 
        // a = a + m[sigma[r][2*i+1]]
        o0 = v4 + m4;
@@ -5644,11 +5756,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        v25 = xor0;
 
        // c = c + d
-       o0 = v22 + v24;
-       o1 = v23 + v25;
-       o1 = o1 + select(0u, 1u, o0 < v22);
-       v22 = o0;
-       v23 = o1;
+       v_2223.x = v22;
+       v_2223.y = v23;
+       v_2425.x = v24;
+       v_2425.y = v25;
+       v_2223 = v_2223 + v_2425 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_2223.x + v_2425.x < v_2223.x);
+       v22 = v_2223.x;
+       v23 = v_2223.y;
+       v24 = v_2425.x;
+       v25 = v_2425.y;
 
        // b = rotr64(b ^ c, 24)
        xor0 = v12 ^ v22;
@@ -5681,11 +5797,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        v25 = (xor1 >> 16u) ^ (xor0 << 16u);
 
        // c = c + d
-       o0 = v22 + v24;
-       o1 = v23 + v25;
-       o1 = o1 + select(0u, 1u, o0 < v22);
-       v22 = o0;
-       v23 = o1;
+       v_2223.x = v22;
+       v_2223.y = v23;
+       v_2425.x = v24;
+       v_2425.y = v25;
+       v_2223 = v_2223 + v_2425 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_2223.x + v_2425.x < v_2223.x);
+       v22 = v_2223.x;
+       v23 = v_2223.y;
+       v24 = v_2425.x;
+       v25 = v_2425.y;
 
        // b = rotr64(b ^ c, 63)
        xor0 = v12 ^ v22;
@@ -5702,11 +5822,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        */
 
        // a = a + b
-       o0 = v4 + v14;
-       o1 = v5 + v15;
-       o1 = o1 + select(0u, 1u, o0 < v4);
-       v4 = o0;
-       v5 = o1;
+       v_45.x = v4;
+       v_45.y = v5;
+       v_1415.x = v14;
+       v_1415.y = v15;
+       v_45 = v_45 + v_1415 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_45.x + v_1415.x < v_45.x);
+       v4 = v_45.x;
+       v5 = v_45.y;
+       v14 = v_1415.x;
+       v15 = v_1415.y;
 
        // // a = a + m[sigma[r][2*i+0]]
        // // skip since adding 0u does nothing
@@ -5738,11 +5862,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        v15 = (xor1 >> 24u) ^ (xor0 << 8u);
 
        // a = a + b
-       o0 = v4 + v14;
-       o1 = v5 + v15;
-       o1 = o1 + select(0u, 1u, o0 < v4);
-       v4 = o0;
-       v5 = o1;
+       v_45.x = v4;
+       v_45.y = v5;
+       v_1415.x = v14;
+       v_1415.y = v15;
+       v_45 = v_45 + v_1415 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_45.x + v_1415.x < v_45.x);
+       v4 = v_45.x;
+       v5 = v_45.y;
+       v14 = v_1415.x;
+       v15 = v_1415.y;
 
        // // a = a + m[sigma[r][2*i+1]]
        // // skip since adding 0u does nothing
@@ -6361,11 +6489,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        v25 = xor0;
 
        // c = c + d
-       o0 = v22 + v24;
-       o1 = v23 + v25;
-       o1 = o1 + select(0u, 1u, o0 < v22);
-       v22 = o0;
-       v23 = o1;
+       v_2223.x = v22;
+       v_2223.y = v23;
+       v_2425.x = v24;
+       v_2425.y = v25;
+       v_2223 = v_2223 + v_2425 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_2223.x + v_2425.x < v_2223.x);
+       v22 = v_2223.x;
+       v23 = v_2223.y;
+       v24 = v_2425.x;
+       v25 = v_2425.y;
 
        // b = rotr64(b ^ c, 24)
        xor0 = v12 ^ v22;
@@ -6401,11 +6533,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        v25 = (xor1 >> 16u) ^ (xor0 << 16u);
 
        // c = c + d
-       o0 = v22 + v24;
-       o1 = v23 + v25;
-       o1 = o1 + select(0u, 1u, o0 < v22);
-       v22 = o0;
-       v23 = o1;
+       v_2223.x = v22;
+       v_2223.y = v23;
+       v_2425.x = v24;
+       v_2425.y = v25;
+       v_2223 = v_2223 + v_2425 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_2223.x + v_2425.x < v_2223.x);
+       v22 = v_2223.x;
+       v23 = v_2223.y;
+       v24 = v_2425.x;
+       v25 = v_2425.y;
 
        // b = rotr64(b ^ c, 63)
        xor0 = v12 ^ v22;
@@ -6422,11 +6558,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        */
 
        // a = a + b
-       o0 = v4 + v14;
-       o1 = v5 + v15;
-       o1 = o1 + select(0u, 1u, o0 < v4);
-       v4 = o0;
-       v5 = o1;
+       v_45.x = v4;
+       v_45.y = v5;
+       v_1415.x = v14;
+       v_1415.y = v15;
+       v_45 = v_45 + v_1415 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_45.x + v_1415.x < v_45.x);
+       v4 = v_45.x;
+       v5 = v_45.y;
+       v14 = v_1415.x;
+       v15 = v_1415.y;
 
        // a = a + m[sigma[r][2*i+0]]
        o0 = v4 + m2;
@@ -6455,11 +6595,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        v15 = (xor1 >> 24u) ^ (xor0 << 8u);
 
        // a = a + b
-       o0 = v4 + v14;
-       o1 = v5 + v15;
-       o1 = o1 + select(0u, 1u, o0 < v4);
-       v4 = o0;
-       v5 = o1;
+       v_45.x = v4;
+       v_45.y = v5;
+       v_1415.x = v14;
+       v_1415.y = v15;
+       v_45 = v_45 + v_1415 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_45.x + v_1415.x < v_45.x);
+       v4 = v_45.x;
+       v5 = v_45.y;
+       v14 = v_1415.x;
+       v15 = v_1415.y;
 
        // a = a + m[sigma[r][2*i+1]]
        o0 = v4 + m8;
@@ -7078,11 +7222,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        v25 = xor0;
 
        // c = c + d
-       o0 = v22 + v24;
-       o1 = v23 + v25;
-       o1 = o1 + select(0u, 1u, o0 < v22);
-       v22 = o0;
-       v23 = o1;
+       v_2223.x = v22;
+       v_2223.y = v23;
+       v_2425.x = v24;
+       v_2425.y = v25;
+       v_2223 = v_2223 + v_2425 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_2223.x + v_2425.x < v_2223.x);
+       v22 = v_2223.x;
+       v23 = v_2223.y;
+       v24 = v_2425.x;
+       v25 = v_2425.y;
 
        // b = rotr64(b ^ c, 24)
        xor0 = v12 ^ v22;
@@ -7118,11 +7266,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        v25 = (xor1 >> 16u) ^ (xor0 << 16u);
 
        // c = c + d
-       o0 = v22 + v24;
-       o1 = v23 + v25;
-       o1 = o1 + select(0u, 1u, o0 < v22);
-       v22 = o0;
-       v23 = o1;
+       v_2223.x = v22;
+       v_2223.y = v23;
+       v_2425.x = v24;
+       v_2425.y = v25;
+       v_2223 = v_2223 + v_2425 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_2223.x + v_2425.x < v_2223.x);
+       v22 = v_2223.x;
+       v23 = v_2223.y;
+       v24 = v_2425.x;
+       v25 = v_2425.y;
 
        // b = rotr64(b ^ c, 63)
        xor0 = v12 ^ v22;
@@ -7139,11 +7291,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        */
 
        // a = a + b
-       o0 = v4 + v14;
-       o1 = v5 + v15;
-       o1 = o1 + select(0u, 1u, o0 < v4);
-       v4 = o0;
-       v5 = o1;
+       v_45.x = v4;
+       v_45.y = v5;
+       v_1415.x = v14;
+       v_1415.y = v15;
+       v_45 = v_45 + v_1415 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_45.x + v_1415.x < v_45.x);
+       v4 = v_45.x;
+       v5 = v_45.y;
+       v14 = v_1415.x;
+       v15 = v_1415.y;
 
        // a = a + m[sigma[r][2*i+0]]
        o0 = v4 + m6;
@@ -7172,11 +7328,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        v15 = (xor1 >> 24u) ^ (xor0 << 8u);
 
        // a = a + b
-       o0 = v4 + v14;
-       o1 = v5 + v15;
-       o1 = o1 + select(0u, 1u, o0 < v4);
-       v4 = o0;
-       v5 = o1;
+       v_45.x = v4;
+       v_45.y = v5;
+       v_1415.x = v14;
+       v_1415.y = v15;
+       v_45 = v_45 + v_1415 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_45.x + v_1415.x < v_45.x);
+       v4 = v_45.x;
+       v5 = v_45.y;
+       v14 = v_1415.x;
+       v15 = v_1415.y;
 
        // // a = a + m[sigma[r][2*i+1]]
        // // skip since adding 0u does nothing
@@ -7789,11 +7949,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        v25 = xor0;
 
        // c = c + d
-       o0 = v22 + v24;
-       o1 = v23 + v25;
-       o1 = o1 + select(0u, 1u, o0 < v22);
-       v22 = o0;
-       v23 = o1;
+       v_2223.x = v22;
+       v_2223.y = v23;
+       v_2425.x = v24;
+       v_2425.y = v25;
+       v_2223 = v_2223 + v_2425 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_2223.x + v_2425.x < v_2223.x);
+       v22 = v_2223.x;
+       v23 = v_2223.y;
+       v24 = v_2425.x;
+       v25 = v_2425.y;
 
        // b = rotr64(b ^ c, 24)
        xor0 = v12 ^ v22;
@@ -7829,11 +7993,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        v25 = (xor1 >> 16u) ^ (xor0 << 16u);
 
        // c = c + d
-       o0 = v22 + v24;
-       o1 = v23 + v25;
-       o1 = o1 + select(0u, 1u, o0 < v22);
-       v22 = o0;
-       v23 = o1;
+       v_2223.x = v22;
+       v_2223.y = v23;
+       v_2425.x = v24;
+       v_2425.y = v25;
+       v_2223 = v_2223 + v_2425 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_2223.x + v_2425.x < v_2223.x);
+       v22 = v_2223.x;
+       v23 = v_2223.y;
+       v24 = v_2425.x;
+       v25 = v_2425.y;
 
        // b = rotr64(b ^ c, 63)
        xor0 = v12 ^ v22;
@@ -7850,11 +8018,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        */
 
        // a = a + b
-       o0 = v4 + v14;
-       o1 = v5 + v15;
-       o1 = o1 + select(0u, 1u, o0 < v4);
-       v4 = o0;
-       v5 = o1;
+       v_45.x = v4;
+       v_45.y = v5;
+       v_1415.x = v14;
+       v_1415.y = v15;
+       v_45 = v_45 + v_1415 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_45.x + v_1415.x < v_45.x);
+       v4 = v_45.x;
+       v5 = v_45.y;
+       v14 = v_1415.x;
+       v15 = v_1415.y;
 
        // // a = a + m[sigma[r][2*i+0]]
        // // skip since adding 0u does nothing
@@ -7886,11 +8058,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        v15 = (xor1 >> 24u) ^ (xor0 << 8u);
 
        // a = a + b
-       o0 = v4 + v14;
-       o1 = v5 + v15;
-       o1 = o1 + select(0u, 1u, o0 < v4);
-       v4 = o0;
-       v5 = o1;
+       v_45.x = v4;
+       v_45.y = v5;
+       v_1415.x = v14;
+       v_1415.y = v15;
+       v_45 = v_45 + v_1415 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_45.x + v_1415.x < v_45.x);
+       v4 = v_45.x;
+       v5 = v_45.y;
+       v14 = v_1415.x;
+       v15 = v_1415.y;
 
        // // a = a + m[sigma[r][2*i+1]]
        // // skip since adding 0u does nothing
@@ -8512,11 +8688,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        v25 = xor0;
 
        // c = c + d
-       o0 = v22 + v24;
-       o1 = v23 + v25;
-       o1 = o1 + select(0u, 1u, o0 < v22);
-       v22 = o0;
-       v23 = o1;
+       v_2223.x = v22;
+       v_2223.y = v23;
+       v_2425.x = v24;
+       v_2425.y = v25;
+       v_2223 = v_2223 + v_2425 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_2223.x + v_2425.x < v_2223.x);
+       v22 = v_2223.x;
+       v23 = v_2223.y;
+       v24 = v_2425.x;
+       v25 = v_2425.y;
 
        // b = rotr64(b ^ c, 24)
        xor0 = v12 ^ v22;
@@ -8549,11 +8729,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        v25 = (xor1 >> 16u) ^ (xor0 << 16u);
 
        // c = c + d
-       o0 = v22 + v24;
-       o1 = v23 + v25;
-       o1 = o1 + select(0u, 1u, o0 < v22);
-       v22 = o0;
-       v23 = o1;
+       v_2223.x = v22;
+       v_2223.y = v23;
+       v_2425.x = v24;
+       v_2425.y = v25;
+       v_2223 = v_2223 + v_2425 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_2223.x + v_2425.x < v_2223.x);
+       v22 = v_2223.x;
+       v23 = v_2223.y;
+       v24 = v_2425.x;
+       v25 = v_2425.y;
 
        // b = rotr64(b ^ c, 63)
        xor0 = v12 ^ v22;
@@ -8570,11 +8754,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        */
 
        // a = a + b
-       o0 = v4 + v14;
-       o1 = v5 + v15;
-       o1 = o1 + select(0u, 1u, o0 < v4);
-       v4 = o0;
-       v5 = o1;
+       v_45.x = v4;
+       v_45.y = v5;
+       v_1415.x = v14;
+       v_1415.y = v15;
+       v_45 = v_45 + v_1415 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_45.x + v_1415.x < v_45.x);
+       v4 = v_45.x;
+       v5 = v_45.y;
+       v14 = v_1415.x;
+       v15 = v_1415.y;
 
        // // a = a + m[sigma[r][2*i+0]]
        // // skip since adding 0u does nothing
@@ -8606,11 +8794,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        v15 = (xor1 >> 24u) ^ (xor0 << 8u);
 
        // a = a + b
-       o0 = v4 + v14;
-       o1 = v5 + v15;
-       o1 = o1 + select(0u, 1u, o0 < v4);
-       v4 = o0;
-       v5 = o1;
+       v_45.x = v4;
+       v_45.y = v5;
+       v_1415.x = v14;
+       v_1415.y = v15;
+       v_45 = v_45 + v_1415 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_45.x + v_1415.x < v_45.x);
+       v4 = v_45.x;
+       v5 = v_45.y;
+       v14 = v_1415.x;
+       v15 = v_1415.y;
 
        // // a = a + m[sigma[r][2*i+1]]
        // // skip since adding 0u does nothing