Replace more scalar addition with vector addition.

diff --git a/src/shaders/compute.wgsl b/src/shaders/compute.wgsl
index ec65fffe826a1c7e8cc59667cde5eaf8f78dd897..3b9c769d5129a9853038a83f98a1ec709f7c8115 100644 (file)
--- a/src/shaders/compute.wgsl
+++ b/src/shaders/compute.wgsl
@@ -400,11 +400,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        */
 
        // a = a + b
-       o0 = v6 + v14;
-       o1 = v7 + v15;
-       o1 = o1 + select(0u, 1u, o0 < v6);
-       v6 = o0;
-       v7 = o1;
+       v_67.x = v6;
+       v_67.y = v7;
+       v_1415.x = v14;
+       v_1415.y = v15;
+       v_67 = v_67 + v_1415 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_67.x + v_1415.x < v_67.x);
+       v6 = v_67.x;
+       v7 = v_67.y;
+       v14 = v_1415.x;
+       v15 = v_1415.y;
 
        // // a = a + m[sigma[r][2*i+0]]
        // // skip since adding 0u does nothing
@@ -436,11 +440,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        v15 = (xor1 >> 24u) ^ (xor0 << 8u);
 
        // a = a + b
-       o0 = v6 + v14;
-       o1 = v7 + v15;
-       o1 = o1 + select(0u, 1u, o0 < v6);
-       v6 = o0;
-       v7 = o1;
+       v_67.x = v6;
+       v_67.y = v7;
+       v_1415.x = v14;
+       v_1415.y = v15;
+       v_67 = v_67 + v_1415 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_67.x + v_1415.x < v_67.x);
+       v6 = v_67.x;
+       v7 = v_67.y;
+       v14 = v_1415.x;
+       v15 = v_1415.y;
 
        // // a = a + m[sigma[r][2*i+1]]
        // // skip since adding 0u does nothing
@@ -1089,11 +1097,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        */
 
        // a = a + b
-       o0 = v6 + v14;
-       o1 = v7 + v15;
-       o1 = o1 + select(0u, 1u, o0 < v6);
-       v6 = o0;
-       v7 = o1;
+       v_67.x = v6;
+       v_67.y = v7;
+       v_1415.x = v14;
+       v_1415.y = v15;
+       v_67 = v_67 + v_1415 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_67.x + v_1415.x < v_67.x);
+       v6 = v_67.x;
+       v7 = v_67.y;
+       v14 = v_1415.x;
+       v15 = v_1415.y;
 
        // // a = a + m[sigma[r][2*i+0]]
        // // skip since adding 0u does nothing
@@ -1125,11 +1137,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        v15 = (xor1 >> 24u) ^ (xor0 << 8u);
 
        // a = a + b
-       o0 = v6 + v14;
-       o1 = v7 + v15;
-       o1 = o1 + select(0u, 1u, o0 < v6);
-       v6 = o0;
-       v7 = o1;
+       v_67.x = v6;
+       v_67.y = v7;
+       v_1415.x = v14;
+       v_1415.y = v15;
+       v_67 = v_67 + v_1415 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_67.x + v_1415.x < v_67.x);
+       v6 = v_67.x;
+       v7 = v_67.y;
+       v14 = v_1415.x;
+       v15 = v_1415.y;
 
        // // a = a + m[sigma[r][2*i+1]]
        // // skip since adding 0u does nothing
@@ -1763,11 +1779,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        */
 
        // a = a + b
-       o0 = v6 + v14;
-       o1 = v7 + v15;
-       o1 = o1 + select(0u, 1u, o0 < v6);
-       v6 = o0;
-       v7 = o1;
+       v_67.x = v6;
+       v_67.y = v7;
+       v_1415.x = v14;
+       v_1415.y = v15;
+       v_67 = v_67 + v_1415 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_67.x + v_1415.x < v_67.x);
+       v6 = v_67.x;
+       v7 = v_67.y;
+       v14 = v_1415.x;
+       v15 = v_1415.y;
 
        // // a = a + m[sigma[r][2*i+0]]
        // // skip since adding 0u does nothing
@@ -1799,11 +1819,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        v15 = (xor1 >> 24u) ^ (xor0 << 8u);
 
        // a = a + b
-       o0 = v6 + v14;
-       o1 = v7 + v15;
-       o1 = o1 + select(0u, 1u, o0 < v6);
-       v6 = o0;
-       v7 = o1;
+       v_67.x = v6;
+       v_67.y = v7;
+       v_1415.x = v14;
+       v_1415.y = v15;
+       v_67 = v_67 + v_1415 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_67.x + v_1415.x < v_67.x);
+       v6 = v_67.x;
+       v7 = v_67.y;
+       v14 = v_1415.x;
+       v15 = v_1415.y;
 
        // // a = a + m[sigma[r][2*i+1]]
        // // skip since adding 0u does nothing
@@ -2440,11 +2464,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        */
 
        // a = a + b
-       o0 = v6 + v14;
-       o1 = v7 + v15;
-       o1 = o1 + select(0u, 1u, o0 < v6);
-       v6 = o0;
-       v7 = o1;
+       v_67.x = v6;
+       v_67.y = v7;
+       v_1415.x = v14;
+       v_1415.y = v15;
+       v_67 = v_67 + v_1415 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_67.x + v_1415.x < v_67.x);
+       v6 = v_67.x;
+       v7 = v_67.y;
+       v14 = v_1415.x;
+       v15 = v_1415.y;
 
        // // a = a + m[sigma[r][2*i+0]]
        // // skip since adding 0u does nothing
@@ -2476,11 +2504,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        v15 = (xor1 >> 24u) ^ (xor0 << 8u);
 
        // a = a + b
-       o0 = v6 + v14;
-       o1 = v7 + v15;
-       o1 = o1 + select(0u, 1u, o0 < v6);
-       v6 = o0;
-       v7 = o1;
+       v_67.x = v6;
+       v_67.y = v7;
+       v_1415.x = v14;
+       v_1415.y = v15;
+       v_67 = v_67 + v_1415 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_67.x + v_1415.x < v_67.x);
+       v6 = v_67.x;
+       v7 = v_67.y;
+       v14 = v_1415.x;
+       v15 = v_1415.y;
 
        // // a = a + m[sigma[r][2*i+1]]
        // // skip since adding 0u does nothing
@@ -3114,11 +3146,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        */
 
        // a = a + b
-       o0 = v6 + v14;
-       o1 = v7 + v15;
-       o1 = o1 + select(0u, 1u, o0 < v6);
-       v6 = o0;
-       v7 = o1;
+       v_67.x = v6;
+       v_67.y = v7;
+       v_1415.x = v14;
+       v_1415.y = v15;
+       v_67 = v_67 + v_1415 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_67.x + v_1415.x < v_67.x);
+       v6 = v_67.x;
+       v7 = v_67.y;
+       v14 = v_1415.x;
+       v15 = v_1415.y;
 
        // // a = a + m[sigma[r][2*i+0]]
        // // skip since adding 0u does nothing
@@ -3150,11 +3186,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        v15 = (xor1 >> 24u) ^ (xor0 << 8u);
 
        // a = a + b
-       o0 = v6 + v14;
-       o1 = v7 + v15;
-       o1 = o1 + select(0u, 1u, o0 < v6);
-       v6 = o0;
-       v7 = o1;
+       v_67.x = v6;
+       v_67.y = v7;
+       v_1415.x = v14;
+       v_1415.y = v15;
+       v_67 = v_67 + v_1415 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_67.x + v_1415.x < v_67.x);
+       v6 = v_67.x;
+       v7 = v_67.y;
+       v14 = v_1415.x;
+       v15 = v_1415.y;
 
        // // a = a + m[sigma[r][2*i+1]]
        // // skip since adding 0u does nothing
@@ -3794,11 +3834,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        */
 
        // a = a + b
-       o0 = v6 + v14;
-       o1 = v7 + v15;
-       o1 = o1 + select(0u, 1u, o0 < v6);
-       v6 = o0;
-       v7 = o1;
+       v_67.x = v6;
+       v_67.y = v7;
+       v_1415.x = v14;
+       v_1415.y = v15;
+       v_67 = v_67 + v_1415 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_67.x + v_1415.x < v_67.x);
+       v6 = v_67.x;
+       v7 = v_67.y;
+       v14 = v_1415.x;
+       v15 = v_1415.y;
 
        // // a = a + m[sigma[r][2*i+0]]
        // // skip since adding 0u does nothing
@@ -3830,11 +3874,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        v15 = (xor1 >> 24u) ^ (xor0 << 8u);
 
        // a = a + b
-       o0 = v6 + v14;
-       o1 = v7 + v15;
-       o1 = o1 + select(0u, 1u, o0 < v6);
-       v6 = o0;
-       v7 = o1;
+       v_67.x = v6;
+       v_67.y = v7;
+       v_1415.x = v14;
+       v_1415.y = v15;
+       v_67 = v_67 + v_1415 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_67.x + v_1415.x < v_67.x);
+       v6 = v_67.x;
+       v7 = v_67.y;
+       v14 = v_1415.x;
+       v15 = v_1415.y;
 
        // a = a + m[sigma[r][2*i+1]]
        o0 = v6 + m6;
@@ -4474,11 +4522,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        */
 
        // a = a + b
-       o0 = v6 + v14;
-       o1 = v7 + v15;
-       o1 = o1 + select(0u, 1u, o0 < v6);
-       v6 = o0;
-       v7 = o1;
+       v_67.x = v6;
+       v_67.y = v7;
+       v_1415.x = v14;
+       v_1415.y = v15;
+       v_67 = v_67 + v_1415 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_67.x + v_1415.x < v_67.x);
+       v6 = v_67.x;
+       v7 = v_67.y;
+       v14 = v_1415.x;
+       v15 = v_1415.y;
 
        // a = a + m[sigma[r][2*i+0]]
        o0 = v6 + m8;
@@ -4507,11 +4559,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        v15 = (xor1 >> 24u) ^ (xor0 << 8u);
 
        // a = a + b
-       o0 = v6 + v14;
-       o1 = v7 + v15;
-       o1 = o1 + select(0u, 1u, o0 < v6);
-       v6 = o0;
-       v7 = o1;
+       v_67.x = v6;
+       v_67.y = v7;
+       v_1415.x = v14;
+       v_1415.y = v15;
+       v_67 = v_67 + v_1415 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_67.x + v_1415.x < v_67.x);
+       v6 = v_67.x;
+       v7 = v_67.y;
+       v14 = v_1415.x;
+       v15 = v_1415.y;
 
        // // a = a + m[sigma[r][2*i+1]]
        // // skip since adding 0u does nothing
@@ -5151,11 +5207,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        */
 
        // a = a + b
-       o0 = v6 + v14;
-       o1 = v7 + v15;
-       o1 = o1 + select(0u, 1u, o0 < v6);
-       v6 = o0;
-       v7 = o1;
+       v_67.x = v6;
+       v_67.y = v7;
+       v_1415.x = v14;
+       v_1415.y = v15;
+       v_67 = v_67 + v_1415 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_67.x + v_1415.x < v_67.x);
+       v6 = v_67.x;
+       v7 = v_67.y;
+       v14 = v_1415.x;
+       v15 = v_1415.y;
 
        // a = a + m[sigma[r][2*i+0]]
        o0 = v6 + m6;
@@ -5184,11 +5244,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        v15 = (xor1 >> 24u) ^ (xor0 << 8u);
 
        // a = a + b
-       o0 = v6 + v14;
-       o1 = v7 + v15;
-       o1 = o1 + select(0u, 1u, o0 < v6);
-       v6 = o0;
-       v7 = o1;
+       v_67.x = v6;
+       v_67.y = v7;
+       v_1415.x = v14;
+       v_1415.y = v15;
+       v_67 = v_67 + v_1415 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_67.x + v_1415.x < v_67.x);
+       v6 = v_67.x;
+       v7 = v_67.y;
+       v14 = v_1415.x;
+       v15 = v_1415.y;
 
        // // a = a + m[sigma[r][2*i+1]]
        // // skip since adding 0u does nothing
@@ -5828,11 +5892,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        */
 
        // a = a + b
-       o0 = v6 + v14;
-       o1 = v7 + v15;
-       o1 = o1 + select(0u, 1u, o0 < v6);
-       v6 = o0;
-       v7 = o1;
+       v_67.x = v6;
+       v_67.y = v7;
+       v_1415.x = v14;
+       v_1415.y = v15;
+       v_67 = v_67 + v_1415 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_67.x + v_1415.x < v_67.x);
+       v6 = v_67.x;
+       v7 = v_67.y;
+       v14 = v_1415.x;
+       v15 = v_1415.y;
 
        // a = a + m[sigma[r][2*i+0]]
        o0 = v6 + m0;
@@ -5861,11 +5929,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        v15 = (xor1 >> 24u) ^ (xor0 << 8u);
 
        // a = a + b
-       o0 = v6 + v14;
-       o1 = v7 + v15;
-       o1 = o1 + select(0u, 1u, o0 < v6);
-       v6 = o0;
-       v7 = o1;
+       v_67.x = v6;
+       v_67.y = v7;
+       v_1415.x = v14;
+       v_1415.y = v15;
+       v_67 = v_67 + v_1415 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_67.x + v_1415.x < v_67.x);
+       v6 = v_67.x;
+       v7 = v_67.y;
+       v14 = v_1415.x;
+       v15 = v_1415.y;
 
        // // a = a + m[sigma[r][2*i+1]]
        // // skip since adding 0u does nothing
@@ -6502,11 +6574,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        */
 
        // a = a + b
-       o0 = v6 + v14;
-       o1 = v7 + v15;
-       o1 = o1 + select(0u, 1u, o0 < v6);
-       v6 = o0;
-       v7 = o1;
+       v_67.x = v6;
+       v_67.y = v7;
+       v_1415.x = v14;
+       v_1415.y = v15;
+       v_67 = v_67 + v_1415 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_67.x + v_1415.x < v_67.x);
+       v6 = v_67.x;
+       v7 = v_67.y;
+       v14 = v_1415.x;
+       v15 = v_1415.y;
 
        // a = a + m[sigma[r][2*i+0]]
        o0 = v6 + m2;
@@ -6535,11 +6611,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        v15 = (xor1 >> 24u) ^ (xor0 << 8u);
 
        // a = a + b
-       o0 = v6 + v14;
-       o1 = v7 + v15;
-       o1 = o1 + select(0u, 1u, o0 < v6);
-       v6 = o0;
-       v7 = o1;
+       v_67.x = v6;
+       v_67.y = v7;
+       v_1415.x = v14;
+       v_1415.y = v15;
+       v_67 = v_67 + v_1415 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_67.x + v_1415.x < v_67.x);
+       v6 = v_67.x;
+       v7 = v_67.y;
+       v14 = v_1415.x;
+       v15 = v_1415.y;
 
        // // a = a + m[sigma[r][2*i+1]]
        // // skip since adding 0u does nothing
@@ -7170,11 +7250,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        */
 
        // a = a + b
-       o0 = v6 + v14;
-       o1 = v7 + v15;
-       o1 = o1 + select(0u, 1u, o0 < v6);
-       v6 = o0;
-       v7 = o1;
+       v_67.x = v6;
+       v_67.y = v7;
+       v_1415.x = v14;
+       v_1415.y = v15;
+       v_67 = v_67 + v_1415 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_67.x + v_1415.x < v_67.x);
+       v6 = v_67.x;
+       v7 = v_67.y;
+       v14 = v_1415.x;
+       v15 = v_1415.y;
 
        // // a = a + m[sigma[r][2*i+0]]
        // // skip since adding 0u does nothing
@@ -7206,11 +7290,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        v15 = (xor1 >> 24u) ^ (xor0 << 8u);
 
        // a = a + b
-       o0 = v6 + v14;
-       o1 = v7 + v15;
-       o1 = o1 + select(0u, 1u, o0 < v6);
-       v6 = o0;
-       v7 = o1;
+       v_67.x = v6;
+       v_67.y = v7;
+       v_1415.x = v14;
+       v_1415.y = v15;
+       v_67 = v_67 + v_1415 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_67.x + v_1415.x < v_67.x);
+       v6 = v_67.x;
+       v7 = v_67.y;
+       v14 = v_1415.x;
+       v15 = v_1415.y;
 
        // // a = a + m[sigma[r][2*i+1]]
        // // skip since adding 0u does nothing
@@ -7859,11 +7947,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        */
 
        // a = a + b
-       o0 = v6 + v14;
-       o1 = v7 + v15;
-       o1 = o1 + select(0u, 1u, o0 < v6);
-       v6 = o0;
-       v7 = o1;
+       v_67.x = v6;
+       v_67.y = v7;
+       v_1415.x = v14;
+       v_1415.y = v15;
+       v_67 = v_67 + v_1415 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_67.x + v_1415.x < v_67.x);
+       v6 = v_67.x;
+       v7 = v_67.y;
+       v14 = v_1415.x;
+       v15 = v_1415.y;
 
        // // a = a + m[sigma[r][2*i+0]]
        // // skip since adding 0u does nothing
@@ -7895,11 +7987,15 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        v15 = (xor1 >> 24u) ^ (xor0 << 8u);
 
        // a = a + b
-       o0 = v6 + v14;
-       o1 = v7 + v15;
-       o1 = o1 + select(0u, 1u, o0 < v6);
-       v6 = o0;
-       v7 = o1;
+       v_67.x = v6;
+       v_67.y = v7;
+       v_1415.x = v14;
+       v_1415.y = v15;
+       v_67 = v_67 + v_1415 + select(vec2<u32>(0u), vec2<u32>(0u, 1u), v_67.x + v_1415.x < v_67.x);
+       v6 = v_67.x;
+       v7 = v_67.y;
+       v14 = v_1415.x;
+       v15 = v_1415.y;
 
        // // a = a + m[sigma[r][2*i+1]]
        // // skip since adding 0u does nothing