Replace more 24-bit scalar rotations with vector rotations.

diff --git a/src/shaders/compute.wgsl b/src/shaders/compute.wgsl
index 06e17f7d8c5811c98c058ea5271c92d75ce1736a..e8750c44b15af788bc3d5de68e1aeab69411e785 100644 (file)
--- a/src/shaders/compute.wgsl
+++ b/src/shaders/compute.wgsl
@@ -461,10 +461,14 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        v31 = v_3031.y;
 
        // b = rotr64(b ^ c, 24)
-       xor0 = v14 ^ v22;
-       xor1 = v15 ^ v23;
-       v14 = (xor0 >> 24u) ^ (xor1 << 8u);
-       v15 = (xor1 >> 24u) ^ (xor0 << 8u);
+       v_2223.x = v22;
+       v_2223.y = v23;
+       v_1415.x = v14;
+       v_1415.y = v15;
+       xor = v_1415 ^ v_2223;
+       v_1415 = vec2((xor.x >> 24u) | (xor.y << 8u), (xor.y >> 24u) | (xor.x << 8u));
+       v14 = v_1415.x;
+       v15 = v_1415.y;
 
        // a = a + b
        v_67.x = v6;
@@ -1224,10 +1228,14 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        v31 = v_3031.y;
 
        // b = rotr64(b ^ c, 24)
-       xor0 = v14 ^ v22;
-       xor1 = v15 ^ v23;
-       v14 = (xor0 >> 24u) ^ (xor1 << 8u);
-       v15 = (xor1 >> 24u) ^ (xor0 << 8u);
+       v_2223.x = v22;
+       v_2223.y = v23;
+       v_1415.x = v14;
+       v_1415.y = v15;
+       xor = v_1415 ^ v_2223;
+       v_1415 = vec2((xor.x >> 24u) | (xor.y << 8u), (xor.y >> 24u) | (xor.x << 8u));
+       v14 = v_1415.x;
+       v15 = v_1415.y;
 
        // a = a + b
        v_67.x = v6;
@@ -1981,10 +1989,14 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        v31 = v_3031.y;
 
        // b = rotr64(b ^ c, 24)
-       xor0 = v14 ^ v22;
-       xor1 = v15 ^ v23;
-       v14 = (xor0 >> 24u) ^ (xor1 << 8u);
-       v15 = (xor1 >> 24u) ^ (xor0 << 8u);
+       v_2223.x = v22;
+       v_2223.y = v23;
+       v_1415.x = v14;
+       v_1415.y = v15;
+       xor = v_1415 ^ v_2223;
+       v_1415 = vec2((xor.x >> 24u) | (xor.y << 8u), (xor.y >> 24u) | (xor.x << 8u));
+       v14 = v_1415.x;
+       v15 = v_1415.y;
 
        // a = a + b
        v_67.x = v6;
@@ -2732,10 +2744,14 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        v31 = v_3031.y;
 
        // b = rotr64(b ^ c, 24)
-       xor0 = v14 ^ v22;
-       xor1 = v15 ^ v23;
-       v14 = (xor0 >> 24u) ^ (xor1 << 8u);
-       v15 = (xor1 >> 24u) ^ (xor0 << 8u);
+       v_2223.x = v22;
+       v_2223.y = v23;
+       v_1415.x = v14;
+       v_1415.y = v15;
+       xor = v_1415 ^ v_2223;
+       v_1415 = vec2((xor.x >> 24u) | (xor.y << 8u), (xor.y >> 24u) | (xor.x << 8u));
+       v14 = v_1415.x;
+       v15 = v_1415.y;
 
        // a = a + b
        v_67.x = v6;
@@ -3498,10 +3514,14 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        v31 = v_3031.y;
 
        // b = rotr64(b ^ c, 24)
-       xor0 = v14 ^ v22;
-       xor1 = v15 ^ v23;
-       v14 = (xor0 >> 24u) ^ (xor1 << 8u);
-       v15 = (xor1 >> 24u) ^ (xor0 << 8u);
+       v_2223.x = v22;
+       v_2223.y = v23;
+       v_1415.x = v14;
+       v_1415.y = v15;
+       xor = v_1415 ^ v_2223;
+       v_1415 = vec2((xor.x >> 24u) | (xor.y << 8u), (xor.y >> 24u) | (xor.x << 8u));
+       v14 = v_1415.x;
+       v15 = v_1415.y;
 
        // a = a + b
        v_67.x = v6;
@@ -4270,10 +4290,14 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        v31 = v_3031.y;
 
        // b = rotr64(b ^ c, 24)
-       xor0 = v14 ^ v22;
-       xor1 = v15 ^ v23;
-       v14 = (xor0 >> 24u) ^ (xor1 << 8u);
-       v15 = (xor1 >> 24u) ^ (xor0 << 8u);
+       v_2223.x = v22;
+       v_2223.y = v23;
+       v_1415.x = v14;
+       v_1415.y = v15;
+       xor = v_1415 ^ v_2223;
+       v_1415 = vec2((xor.x >> 24u) | (xor.y << 8u), (xor.y >> 24u) | (xor.x << 8u));
+       v14 = v_1415.x;
+       v15 = v_1415.y;
 
        // a = a + b
        v_67.x = v6;
@@ -5030,10 +5054,14 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        v31 = v_3031.y;
 
        // b = rotr64(b ^ c, 24)
-       xor0 = v14 ^ v22;
-       xor1 = v15 ^ v23;
-       v14 = (xor0 >> 24u) ^ (xor1 << 8u);
-       v15 = (xor1 >> 24u) ^ (xor0 << 8u);
+       v_2223.x = v22;
+       v_2223.y = v23;
+       v_1415.x = v14;
+       v_1415.y = v15;
+       xor = v_1415 ^ v_2223;
+       v_1415 = vec2((xor.x >> 24u) | (xor.y << 8u), (xor.y >> 24u) | (xor.x << 8u));
+       v14 = v_1415.x;
+       v15 = v_1415.y;
 
        // a = a + b
        v_67.x = v6;
@@ -5790,10 +5818,14 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        v31 = v_3031.y;
 
        // b = rotr64(b ^ c, 24)
-       xor0 = v14 ^ v22;
-       xor1 = v15 ^ v23;
-       v14 = (xor0 >> 24u) ^ (xor1 << 8u);
-       v15 = (xor1 >> 24u) ^ (xor0 << 8u);
+       v_2223.x = v22;
+       v_2223.y = v23;
+       v_1415.x = v14;
+       v_1415.y = v15;
+       xor = v_1415 ^ v_2223;
+       v_1415 = vec2((xor.x >> 24u) | (xor.y << 8u), (xor.y >> 24u) | (xor.x << 8u));
+       v14 = v_1415.x;
+       v15 = v_1415.y;
 
        // a = a + b
        v_67.x = v6;
@@ -6550,10 +6582,14 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        v31 = v_3031.y;
 
        // b = rotr64(b ^ c, 24)
-       xor0 = v14 ^ v22;
-       xor1 = v15 ^ v23;
-       v14 = (xor0 >> 24u) ^ (xor1 << 8u);
-       v15 = (xor1 >> 24u) ^ (xor0 << 8u);
+       v_2223.x = v22;
+       v_2223.y = v23;
+       v_1415.x = v14;
+       v_1415.y = v15;
+       xor = v_1415 ^ v_2223;
+       v_1415 = vec2((xor.x >> 24u) | (xor.y << 8u), (xor.y >> 24u) | (xor.x << 8u));
+       v14 = v_1415.x;
+       v15 = v_1415.y;
 
        // a = a + b
        v_67.x = v6;
@@ -7316,10 +7352,14 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        v31 = v_3031.y;
 
        // b = rotr64(b ^ c, 24)
-       xor0 = v14 ^ v22;
-       xor1 = v15 ^ v23;
-       v14 = (xor0 >> 24u) ^ (xor1 << 8u);
-       v15 = (xor1 >> 24u) ^ (xor0 << 8u);
+       v_2223.x = v22;
+       v_2223.y = v23;
+       v_1415.x = v14;
+       v_1415.y = v15;
+       xor = v_1415 ^ v_2223;
+       v_1415 = vec2((xor.x >> 24u) | (xor.y << 8u), (xor.y >> 24u) | (xor.x << 8u));
+       v14 = v_1415.x;
+       v15 = v_1415.y;
 
        // a = a + b
        v_67.x = v6;
@@ -8079,10 +8119,14 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        v31 = v_3031.y;
 
        // b = rotr64(b ^ c, 24)
-       xor0 = v14 ^ v22;
-       xor1 = v15 ^ v23;
-       v14 = (xor0 >> 24u) ^ (xor1 << 8u);
-       v15 = (xor1 >> 24u) ^ (xor0 << 8u);
+       v_2223.x = v22;
+       v_2223.y = v23;
+       v_1415.x = v14;
+       v_1415.y = v15;
+       xor = v_1415 ^ v_2223;
+       v_1415 = vec2((xor.x >> 24u) | (xor.y << 8u), (xor.y >> 24u) | (xor.x << 8u));
+       v14 = v_1415.x;
+       v15 = v_1415.y;
 
        // a = a + b
        v_67.x = v6;
@@ -8842,10 +8886,14 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        v31 = v_3031.y;
 
        // b = rotr64(b ^ c, 24)
-       xor0 = v14 ^ v22;
-       xor1 = v15 ^ v23;
-       v14 = (xor0 >> 24u) ^ (xor1 << 8u);
-       v15 = (xor1 >> 24u) ^ (xor0 << 8u);
+       v_2223.x = v22;
+       v_2223.y = v23;
+       v_1415.x = v14;
+       v_1415.y = v15;
+       xor = v_1415 ^ v_2223;
+       v_1415 = vec2((xor.x >> 24u) | (xor.y << 8u), (xor.y >> 24u) | (xor.x << 8u));
+       v14 = v_1415.x;
+       v15 = v_1415.y;
 
        // a = a + b
        v_67.x = v6;