Replace more 24-bit scalar rotations with vector rotations.

diff --git a/src/shaders/compute.wgsl b/src/shaders/compute.wgsl
index 2981957f26f88c5a599b718232b69370686e51f3..a8e99a949d62222f9a622c6309acb453097d61a0 100644 (file)
--- a/src/shaders/compute.wgsl
+++ b/src/shaders/compute.wgsl
@@ -752,10 +752,14 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        v27 = v_2627.y;
 
        // b = rotr64(b ^ c, 24)
-       xor0 = v14 ^ v16;
-       xor1 = v15 ^ v17;
-       v14 = (xor0 >> 24u) ^ (xor1 << 8u);
-       v15 = (xor1 >> 24u) ^ (xor0 << 8u);
+       v_1617.x = v16;
+       v_1617.y = v17;
+       v_1415.x = v14;
+       v_1415.y = v15;
+       xor = v_1415 ^ v_1617;
+       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_45.x = v4;
@@ -1527,10 +1531,14 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        v27 = v_2627.y;
 
        // b = rotr64(b ^ c, 24)
-       xor0 = v14 ^ v16;
-       xor1 = v15 ^ v17;
-       v14 = (xor0 >> 24u) ^ (xor1 << 8u);
-       v15 = (xor1 >> 24u) ^ (xor0 << 8u);
+       v_1617.x = v16;
+       v_1617.y = v17;
+       v_1415.x = v14;
+       v_1415.y = v15;
+       xor = v_1415 ^ v_1617;
+       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_45.x = v4;
@@ -2293,10 +2301,14 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        v27 = v_2627.y;
 
        // b = rotr64(b ^ c, 24)
-       xor0 = v14 ^ v16;
-       xor1 = v15 ^ v17;
-       v14 = (xor0 >> 24u) ^ (xor1 << 8u);
-       v15 = (xor1 >> 24u) ^ (xor0 << 8u);
+       v_1617.x = v16;
+       v_1617.y = v17;
+       v_1415.x = v14;
+       v_1415.y = v15;
+       xor = v_1415 ^ v_1617;
+       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_45.x = v4;
@@ -3062,10 +3074,14 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        v27 = v_2627.y;
 
        // b = rotr64(b ^ c, 24)
-       xor0 = v14 ^ v16;
-       xor1 = v15 ^ v17;
-       v14 = (xor0 >> 24u) ^ (xor1 << 8u);
-       v15 = (xor1 >> 24u) ^ (xor0 << 8u);
+       v_1617.x = v16;
+       v_1617.y = v17;
+       v_1415.x = v14;
+       v_1415.y = v15;
+       xor = v_1415 ^ v_1617;
+       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_45.x = v4;
@@ -3843,10 +3859,14 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        v27 = v_2627.y;
 
        // b = rotr64(b ^ c, 24)
-       xor0 = v14 ^ v16;
-       xor1 = v15 ^ v17;
-       v14 = (xor0 >> 24u) ^ (xor1 << 8u);
-       v15 = (xor1 >> 24u) ^ (xor0 << 8u);
+       v_1617.x = v16;
+       v_1617.y = v17;
+       v_1415.x = v14;
+       v_1415.y = v15;
+       xor = v_1415 ^ v_1617;
+       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_45.x = v4;
@@ -4624,10 +4644,14 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        v27 = v_2627.y;
 
        // b = rotr64(b ^ c, 24)
-       xor0 = v14 ^ v16;
-       xor1 = v15 ^ v17;
-       v14 = (xor0 >> 24u) ^ (xor1 << 8u);
-       v15 = (xor1 >> 24u) ^ (xor0 << 8u);
+       v_1617.x = v16;
+       v_1617.y = v17;
+       v_1415.x = v14;
+       v_1415.y = v15;
+       xor = v_1415 ^ v_1617;
+       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_45.x = v4;
@@ -5396,10 +5420,14 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        v27 = v_2627.y;
 
        // b = rotr64(b ^ c, 24)
-       xor0 = v14 ^ v16;
-       xor1 = v15 ^ v17;
-       v14 = (xor0 >> 24u) ^ (xor1 << 8u);
-       v15 = (xor1 >> 24u) ^ (xor0 << 8u);
+       v_1617.x = v16;
+       v_1617.y = v17;
+       v_1415.x = v14;
+       v_1415.y = v15;
+       xor = v_1415 ^ v_1617;
+       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_45.x = v4;
@@ -6168,10 +6196,14 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        v27 = v_2627.y;
 
        // b = rotr64(b ^ c, 24)
-       xor0 = v14 ^ v16;
-       xor1 = v15 ^ v17;
-       v14 = (xor0 >> 24u) ^ (xor1 << 8u);
-       v15 = (xor1 >> 24u) ^ (xor0 << 8u);
+       v_1617.x = v16;
+       v_1617.y = v17;
+       v_1415.x = v14;
+       v_1415.y = v15;
+       xor = v_1415 ^ v_1617;
+       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_45.x = v4;
@@ -6940,10 +6972,14 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        v27 = v_2627.y;
 
        // b = rotr64(b ^ c, 24)
-       xor0 = v14 ^ v16;
-       xor1 = v15 ^ v17;
-       v14 = (xor0 >> 24u) ^ (xor1 << 8u);
-       v15 = (xor1 >> 24u) ^ (xor0 << 8u);
+       v_1617.x = v16;
+       v_1617.y = v17;
+       v_1415.x = v14;
+       v_1415.y = v15;
+       xor = v_1415 ^ v_1617;
+       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_45.x = v4;
@@ -7712,10 +7748,14 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        v27 = v_2627.y;
 
        // b = rotr64(b ^ c, 24)
-       xor0 = v14 ^ v16;
-       xor1 = v15 ^ v17;
-       v14 = (xor0 >> 24u) ^ (xor1 << 8u);
-       v15 = (xor1 >> 24u) ^ (xor0 << 8u);
+       v_1617.x = v16;
+       v_1617.y = v17;
+       v_1415.x = v14;
+       v_1415.y = v15;
+       xor = v_1415 ^ v_1617;
+       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_45.x = v4;
@@ -8490,10 +8530,14 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        v27 = v_2627.y;
 
        // b = rotr64(b ^ c, 24)
-       xor0 = v14 ^ v16;
-       xor1 = v15 ^ v17;
-       v14 = (xor0 >> 24u) ^ (xor1 << 8u);
-       v15 = (xor1 >> 24u) ^ (xor0 << 8u);
+       v_1617.x = v16;
+       v_1617.y = v17;
+       v_1415.x = v14;
+       v_1415.y = v15;
+       xor = v_1415 ^ v_1617;
+       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_45.x = v4;
@@ -9265,10 +9309,14 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        v27 = v_2627.y;
 
        // b = rotr64(b ^ c, 24)
-       xor0 = v14 ^ v16;
-       xor1 = v15 ^ v17;
-       v14 = (xor0 >> 24u) ^ (xor1 << 8u);
-       v15 = (xor1 >> 24u) ^ (xor0 << 8u);
+       v_1617.x = v16;
+       v_1617.y = v17;
+       v_1415.x = v14;
+       v_1415.y = v15;
+       xor = v_1415 ^ v_1617;
+       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_45.x = v4;