Replace some 32-bit scalar rotations with vector rotations.

diff --git a/src/shaders/compute.wgsl b/src/shaders/compute.wgsl
index 48cfd75f09d9d5f0f8f17708511db0c59595a93f..2349da9d1a7f0ef339efe59061e48ee5cff8b6b0 100644 (file)
--- a/src/shaders/compute.wgsl
+++ b/src/shaders/compute.wgsl
@@ -145,10 +145,9 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        v1 = o1;
 
        // d = rotr64(d ^ a, 32)
-       xor0 = v24 ^ v0;
-       xor1 = v25 ^ v1;
-       v24 = xor1;
-       v25 = xor0;
+       v_2425 = v_2425 ^ v_01;
+       v24 = v_2425.y;
+       v25 = v_2425.x;
 
        // c = c + d
        v_1617.x = v16;
@@ -878,10 +877,9 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        // // skip since adding 0u does nothing
 
        // d = rotr64(d ^ a, 32)
-       xor0 = v24 ^ v0;
-       xor1 = v25 ^ v1;
-       v24 = xor1;
-       v25 = xor0;
+       v_2425 = v_2425 ^ v_01;
+       v24 = v_2425.y;
+       v25 = v_2425.x;
 
        // c = c + d
        v_1617.x = v16;
@@ -1608,10 +1606,9 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        // // skip since adding 0u does nothing
 
        // d = rotr64(d ^ a, 32)
-       xor0 = v24 ^ v0;
-       xor1 = v25 ^ v1;
-       v24 = xor1;
-       v25 = xor0;
+       v_2425 = v_2425 ^ v_01;
+       v24 = v_2425.y;
+       v25 = v_2425.x;
 
        // c = c + d
        v_1617.x = v16;
@@ -2329,10 +2326,9 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        // // skip since adding 0u does nothing
 
        // d = rotr64(d ^ a, 32)
-       xor0 = v24 ^ v0;
-       xor1 = v25 ^ v1;
-       v24 = xor1;
-       v25 = xor0;
+       v_2425 = v_2425 ^ v_01;
+       v24 = v_2425.y;
+       v25 = v_2425.x;
 
        // c = c + d
        v_1617.x = v16;
@@ -3059,10 +3055,9 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        // // skip since adding 0u does nothing
 
        // d = rotr64(d ^ a, 32)
-       xor0 = v24 ^ v0;
-       xor1 = v25 ^ v1;
-       v24 = xor1;
-       v25 = xor0;
+       v_2425 = v_2425 ^ v_01;
+       v24 = v_2425.y;
+       v25 = v_2425.x;
 
        // c = c + d
        v_1617.x = v16;
@@ -3804,10 +3799,9 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        v1 = o1;
 
        // d = rotr64(d ^ a, 32)
-       xor0 = v24 ^ v0;
-       xor1 = v25 ^ v1;
-       v24 = xor1;
-       v25 = xor0;
+       v_2425 = v_2425 ^ v_01;
+       v24 = v_2425.y;
+       v25 = v_2425.x;
 
        // c = c + d
        v_1617.x = v16;
@@ -4537,10 +4531,9 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        // // skip since adding 0u does nothing
 
        // d = rotr64(d ^ a, 32)
-       xor0 = v24 ^ v0;
-       xor1 = v25 ^ v1;
-       v24 = xor1;
-       v25 = xor0;
+       v_2425 = v_2425 ^ v_01;
+       v24 = v_2425.y;
+       v25 = v_2425.x;
 
        // c = c + d
        v_1617.x = v16;
@@ -5267,10 +5260,9 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        // // skip since adding 0u does nothing
 
        // d = rotr64(d ^ a, 32)
-       xor0 = v24 ^ v0;
-       xor1 = v25 ^ v1;
-       v24 = xor1;
-       v25 = xor0;
+       v_2425 = v_2425 ^ v_01;
+       v24 = v_2425.y;
+       v25 = v_2425.x;
 
        // c = c + d
        v_1617.x = v16;
@@ -5997,10 +5989,9 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        // // skip since adding 0u does nothing
 
        // d = rotr64(d ^ a, 32)
-       xor0 = v24 ^ v0;
-       xor1 = v25 ^ v1;
-       v24 = xor1;
-       v25 = xor0;
+       v_2425 = v_2425 ^ v_01;
+       v24 = v_2425.y;
+       v25 = v_2425.x;
 
        // c = c + d
        v_1617.x = v16;
@@ -6727,10 +6718,9 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        // // skip since adding 0u does nothing
 
        // d = rotr64(d ^ a, 32)
-       xor0 = v24 ^ v0;
-       xor1 = v25 ^ v1;
-       v24 = xor1;
-       v25 = xor0;
+       v_2425 = v_2425 ^ v_01;
+       v24 = v_2425.y;
+       v25 = v_2425.x;
 
        // c = c + d
        v_1617.x = v16;
@@ -7463,10 +7453,9 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        v1 = o1;
 
        // d = rotr64(d ^ a, 32)
-       xor0 = v24 ^ v0;
-       xor1 = v25 ^ v1;
-       v24 = xor1;
-       v25 = xor0;
+       v_2425 = v_2425 ^ v_01;
+       v24 = v_2425.y;
+       v25 = v_2425.x;
 
        // c = c + d
        v_1617.x = v16;
@@ -8196,10 +8185,9 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
        // // skip since adding 0u does nothing
 
        // d = rotr64(d ^ a, 32)
-       xor0 = v24 ^ v0;
-       xor1 = v25 ^ v1;
-       v24 = xor1;
-       v25 = xor0;
+       v_2425 = v_2425 ^ v_01;
+       v24 = v_2425.y;
+       v25 = v_2425.x;
 
        // c = c + d
        v_1617.x = v16;