From: Chris Duncan <chris@zoso.dev>
Date: Tue, 14 Jan 2025 18:49:23 +0000 (-0800)
Subject: Replace more 32-bit scalar rotations with vector rotations.
X-Git-Url: https://zoso.dev/?a=commitdiff_plain;h=0aeccea1eced8156586c73457aa9faad845fcb1b;p=nano-pow.git

Replace more 32-bit scalar rotations with vector rotations.
---

diff --git a/src/shaders/compute.wgsl b/src/shaders/compute.wgsl
index ad68c43..e89e4f3 100644
--- a/src/shaders/compute.wgsl
+++ b/src/shaders/compute.wgsl
@@ -707,10 +707,13 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	// v5 = o1;
 
 	// d = rotr64(d ^ a, 32)
-	xor0 = v26 ^ v4;
-	xor1 = v27 ^ v5;
-	v26 = xor1;
-	v27 = xor0;
+	v_45.x = v4;
+	v_45.y = v5;
+	v_2627.x = v26;
+	v_2627.y = v27;
+	v_2627 = v_2627 ^ v_45;
+	v26 = v_2627.y;
+	v27 = v_2627.x;
 
 	// c = c + d
 	v_1617.x = v16;
@@ -1456,10 +1459,13 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	// v5 = o1;
 
 	// d = rotr64(d ^ a, 32)
-	xor0 = v26 ^ v4;
-	xor1 = v27 ^ v5;
-	v26 = xor1;
-	v27 = xor0;
+	v_45.x = v4;
+	v_45.y = v5;
+	v_2627.x = v26;
+	v_2627.y = v27;
+	v_2627 = v_2627 ^ v_45;
+	v26 = v_2627.y;
+	v27 = v_2627.x;
 
 	// c = c + d
 	v_1617.x = v16;
@@ -2196,10 +2202,13 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	// v5 = o1;
 
 	// d = rotr64(d ^ a, 32)
-	xor0 = v26 ^ v4;
-	xor1 = v27 ^ v5;
-	v26 = xor1;
-	v27 = xor0;
+	v_45.x = v4;
+	v_45.y = v5;
+	v_2627.x = v26;
+	v_2627.y = v27;
+	v_2627 = v_2627 ^ v_45;
+	v26 = v_2627.y;
+	v27 = v_2627.x;
 
 	// c = c + d
 	v_1617.x = v16;
@@ -2939,10 +2948,13 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v5 = o1;
 
 	// d = rotr64(d ^ a, 32)
-	xor0 = v26 ^ v4;
-	xor1 = v27 ^ v5;
-	v26 = xor1;
-	v27 = xor0;
+	v_45.x = v4;
+	v_45.y = v5;
+	v_2627.x = v26;
+	v_2627.y = v27;
+	v_2627 = v_2627 ^ v_45;
+	v26 = v_2627.y;
+	v27 = v_2627.x;
 
 	// c = c + d
 	v_1617.x = v16;
@@ -3694,10 +3706,13 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	// v5 = o1;
 
 	// d = rotr64(d ^ a, 32)
-	xor0 = v26 ^ v4;
-	xor1 = v27 ^ v5;
-	v26 = xor1;
-	v27 = xor0;
+	v_45.x = v4;
+	v_45.y = v5;
+	v_2627.x = v26;
+	v_2627.y = v27;
+	v_2627 = v_2627 ^ v_45;
+	v26 = v_2627.y;
+	v27 = v_2627.x;
 
 	// c = c + d
 	v_1617.x = v16;
@@ -4449,10 +4464,13 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	// v5 = o1;
 
 	// d = rotr64(d ^ a, 32)
-	xor0 = v26 ^ v4;
-	xor1 = v27 ^ v5;
-	v26 = xor1;
-	v27 = xor0;
+	v_45.x = v4;
+	v_45.y = v5;
+	v_2627.x = v26;
+	v_2627.y = v27;
+	v_2627 = v_2627 ^ v_45;
+	v26 = v_2627.y;
+	v27 = v_2627.x;
 
 	// c = c + d
 	v_1617.x = v16;
@@ -5195,10 +5213,13 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	// v5 = o1;
 
 	// d = rotr64(d ^ a, 32)
-	xor0 = v26 ^ v4;
-	xor1 = v27 ^ v5;
-	v26 = xor1;
-	v27 = xor0;
+	v_45.x = v4;
+	v_45.y = v5;
+	v_2627.x = v26;
+	v_2627.y = v27;
+	v_2627 = v_2627 ^ v_45;
+	v26 = v_2627.y;
+	v27 = v_2627.x;
 
 	// c = c + d
 	v_1617.x = v16;
@@ -5941,10 +5962,13 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	// v5 = o1;
 
 	// d = rotr64(d ^ a, 32)
-	xor0 = v26 ^ v4;
-	xor1 = v27 ^ v5;
-	v26 = xor1;
-	v27 = xor0;
+	v_45.x = v4;
+	v_45.y = v5;
+	v_2627.x = v26;
+	v_2627.y = v27;
+	v_2627 = v_2627 ^ v_45;
+	v26 = v_2627.y;
+	v27 = v_2627.x;
 
 	// c = c + d
 	v_1617.x = v16;
@@ -6687,10 +6711,13 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v5 = o1;
 
 	// d = rotr64(d ^ a, 32)
-	xor0 = v26 ^ v4;
-	xor1 = v27 ^ v5;
-	v26 = xor1;
-	v27 = xor0;
+	v_45.x = v4;
+	v_45.y = v5;
+	v_2627.x = v26;
+	v_2627.y = v27;
+	v_2627 = v_2627 ^ v_45;
+	v26 = v_2627.y;
+	v27 = v_2627.x;
 
 	// c = c + d
 	v_1617.x = v16;
@@ -7433,10 +7460,13 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	v5 = o1;
 
 	// d = rotr64(d ^ a, 32)
-	xor0 = v26 ^ v4;
-	xor1 = v27 ^ v5;
-	v26 = xor1;
-	v27 = xor0;
+	v_45.x = v4;
+	v_45.y = v5;
+	v_2627.x = v26;
+	v_2627.y = v27;
+	v_2627 = v_2627 ^ v_45;
+	v26 = v_2627.y;
+	v27 = v_2627.x;
 
 	// c = c + d
 	v_1617.x = v16;
@@ -8185,10 +8215,13 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	// v5 = o1;
 
 	// d = rotr64(d ^ a, 32)
-	xor0 = v26 ^ v4;
-	xor1 = v27 ^ v5;
-	v26 = xor1;
-	v27 = xor0;
+	v_45.x = v4;
+	v_45.y = v5;
+	v_2627.x = v26;
+	v_2627.y = v27;
+	v_2627 = v_2627 ^ v_45;
+	v26 = v_2627.y;
+	v27 = v_2627.x;
 
 	// c = c + d
 	v_1617.x = v16;
@@ -8934,10 +8967,13 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	// v5 = o1;
 
 	// d = rotr64(d ^ a, 32)
-	xor0 = v26 ^ v4;
-	xor1 = v27 ^ v5;
-	v26 = xor1;
-	v27 = xor0;
+	v_45.x = v4;
+	v_45.y = v5;
+	v_2627.x = v26;
+	v_2627.y = v27;
+	v_2627 = v_2627 ^ v_45;
+	v26 = v_2627.y;
+	v27 = v_2627.x;
 
 	// c = c + d
 	v_1617.x = v16;