From 7a36229a8f1c6125c54dcc2f25c688e94af640e2 Mon Sep 17 00:00:00 2001
From: Chris Duncan <chris@zoso.dev>
Date: Tue, 14 Jan 2025 13:41:45 -0800
Subject: [PATCH] The 32-bit rotate was not correctly or completely implemented
 like the other rotations due to getting fooled by the simplicity. Start
 fixing starting with v[24-25].

---
 src/shaders/compute.wgsl | 84 +++++++++++++++++++++++-----------------
 1 file changed, 48 insertions(+), 36 deletions(-)

diff --git a/src/shaders/compute.wgsl b/src/shaders/compute.wgsl
index 0808814..b89fe92 100644
--- a/src/shaders/compute.wgsl
+++ b/src/shaders/compute.wgsl
@@ -108,9 +108,10 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	// d = rotr64(d ^ a, 32)
 	v24 = v_2425.x;
 	v25 = v_2425.y;
-	v_2425 = v_2425 ^ v_01;
-	v24 = v_2425.y;
-	v25 = v_2425.x;
+	xor = v_2425 ^ v_01;
+	v_2425 = vec2(xor.y, xor.x);
+	v24 = v_2425.x;
+	v25 = v_2425.y;
 
 	// c = c + d
 	v_2425.x = v24;
@@ -574,9 +575,10 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	// d = rotr64(d ^ a, 32)
 	v24 = v_2425.x;
 	v25 = v_2425.y;
-	v_2425 = v_2425 ^ v_01;
-	v24 = v_2425.y;
-	v25 = v_2425.x;
+	xor = v_2425 ^ v_01;
+	v_2425 = vec2(xor.y, xor.x);
+	v24 = v_2425.x;
+	v25 = v_2425.y;
 
 	// c = c + d
 
@@ -1364,9 +1366,10 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	// d = rotr64(d ^ a, 32)
 	v24 = v_2425.x;
 	v25 = v_2425.y;
-	v_2425 = v_2425 ^ v_01;
-	v24 = v_2425.y;
-	v25 = v_2425.x;
+	xor = v_2425 ^ v_01;
+	v_2425 = vec2(xor.y, xor.x);
+	v24 = v_2425.x;
+	v25 = v_2425.y;
 
 	// c = c + d
 
@@ -2154,9 +2157,10 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	// d = rotr64(d ^ a, 32)
 	v24 = v_2425.x;
 	v25 = v_2425.y;
-	v_2425 = v_2425 ^ v_01;
-	v24 = v_2425.y;
-	v25 = v_2425.x;
+	xor = v_2425 ^ v_01;
+	v_2425 = vec2(xor.y, xor.x);
+	v24 = v_2425.x;
+	v25 = v_2425.y;
 
 	// c = c + d
 
@@ -2943,9 +2947,10 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	// d = rotr64(d ^ a, 32)
 	v24 = v_2425.x;
 	v25 = v_2425.y;
-	v_2425 = v_2425 ^ v_01;
-	v24 = v_2425.y;
-	v25 = v_2425.x;
+	xor = v_2425 ^ v_01;
+	v_2425 = vec2(xor.y, xor.x);
+	v24 = v_2425.x;
+	v25 = v_2425.y;
 
 	// c = c + d
 
@@ -3732,9 +3737,10 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	// d = rotr64(d ^ a, 32)
 	v24 = v_2425.x;
 	v25 = v_2425.y;
-	v_2425 = v_2425 ^ v_01;
-	v24 = v_2425.y;
-	v25 = v_2425.x;
+	xor = v_2425 ^ v_01;
+	v_2425 = vec2(xor.y, xor.x);
+	v24 = v_2425.x;
+	v25 = v_2425.y;
 
 	// c = c + d
 
@@ -4517,9 +4523,10 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	// d = rotr64(d ^ a, 32)
 	v24 = v_2425.x;
 	v25 = v_2425.y;
-	v_2425 = v_2425 ^ v_01;
-	v24 = v_2425.y;
-	v25 = v_2425.x;
+	xor = v_2425 ^ v_01;
+	v_2425 = vec2(xor.y, xor.x);
+	v24 = v_2425.x;
+	v25 = v_2425.y;
 
 	// c = c + d
 
@@ -5305,9 +5312,10 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	// d = rotr64(d ^ a, 32)
 	v24 = v_2425.x;
 	v25 = v_2425.y;
-	v_2425 = v_2425 ^ v_01;
-	v24 = v_2425.y;
-	v25 = v_2425.x;
+	xor = v_2425 ^ v_01;
+	v_2425 = vec2(xor.y, xor.x);
+	v24 = v_2425.x;
+	v25 = v_2425.y;
 
 	// c = c + d
 
@@ -6093,9 +6101,10 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	// d = rotr64(d ^ a, 32)
 	v24 = v_2425.x;
 	v25 = v_2425.y;
-	v_2425 = v_2425 ^ v_01;
-	v24 = v_2425.y;
-	v25 = v_2425.x;
+	xor = v_2425 ^ v_01;
+	v_2425 = vec2(xor.y, xor.x);
+	v24 = v_2425.x;
+	v25 = v_2425.y;
 
 	// c = c + d
 
@@ -6881,9 +6890,10 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	// d = rotr64(d ^ a, 32)
 	v24 = v_2425.x;
 	v25 = v_2425.y;
-	v_2425 = v_2425 ^ v_01;
-	v24 = v_2425.y;
-	v25 = v_2425.x;
+	xor = v_2425 ^ v_01;
+	v_2425 = vec2(xor.y, xor.x);
+	v24 = v_2425.x;
+	v25 = v_2425.y;
 
 	// c = c + d
 
@@ -7671,9 +7681,10 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	// d = rotr64(d ^ a, 32)
 	v24 = v_2425.x;
 	v25 = v_2425.y;
-	v_2425 = v_2425 ^ v_01;
-	v24 = v_2425.y;
-	v25 = v_2425.x;
+	xor = v_2425 ^ v_01;
+	v_2425 = vec2(xor.y, xor.x);
+	v24 = v_2425.x;
+	v25 = v_2425.y;
 
 	// c = c + d
 
@@ -8455,9 +8466,10 @@ fn main(@builtin(global_invocation_id) id: vec3<u32>) {
 	// d = rotr64(d ^ a, 32)
 	v24 = v_2425.x;
 	v25 = v_2425.y;
-	v_2425 = v_2425 ^ v_01;
-	v24 = v_2425.y;
-	v25 = v_2425.x;
+	xor = v_2425 ^ v_01;
+	v_2425 = vec2(xor.y, xor.x);
+	v24 = v_2425.x;
+	v25 = v_2425.y;
 
 	// c = c + d
 
-- 
2.34.1