Updates for puzzle 23 (#226)

Additional fixes following #203 for puzzle 23.

Author: mdewing

book/src/puzzle_23/elementwise.md

@@ -90,7 +90,7 @@ This performs element-wise addition across the entire SIMD vector (if supported)
 ### 5. **SIMD storing**
 
 ```mojo
-output.store[simd_width](idx, 0, result)  # Store 4 results at once (GPU-dependent)
+output.store[simd_width](Index(idx), result)  # Store 4 results at once (GPU-dependent)
 ```
 
 Writes the entire SIMD vector back to memory in one operation.
@@ -237,7 +237,7 @@ idx = indices[0]                                  # Linear index: 0, 4, 8, 12...
 a_simd = a.aligned_load[simd_width](Index(idx))       # Load: [a[0:4], a[4:8], a[8:12]...] (4 elements per load)
 b_simd = b.aligned_load[simd_width](Index(idx))       # Load: [b[0:4], b[4:8], b[8:12]...] (4 elements per load)
 ret = a_simd + b_simd                             # SIMD: 4 additions in parallel (GPU-dependent)
-output.store[simd_width](Index(global_start), ret)     # Store: 4 results simultaneously (GPU-dependent)
+output.store[simd_width](Index(idx), ret)     # Store: 4 results simultaneously (GPU-dependent)
 ```
 
 **Execution Hierarchy Visualization:**

book/src/puzzle_23/gpu-thread-vs-simd.md

@@ -38,10 +38,10 @@ Each GPU thread can process multiple data elements simultaneously using **SIMD (
 
 ```mojo
 # Within one GPU thread:
-a_simd = a.load[simd_width](idx, 0)      # Load 4 floats simultaneously
-b_simd = b.load[simd_width](idx, 0)      # Load 4 floats simultaneously
+a_simd = a.load[simd_width](Index(idx))      # Load 4 floats simultaneously
+b_simd = b.load[simd_width](Index(idx))      # Load 4 floats simultaneously
 result = a_simd + b_simd                 # Add 4 pairs simultaneously
-output.store[simd_width](idx, 0, result) # Store 4 results simultaneously
+output.store[simd_width](Index(idx), result) # Store 4 results simultaneously
 ```
 
 ## Pattern comparison and thread-to-work mapping

book/src/puzzle_23/tile.md

@@ -84,10 +84,10 @@ This `@parameter` loop unrolls at compile-time for optimal performance.
 ### 4. **SIMD operations within tile elements**
 
 ```mojo
-a_vec = a_tile.load[simd_width](i, 0)  # Load from position i in tile
-b_vec = b_tile.load[simd_width](i, 0)  # Load from position i in tile
+a_vec = a_tile.load[simd_width](Index(i))  # Load from position i in tile
+b_vec = b_tile.load[simd_width](Index(i))  # Load from position i in tile
 result = a_vec + b_vec                 # SIMD addition (GPU-dependent width)
-out_tile.store[simd_width](i, 0, result)  # Store to position i in tile
+out_tile.store[simd_width](Index(i), result)  # Store to position i in tile
 ```
 
 ### 5. **Thread configuration difference**
@@ -232,10 +232,10 @@ Tile 31 (thread 31): [992, 993, ..., 1023] ← Elements 992-1023
 ```mojo
 @parameter
 for i in range(tile_size):
-    a_vec = a_tile.load[simd_width](i, 0)
-    b_vec = b_tile.load[simd_width](i, 0)
+    a_vec = a_tile.load[simd_width](Index(i))
+    b_vec = b_tile.load[simd_width](Index(i))
     ret = a_vec + b_vec
-    out_tile.store[simd_width](i, 0, ret)
+    out_tile.store[simd_width](Index(i), ret)
 ```
 
 **Why sequential processing?**

book/src/puzzle_23/vectorize.md

@@ -68,8 +68,8 @@ This calculates the exact global position for each SIMD vector within the chunk.
 ### 3. **Direct tensor access**
 
 ```mojo
-a_vec = a.load[simd_width](global_start, 0)     # Load from global tensor
-output.store[simd_width](global_start, 0, ret)  # Store to global tensor
+a_vec = a.aligned_load[simd_width](Index(global_start))     # Load from global tensor
+output.store[simd_width](Index(global_start), ret)  # Store to global tensor
 ```
 
 Note: Access the original tensors, not the tile views.