Correct mma_sp_m16n8k16 implementation

eschnett · Jan 16, 2024 · a189cf4 · a189cf4
1 parent 119f40b
commit a189cf4
Showing 1 changed file with 1 addition and 1 deletion.
diff --git a/src/IndexSpaces.jl b/src/IndexSpaces.jl
@@ -2010,7 +2010,7 @@ CUDA.@device_override function mma_sp_m16n8k16(
         C[1].val % Int32,
         C[2].val % Int32,
         e.val % Int32,
-        F % Int32,
+        f % Int32,
     )
     return (Float16x2(D[1] % UInt32), Float16x2(D[2] % UInt32))
 end