Improve the documentation for the Kernel trait

In particular, note the caveat that the input may be uninitialized if `beta` is
zero. This is an aspect that ought to be modeled in the types but currently is
not.

src/gemm/kernels.rs

@@ -317,9 +317,14 @@ unsafe fn simd_gemm<S: SimdFloat, const MR: usize, const NR_REGS: usize>(
     }
 }
 
-/// Kernel that computes a small tile of a matrix multiplication output.
+/// Kernel that computes a small tile of a general matrix multiplication (GEMM) or
+/// general matrix-vector multiplication (GEMV).
 ///
-/// The kernel corresponds to Loop 6 (the "microkernel") in Page 4 of [^1]. The
+/// This trait is an interface for the architecture-specific inner loop for matrix
+/// multiplication and matrix-vector multiplication, as well as the methods that
+/// pack the input matrices into a format that is efficient for the kernel to use.
+///
+/// The GEMM kernel corresponds to Loop 6 (the "microkernel") in Page 4 of [^1]. The
 /// tile size depends upon the kernel and is specified by the `MR` and `NR`
 /// associated constants. See Section 3.2 [^1] for theory behind choosing the
 /// `MR` and `NR` values.
@@ -365,14 +370,21 @@ pub unsafe trait Kernel<LhsT, RhsT, OutT>: Sync {
         cols: Range<usize>,
     );
 
-    /// Compute a tile of the output matrix. The output is stored in row-major
-    /// order with `MR` rows and `NR` columns, a row stride of `tile_row_stride`
-    /// and column stride of 1.
+    /// Compute a tile of the output matrix.
+    ///
+    /// The output is stored in row-major order with `MR` rows and `NR` columns,
+    /// a row stride of `tile_row_stride` and column stride of 1.
+    ///
+    /// The `a` and `b` inputs are the input matrices packed by the `pack_a_block`
+    /// and `pack_b_block` methods. The `depth` input specifies the number of
+    /// columns of A and rows of B that are in the packed inputs.
     ///
     /// # Safety
     ///
     /// The caller must ensure that `tile_ptr` points to a buffer of the correct
-    /// size.
+    /// size. If `beta` is zero then the output may be uninitialized and must not
+    /// be read by the implementation. If `beta` is non-zero then the output must
+    /// be initialized and the implementation will read from it.
     unsafe fn kernel(
         &self,
         tile_ptr: *mut OutT,
@@ -398,7 +410,7 @@ pub unsafe trait Kernel<LhsT, RhsT, OutT>: Sync {
     ///
     /// # Safety
     ///
-    /// The caller must ensure that the kernel is supported on the current
-    /// system.
+    /// If `beta` is non-zero then the output may be uninitialized and must not be
+    /// read by the implementation.
     fn gemv_kernel(&self, out: &mut [OutT], a: &[LhsT], b: Matrix<RhsT>, alpha: f32, beta: OutT);
 }