matrix: directly implement dot product to avoid a copy

fixed_wide_vectors/src/macros/fixed_wide.rs

@@ -83,17 +83,91 @@ macro_rules! impl_wide_vector_operations {
 	};
 }
 
+#[doc(hidden)]
+#[macro_export(local_inner_macros)]
+macro_rules! impl_matrix_wide_mul_transpose_helper {
+	(
+		$lhs_axis:expr, $wide_mul:ident, $rhs:ident,
+		($struct: ident { $($field: ident), + }),
+		($from_struct: ident { $($from_field: ident), + }),
+		$static_field: ident
+	) => {
+		$crate::sum_repeating!(
+			$( + $lhs_axis.$field.$wide_mul($rhs.$from_field.$static_field) ) +
+		)
+	}
+}
+#[doc(hidden)]
+#[macro_export(local_inner_macros)]
+macro_rules! impl_matrix_wide_mul_inner {
+	(
+		// MatY<VecX>.MatX<VecZ> = MatY<VecZ>
+		$lhs:ident, $lhs_field_outer:ident, $wide_mul:ident, $rhs:ident,
+		$struct_inner_thru: tt, //VecX
+		($struct_inner: ident { $($field_inner: ident), + }), //VecX
+		($rhs_struct_inner: ident { $($rhs_field_inner: ident), + }), //VecZ
+		$rhs_outer: tt //MatX
+	) => {
+		$rhs_struct_inner {
+			$(
+				//directly dot product to avoid a copy
+				$rhs_field_inner: $crate::impl_matrix_wide_mul_transpose_helper!{
+					//lhs_axis
+					$lhs.$lhs_field_outer,$wide_mul,$rhs,
+					$struct_inner_thru, //VecZ
+					$rhs_outer, //MatX
+					$rhs_field_inner
+				}
+			), +
+		}
+	}
+}
+#[doc(hidden)]
+#[macro_export(local_inner_macros)]
+macro_rules! impl_matrix_wide_mul_outer {
+	(
+		// MatY<VecX>.MatX<VecZ> = MatY<VecZ>
+		$lhs:ident, $wide_mul:ident, $rhs:ident,
+		//result matrix shape
+		($struct_outer: ident { $($field_outer: ident), + }),//MatY
+		$rhs_struct_inner: tt,//VecZ
+		//inner loop shape
+		$struct_inner: tt,//VecX
+		$rhs_matrix: tt//MatX
+
+	) => {
+		$struct_outer {
+			$(
+				$field_outer: $crate::impl_matrix_wide_mul_inner!{
+					$lhs, $field_outer, $wide_mul, $rhs,
+					$struct_inner, //VecX
+					$struct_inner, //VecX
+					$rhs_struct_inner, //VecZ
+					$rhs_matrix //MatX
+				}
+			), +
+		}
+	}
+}
 
 // Notes:
 // Mat3<Vec2>.dot(Vec2) -> Vec3
+// lhs.dot(rhs) -> out
+// lhs = Mat3<Vec4>
+// rhs = Mat4<Vec2>
+// out = Mat3<Vec2>
 // Mat3<Vec4>.dot(Mat4<Vec2>) -> Mat3<Vec2>
-// mat.mat can be implemented off the back of mat.vec
+// how to matrix multiply:
+// RHS TRANSPOSE
+// Mat4<Vec2> -> Mat2<Vec4>
+// rhs_t = Mat2<Vec4>
+// inner loop:
+// out[y][x] = lhs[y].dot(rhs_t[x])
 
 #[doc(hidden)]
 #[macro_export(local_inner_macros)]
 macro_rules! impl_matrix_wide_mul {
 	(
-		//TODO: Fixed point impls
 		($struct_outer: ident { $($field_outer: ident), + }, $vector_outer: ident { $($vector_field_outer: ident), + }, $size_outer: expr),
 		($struct_inner: ident { $($field_inner: ident), + }, $matrix_inner: ident { $($matrix_field_inner: ident), + }, $size_inner: expr),
 		($rhs_struct_inner: ident { $($rhs_field_inner: ident), + }, $rhs_matrix_inner: ident { $($rhs_matrix_field_inner: ident), + }, $rhs_size_inner: expr),
@@ -103,13 +177,15 @@ macro_rules! impl_matrix_wide_mul {
 			paste::item!{
 			#[inline]
 			pub fn [<wide_dot_ $size_outer x $size_inner _ $size_inner x $rhs_size_inner _ $lhs _ $rhs>](self,rhs:$matrix_inner<$rhs_struct_inner<fixed_wide::fixed::Fixed<{$rhs},{$rhs*32}>>>)->$struct_outer<$rhs_struct_inner<fixed_wide::fixed::Fixed<{$lhs+$rhs},{($lhs+$rhs)*32}>>>{
-				//just made this up, don't trust it
-				let tr=rhs.transpose();
-				//TODO: use a macro expansion instead of transpose and map
-				self.map(|axis|
-					tr.map(|trax|
-						axis.[<wide_dot_ $lhs _ $rhs>](trax)
-					).to_vector()
+				$crate::impl_matrix_wide_mul_outer!(
+					//constituent idents
+					self,[<wide_mul_ $lhs _ $rhs>],rhs,
+					//result matrix shape
+					($struct_outer { $($field_outer), + }),
+					($rhs_struct_inner { $($rhs_field_inner), + }),
+					//inner loop shape
+					($struct_inner { $($field_inner), + }),
+					($matrix_inner { $($matrix_field_inner), + })
 				)
 			}
 			}