Emu

Emu is a procedural macro for writing numerical computation scripts. Emu translates your code to clean, compact OpenCL code at compile time and stores the translated code in the EMU global constant. You can then use OpenCL bindings such as ocl or rust-opencl to run code from your EMU on pretty much any GPU or CPU.

emu! {
	// adds a scalar to elements of a buffer
	add(global_buffer [f32], scalar f32) {
		global_buffer[get_global_id(0)] += scalar;
	}

	// multiplies elements of a buffer by a scalar
	multiply(global_buffer [f32], scalar f32) {
		global_buffer[get_global_id(0)] *= scalar;
	}

	// multiplies 2 matrices
	// n is the dimension of the matrices
	// a and b are the matrices to be multiplied, c is the result
	multiply_matrices(n i32, global_a [f32], global_b [f32], global_c [f32]) {
		// indices of cells to multiply
		let i: i32 = get_global_id(0);
		let j: i32 = get_global_id(1);

		// execute step of multiplication
		for k in 0..n {
			global_c[i * n + j] += global_a[i * n + k] * global_b[k * n + j];
		}
	}
}