Data processor with improved loop handling utilizing improved register

Abstract

A plurality of physical registers greater in number than can be addressed by instructions are provided. The physical registers are divided into a plurality of partial groups or windows. Each of the physical registers is specified by a physical register number determined by a combination of the number of the window and a register number contained in the window. The window number is indicated by a current floating point window pointer stored in a current floating point window pointer register. In executing a vector calculation program having repetitive loops of an instruction sequence, the window is changed for each loop by changing the value of the current floating point window pointer. Vector data to be calculated at the i-th loop is read from a main memory at the loop before the i-th loop. The calculated result of the vector data at the i-th loop is stored in the main memory at the loop after the i-th loop, thereby eliminating performance degradation to be caused by a slow data read time and improving the operation speed.