An Ising machine stands as a compelling solution for large-scale combinatorial optimization problems, leveraging its unconventional computational architecture. We propose an optical parallel feedback system, employing mutual injection locking of vertical-cavity surface-emitting lasers (VCSELs) and programmable spatial light modulators (SLMs), specifically aimed at removing electrical bottlenecks to achieve all-optical computation. Through simulation results, the ability to solve a 16-bit square-lattice MAXCUT optimization problem is analyzed. As a proof-of-concept, we demonstrate an optical injection lock using single-mode master and slave VCSELs. Remarkably, we achieve polarization switching with less than 1% of the master VCSEL's injection power. Building upon this modest achievement, we outline a roadmap envisioning the integration of advanced nanophotonic devices, particularly metasurface-enabled SLMs and VCSELs. This strategic evolution targets scalability and programmability while upholding an all-optical computing paradigm.