A desired property of secure programs is control flow integrity (CFI): an attacker must not be able to alter how instructions are chained as specified in the program. Numerous techniques try to achieve this property with various trade-offs. But to achieve fine-grained CFI, one is required to extract a precise control flow graph (CFG), describing how instructions are chained together. Unfortunately it is not achievable in general. In this paper, we propose a way to overcome this impossibility result by restricting the instruction set architecture (ISA) semantics. We show that forbidding indirect jumps unlocks a precise CFG extraction for all acceptable programs. We discuss the implications and limitations of the new semantics and argue for the adoption of restricted ISAs for security-related computation.