MULH is not supported but there's nothing preventing the toolchain from generating them right now (except that it's rarely a useful instruction). Structs, however, sometimes benefit from these optimizations. This is not an issue suggesting to implement MULH, but instead to modify the GCC manycore profile such that MULH is very very low. This should influence the compiler to choose MUL + shift instead of MULH, since there is no unique functionality that MULH provides
https://github.com/bespoke-silicon-group/bsg_replicant/issues/801
MULH is not supported but there's nothing preventing the toolchain from generating them right now (except that it's rarely a useful instruction). Structs, however, sometimes benefit from these optimizations. This is not an issue suggesting to implement MULH, but instead to modify the GCC manycore profile such that MULH is very very low. This should influence the compiler to choose MUL + shift instead of MULH, since there is no unique functionality that MULH provides