require_rv32; P_NARROW_RD_RS1_ULOOP(8, 16, { uint32_t shamt = P_UFIELD(RS2, 0, 8) & 0xF; uint32_t result; if (shamt == 0) { result = p_rs1; } else { uint32_t shifted = p_rs1 >> shamt; uint32_t roundbit = (p_rs1 >> (shamt - 1)) & 1; result = shifted + roundbit; } p_rd = P_USAT_FULL(8, (sreg_t)result); })