Note: emulating "store byte" needs 17 instructions in the general case. Can be shortened if you know in advance which byte in the word will be replaced. //

Note: emulating "store byte" needs 17 instructions in the general case. Can be shortened if you know in advance which byte in the word will be replaced. // Implement store byte using only addi, add, nand, sll, sra, jal, jalr, blt, lw, sw // void store_byte(void *p, char val); .macro nand dst,src1,src2 and \dst,\src1,\src2 xori \dst,\dst,-1 .endm .globl store_byte store_byte: // preparation, can be shortened if e.g. byte offset is known addi t0,x0,3 addi t1,x0,-1 nand t2,a0,t0 nand t2,t2,t1 // byte offset sll t2,t2,t0 // bit offset nand t0,t0,t1 nand a0,a0,t0 nand a0,a0,t1 // word ptr addi t0,x0,255 sll t0,t0,t2 // mask // the actual work sll a1,a1,t2 // shifted val nand a1,a1,t0 // ~val nand t0,t0,t1 // ~mask lw a3,(a0) nand a3,a3,t0 // ~word, dst field all 1s nand a3,a3,a1 // insert val sw a3,(a0) jalr x0,(x1) Having access to the full RV32I instruction set, minus `sb`, shortens the preparation from 10 instructions to 5, but the actual work stays at 7 instructions.