Light Utilization with Multicycle Operational Stages (LUMOS) RISC-V Processor Core
Name : Ehsan Seyfibonab
Student number : 97412229
Date : 07/02/2024
product[WIDTH + FBITS - 1 : FBITS]
li sp, 0x3C00
This instruction loads the immediate value 0x3C00 into the stack pointer register sp, setting the stack pointer to point to memory address 0x3C00
addi gp, sp, 392
This instruction adds the immediate value 392 to the stack pointer sp and stores the result in the global pointer register gp. It moves the global pointer to a location 392 bytes above the stack pointer.
loop:
This indicates the beginning of a loop labeled loop
flw f1, 0(sp)
Load a single-precision floating-point value from memory at the address pointed to by the stack pointer sp and store it in floating-point register f1
flw f2, 4(sp)
Load a single-precision floating-point value from memory at the address 4 bytes above the stack pointer sp and store it in floating-point register f2.
fmul.s f10, f1, f1
Multiply the floating-point values in registers f1 and f1, storing the result in floating-point register f10.
fmul.s f20, f2, f2
Multiply the floating-point values in registers f2 and f2, storing the result in floating-point register f20.
fadd.s f30, f10, f20
Add the floating-point values in registers f10 and f20, storing the result in floating-point register f30.
fsqrt.s x3, f30
Compute the square root of the floating-point value in register f30 and store the result in the x3 register.
fadd.s f0, f0, f3
Add the floating-point value in register f3 to the floating-point value in register f0, storing the result in register f0.
addi sp, sp, 8
Increment the stack pointer sp by 8 bytes, effectively moving it to the next set of floating-point values in memory.
blt sp, gp, loop
Branch to the loop label if the value in the stack pointer sp is less than the value in the global pointer gp, effectively looping back to load more floating-point values.
ebreak
This instruction triggers a breakpoint exception, which is commonly used for debugging purposes to pause program execution.
Overally this code loads x axis and y axis values at each iteration and calculates distance for each set And then it increases the stack pointer and continues this iteration.
First I defined 6 states for different situations in multiplication. First stage is the initializatio stage where I set partialproducts to be 0 and multiplierCircuitInput1 and multiplierCircuitInput2 to be the first 16 bit of input operands. At the end I set the next state to be FIRSTMUL. At FIRSTMUL state I store multiplierCircuitResult to partialProduct1. And the I set multiplierCircuitInput1 to be the last 16 bit of operand_1 and multiplierCircuitInput2 to be the first 16 bits of operand_2. And the I set the next state to be SECONDMUL. At SECONDMUL state I store multiplierCircuitResult to partialProduct2 with 16 bit shift to left. And the I set multiplierCircuitInput1 to be the first 16 bit of operand_1 and multiplierCircuitInput2 to be the last 16 bits of operand_2. And the I set the next state to be THIRDMUL. At SECONDMUL state I store multiplierCircuitResult to partialProduct3 with 16 bit shift to left. And the I set multiplierCircuitInput1 to be the last 16 bit of operand_1 and multiplierCircuitInput2 to be the last 16 bits of operand_2. And the I set the next state to be FOURTHMUL. At SECONDMUL state I store multiplierCircuitResult to partialProduct3 with 32 bit shift to left. And the I set the next state to be ADD. At ADD state I add the four partial products and store the result in product and I set product_ready to 1 and state to INITIALSTATE.
In this block radicant is the register to hold radicant which appends the next two bits of operand to itself.
temp_root is the register which will substracted from radicant. This register is the result with 01 appending to the right of it.
We will have (WIDTH + FBITS) / 2 iterations and the final_result will be the root register and root_ready will be 1.