# Homework3: SoftCPU
## Setup
* download [riscv-none-embed-gcc](https://xpack.github.io/riscv-none-embed-gcc/)
* set path

 cd $HOME
wget https://github.com/xpack-dev-tools/riscv-none-embed-gcc-xpack/releases/download/v10.2.0-1.2/xpack-riscv-none-embed-gcc-10.2.0-1.2-linux-x64.tar.gz
tar zxvf xpack-riscv-none-embed-gcc-10.2.0-1.2-linux-x64.tar.gz
cd riscv-none-embed-gcc
echo "export PATH=`pwd`/bin:$PATH" > setenv
cd $HOME
source riscv-none-embed-gcc/setenv

* download other thins we need

sudo apt-get install lcov
sudo apt-get install ccache

* and to git clone [srv32](https://github.com/sysprog21/srv32)
* build simulation through Makefile

cd srv32/tools
make
cd ../sim
make

## Origin Code
* cause the compiler can't recoginized booling function , so I change the original code in assignment1 of bool to be int .

#include <stdio.h>

int isPerfectSquare(int num){

    int k = 1;
    while( num > 0)
    {
       num  = num - k;
       k = k +  2;
    }
    return num == 0;	
}
int main(void){
	int number1 = 14 , number2 = 16;
	printf("is %d a perfect square number?%d\n",number1,isPerfectSquare(number1));
	printf("is %d a perfect square number?%d\n",number2,isPerfectSquare(number2));
}

and the result is below:

## Optimization

* because the origin code are already simple,so it's hard to optimize it by chaging the whole architecture
* In my first try,I seek a solution for another way : **binary search**

#include <stdio.h>


int isPerfectSquare(int num){

int m;
int l;
int r = num;
int sq;
while(l <= r){
   m = (l + r) /2;
   sq = m * m ;
   if(sq < num ){
   	l = m +1;
   }
   else if(sq > num){
   	r = m -1;
   }
   else{
   	return 1;
   }
	    
}
return 0;

}


int main(void){
	int number1 = 14 , number2 = 16;
	printf("is %d a perfect square number?%d\n",number1,isPerfectSquare(number1));
	printf("is %d a perfect square number?%d\n",number2,isPerfectSquare(number2));
}


* but the number of cycles getting higher...

So I keep finding another optimization, and I get a new way to reduce clock cycles(although just a little) : **Newton's method**

int isPerfectSquare(int num){
    double x = 1;
    int i;
    for(i=0;i<3;i++ )
	    x = 0.5 * (x + num /  x);
    x = (int)x;
    if((x*x) == num) ret
    else return 0;
}

* because in my example the 2 inputs are little so I set the slack variable i to the range [0,2], and it can work until the number > 36

## Simple 3-stage pipeline RISC-V processor
* **3-stage**
* **full forward**
* **branch penalty is 2**
* **SRV32 pipeline architecture**:


## Waveform
* first, open gtkwave after installed it thtough:

sudo apt install gtkwave
gtkwave

* then we get in riscv , and cause srv32 have only three stages, so I just pick up if_pc/ex_pc/wb_pc


* see `sim/trace.log`
* take lw instruction for example

00000250 00052783 read 0x000213f0 => 0x00020588, x15 (a5) <= 0x00020588

* pc is ```000002500``` ,instr is ```0001B517```
* read data of lw is 0x00020588
* imem_rdata is 0x00052783 ,correct

* we can see in 2 clks later(i.e. lw instruction in WB stage) the dmem_rdata signal be ```0x00000000``` same as the value we want to read