Assignment3: SoftCPU

tags: Computer Architecture 2021

Link

Assignment 1
Assignment 3 建置教學

Environment

• srv32 : A simple RISC-V 3-stage pipeline processor

Setting up the Environment

After finishing setting up the Environment,I try to make in the tools folder and sim folder.

$ cd srv32/sim/
$ make

Unfortunately, some errors occured when I make in the sim folder.

/home/xueyang/srv32/sim/../verilator/bin/verilator -O3 -cc -Wall -Wno-STMTDLY -Wno-UNUSED +define+MEMSIZE=128    --trace-fst --Mdir sim_cc --build --exe sim_main.cpp getch.cpp -o sim -f filelist.txt  ../rtl/top.v
make[1]: Entering directory '/home/xueyang/srv32/sim/sim_cc'
ccache g++  -I.  -MMD -I/home/xueyang/srv32/sim/../verilator/include -I/home/xueyang/srv32/sim/../verilator/include/vltstd -DVM_COVERAGE=0 -DVM_SC=0 -DVM_TRACE=1 -DVM_TRACE_FST=1 -faligned-new -fcf-protection=none -Wno-bool-operation -Wno-sign-compare -Wno-uninitialized -Wno-unused-but-set-variable -Wno-unused-parameter -Wno-unused-variable -Wno-shadow      -std=gnu++14 -Os -c -o getch.o ../getch.cpp
g++: error: unrecognized command line option ‘-faligned-new’; did you mean ‘-falign-jumps’?
g++: error: unrecognized command line option ‘-fcf-protection=none’; did you mean ‘-flto-partition=none’?
make[1]: *** [Vriscv.mk:63: getch.o] Error 1
make[1]: Leaving directory '/home/xueyang/srv32/sim/sim_cc'
%Error: make -C sim_cc -f Vriscv.mk exited with 2
%Error: Command Failed /home/xueyang/srv32/sim/../verilator/bin/verilator_bin -O3 -cc -Wall -Wno-STMTDLY -Wno-UNUSED +define+MEMSIZE=128 --trace-fst --Mdir sim_cc --build --exe sim_main.cpp getch.cpp -o sim -f filelist.txt ../rtl/top.v
make: *** [Makefile:61: sim] Error 2

So, I find the -faligned-new and -fcf-protection=none in the /srv32/verilator/include/verilated.mk

$ grep -r faligned-new

verilator/include/verilated.mk:CFG_CXXFLAGS_NO_UNUSED =  -faligned-new -fcf-protection=none -Wno-bool-operation -Wno-sign-compare -Wno-uninitialized -Wno-unused-but-set-variable -Wno-unused-parameter -Wno-unused-variable -Wno-shadow

Now, it can work smoothly.

Requirements 1

The hw.c is below.

#include<stdio.h>
#include<stdint.h>
 
int number_of_1bits(uint32_t  n){
    int count = 0;
    while (n != 0){
        n = n & (n - 1);
        count++;
    }
    return count;
}

int main()
{
	uint32_t input = 24;
	int ans = number_of_1bits(input);
	printf("%d\n",ans);
	return 0;
}

The Makefile is below.

include ../common/Makefile.common

EXE      = .elf
SRC      = hw3.c
CFLAGS  += -L../common
LDFLAGS += -T ../common/default.ld
TARGET   = hw3
OUTPUT   = $(TARGET)$(EXE)

.PHONY: all clean

all: $(TARGET)

$(TARGET): $(SRC)
	$(CC) $(CFLAGS) -o $(OUTPUT) $(SRC) $(LDFLAGS)
	$(OBJCOPY) -j .text -O binary $(OUTPUT) imem.bin
	$(OBJCOPY) -j .data -O binary $(OUTPUT) dmem.bin
	$(OBJCOPY) -O binary $(OUTPUT) memory.bin
	$(OBJDUMP) -d $(OUTPUT) > $(TARGET).dis
	$(READELF) -a $(OUTPUT) > $(TARGET).symbol

clean:
	$(RM) *.o $(OUTPUT) $(TARGET).dis $(TARGET).symbol [id]mem.bin memory.bin

I found that printf must contain \n in the end, otherwise it can't work fine. Also, we should make at the root directory.

I found that dhrystone and cormark is in the sw folder, so I create a foler to store hw.c and Makefile. At the root directory, I can use make hw3 and the result is following.

Requirements 2

0000003c <number_of_1bits>:
      3c:       00050793                mv      a5,a0
      40:       00000513                li      a0,0
      44:       00078c63                beqz    a5,5c <number_of_1bits+0x20>
      48:       fff78713                addi    a4,a5,-1
      4c:       00e7f7b3                and     a5,a5,a4
      50:       00150513                addi    a0,a0,1
      54:       fe079ae3                bnez    a5,48 <number_of_1bits+0xc>
      58:       00008067                ret
      5c:       00008067                ret

There are RAW data hazard as below.But I found that number_of_1bits function is not executed in the trace.log.

IF/ID	EX	WB
beqz a5,address	addi a4,a5,-1	and a5,a5,a4

we can found that branch penalty is 2 cycle.

Requirements 3

I have to check this note that jserv comment on my Lab1, so I decide to write a new program which calculating 2 bits at a time. Thw following is my new program.

#include<stdio.h>
#include<stdint.h>

int number_of_1bits(uint32_t n){
    uint32_t mask[] = {0x55555555,0x33333333,0x0F0F0F0F,0x00FF00FF,0x0000FFFF};
    for(int i = 0,j = 1;i < 5; i++,j <<= 1){
        n = (n & mask[i]) + ((n >> j) & mask[i]);
    }
    return n;
}

int main()
{
	uint32_t input = 0xFFFFFFFF;
	int ans = number_of_1bits(input);
	printf("%d\n",ans);
	return 0;
}

Before I make, I expect that the cycles count of the new program will be much less than that of the version 1 program.The following is new program information.

The following is first written.

Based on the same input value(0xFFFFFFFF),the result is not as I expected.
I modify the code of version 2. The following is a code of modified version 2.

#include<stdio.h>
#include<stdint.h>

int number_of_1bits(uint32_t n){
    uint32_t mask[] = {0x55555555,0x33333333,0x0F0F0F0F,0x00FF00FF,0x0000FFFF};
    n = (n & mask[0]) + ((n >> 1) & mask[0]);
    n = (n & mask[1]) + ((n >> 2) & mask[1]);
    n = (n & mask[2]) + ((n >> 4) & mask[2]);
    n = (n & mask[3]) + ((n >> 8) & mask[3]);
    n = (n & mask[4]) + ((n >> 16) & mask[4]);
    return n;
}

int main()
{
	uint32_t input = 0xFFFFFFFF;
	int ans = number_of_1bits(input);
	printf("%d\n",ans);
	return 0;
}

I do the loop unrolling for number_of_1bits function

Now,it look the same as version 1.Also,I found that the number of instruction of version 2 that it does not do the loop unrolling is 1541.After doing the loop unrolling, the number of instruction is 1510. It reduces to 31 intructions and 35 cycles.

Requirement 4

This assignment gave me a better understanding of hardware operation. Using the GTKWave sofrware,I can choose signal wire that I want to observe.From this assignment 3,I learned srv32 how to work on 3 pipelines and how to handle data harzard.