[CMSC 313 Home] | [Syllabus] | [Homework] | [Projects] | [Lecture Notes] | [Printable all notes] | [Files] | [NASM resource] |
The project is to be submitted on linux.gl.umbc.edu as submit cs313_squire proj1 convert.asm submit cs313_squire proj2 math_64.asm submit cs313_squire proj3 plotc.asm submit cs313_squire proj4 proj4.vhdl or submit cs313_squire proj4 proj4.v submit cs313_squire proj5 proj5.vhdl or submit cs313_squire proj5 proj5.v To see what is submitted submitls cs313_squire proj1 To delete a file that was submitted submitrm cs313_squire proj1 convert.asm
From anywhere you can reach the internet: ssh your-username@linux.gl.umbc.edu # or use putty, etc. your-password mkdir cs313 # or whatever directory name you want, only once cd cs313 # every time you log in for CMSC 313 # Get some sample files. (some not needed until later) cp /afs/umbc.edu/users/s/q/squire/pub/download/hello_64.asm . cp /afs/umbc.edu/users/s/q/squire/pub/download/intarith_64.asm . cp /afs/umbc.edu/users/s/q/squire/pub/download/fltarith_64.asm . cp /afs/umbc.edu/users/s/q/squire/pub/download/xor.circ . # be sure to type the final space dot # you can type in the command lines or get these Makefile's cp /afs/umbc.edu/users/s/q/squire/pub/download/Makefile_nasm Makefile # test compile hello_64.asm nasm -f elf64 hello_64.asm # or just make (you have Makefile) gcc -m64 -o hello_64 hello_64.o ./hello_64 > hello_64.out cat hello_64.out ls -ltr # see files you have Or type make # or make -f Makefile_nasm if copied that way ls -ltr If this did not work, see Help Desk, T.A. or instructor
Write and submit a NASM assembly language program "convert.asm" that implements the number conversions like you did for Homework 1. The files integer print and test ifint_64.asm integer arithmetic intarith_64.asm floating print and test ifflt_64.asm floating srithmetic fltarith_64.asm will be helpful. You start with two constants in the .data section dec1: db '1','2','6','.','3','7','5',0 bin1: dq 01010110110111B ; 10101101.10111 note where binary point should be You convert dec1 to a string of characters that is the binary representation of 126.3750 with a binary point and three bits to the right of the binary point. OK to just print integer bits on one line, then fraction bits on next line. Print your characters with printf. Look up how to compute in nasm, intarith.asm for converting dec1 '1','2','3' fltarith.asm for converting dec1 '.','3','7','5','0' Remember '1' is ASCII, also times 100 mov rax,0 mov al,[dec1] ; ASCII 1 sub rax,48 ; now have binary 1 imul qword [a100] ; now have binary 100 add 2*10+6 [dec1+1]*10+[dec1+2] mov [sum], rax mov rax,0 mov al,[dec1+1] ; ASCII 2 sub rax,48 ; now have binary 2 imul qword [a10] add rax, [sum] ; now have binary 120 mov [sum], rax When you get binary 126, in [sum] 1111110 now print the bits, do not need leading zeros. Then do .375, ok to get integer 375, float it, 375.0 and divide by float 1000.0 to get float 0.375 and use lecture 1. You convert bin1 to a string of characters that is the decimal representation of 10101101.10111. Print your characters (string) with printf or a kernel call. Use %ld to print 64 bit integers, or use %c to print characters. mov rax, [bin1] shr rax, 5 ; shift off .10111 5 bits '.' not stored mov [whole], rax ; save 10101101 can print with %ld mov rax, [bin1] and rax, 31 ; 31 in binary is 11111 save the 10111 fraction mov [frack], rax mov rax, [whole] and rax, 1 ; save only bottom bit mov [bit1], rax ; or [bit] using resb 7 mov rax, [whole] shr rax, 1 and rax, 1 ; save only second bit mov [bit2], rax ; or [bit+1] using resb 7 mov rax, [whole] shr rax, 2 and rax, 1 ; save only third bit mov [bit3], rax ; or [bit+2] using resb 7 print bits each a %ld then do fraction top bit is 0.5, next bit is 0.25, next bit is 0.124 1/2 1/4 1/8 etc. You may use any method of your choice, and you may print results as four numbers: '1','2','6' as 1's and 0's, binary. '.','3','7','5' as '.' 1's and 0's binary. 010101101 as a decimal number, integer .10111 as a decimal number, .dddd decimal fraction. submit your file, when it is working correctly, submit cs313_squire proj1 convert.asm Your file must assemble with no errors and execute with the commands: nasm -f elf64 convert.asm gcc -m64 -o convert convert.o ./convert # ./ needed if '.' not first in PATH Then submit cs313_squire proj1 convert.asm Note: '1' is an ASCII character. Subtract 48 from an ASCII character to get a binary number. Add 48 to a binary number in the range 0 to 9 to get the ASCII character '0' to '9'. '1','2','6' is 1*100 + 2*10 + 6 = 126, binary in a register. See horner_64.asm for sample loops. and loopint_64.asm another sample. You do not have to use loops, you can solve just specific problem. It is OK to process and print one character or digit at a time. A snippet of sample code for printing in Nasm: dec1: db '1','2','6','.','3','7','5', 0 fmt_char: db "%c",0 ; no '\n' thus no 10 fmt_dig: db "%1ld",0 ; print just one digit, e.g. 0 or 1 fmt_end: db 10, 0 ; just end line mov rdi,fmt_char ; print a single character mov rax, 0 ; be safe, zero all rax mov al, [dec1] ; byte into bottom of rax mov rsi, rax ; must go 64-bit to 64-bit mov rax, 0 ; no float call printf mov rdi,fmt_dig ; print a single character as digit mov rax, 0 ; be safe, zero all rax mov al, [dec1+1] ; next byte into bottom of rax sub rax, 48 ; change character digit to number ; imul rax, 10 ; '2' is 20 need to add up 1*100+2*10+4 mov rsi, rax ; must go 64-bit to 64-bit mov rax, 0 ; no float call printf mov rdi,fmt_end ; print end of line mov rax, 0 ; no float call printf Note: and rax,1 ; print with %1ld, prints bottom bit as 0 or 1 ; shr rax to get the bit you want Hint, C code, for converting .375 to .011 frac_bin.c frac_bin.out Beware rounding when storing double as integer. May need fld, fld, compp as in ifflt_64.asm One possible way to print first two binary numbers: dec1: db '1','2','6','.','3','7','5',0 fmt1: db "126 = %ld%ld%ld%ld",0 ; format for 126, first 4 bits fmt2: db "%ld%ld%ld", 10, 0 ; format for last 3 bits fmt3: db ".375 = %ld%ld%ld",10,0 ; format for .375 bitw: resq 7 ; reserve 7 64-bit word for bits 1 or 0 bitf: resq 3 ; reserve 3 words for bits 1 or 0 ; compute bits for 126 into bitw top bit first ; compute bits for .375 into bitf top bit first mov rdi, fmt1 ; first arg, format mov rsi, [bitw] ; second arg mov rdx, [bitw+1*8] ; third arg mov rcx, [bitw+2*8] ; fourth arg mov r8, [bitw+3*8] ; fifth arg mov r9, [bitw+4*8] ; sixth arg mov rax, 0 ; no xmm used call printf ; Call C function mov rdi, fmt2 ; first arg, format mov rsi, [bitw+5*8] ; seventh arg mov rdx, [bitw+6*8] ; eighth arg mov rax, 0 ; no xmm used call printf ; Call C function mov rdi, fmt3 ; first arg, format mov rsi, [bitf] ; second arg mov rdx, [bitf+8] ; third arg mov rcx, [bitf+2*8] ; forth arg mov rax, 0 ; no xmm used call printf ; Call C function for bin1 use %ld for integer part, %f for fraction Partial credit: 25% for decimal integer to binary 25% for decimal fraction to binary 25% for binary integer to decimal 25% for binary fraction to decimal Zero points if your convert.asm does not compile, if your convert.asm just prints the answers without doing the conversion. if two or more convert.asm are copied
Write and submit NASM assembly language functions that implement the given "C" functions in math_64.c The main program test_math_64.c that does not know how the functions are implemented. The test program is test_math_64.c The .h file with function prototypes is math_64.h Your correct output should be test_math_64.chk Note: There is zero credit when math_64.asm does not compile without errors. Your file must assemble with no errors and execute on linux.gl.umbc.edu with the commands: nasm -g -f elf64 math_64.asm gcc -g3 -m64 -o test_math_64 test_math_64.c math_64.o ./test_math_64 > test_math_64.out cat test_math_64.out Then submit cs313_squire proj2 math_64.asm OK to use printf for debug print to help in development. Please comment out debug print before submit. OK to have math_64.c as long comment in your math_64.asm For debugging due to segfault: gdb test_math_64 break main run step step keep stepping until segfault, thus see where you have a bug nexti use in place of step to step one instruction at a time or, if it runs to segfault, backtrace disassemble will show you address of instructions q you need q for quit, then y for yes to quit gdb yYour project is to convert math_64.c to math_64.asm
You may use pre_math_64.asm renamed to math_64.asm as a start. Compile and run to be sure compilation and execution are working, then add project code. Using my pre_math_64.asm as your math_64.asm in dot: mov rcx, [n] ; number of items n..1, need -8 for n-1..0 mov rax, [x] ; address of x array need *8 for each i mov rbx, [y] ; address of y array dotloop: fld qword [rax+rcx*8-8] ; quad word, n-1 is last subscript fmul qword [rbx+rcx*8-8] ; now have x[i]*y[i] in C in cross: Note addresses passed, so to do z[0] = 0.0 zero: dq 0.0 fld qword [zero] ; value of zero loaded mov rcx, [z] ; z: has address of callers z array fstp qword [rcx] ; can not say qword [[z]] ; add 8 to rax for next value ; rax has base address of x ; rbx has base address of y ; rcx has base address of z mov r8,[m] ; loop m=0 .. m<n mov r9,[j] ; inner loop j=0 .. j<n mov r10,[k] ; computed if ... fld qword [rax+8*r9] ; x[j] fmul qword [rbx+8*r10] ; x[j]*y[k] fadd qword [rcx+8*r8] ; z[m]+x[j]*y[k] fstp qword [rcx+8*r8] ; z[m] = z[m]+x[j]*y[k] fld qword [sign] fchs fstp qword [sign] ; sign = -sign All referenced files may be copied to your directory using: Replace xxx.x with the file you want. cp /afs/umbc.edu/users/s/q/squire/pub/download/xxx.x . when working: submit cs313_squire proj2 math_64.asm I do not know how to use .asm functions with other languages, yet for those who are not familiar with "C", here are more languages with math_64.??? and calling program in that language. math_64.py3 test_math_64_1.py3 tesst_math_64_1_py3.out test_math_64.py3 test_math_64_py3.out math_64.cpp test_math_64.cpp test_math_64_cpp.out test_math_64.java dot cross in here test_math_64_java.out Here is a very large conversion of matrix times vector multiply from C to .asm Large because I used much debug print, almost every access and almost every operation printed with debug print. I suggest lots of debug print while developing a program. Then, after it works, comment out debug print rather than delete debug print. Thus, 10 years later when you want to make changes, you can easily turn debug print back on. mat_vec.c source code to convert mat_vec_c.out to check .asm The .asm as developed with much debug print: mat_vec_64.asm source code converted mat_vec_64.out to check .asm Finally, the last lines agree, many segfault during conversion. Then comment out debug print and run again: mat_vec_64n.asm source code converted mat_vec_64n.out to check .asm Finally finally make smaller by deleting comments: mat_vec_64f.asm source code converted mat_vec_64f.out to check .asm Then, with confidence matrix time matrix was programmed: mat_mat.c source code to convert mat_mat_c.out to check .asm mat_mat_64.asm source code converted mat_mat_64.out to check .asm With no debug print
You are to write a program that does NOT use "C" functions or libraries. This project is based on lecture 9. You may use system calls or BIOS calls from Lecture 9 to implement the program. See hellos_64.asm for compiling, _start To print a character from a2 array at index row=i, col=j ; array of bytes, characters size 41*21: ; ; 0 40 column subscripts ; ------- +1.0 ; 0 | *** | ; rows | * * | 0.0 ; index 20 |* *| ; ------- -1.0 ; -pi 0 +pi ; a2: resb 41*21 ; in section .bss ; in section .txt mov rax, [i] ; a2+i*ncol+j is byte imul rax, [ncol] add rax, [j] add rax, a2 mov rsi, rax ; address of character to print mov rax, 1 ; system call 1 is write mov rdi, 1 ; file handle 1 is stdout mov rdx, 1 ; number of bytes syscall ; invoke operating system to do the write At end of j loop, just set rsi, a10 address of character 10 is line feed. copy lines above To compile and run your program, use: nasm -g -f elf64 plotc.asm ld -o plotc plotc.o ./plotc You only need to print one character at a time, rdx, 1 in syscall. Print 10, '\n' at end of each line, end of j print loop. Your program is to make a simple character plot of cos(x) for x from -Pi to Pi, -3.14159 to 3.14159 in 41 steps, dx = 0.15708 Use 21 rows, middle row for cos(0.0) = 1.0, top row for cos(Pi/2) = 0.0, bottom row for cos(-Pi)=cos(Pi) = -1.0 For each column plotting an '*' at row k = int(20.0 - (y+1)*10.0) A very small version of the plot would look like: * 9 columns, 7 rows * * * * * * * * Compute cos(x) in your program y = cos(x) = 1 - x^2/2! + x^4/4! - x6^/6! + x^8/8! OK to use code from horner_64.asm float af: dq 1.0, 0.0, -0.5, 0.0, 0.041667, 0.0, -0.001389, 0.0. 0.000025 N: dq 8 XF: dq 0.0 This computes YF = cos(XF) mov rcx,[N] ; loop iteration count initialization, n fld qword [af+8*rcx]; accumulate value here, get coefficient a_n h5loop: fmul qword [XF] ; * XF fadd qword [af+8*rcx-8] ; + aa_n-i loop h5loop ; decrement rcx, jump on non zero fstp qword [Y] ; store Y Then compute kf = 20.0 - (Y+1.0)*10.0 floating point Then store k as integer: fistp qword [k] Then compute double subscript, integer, k*ncol+j in rax Then store star: mov bl, [star] mov [a2+rax], bl Note: For printing mov rsi, rax // syscall (rcx for int) add rsi, a2 // not [a2+rax] need address If it runs to your satisfaction, Then submit cs313_squire proj3 plotc.asm The program in "C" is See plotc_64.c for possible method See plotc_64.outc "C" output See plotc.chk Nasm output See hornerc_64.asm for computing cos(x) // plotc_64.c simple plot of cos(x) #include <stdio.h>> #define ncol 41 #define nrow 21 int main(int argc, char *srgv[]) { char points[nrow][ncol]; // char == byte char point = '*'; char space = ' '; long int i, j, k, rcx; double af[] = {1.0, 0.0, -0.5, 0.0, 0.041667, 0.0, -0.001389, 0.0, 0.000025}; long int n = 8; double xx, y; double dx = 0.15708; // 6.2832/40.0 // clear points to space ' ' for(i=0; i0; rcx--) y = y*xf + af[i-1]; k = 20 - (y+1.0)*10.0; // scale 1.0 to -1.0, 0 to 20 // k = (int)(20.0+(y+1.0)*(-10.0) in assembly language printf("x=%f, y=%f, k=%d \n", x, y, k); fflush(stdout); points[k][j] = point; xf = xf + dx; // XF = XF + DX0 in assembly language } // print points for(i=0; i Nasm code for loops to clear and print array of characters array2_64.asm sample code array2_64.out output snippet of code, double loop, to clear array (ultra conservative, keeping i and j in memory) These 3 lines of "C" code become many lines of assembly // clear points to space ' ' for(i=0; i<nrow; i++) for(j=0; j<ncol; j++) points[i][j] = space; section .bss ; ncol=7, nrow=5 for demo a2: resb 21*41 ; two dimensional array of bytes i: resq 1 ; row subscript j: resq 1 ; col subscript k: resq 1 ; row subscript computed SECTION .text ; Code section. just snippet ; clear a2 to space mov rax,0 ; i=0 for(i=0; mov [i],rax loopi: mov rax,[i] ; reload i, rax may be used mov rbx,0 ; j=0 for(j=0; mov [j],rbx loopj: mov rax,[i] ; reload i, rax may be used mov rbx,[j] ; reload j, rbx may be used imul rax,[ncol] ; i*ncol add rax, rbx ; i*ncol + j mov dl, [spc] ; need just character, byte mov [a2+rax],dl ; store space mov rbx,[j] inc rbx ; j++ mov [j],rbx cmp rbx,[ncol] ; j<ncol jne loopj mov rax,[i] inc rax ; i++ mov [i],rax cmp rax,[nrow] ; i<nrow jne loopi ; end clear a2 to space ; j = 0; ; XF = X0; fld qword [X0] fstp qword [Xf] From horner_64.asm use fld qword [X0] fstp qword [XF] mov rax, 0 mov [j], rax ; j = 0 cos: mov rcx,[N] ; loop iteration count initialization, n fld qword [af+8*rcx]; accumulate value here, get coefficient a_n h5loop: fmul qword [XF] ; * XF fadd qword [af+8*rcx-8] ; + aa_n-i loop h5loop ; decrement rcx, jump on non zero fstp qword [Y] ; store Y ; k = 20.0 + (Y+1.0)*(-10.0) fistp qword [k] ; rax gets k * ncol + j ; put "*" in dl, then dl into [a2+rax] ; XF = XF + DX0; ; j = j+1; ; if(j != ncol) go to cos jloop ; copy clear a2 to space ; in jloop renamed, use syscall print from hellos_64.asm ; add rax,a2 replaces dl stuff ; mov rsi, rax (moved up) replaces mov rsi, msg ; replace any len with 1 ; after jloop insert line feed lf: db 10 ; mov rsi, lf in lpace of mov rsi, rax ; use exit code from hellos_64.asm ; no push or pop rbx in .data af: dq 1.0, 0.0, -0.5 ; coefficients of cosine polynomial, a_0 first dq 0.0, 0.041667, 0.0, -0.001389, 0.0, 0.000025 XF: dq 0.0 ; computed compute Y = cos(XF) Y: dq 0.0 ; computed N: dq 8 ; power of polynomial X0: dq -3.14159 ; start XF DX0: dq 0.15708 ; increment for XF ncol-1 times one: dq 1.0 nten: dq -10.0 twenty dq 20.0 a10 db 10,0 ; need address of a10 for linefeed ncol: dq 41 nrow: dq 21 spc: db ' ' star: db '*' Your plotc.asm can NOT use printf or any "C" functions. Thus you use global _start and _start: in place of global main and main: ; compile using nasm -g -f elf64 plotc.asm <-- submit plotc.asm ; ld -o plotc plotc.o # not gcc ; ./plotc > plotc.out ; cat plotc.out To do printout, use structure of clear to space, inserting: mov rax,[i] imul rax,[ncol] ; i*ncol add rax, [j] add rax, a2 ; a2 + i*ncol + j ; SYSCALL PRINT mov rsi, rax ; address of character to output mov rax, 1 ; system call 1 is write mov rdi, 1 ; file handle 1 is stdout mov rdx, 1 ; number of bytes syscall Your plotc.asm should contain, in this order: ; plotc.asm your name would be nice section .data ; stuff from above section .bss ; stuff from above section .text global _start _start: push rbp ; iloop ; jloop ; blank=space a2 array ; end jloop ; end iloop ; XF = 0.0 ; jloop ; compute cosine and put * in a2 array ; end jloop ; iloop ; jloop ; print a2 array ; end jloop ; end iloop pop rbp mov rax,60 mov rdi,0 syscall ; done, exit ; end of plotc.asm when working: submit cs313_squire proj3 plotc.asm
Use ghdl or iverilog: (Cadence VHDL and Verilog license expired) to use ghdl on linux.gl.umbc.edu, use Makefile_g Makfile_g to use iverilog on linux.gl.umbc.edu, use Makefile_ive Makfile_ive Both these Makefiles do both proj4 and proj5 For ghdl: Use proj4.vhdl as the start of project 4. Everything has been provided to build and test a 4-bit times 4-bit unsigned parallel multiply. In order to have less VHDL, a "madd4" entity was created. The multiplier can now be built from exactly four of the "madd4" entities. (Slightly different from multiplier used in the lecture.) The first "madd4" is in the file. You must code the three remaining "madd4" and code the "dot" merge of "cout" with the top three bits of the "sum", and the product bits "p". Notes: Each box is a madd4 entity. The boxes should be labeled a0:, a1:, a2: and a3:. The cout signals are named c(0), c(1), c(2) and c(3). The sum signals are named s0, s1, s2, p(6 downto 3). The dot where three wires join the cout wire is coded in VHDL as s0s <= c(0) & s0(3 downto 1); The s0s 4-bit signal goes into the madd4 'b' input. The first 'b' input must be four zero bits, signal zero4. The low order product bit, p(0) is the bottom bit of s0 and is coded in VHDL as p(0) <= s0(0); You need to type source vhdl_cshrc once per log on. You need first to follow vhdl instructions below on cs313.tar. For Verilog: Use proj4.v as the start of project 4. Note some different signal names are used, s0s is b1 and made with assign statements. This is a modification of mul4.v Fill in module madd4 using four madd modules. Then instantiate four madd4 to build the circuit. Your output should have correct 1 or 0 in place of "z" proj4_v.chk Other sample Verilog files add4.v mul4.v
cp /afs/umbc.edu/users/s/q/squire/pub/download/Makefile_g . cp /afs/umbc.edu/users/s/q/squire/pub/download/Makefile_ive . Now work project 4: Run the following commands to test your proj4.vhdl or proj4.v: make -f Makefile_g make -f Makefile_ive You do the submit, submit cs313_squire proj4 proj4.vhdl or submit cs313_squire proj4 proj4.v check your products by hand or by .chk proj4_g4.chk proj4_v4.chk
Use ghdl or iverilog: (Cadence VHDL and Verilog license expired) to use ghdl on linux.gl.umbc.edu, use Makefile_g Makfile_g to use iverilog on linux.gl.umbc.edu, use Makefile_ive Makfile_ive Both these Makefiles do both proj4 and proj5 Finish up the design and finish up the implementation of a six bit spin lock. You are given a starter VHDL file proj5.vhdl Or, use the given starter Verilog file proj5.v The spin lock is given by Use names A, B, C, D, E, F for the spin lock, there is debug print in proj5.vhdl and proj5.v for testing. Initialize all D flip flops to '0' except set A to '1'. Be sure to compute "activate" along with the Ain, Bin, etc. The test input has the name "rcvr" and has 10 entries. The code to be detected is 6 bits in the middle. The entity dff1 in VHDL, module dff6 in verilog, is used by the spin lock is ready to use in proj5.vhdl. The circuit symbol is: The module dff6 that is used by the spin lock is ready to use in proj5.v. similar circuit symbol. Your project is to finish the VHDL or verilog code for the spin lock. Look for "???" See lecture notes Lect 23 for method of converting a sequential circuit to digital logic. The lecture notes have legal VHDL statements, e.g Ain <= ... ; The Verilog uses Ain = ...; Code the digital logic in VHDL and add the VHDL statements into proj5.vhdl Copy files into your vhdl directory with the following commands: cp /afs/umbc.edu/users/s/q/squire/pub/download/proj5.vhdl . Make changes the run: make -f Makefile_g For Verilog Copy files into your vhdl directory with the following commands: cp /afs/umbc.edu/users/s/q/squire/pub/download/proj5.v . Run with make -f Makefile_ive Your output should have i=7 activate=1 proj5v.chk proj5_vhdl.chk in ghdl you get an error message. Then submit cs313_squire proj5 proj5.vhdl or submit cs313_squire proj5 proj5.v
Last updated 11/27/2019