A Laboratory Manual for the SPARC

Arthur B. Maccabe
Jeff Vandyke
Department of Computer Science
The University of New Mexico
Albuquerque, NM 87131

Copyright c 1993, 1994, 1995, 1996

Permission is granted to make copies for classroom use

January 16, 1996

Introduction
This laboratory manual was developed to provide a hands on introduction to the
SPARC architecture. The labs are based on ISEM, an instructional SPARC emulator developed at the University of New Mexico.
The ISEM package is available via anonymous ftp. To obtain a copy ftp to cs.unm.edu
and cd to pub/ISEM. The README file in this directory should provide you with the
information needed to obtain a working copy for your environment. ISEM currently runs
on most Unix boxes., There are plans to port ISEM to the DOS/Windows environment as
well as the Mac. If you have any difficulty getting a copy of ISEM or would like more
information regarding the status of the ports, send email to [email protected].
In addition to an instruction set emulator for the SPARC, the ISEM package includes
emulations for several devices (a character mapped display, a bitmapped display, a UART,
etc.), an assembler, and a linker. The assember is a slightly modified version of the GNU
assembler (gas version 2.1.1). The primary modification is the addition of several synthetic
operations to support loads and stores from/to arbitrary locations in memory. These operations are described in the first few laboratory write ups.
The lab maunal is complete in that it covers all of the SPARC operations and instruction
formats. As such, students should not require individual copies of The SPARC Architecture Manual. However, we have found it useful to have copies of the SPARC Architecture
Manual available to students on a reference basis.
This lab manual has been designed to accompany Computer Systems: Architecture, Organization, and Programming by Maccabe (Richard D. Irwin, 1993). However, the manual does
not directly reference the text and, as such, could be used with other text books.

Laboratory 1
Using ISEM (the Instructional
SPARC Emulator)
1.1 Goal
To describe the translation of assembly language programs and introduce basic features of
ISEM, the instructional SPARC emulator.

1.2 Objectives
After completing this lab, you will be able to:







Assemble and link SPARC programs,

Load SPARC programs,
Run SPARC programs,
Examine/modify memory locations, and
Examine/modify registers.

1.3 Discussion
In this lab we describe the translation process and introduce the basic features of ISEM.
We begin by describing the translation process: the steps used to translate an assembly
language program into an executable program. After describing the translation process,
we describe how you can execute and test executable programs using ISEM.
The first thing you need is a simple SPARC program. Figure 1.1 illustrates a simple
SPARC program. We will use this program in the remainder of this lab.
Activity 1.1

Using a text editor, enter the program shown in Figure 1.1 into a file called foo.s.

1.3.1 Assembling and Linking Programs

Once you have a SPARC program, you will need to assemble and link your program before you can run it using ISEM. The assembler translates assembly language programs into
object program. The linker transforms object programs into executable programs that you can
execute using ISEM
To assemble a SPARC program, you need to specify the file that holds the assembly
language program and the file that will hold the resulting object program. Assembly commands start with the name of the assembler, isem-as, followed by (optional) listing specifications, followed by an output file specification, followed by the name of the source file.
1

Lab 1. Using ISEM (the Instructional SPARC Emulator)

x:
y:
z:

.data
.word 0x42
.word 0x20
.word 0

!
!
!
!

variables
initialize x to 0x42
initialize y to 0x20
initialize z to 0

.text

! instructions

start:

end:

set
ld
set
ld
add
add
set
st

x, %r2
[%r2], %r2
y, %r3
[%r3], %r3
%r2, %r2, %r2
%r2, %r3, %r2
z, %r3
%r2, [%r3]

ta

!
!
!
!
!
!
!
!

&x --> %r2

x --> %r2
&y --> %r3
y --> %r3
r2 + r2 --> %r2
r2 + r3 --> %r2
&z --> %r3
r2 --> x

Figure 1.1 A sample SPARC program implementing z

= 2x + y

Hexadecimal ISEM reports all of its results in hexadecimal. To simplify your interaction
with ISEM, we will use hexadecimal notation in our programs. The program shown in
Figure 1.1 uses the hexadecimal constants: 0x42 and 0x20.

We will use the listing specification ;als (to generate a source listing and a list of symbols). The output specification consists of a -o followed by the name of the output file.
Figure 1.2 illustrates the interaction that results when you assemble the program in foo.s.
Activity 1.2

Assemble the program in foo.s, placing the object code in the file foo.o.

The linker is called isem-ld. Linker commands start with the name of the linker, isem-ld,
followed by an output specification, followed by the name of the file containing the object
program.
Activity 1.3

Link the object code in foo.o, placing the executable program in the file foo.

1.3.2 Testing Your Program

Once you have assembled and linked your assembly language program, you can execute
your program using ISEM. To start ISEM, you need to issue the ISEM command, isem.
When you start ISEM, you will see an introductory message followed by the ISEM
prompt (ISEM>). When you see this prompt, you can issue an ISEM command. Figure 1.3 illustrates the interaction you should expect to see when you start up ISEM.
The load command
The load command is used to load executable programs into the ISEM environment. The
load command consists of the name of the command (load) followed by the name of a file
containing an executable program.
Figure 1.4 illustrates the ISEM load command. In examining this interaction, note that
ISEM tells you where it loaded the program text (instructions) and data (variables). ISEM

Lab 1. Using ISEM (the Instructional SPARC Emulator)

; ;

% isem-as als o foo.o foo.s

SPARC GAS foo.s

page 1

1
2 0030 00000042

x:

.data
.word 0x42

! variables
! initialize x

3 0034 00000020

y:

.word 0x20

! initialize y

4 0038 00000000

z:

.word 0

! initialize z

.text

! instructions

to 0x42
to 0x20
to 0
5
6
7
8
8
9
10
10
11
12

start:
0000 05000000
8410A000
0008 C4008000
000c 07000000
8610E000
0014 C600C000
0018 84008002

set

x, %r2

! &x --> %r2

ld
set

[%r2], %r2
y, %r3

! x --> %r2
! &y --> %r3

ld
add

[%r3], %r3
%r2, %r2, %r2

! y --> %r3
! r2 + r2 -->

13 001c 84008003

add

%r2, %r3, %r2

! r2 + r3 -->

14 0020 07000000
14
8610E000
15 0028 C420C000
16 002c 91D02000
17

set

z, %r3

! &z --> %r3

st
ta

%r2, [%r3]
0

! r2 --> x

%r2
%r2

SPARC GAS

end:

foo.s

DEFINED SYMBOLS
foo.s:2
foo.s:3
foo.s:4
foo.s:7
foo.s:17

page 2

2:00000030
2:00000034
2:00000038
1:00000000
1:00000030

x
y
z
start
end

UNDEFINED SYMBOLS

Figure 1.2 Illustrating the isem-as command.

File Names Traditionally, file name suffixes indicate the type of program stored in a
file. Files that contain assembly language programs have a suffix of .s. Files that contain
object programs have a suffix of .o. Files that contain executable programs do not usually
have a suffix.
In the preceding paragraphs, we used the files foo.s (for the assembly language program), foo.o (for the object program), and foo (for the executable program).

also tells you the current value of the program counter (PC) and the next program counter
(nPC). Finally, ISEM shows you the next instruction that it will execute (i.e., the instruction
pointed to by the PC).

Lab 1. Using ISEM (the Instructional SPARC Emulator)

% isem
Instructional SPARC Emulator
Copyright 1993 - Computer Science Department
University of New Mexico
ISEM comes with ABSOLUTELY NO WARRANTY
ISEM Ver 1.00a : Mon Nov 1 20:25:01 MST 1993

Figure 1.3 Illustrating the isem command.

>

ISEM
load foo
Loading File: foo
2000 bytes loaded into Text region at address 8:0
2000 bytes loaded into Data region at address a:2000
PC: 08:00000020

nPC: 00000024

PSR: 0000003e

N:0 Z:0 V:0

C:0
start

sethi

0x8, %g2

Figure 1.4 Illustrating the load command.

Activity 1.4

Start ISEM and load the file foo.

Note that the instruction (sethi 0x8, %g2) doesnt look like the first instruction in the
sample program (set x, %r2). We will discuss the reason for this when we consider synthetic
operations in Lab 9. For now, it is sufficient to know that set instruction may be implemented
using two instructions: a sethi instruction followed by an or instruction.
The trace command
You can execute your program, one instruction at a time, using the trace command. The
trace command executes a single instruction and reports the values stored in the registers,
followed by the next instruction to be executed.
Figure 1.5 illustrates three successive executions of the trace command. Note that register 2 (first row, third column) now has the value 0x00000042the value used in the initialization of x.
To complete the execution of the sample program, you need to issue nine more trace
commands (a total of twelve trace commands). As you issue trace commands, note how
the values in registers 2 and 3 change. When you have executed all of the instructions in
the sample program, ISEM will print the message Program exited normally. Figure 1.6
illustrates the execution of the last two trace commands.
Activity 1.5

Execute the sample program by issuing twelve trace commands.

The dump command

The trace command is useful because it lets you see how each instruction affects the registers when it is executed. You can also examine the contents of memory using the dump
command. You can issue a dump command any time you see the ISEM> prompt.
To use the dump command, you need to specify the range of memory values that you
want to examine. A range of memory locations is specified using two memory addresses

Lab 1. Using ISEM (the Instructional SPARC Emulator)

>

ISEM
trace
----0--- ----1------7--G 00000000 00000000
00000000
O 00000000 00000000
00000000
L 00000000 00000000
00000000
I 00000000 00000000
00000000
PC: 08:00000024
C:0

----2--- ----3--- ----4--- ----5--- ----6--00002000 00000000 00000000 00000000 00000000

00000000 00000000 00000000 00000000 00000000
00000000 00000000 00000000 00000000 00000000
00000000 00000000 00000000 00000000 00000000
nPC: 00000028

start+04 :

or

PSR: 0000003e

N:0 Z:0 V:0

%g2, 0x60, %g2

>

ISEM
trace
----0--- ----1------7--G 00000000 00000000
00000000
O 00000000 00000000
00000000
L 00000000 00000000
00000000
I 00000000 00000000
00000000
PC: 08:00000028
C:0

----2--- ----3--- ----4--- ----5--- ----6--00002060 00000000 00000000 00000000 00000000

00000000 00000000 00000000 00000000 00000000
00000000 00000000 00000000 00000000 00000000
00000000 00000000 00000000 00000000 00000000
nPC: 0000002c

start+08 :

ld

PSR: 0000003e

N:0 Z:0 V:0

[%g2], %g2

>

ISEM
trace
----0--- ----1------7--G 00000000 00000000
00000000
O 00000000 00000000
00000000
L 00000000 00000000
00000000
I 00000000 00000000
00000000
PC: 08:0000002c
C:0

----2--- ----3--- ----4--- ----5--- ----6--00000042 00000000 00000000 00000000 00000000

00000000 00000000 00000000 00000000 00000000
00000000 00000000 00000000 00000000 00000000
00000000 00000000 00000000 00000000 00000000
nPC: 00000030

start+0c :

sethi

PSR: 0000003e

N:0 Z:0 V:0

0x8, %g3

Figure 1.5 Illustrating the trace command.

separated by a comma. Memory addresses can be specified using an integer value or the
name of a label. For example, to see the final value stored in the memory location associated
with the label z you could use the range z, z.
Figure 1.7 illustrates the dump command. The dump command reports memory values
in pairs of hexadecimal digits. Each word of memory is 32 bits and, as such, requires four

Lab 1. Using ISEM (the Instructional SPARC Emulator)

>

ISEM
trace
----0--- ----1------7--G 00000000 00000000
00000000
O 00000000 00000000
00000000
L 00000000 00000000
00000000
I 00000000 00000000
00000000
PC: 08:0000004c
C:0

----2--- ----3--- ----4--- ----5--- ----6--000000a4 00002068 00000000 00000000 00000000

00000000 00000000 00000000 00000000 00000000
00000000 00000000 00000000 00000000 00000000
00000000 00000000 00000000 00000000 00000000
nPC: 00000050

start+2c :

ta

PSR: 0000003e

N:0 Z:0 V:0

[%g0 + 0x0]

>

ISEM
trace
Program exited normally.
----0--- ----1--- ----2--- ----3------7--G 00000000 00000000 000000a4 00002068
00000000
O 00000000 00000000 00000000 00000000
00000000
L 00000000 0000004c 00000050 00000000
00000000
I 00000000 00000000 00000000 00000000
00000000
PC: 09:00000800
nPC: 00000804
C:0
end+7b0 :

jmpl

----4--- ----5--- ----6--00000000 00000000 00000000

00000000 00000000 00000000
00000000 00000000 00000000
00000000 00000000 00000000
PSR: 0000009d

N:0 Z:0 V:0

[%l2], %g0

Figure 1.6 Completing execution of the sample program.

pairs of hexadecimal digits. In examining Figure 1.7, note that z holds the value 0xa4
(the final value in register %r2).

>

ISEM
dump z,z
0a:00002068 00 00 00 a4 00 00 00 00

........

Figure 1.7 Illustrating the dump command.

Activity 1.6

Use the dump command to examine the values stored in the memory locations x, y, and z.

Memory Addresses When you are interacting with ISEM, memory addresses can be
specified using integer constants or the labels defined in an assembly language program.

The symb command

Because memory addresses can be specified using the labels defined in an executable program, you may be interested in knowing which labels have defined values. After you

Lab 1. Using ISEM (the Instructional SPARC Emulator)

have loaded an executable program, you can use the symb command to display the values
defined by the program. Figure 1.8 illustrates the symb command.

>

ISEM
symb
Symbol List
end
start
x
y
z

:
:
:
:
:

00000050
00000020
00002060
00002064
00002068

Figure 1.8 Illustrating the symb command.

Activity 1.7

Use the symb command to examine the labels defined by the sample program.

The edit command

The edit command sets the values stored in memory locations. Each edit command requires
two arguments, the memory location to edit and the value to store in the memory location.
For example, you could use the command edit x 0x20 to set the value of the memory
location labeled x to the value 0x20.
Activity 1.8

Use the edit command to change the values associated with x and y.

The reg command

You can use the reg command to set the values in the registers. The reg command can
have zero, one or two arguments. The first argument (if present) must be the name of a
SPARC register. The second argument (if present) must be a value. With zero arguments
the reg command prints the contents of all of the SPARC registers. With one argument
the reg command prints the value of the specified register. With three arguments the reg
command sets the register specified by the first argument to the value specified by the
second argument.
Table 1.1 summarizes the names of the SPARC registers. For now, you only need to be
familiar with the integer registers and the program counters. We will consider the remaining registers in later labs.
Table 1.1 SPARC Registers
Group name
Integer registers
Program counters
Multiply/divide register
Floating point registers
Floating point queue
Floating point status register
Processor status register
Window invalid mask
Trap base register

Activity 1.9

Register names
%r0%r31
%pc, %npc
%y
%f0%f31
%fq
%fsr
%psr
%wim
%tbr

Use the reg command to examine the values of the registers.

Lab 1. Using ISEM (the Instructional SPARC Emulator)

For example, the command reg %pc start resets the PC to the start of the program.
You can use this command when you want to rerun the sample program.
The run command
You can use the run command to execute your program, starting with the instruction
pointed to by the PC. This command does not take any arguments and executes instructions until it encounters a breakpoint, an illegal instruction, or a program termination instruction (ta 0).
Activity 1.10 Use the reg command to reset the value of %pc. Then use the run command to rerun the
sample program.

Figure 1.9 illustrates the run command. This interaction starts by setting the %pc and
then issuing the run command. Note that the run command produces an error message.
In this case, the run command stopped executing program instructions because it encountered an illegal instruction. Whenever you load a program, ISEM makes sure that there is
an illegal instruction following the last instruction in your program.

>

ISEM
reg %pc start
Register: %pc = 20

>

ISEM
run
Program exited normally.

Figure 1.9 Illustrating the reg and run commands.

The break command

You can use the break command to set breakpoints (the run command terminates when
it encounters a breakpoint). To set a breakpoint, you can issue a break command with
a single argument, the address of the breakpoint. After you have set a breakpoint, the
run command will stop executing your program just after it executes the instruction at the
specified address.
Figure 1.10 illustrates the use of the break command. In this case, the breakpoint stops
the program 20 bytes after the start labeljust before the add %r2, %r2, %r2 instruction.
Note that ISEM reports the breakpoint address in hexadecimal.

>

ISEM
reg %pc start
Register: %pc = 20

>

ISEM

break start+20

>

run
ISEM
Breakpoint encountered at start+14

Figure 1.10 Illustrating the break command.

The help command

The help command may be the most important command provided by ISEM. The help
command takes a single, optional argument. When it is supplied, the argument is the
name of the item that you would like more information about. Without any arguments,
the help command tells you the items that help knows about.

Lab 1. Using ISEM (the Instructional SPARC Emulator)

The quit command

When you are done interacting with ISEM, you can issue the quit command.

1.3.3 Memory regions

The ISEM/SPARC architecture provides two separate regions of memory: one for text
(code) and another for data. This means that there are two memory locations with the
address 100, one in the program memory and another in the data memory. ISEM is reasonably intelligent about when it uses each of these memories. ISEM always fetches instructions from the program memory and the load and store operations always refer to the data
memory. In addition, the load command always loads programs into the program memory
and the dump and edit commands use the data memory by default.

1.4 Summary
In this lab we have described the steps in the translation process and introduced that basic
functions provided by ISEM.
The ISEM assembler, isem-as, translates an assembly language program into object
code. The linker, isem-ld, translates an object code program into an executable program.
Figure 1.11 summarizes the steps uses to translate an assembly language program into an
executable program.

'& %$ '& $% '

$
&%
foo.s

foo.o

assembly
code

isem-as

foo

object
code

;als ;o foo.o foo.s

isem-ld

executable
code

;o foo foo.o

Figure 1.11 Translation steps

When you have an executable program, you can use ISEM to execute and test your
program. Figure 1.12 illustrates the basic components of ISEM and shows the commands
that manipulate these components. Table 1.2 summarizes the ISEM commands that we
have discussed in this lab.

1.5 Review Questions

1. What is the name of the tool that translates an assembly language program into object
code? What is the name of the tool that translates object code into an executable
program?
2. What does the trace command do?
3. What does the dump command do?
4. What does the symb command do?
5. What does the edit command do?
6. What does the reg command do?
7. What does the run command do?
8. What is a breakpoint? Explain how to set a breakpoint in ISEM.

10

Lab 1. Using ISEM (the Instructional SPARC Emulator)

Registers

Memory
Symbol
table
Data

OC

CC

symb

Text
(code)

%r0 (zero)
%r1 (reserved)
%r2
%r3
..
.

C
1
CC ;;;
6
C
;
?
dump
load
edit

%r30
%r31
%pc

6
?
reg

Figure 1.12 The components of ISEM

Table 1.2 The commands of ISEM
Syntax
break address
dump [address [, address]]
edit address value
help [topic]
load filename
quit
symb
reg [register [, value]]
run
trace

Meaning
Set an execution breakpoint.
Display the contents of memory
Set the contents of a memory location.
Provide information about an isem topic or command.
Load an executable program.
Exit ISEM.
Display the symbol table.
Display or set the contents of a register.
Execute instructions to a breakpoint or illegal instruction.
Execute the next instruction.

Notes:
Items in square brackets ([ . . . ]) are optional.
address
a memory address (may be a label or number
value
an integer value (0x. . . means hexadecimal notation)
filename
the name of a file
register
the name of a register

1.6 Exercises
1. After you have successfully assembled the sample program, perform the following
modifications to the second line of the program (x: .word 0x42). In each case, you
should start with the original sample program and you should write down the error
message (if any) produced by the assembler.
a. remove the : following x,
b. remove the . preceding word.
2. After you have successfully assembled the sample program, perform the following
modifications to the eleventh line of the program (ld [%r3], %r3). In each case, you
should start with the original sample program and you should write down the error

Lab 1. Using ISEM (the Instructional SPARC Emulator)

11

message (if any) produced by the assembler.

a. remove the [ and ] surrounding first %r3,
b. remove the % preceding the first r3.
3. Load the sample program into ISEM, run it, and then issue the following ISEM commands. In each case, describe the output produced by ISEM and explain why ISEM
produced the results that it produced (you may need to use the help command to get
further information about the commands).
a.
b.
c.
d.
e.

dump x,z
dump z,x
dump z
dump x
dump start,end

4. In examining Figure 1.4, note that the value of the PC is given as 08:00000020. What
does the 08: mean?
5. In examining Figure 1.7, note that the address for z is given as 0a:00002068. What
does the 0a: mean?

12

Lab 1. Using ISEM (the Instructional SPARC Emulator)

Laboratory 2
Assembly Language Programming
2.1 Goal
To introduce the fundamentals of assembly language programming.

2.2 Objectives
After completing this lab, you will be able to write assembly language programs that use:









The .text and .data assembler directives,

The .word assembler directive,
The integer registers (%r0%r31),
The (synthetic) set operation,
The load and store operations,
The signed integer addition and subtraction operations, and
The (synthetic) mov operation.

2.3 Discussion
In this lab we introduce the fundamentals of SPARC assembly language programming. In
particular, we consider basic assembler directives, register naming conventions, the (synthetic) load and store operations, the integer addition and subtraction operations, and the
(synthetic) register copy and register set operations. We begin by considering the structure
of assembly language programs.

2.3.1 Assembly language

Assembly language programs are line-oriented. That is, the assembler translates an assembly language program one line at a time. The assembler recognizes four types of lines:
empty lines, label definition lines, directive lines, and instruction lines.





A line that only has spaces or tabs (i.e., white space) is an empty line. Empty lines are
ignored by the assembler.
A label definition line consists of a label definition. A label definition consists of an
identifier followed by a colon (:). As in most programming languages, an identifier
must start with a letter (or an underscore) and may be followed by any number of
letters, underscores, and digits.
A directive line consists of an optional label definition, followed by the name of an
assembler directive, followed by the arguments for the directive. In this lab we will
consider three assembler directives: .data, .word, and .text.
13

14

Lab 2. Assembly Language Programming

An instruction line consists of an optional label definition, followed by the name of

an operation, followed by the operands. In this lab we will consider five operations:
load, store, set, add, and sub.

Every line can conclude with a comment. Comments begin with the character !.
Whenever it encounters a !, the assembler ignores the ! and the remaining characters
on the line.
Activity 2.1 Consider the SPARC program presented in Figure 1.1. For each nonempty line in the program,
identify any labels defined and identify any assembler directives and assembly language instructions.

2.3.2 Directives
In this lab we introduce three directives: .data, .text, and .word. The first two (.data and
.text) are used to separate variable declarations and assembly language instructions. The
.word directive is used to allocate and initialize space for a variable.
Each group of variable declarations should be preceded by a .data directive. Each
group of assembly language instructions should be preceded by a .text directive. Using
these directives, you could mix variable declarations and assembly language instructions;
however, for the present, your assembly language programs should consist of a group of
variable declarations followed by a group of assembly language instructions.
A variable declaration starts with a label definition (the name of the variable), followed
by a .word directive, followed by the initial value for the variable. The assembler supports
a fairly flexible syntax for specifying the initial value. For now, we will use simple integer values to initialize our variables. By default, the assembler assumes that numbers are
expressed using decimal notation. You can use hexadecimal notation if you use the 0x
prefix. Example 2.1 illustrates a group of variable declarations.

xy

Example 2.1 Give directives to allocate space for three variables, , , and . You should initialize these
variables to decimal 23, hexadecimal 3fce, and decimal 42, respectively.

x:
y:
z:

.data
.word
.word
.word

23
0x3fce
42

!
!
!
!

start
int x
int y
int z

a
=
=
=

group of variable declarations

23;
0x3fce;
42;

2.3.3 Labels
In an assembly language program, a label is simply a name for an address. For example,
given the declarations shown in Example 2.1, x is a name for the address of a memory
location that was initialized to 23. On the SPARC an address is a 32-bit value. As such,
labels are 32-bit values when they are used in assembly language programs.

2.3.4 Integer registers

The SPARC integer unit provides thirty-two general purpose registers. Each integer register holds 32-bits. The integer registers are called %r0 through %r31. In addition to the
names %r0 through %r31, the integer registers have alternate names (aliases) as shown in
Table 2.1.
The letter used in each group of aliases (g, o, l, or i) denotes the name for the group of
registers. The group names are related to procedure calling conventions. We will discuss

Lab 2. Assembly Language Programming

15

Table 2.1 Aliases for the integer registers

Integer registers
%r0%r7
%r8%r15
%r16%r23
%r24%r31

Alternate names
%g0%g7
%o0%o7
%l0%l7
%i0%i7

Group name
Global registers
Output registers
Local registers
Input registers

the meanings of these group names when we consider register windows in Lab 11. In the
meantime, we will use the %r names in our assembly language programs. As you may
have noted, ISEM uses the alternate names when it reports the contents of the registers
and when it shows the next instruction to execute.
Register Names Register names on the SPARC always start with a percent sign (%).
For example, the integer registers are named %r0 through %r31.

2.3.5 %r0
The value stored in %r0 is always zero and cannot be altered. If an instruction specifies
%r0 is used as the destination, the result is simply discarded. It is not an error to execute
an instruction that specifies %r0 as the destination for the result; however, the contents of
%r0 will not be altered when this instruction is executed.
%r0 Register %r0 always holds the value zero. The value stored in this register cannot
be altered.

2.3.6 The set operation

The set operation can be used to load a 32-bit signed integer constant into a register. Every
set instruction has two operands: the 32-bit value followed by the destination register.
Table 2.2 summarizes the set operation.
Table 2.2 The set operation
Operation
register set
Notes:
siconst32
rd

Assembler syntax
set siconst32 , rd

Operation implemented
reg[rd] = siconst32

a 32-bit signed integer constant (can be specified by a label)

the destination register

Destination last In SPARC assembly language instructions, the destination is specified

as the last operand.

16

Lab 2. Assembly Language Programming

Example 2.2 Using set instructions, write code that will load the value 0x42 into register %r2 and the
address of x (from Example 2.1) into register %r3.
set
set

0x42, %r2
x, %r3

2.3.7 The load and store operations

The SPARC is based on the load/store architecture. This means that registers are used
as the operands for all data manipulation operations. The operands for these operations
cannot be in memory locations. Table 2.3 summarizes simple versions of the load and store
operations. (We will cover these operations in more detail in later labs.)
Table 2.3 The load and store operations
Operation
load word
store word

Assembler syntax
ld [rs], rd
st rs, [rd]
Notes:
rd
rs

Operation implemented
reg[rd] = memory[reg[rs]]
memory[reg[rd]] = reg[rs]

the destination register

the source register

Source and destination registers When we introduce assembly language syntax, the
names rs and rd are used to denote source and destination registers, respectively. When an
instruction uses multiple source registers, we use subscripts to distinguish these registers.

2.3.8 The addition and subtraction operations

The SPARC uses 2s complement representation for signed integer values. Signed additions
and subtractions are performed using 32-bit arithmetic (the source and destination values
are 32 bits).
Table 2.4 summarizes the signed addition and subtraction operations provided by the
SPARC. The SPARC provides two instruction formats for each of the arithmetic operations. Both formats use three explicit operandstwo source operands, and a destination
operand. In the first format both of the source operands are in registers. In the second
format, one of the source operands is in a register while the other is a small constant value.
This constant value may be negative or positive; however, its 2s complement representation must fit in 13 bits. Example 2.3 presents a SPARC assembly language program that
illustrates variable declarations and the operations (load, store, add, and sub) that we have
described in this lab.
Integer Constants We use the name siconstn to denote a signed integer constant in assembly language syntax. The subscript indicates, , the number of bits used in the 2s
complement representation of this value.

Lab 2. Assembly Language Programming

17

Table 2.4 The addition and subtraction operations

Operation
integer addition
integer subtraction
Note:
siconst13

Assembler syntax
add rs1 ,rs2 , rd
add rs, siconst13 , rd
sub rs1 , rs2 , rd
sub rs, siconst13 , rd

Operation implemented
reg[rd] = reg[rs1 ] + reg[rs2 ]
reg[rd] = reg[rs] + siconst13
reg[rd] = reg[rs1 ] reg[rs2 ]
reg[rd] = reg[rs] siconst13

;
;

a 13-bit (2s complement) signed integer constant.

2.3.9 Program termination

Programs to be run in the ISEM environment should terminate their execution by executing
the instruction ta 0. Whenever this instruction is executed, ISEM will stop executing
instructions and print the message Program exited normally.
Example 2.3

Write a SPARC assembly language program to evaluate the statement

a:
b:
c:
d:

start:

end:

.data
.word
.word
.word
.word

0x42
0x43
0x44
0x45

.text
set
ld
set
ld
set
ld
set
ld

a, %r1
[%r1],
b, %r1
[%r1],
c, %r1
[%r1],
d, %r1
[%r1],

add
sub
sub
set
st

%r2, %r3, %r2

%r4, %r5, %r3
%r2, %r3, %r2
a, %r1
%r2, [%r1]

ta

a = (a + b) ; (c ; d).

%r2

--> %r2

%r3

--> %r3

%r4

--> %r4

%r5

--> %r5

!
!
!

a + b --> %r2
c ; d --> %r3
(a + b) ; (c ; d)

--> %r2

(a + b) ; (c ; d)

--> a

Activity 2.2 Using a text editor, enter the program shown in Example 2.3 into a file, assemble it, link it, and
test it using isem.

2.3.10 The mov operation

We conclude this lab by considering another (synthetic) operation: mov. The mov operation is used to copy the value stored in one register to another register. This operation can
also be used to load a small integer value into a register. Table 2.5 summarizes the mov
operations.

18

Lab 2. Assembly Language Programming

Table 2.5 The register copy and register set operations

Operation
register copy
load constant

Assembler syntax
mov rs, rd
mov siconst13 , rd

Operation implemented
reg[rd] = reg[rs]
reg[rd] = signextend(siconst13 )

Because you can always use the set operation to load a 13-bit value to an integer register,
the second version of the mov operation is redundant for integer registers. However, as we
will discuss, this version of the mov operation is used to load the other state registers on
the SPARC.

2.4 Summary
In this lab we have introduced the basics of SPARC assembly language programming. We
began by considering the structure of an assembly language program. We then considered
the names and uses of the integer registers. We then introduced three assembler directives:
.text, .data, and .word. The first two (.text and .data) are used to identify sections of an assembly language program. The last two (.data and .word) are used to declare and initialize
variables. We will consider additional assembler directives in later labs. We concluded the
lab by introducing six assembly language operations: set, load, store, add, sub, and mov.
Figure 2.1 provides a graphical illustration for several of the operations that we have
introduced in this lab. In particular, this figure illustrates the data paths used in the load,
store, addition, and subtraction operations.
Memory

text

data

* Instruction Register
siconst
ifetch


?

6
6
?
?
store

Integer
@@add/sub;;
registers
load (%r0%r31)


13

Figure 2.1 Illustrating the load, store, add, and subtract operations
The set and mov operations are synthetic (or pseudo) operations. That is, these operations are not really SPARC operations. Instead, the assembler translates these operations
into one or more SPARC operations when it assembles your program. We will consider
synthetic operations in Lab 9

2.5 Review Questions

1. Explain the difference between a directive, an operation, and an instruction. Give an
example of each.

Lab 2. Assembly Language Programming

19

2. How are the integer registers named on the SPARC?

3. How many integer registers are there on the SPARC?
4. For each of the integer registers that have special attributes, explain the special attributes.
5. What does siconst13 mean when used to specify assembly language syntax?
6. The subtraction operation has two source operands. Which operand is subtracted
from the other?
7. Describe when you would use the set operation.

2.6 Exercises
1. Suppose that the SPARC did not provide a (synthetic) register copy operation, explain
how you could emulate this operation.
2. For each of the following statements, write, assemble, and test a SPARC assembly
language fragment that implements the statement. Be certain to declare and initialize
all variables in your assembly language programs.
a.
b.
c.
d.
e.

a = c + d.
a = (c + d) ; (c + b + d ; e).
a = (d ; 13) + (a + 23).
a = d + 9832.
a = 87765 ; c.

20

Lab 2. Assembly Language Programming

Laboratory 3
Implementing Control Structures
3.1 Goal
To cover the implementation of control structures using the SPARC instruction set.

3.2 Objectives
After completing this lab, you will be able to write assembly language programs that use:






The condition code register,

Operations that set the condition code register,
The conditional and unconditional branching operations of the SPARC, and
Conditional nullification.

3.3 Discussion
In this lab we introduce a subset of the SPARC branching operations. In particular, we
introduce the operations that provide conditional and unconditional branching based on
the bits in a condition code register.
We begin by considering the bits in the condition code register of a SPARC processor.
After introducing these bits, we consider the operations that affect the bits in the condition
code register. We then consider the conditional and unconditional branching operations
that use the bits in the condition code register to control branching. Next, we introduce
nullification (annulment) in the branching operations of the SPARC. We conclude by considering several examples to illustrate the SPARC operations and by introducing the compare operation provided by the SPARC.

3.3.1 The condition code register

The condition code register on the SPARC has four bits: Z (Zero), N (Negative), C (Carry),
and V (oVerflow). The standard arithmetic operations (e.g., addition and subtraction) do
not update the bits in the condition code register. Instead, there are special operations that
update the condition code register. Table 3.1 summarizes a collection of operations that
update the bits in the condition code register. The names for these operations have a suffix
of cc to indicate that they update the bits in the condition code register.
In most cases, the effect that an operation has on the condition codes is just what you
would expect. Most of these operations set the Z bit when the result of the operation is
zero, and clear this bit when the result is nonzero. Similarly, most of these operations set
the N bit when the result of the operation is negative, and clear this bit when the result
is nonnegative. The V bit is usually set when the (signed integer) result of the operation
21

22

Lab 3. Implementing Control Structures

Table 3.1 Updating the condition code register

Operation
Signed integer addition
Signed integer subtraction

Operation name
addcc
subcc

cannot be stored in 32 bits, and cleared when the result can be stored in 32 bits. Finally, the
C bit is set when the operation generates a carry out of the most significant bit, and cleared
otherwise.
In most contexts, you will be most interested in the N and Z bits of the condition code
register and we will emphasize these bits in the remainder of this lab. We will consider the
remaining bits in the condition code register (the C and V bits) at greater length in Lab 13.

3.3.2 Branching operations

The SPARC provides 16 basic branching operations. These operations are summarized in
Table 3.2. Note that the first two operations, ba (branch always) and bn (branch never), are
unconditionalthe operation specifies whether the branch is taken. The remaining operations are conditional branching operations. When these operations are used, the branch
is only taken when the specified condition in met. In last column of Table 3.2 we use a
boolean expression involving the bits of the condition code register to specify the condition. The condition is satisfied if the boolean expression results in the value 1; otherwise
(if the expression results in 0), the condition is not satisfied and the processor continues
with sequential execution of instructions. The target specified in an assembly language
instruction is a label defined by the program.
Table 3.2 Branching operations on the SPARC
Operation
Branch always
Branch never
Branch not equal
Branch equal
Branch greater
Branch less or equal
Branch greater or equal
Branch less
Branch greater, unsigned
Branch less or equal, unsigned
Branch carry clear
Branch carry set
Branch positive
Branch negative
Branch overflow clear
Branch overflow set

Assembler syntax
ba
target
bn
target
bne
target
be
target
bg
target
ble
target
bge
target
bl
target
bgu
target
bleu
target
bcc
target
bcs
target
bpos target
bneg target
bvc
target
bvs
target

Branch condition
1 (always)
0 (never)
not Z
Z
not (Z or (N xor V))
Z or (N xor V)
not (N xor V)
N xor V
not (C or Z)
C or Z
not C
C
not N
N
not V
V

In addition to the operation names defined in Table 10, the SPARC defines several synonyms for these operations. These synonyms are summarized in Table 3.3.
Like most RISC machines, the SPARC uses a branch delay slot. By default, the instruction following a branch instruction is executed whenever the branch instruction is

Lab 3. Implementing Control Structures

23

Table 3.3 Synonyms for branching operations

Operation
Branch nonzero
Branch zero
Branch greater or equal, unsigned
Branch less, unsigned

Operation name
bnz
bz
bgeu
blu

Synonym for
bne
be
bcc
bcs

executed.
SPARC assemblers provide a special (synthetic) operation, nop, for situations when it is
not convenient to put a useful instruction in the delay slot of a branch instruction. In assembly language a nop instruction has no operands (i.e., a nop instruction is fully specified
by the name of the operation). When a nop instruction is executed, it does not alter any
of the registers or values stored in memory. However, the use of nop instructions causes
the processor to execute more instructions and, as such, increases the time required to execute the program. Example 3.1 illustrates the conditional and unconditional branching
operations.
Example 3.1 Translate the following C code fragment into SPARC assembly language.
int temp;
int x = 0;
int y = 0x9;
int z = 0x42;
temp = y;
while( temp > 0 )
x = x + z;
temp = temp - 1;

To simplify the translation, we fill the branch delay slots with nop instructions.
.data
x:
.word
0
y:
.word
0x9
z:
.word
0x42

start:

top:
test:

end:

.text
set
ld
set
ld
mov
add
ba
nop
add
subcc
bg
nop
set
st
ta

y, %r1
[%r1], %r2
z, %r1
[%r1], %r3
%r0, %r4
%r2, 1, %r2
test
%r4, %r3, %r4
%r2, 1, %r2
top

x, %r1
%r4, [%r1]
0

! well use %r2 for temp

! well use %r3 for z
! well use %r4 for x
!
!
!
!
!
!
!

set up for decrement

test the loop condition
BRANCH DELAY SLOT
x + z --> x
temp - 1 --> temp
temp > 0 ?
BRANCH DELAY SLOT

! store x

24

Lab 3. Implementing Control Structures

Activity 3.1 After each trace command, ISEM reports the values of the bit in the condition code register.
Type the program shown Example 3.1 into a file, assemble it, link it, and load it into ISEM. Trace the program
execution, noting how each instruction affects the bits in the SPARC condition code register.

The SPARC keeps track of the instructions to execute using two program counters: PC,
and nPC. The first program counter, PC, holds the address of the next instruction to execute. The second program counter, nPC, holds the next value for PC. Usually, the SPARC
updates the program counters at the end of each instruction execution by assigning the
current value of nPC to PC, and adding 4 to the value of nPC. When it executes a branching operation, the SPARC assigns the current value of nPC to PC and then updates the
value of nPC. If the branch is taken, nPC is assigned the value of the target specified in the
instruction; otherwise, nPC is incremented by 4. The branch delay slot arises because the
PC is assigned the old value of nPC (before nPC is assigned the target of the branch).
Activity 3.2 After each trace command, ISEM reports the values of PC and nPC. Run the program shown
in Example 3.1 noting the changes to PC and nPC.

3.3.3 Nullification
Every branching instruction can specify that the affect of the instruction in the branch delay slot is to be nullified (annulled in SPARC terminology) if the branch specified by the
conditional branching instruction is not taken. In assembly language, this conditional nullification is specified by appending a suffix of ,a to the name of the branching operation.
Example 3.2 illustrates conditional nullification.
Example 3.2 Rewrite the code fragment shown in Example 3.1 so that the code has meaningful instructions
in the branch delay slots.
x:
y:
z:

start:

top:

end:

.data
.word
.word
.word

0
0x9
0x42

.text
set
ld
set
ld
mov

y, %r1
[%r1], %r2
z, %r1
[%r1], %r3
%r0, %r4

add
subcc
bg,a
add

%r2, 1, %r2
%r2, 1, %r2
top
%r4, %r3, %r4

!
!
!
!

set
st
ta

x, %r1
%r4, [%r1]
0

! store x

! well use %r2 for temp

! well use %r3 for z
! well use %r4 for x
set up for decrement
temp - 1 --> temp
temp > 0 ?
x + z --> x

Lab 3. Implementing Control Structures

25

3.3.4 The (synthetic) integer comparison operation

Assemblers for the SPARC provide a synthetic integer comparison operation. You can use
this operation when the data manipulation operations do not establish the needed values
in the condition code register. Table 3.4 summarizes the integer comparison operation. This
operation can be used to compare the contents of two registers or to compare the contents
of a register with a small integer constant.
Table 3.4 Signed integer comparison
Operation
integer comparison

Assembler syntax
cmp sr1 , sr2
cmp sr, siconst13

Operation implemented
reg[sr1 ] reg[sr2 ]
reg[dr] = reg[sr] siconst13

3.3.5 Delayed control-transfer couples

When a branch instruction is in the delay slot of another branch instruction, the pair of
branch instructions is called a delayed control-transfer couple. If you use a delayed
control-transfer couple on the SPARC, the first branch operation should be an unconditional branch; otherwise, the sequence of instructions executed when the delayed controltransfer couple is executed is not defined. We will consider delayed control-transfer couples in greater depth when we consider traps and exceptions in Lab 16.

3.4 Summary
In this lab we have introduced the condition code register, the basic branching operations,
and the integer comparison operation. The branching operations include two unconditional branch operations (ba and bn) and a host of conditional branching operations. The
SPARC branching operations have a branch delay slot. That is, the instruction following a branch instruction is executed whenever the branch instruction is executed. The
SPARC provides conditional annulment of the instruction in the branch delay slot. When
the branch operation specifies annulment (using the operator suffix ,a), the affects of the
instruction are canceled (note, the instruction is executed, but the execution has no affect).

3.5 Review Questions

1. What are the bits in the condition code register.
2. Name two operations that affect the bits in the condition code register and explain
how they affect these bits.
3. What are the two program counters on the SPARC. Explain how these program counters are used.
4. Is the affect of the instruction in the delay slot of an annulled branch canceled when
the branch is take or when the branch is not taken? Explain why the designers of the
SPARC chose the this version of nullification.

26

Lab 3. Implementing Control Structures

3.6 Exercises
1. Suppose that your assembler did not provide an integer comparison operation. Explain how you could implement this operation using the other SPARC operations
that we have considered in this and previous labs.
2. Consider the SPARC code presented in Example 3.2. Currently, the loop is executed
y times. If y is larger than z it would be better to execute the loop z times.
Rewrite the code shown in Example 3.2 to take advantage of this observation.
3. Write a SPARC program that has four variables: x  , y > , z , and w. Your program
should assign the quotient of x=y to z and the remainder to w. (You should write this
code using the operations presented in this and previous labs. Do not use the SPARC
integer multiplication or division operations.
4. Write a SPARC program that will compute the greatest common divisor of a and b
and assign this value to c.

Laboratory 4
Multiplication and Division
4.1 Goal
To cover the SPARC operations related to multiplication and division.

4.2 Objectives
After completing this lab, you will be able to write assembly language programs that use:

The signed and unsigned multiplication and division operations,

4.3 Discussion
In this lab we consider the SPARC operations related to integer multiplication and division.
We begin by considering the signed integer multiplication and division operations.

4.3.1 The multiplication and division operations

The integer multiplication operations multiply two 32-bit source values and produce a 64bit result. The most significant 32 bits of the result are stored in the Y register (%y) and the
remaining 32 bits are stored in one of the integer registers. Figure 4.1 illustrates the use of
the Y register during a multiplication operation.
The integer division operations divide a 32-bit value into a 64-bit value and produce
a 32-bit result. The Y register provides the most significant 32 bits of the 64-bit dividend.
One of the source values provides the least significant 32 bits, while the other provides
the 32 bit divisor. A implementation of the SPARC may optionally store the remainder in
the Y register. ISEM does not store the remainder in the Y register, so we will adopt this
convention in our presentation. Figure 4.2 illustrates the use of the Y register during the
division operation.
Table 4.1 summarizes the assembly language syntax for the integer multiplication and
division operations provided by the SPARC. Like the addition and subtraction operations,
the multiplication and division operations have two assembly language formats: one that
uses registers for both source operands and another that uses a register and a small constant
value for the source operands.
Table 4.2 summarizes the names for the signed and unsigned integer multiplication and
division operations. Note that each operation has two SPARC operations: one that affects
bits in the condition code register (e.g., smulcc), and another that does not (e.g., smul).
Example 4.1 illustrates the use of these operations.

27

28

Lab 4. Multiplication and Division

Instruction register
iconst13

?
?
@@ smul/umul ;;
@
;

Y register

most significant
32 bits

Integer
registers
(%r0%r31)

least significant
32 bits

Figure 4.1 Integer multiplication

Instruction register
iconst13
least significant
32 bits
most significant
32 bits

Y register

??
?
@@ sdiv/udiv ;;
@
;
quotient

Integer
registers
(%r0%r31)

Figure 4.2 Integer division

Table 4.1 Assembly language formats for the integer multiplication and division operations
Operation
integer multiplication
integer division

Assembler syntax
mul-op rs1 ,rs2 , rd
mul-op rs, iconst13 , rd
div-op
rs1 , rs2 , rd
div-op
rs, iconst13 , rd

Operation implemented

f%y, reg[rd]g = reg[rs1]  reg[rs2]

f%y, reg[rd]g = reg[rs]  iconst13
reg[rd] = f%y, reg[rs1 ]g = reg[rs2 ]
reg[rd] = f%y, reg[rs]g = iconst13

Notes:

 iconst13 denotes an integer constant. This constant is signed when it used with a signed
operation (e.g., smul) and unsigned when it is used with an unsigned operation (e.g., umul).
The value must be represented in 13 bits.
 fx, yg denotes a 64-bit value (or storage location) constructed from two 32-bit values x and
y. The first of these values, x, is the most significant.

Lab 4. Multiplication and Division

29

Table 4.2 The signed and unsigned integer multiplication and division operations
Operation
signed integer multiplication
unsigned integer multiplication
signed integer division
unsigned integer division

Operation names
smul
smulcc
umul umulcc
sdiv
sdivcc
udiv
udivcc

a = (a  b)=c

Example 4.1 Write a SPARC program to evaluate the statement

. In writing this code you
should assume that , , and are signed integer values and that all results can be represented in 32 bits.

ab

a:
b:
c:

start:

end:

.data
.word
.word
.word

0x42
0x43
0x44

.text
set
ld
set
ld
set
ld

a, %r1
[%r1], %r2
b, %r1
[%r1], %r3
c, %r1
[%r1], %r4

smul
sdiv

%r2, %r3, %r2

%r2, %r4, %r2

! a* b --> %y, %r2

! %y, %r2 / c --> %r2

set
st

a, %r1
%r2, [%r1]

! %r2 --> a

ta

The signed and unsigned operations are distinguished by the way they interpret their
operands. The signed operations interpret their source operands as signed integers and
produce signed integer results. The unsigned operations interpret their source operands
as unsigned integers and produce unsigned integer results.

4.3.2 Updating the condition code bits

The operations that have names ending in cc update the bits in the condition code register. The integer multiplication operations (smulcc and umulcc) always clear the V (overflow) and C (carry) bits of the condition code register. In addition, these operations update
the N (negative) and Z (zero) bits of the condition code register. Although the multiplication operations produce a 64-bit result, updates to the N and Z flags are only based on the
least significant 32 bits of the result. Like the multiplication operations, the division operations (sdivcc and udivcc) also clear the C bit in the condition code register. In addition, the
division operations update the N, Z, and V bits in the condition code register based on the
value of the 32-bit result.

30

Lab 4. Multiplication and Division

4.3.3 Examining and setting the Y register

Note that the code presented in Example 4.1 does not need to examine or set the value in
the Y register. The multiplication sets the Y register and the division uses the value set by
the multiplication. In many other cases you will need to examine or set the contents of
the Y register. In particular, you may need to examine the contents of the Y register after
a multiplication or set the contents of the Y register before a division. You can use the
(synthetic) mov operation introduced in Lab 2 to copy the contents of the Y register to an
integer register or vice versa. You can also use the mov operation to set the contents of the
Y register. Table 4.3 summarizes the mov operation as it applies to the Y register. The first
format copies the contents of the Y register to an integer register. The second format copies
the contents of an integer register to the Y register. The third instruction format stores a
small integer constant into the Y register.
Table 4.3 The mov operation applied to the Y register.
Operation
register copy
register set

Assembler syntax
mov %y, rd
mov rs, %y
mov siconst13 , %y

Operation implemented
reg[rd] = reg[%y]
reg[%y] = reg[rs]
reg[%y] = iconst13

When you store a value into the Y register (using a mov instruction), it takes three
instruction cycles before the Y register is actually updated. This means that you need to
make sure there are at least three instructions between an instruction that uses the Y register
as a destination and an instruction that uses the value stored in the Y register.
Writing to the Y register Remember, always make sure that there are at least three instructions between any instruction that writes to the %y register and an instruction that
uses the value in the %y register.

a = (a+ b)=c

Example 4.2 Write a SPARC assembly language fragment to evaluate the statement
. Again,
you should assume that , , and are signed integers and that all results can be represented in 32 bits.

ab

a:
b:
c:

.data
.word
.word
.word

0x42
0x43
0x44

start:

.text
mov

%r0, %y

! clear the Y register -- THERE MUST BE

AT
! LEAST 3 INSTRUCTIONS BETWEEN THE MOV
AND
! SDIV INSTRUCTIONS
set
ld
set
ld
set
ld

a, %r1
[%r1], %r2
b, %r1
[%r1], %r3
c, %r1
[%r1], %r4

Lab 4. Multiplication and Division

end:

31

add
sdiv

%r2, %r3, %r2

%r2, %r4, %r2

! a + b --> %r2
! %r2 / c --> %r2

set
st
ta

a, %r1
%r2, [%r1]
0

! %r2 --> a

4.3.4 The multiply step operation

Prior to Version 8 the SPARC did not have integer multiplication or division operations.
These operations had to be performed using more primitive operations. To simplify integer
multiplication, earlier versions of the SPARC provided a multiply step operation, mulscc.
We will consider this operation in Lab 13 when take a closer look at integer arithmetic on
the SPARC.

4.4 Summary
In this lab we have covered the integer multiplication and division operations provided
by the SPARC. As with the other arithmetic operation (add and sub), there are versions of
the multiplication and division operations that update the condition code bits and other
multiplication and division operations that do not alter the condition code bits.
All of the multiplication and division operations operations use a special purpose register, the Y register (%y). You can use the mov operation to examine and set the contents of
the Y register.

4.5 Review Questions

1. How is the Y register used in the integer multiplication operations?
2. How is the Y register used in the integer division operations?

4.6 Exercises

32

Lab 4. Multiplication and Division

Laboratory 5
Bit Manipulation and Character
I/O
5.1 Goal
To cover the bit manipulation operations provided by the SPARC and the character I/O
traps provided by ISEM.

5.2 Objectives
After completing this lab, you will be able to write assembly language programs that use:





The bitwise operations,

The shift operations, and
The character I/O traps provided by ISEM.

5.3 Discussion
In this lab we introduce the bit manipulation operations of the SPARC. In particular, we
consider the logical operations and, or, and xor. In addition, we consider the synthetic
operation not. We follow this with a discussion of the shift operations sll, srl, and sra. The
lab ends with a short discussion of the character I/O facilities provided by ISEM.

5.3.1 Bitwise operations

Table 5.1 summarizes the bitwise operations of the SPARC. Like the other data manipulation operations (e.g., add, sub, smul, sdiv, etc.), there are two instruction formats for each
of the bitwise operations. Both formats use three explicit operandstwo source operands
and a destination operand. In the first format, both of the source operands are in registers. In the second format, one of the source operands is in a register, the other is a small
constant value. This constant may be positive or negative; however its 2s complement
representation must fit in 13 bitsthe SPARC does sign extend this value. Examples 5.1
and 5.2 illustrate uses of the bitwise operations.
Example 5.1
a:
b:
c:
d:

Write a SPARC program to evaluate the statement:

.data
.word
.word
.word
.word

0x42
0x43
0x44
0x45

33

a = (a&b) ^ (cjd).

34

Lab 5. Bit Manipulation and Character I/O

Table 5.1 Bitwise operations

Operation
and
or
exclusive or
and not
or not
exclusive nor

start:

end:
Example 5.2
bit).

Assembler syntax
and
rs1 , rs2 , rd
and
rs1 , siconst13 , rd
or
rs1 , rs2 , rd
or
rs1 , siconst13 , rd
xor
rs1 , rs2 , rd
xor
rs1 , siconst13 , rd
andn rs1 , rs2 , rd
andn rs1 , siconst13 , rd
orn
rs1 , rs2 , rd
orn
rs1 , siconst13 , rd
xnor
rs1 , rs2 , rd
xnor
rs1 , siconst13 , rd

Operation implemented
reg[rd] = reg[rs1 ] & reg[rs2 ]
reg[rd] = reg[rs1 ] & siconst13
reg[rd] = reg[rs1 ] reg[rs2 ]
reg[rd] = reg[rs1 ] siconst13
reg[rd] = reg[rs1 ] reg[rs2 ]
reg[rd] = reg[rs1 ] siconst13
reg[rd] = reg[rs1 ] & reg[rs2]
reg[rd] = reg[rs1 ] & siconst13
reg[rd] = reg[rs1 ]
reg[rs2 ]
reg[rd] = reg[rs1 ]
siconst13
reg[rd] = reg[rs1 ]
reg[rs2]
reg[rd] = reg[rs1 ]
siconst13

j
j
^
^



j
j
^
^

.text
set
ld
set
ld
set
ld
set
ld

a, %r1
[%r1],
b, %r1
[%r1],
c, %r1
[%r1],
d, %r1
[%r1],

and
or
xor

%r2, %r3, %r2

%r4, %r5, %r4,
%r2, %r4, %r2

! a & b --> %r2

! c d --> %r4
! %r2
%r4 --> r2

set
st
ta

a, %r1
%r2, [%r1]
0

! %r2 --> a

%r2

! a --> %r2

%r3

! b --> %r3

%r4

! c --> %r4

%r5

! d --> %r5

Write a SPARC program to clear bits 5 through 12 of the word

n:

start:

n (bit 0 is the least significant

.data
.word

0xaaaaaaaa

.text
set
ld

n, %r1
[%r1], %r2

! n --> %r2

andn

%r2, 0x1fe0, %r2

! %r2 &

0x1fe0

--> %r2

st
%r2, [%r1]
! %r2 --> n
end:
ta
0
In this case, the mask (0x1fe0) can be represented using 12 bits and, as such, we can use an andn
instruction with an immediate value.
Activity 5.1

Write a SPARC program to clear bits 5 through 13 of the word .

Lab 5. Bit Manipulation and Character I/O

35

5.3.2 Condition codes

Like the other data manipulation operations, there are separate bitwise operations that
update the condition code register after they perform the specified operation. Table 5.2
summarizes the collection of bitwise operations that set the condition codes. The names for
these operations have a suffix of cc to indicate that they update the bits in the condition
code register.
Table 5.2 Bitwise operationscondition code versions
Operation
and
or
exclusive or
and not
or not
exclusive nor

Operation name
andcc
orcc
xorcc
andncc
orncc
xnorcc

In all cases, the V and C flags are cleared.

If the result is zero, the Z flag is set, otherwise Z is cleared.
The N flag is set to the most significant bit in the result.

5.3.3 Synthetic operation not

Bitwise inversion is provided by a synthetic operation. Table 5.3 summarizes this operation. Bitwise inversion does not affect the bits in the condition code register.
Table 5.3 Bitwise inversion
Operation
not

Assembler syntax
not rs, rd
not rd

Operation implemented
reg[rd] = reg[rs]
reg[rd] = reg[rd]




5.3.4 Shift operations

Table 5.4 summarizes the shift operations of the SPARC. The SPARC provides two instruction formats for each of the shift operations. The shift operations provide a mechanism for
moving every bit in a value to the left or right by a specified number of positions. The shift
operations do not affect the bits in the condition code register.
Table 5.4 The shift operations
Operation
left shift logical
right shift logical
right shift arithmetic

Assembler syntax
sll
rs1 , rs2 , rd
sll
rs1 , siconst13 , rd
srl
rs1 , rs2 , rd
srl
rs1 , siconst13 , rd
sra rs1 , rs2 , rd
sra rs1 , siconst13 , rd

Operation implemented
reg[rd] = reg[rs1 ]
reg[rs2 ]
reg[rd] = reg[rs1 ]
siconst13
reg[rd] = reg[rs1 ]
reg[rs2 ]
reg[rd] = reg[rs1 ]
siconst13
reg[rd] = reg[rs1 ]31 reg[rs1 ]
reg[rd] = reg[rs1 ]31 reg[rs1 ]

<<
<<
>>
>>
j
j

>> reg[rs2]
>> siconst13

36

Lab 5. Bit Manipulation and Character I/O

As shown in Table 5.4, the shift operations have two source operands. The first operand
specifies the value to be shifted. This value must be stored in an integer register. The second source operand specifies the amount of the shift. This operand may be stored in an
integer register or it may be a small constant value. The processor only uses the least significant 5 bits of the second source operand (and ignores the remaining bits of this operand).
Example 5.3 illustrates the shift operations.
Example 5.3 Write a SPARC program to count the number of bits that are set (i.e., 1) in the memory location
(a variable). The result should be stored in %r2. In writing this code, you may use any of the remaining
registers as temporaries.

n:

start:

loop:

cont:

.data
.word 0xaaaaaaaa
.text
set
ld
clr
andcc
be
nop
inc
srl

n, %r1
[%r1], %r3
%r2
%r3, 1, %r0
cont
%r2
%r3, 1, %r3

! clear the result

!
!
!
!
!

check the lowest bit

skip if zero
(delay slot)
increment the 1s count
shift data right, creating a new low

bit

end:

cmp
bne
nop

%r3, 0
loop

ta

! if %r3 == 0, were done

5.3.5 Character data and character I/O in ISEM

When you use isem-as, character values are delimited using single quotes. For example,
B denotes the value associated with the letter B. When it encounters a character value,
isem-as converts the character value into its ASCII representation. As such, writing B in
an assembly language program is equivalent to writing 0x42 (or 66). You should use the
notation that best expresses the intent of your code.
ISEM provides the user with a primitive form of character input/output using the trap
mechanism. We will discuss the trap mechanism in detail in Lab 16. For now, we note
that the following instructions can be used to get and put characters from the standard I/O
devices.
ta
ta

1
2

! %r8 --> putchar

! getchar --> %r8

The first of these instructions prints the character in the least significant byte of register
%r8 (= %o0) to standard output and the second reads a character from standard input and
places the result in the least significant byte of %r8, clearing the most significant 24 bits of
this register. Example 5.4 illustrates the use of these I/O instructions.
Example 5.4 Write a SPARC/ISEM program that reads a one digit number, adds five to the number and
prints the result.
start:

.text
ta

! digit --> %r8

Lab 5. Bit Manipulation and Character I/O

37

sub
add
cmp
ble
nop

%r8, 0, %r8
%r8, 5, %r8
%r8, 9
one_dg

! convert from digit to number

! add 5
! is the result greater than 9?

mov
set
ta
mov
sub

%r8,
1,
1
%r7,
%r8,

!
!
!
!

add
ta

%r8, 0, %r8
1

ta

%r7
%r8
%r8
10, %r8

one_dg:

copy %r8 into %r7

the most significant digit must be 1
1 --> putchar
restore %r8

! %r8 holds a number less than 10

! convert number to digit
! print the least significant digit

5.4 Summary
In this lab we have introduced the bitwise and shift operations provided by the SPARC. In
addition, we have introduced the notation used for character data and primitive mechanisms for character input and output. Example 5.5 illustrates all of these operations.
Example 5.5 Write a SPARC/ISEM program to print the binary representation of the unsigned integer in
memory location . In writing the code, you may use any of the registers as temporaries.

n:

start:

loop:

print:

end:

.data
.word

0xaaaaaaaa

.text
set
ld
set

n, %r1
[%r1], %r2
1
31, %r3

set

32, %r4

andcc
be
set

%r2, %r3, %g0

print
0, %r8

! check for one

!
branch on zero
! (delay slot)

set

1, %r8

! must have been a one

ta

! print %r8

deccc
bg
srl

%r4
loop
%r3, 1, %r3

! decrement count
! continue until count == 0
! shift mask right (bd)

ta

<<

! initialize the mask to start

!
with the most significant bit
! number of bits to print

38

Lab 5. Bit Manipulation and Character I/O

5.5 Exercises
1. Write a SPARC program that compares the memory contents of the word pointed to
by register r to the contents of register r on a bit-by-bit basis. For all bits i, if
the value of bit i of r is smaller than the value of bit i of r , set bit i of r to 1;
otherwise, set bit i of r to 0.
2. Write a SPARC program that distinguishes ASCII-coded hexadecimal digits from
other bytes and which converts valid digit codes to the corresponding hexadecimal
value. Only valid digit codes are to be converted. Leave invalid characters unaltered. Suppose that register r holds the character on program entry, and that r
holds the result on program exit. Use register r to report on validity. If valid, r
must equal 0; if invalid, r must equal 1.

%2

%3

[% 2]
%3

%2

%3

%2
%3

%3

%3

%3

Valid Digits, Their Codes and Values

Digit
0
..
.

Digit
Code
0x30
..
.

Hex
Value
0x0
..
.

9
a,A
..
.

0x39
0x61,0x41
..
.

0x9
0xa
..
.

f,F

0x66,0x46

0xf

3. Write a SPARC program to print the hexadecimal representation of an unsigned integer. The unsigned integer should be named n and declared in the data segment. Your
program should finish by printing a newline.

Laboratory 6
Assembler Directives, Assembler
Expressions, and Addressing
Modes
6.1 Goal
To cover several assembler directives and assembler expressions provided by the GNU
assembler (gas) and the SPARC addressing modes.

6.2 Objectives
After completing this lab, you will be able to write assembly language programs that use:






The assembler directives provided by the GNU assembler,

Assemble expressions,
The SPARC addressing modes in load and store instructions, and
And use the SETHI instruction.

6.3 Discussion
In this lab we introduce three new assembler directives: a directive to allocate space, a
directive to define symbolic constants, and a directive to include header files. After we
describing these directives, we discuss assembler expressions and introduce the distinction between relocatable values and absolute values. We conclude this lab by discussing the
memory addressing modes provided by the SPARC and the SETHI instruction.

6.3.1 Assembler directives

In Lab 2 we introduced three assembler directives: .data, .text, and .word. In this Lab, we
introduce three more directives: .skip, .set, and .include. The .skip directive is used to allocate
space. The .set directive is used to define a symbolic constant. The .include directive is used
to include source (header) files.
You can use the .skip directive to allocate space in the current assembler segment (data
or text). This directive takes one or two arguments. The first argument specifies the number
of bytes to skip in the current assembler segment. The second argument specifies the value
to be deposited in the skipped bytes. If the second argument is omitted, it is assumed to be
zero.
You can define symbolic constants using the .set directive. This directive takes two arguments. The first argument is the name of the symbol to be defined. The second argument
39

40

Lab 6. Assembler Directives, Assembler Expressions, and Addressing Modes

is an expression that defines the value of the symbol. This directive can be written using
standard directive syntax (e.g., .set symbol, expression), or it can be written using the infix
= operator (e.g., symbol = expression).
In many cases, you will want to collect a group of definitions for symbolic constants into
a header file that can be included in several different programs or modules. (By including
the same file in each of the programs or modules, you can be sure that all of the programs
and modules use the same values for the symbolic constants.) The .include directive supports this style of programming. This directive takes a single argument, a string that gives
the name of the file to include. The code from the included file logically replaces the .include
directive. When the assembler is finished processing the included file, it resumes after the
.include directive in the original file.
Table 6.1 summarizes the directives that we have introduced in this section.
Table 6.1 Assembler directives
Operation
Allocate space
Symbolic constant
Include file

Assembler syntax
.skip
n
.set
symbol, expression
symbol = expression
.include filename

6.3.2 Assembler Expressions

The .set and .skip directives use assembler expressions. In addition, you can use assembler
expressions whenever you use a constant value in an assembly language instruction. Table 6.2 summarizes the operators that you can use constructing expressions. The operands
can be expressions (using parentheses to override precedence), symbols (defined as labels
or using the .set directive), or numbers.
Table 6.2 Assembler expressions
Operator

;
+

=
<<
>>
j
&

+
;

Operation
Unary minus
Unary plus
Multiplication
Division
Left shift
Right shift
Bitwise inclusive or
Bitwise and
Bitwise exclusive or
Bitwise and not
Addition
Binary subtraction

Precedence
highest
high middle

low middle

lowest

When considering assembler expressions, it is useful to distinguish between relocatable

values and absolute values. Labels are the simplest examples of relocatable values. They
are relocatable because their final values depend on where your program is loaded into
memory. Numbers are the simplest examples of absolute values. Absolute values do not
depend on where your program is loaded into memory.

Lab 6. Assembler Directives, Assembler Expressions, and Addressing Modes

41

You can use absolute values with any of the operators. If all of the operands for an
operator are absolute values, the expression using the operator is an absolute value.
You can only use relocatable values in expressions using the binary addition and subtraction operators. When you use binary addition, at most one operand can be a relocatable
value, the other operand must be an absolute value. If one operand is relocatable, the value
of the expression is relocatable.
When you use binary subtraction, you cannot subtract a relocatable value from an absolute value. When subtract an absolute value from a relocatable value, the result is a relocatable value. When you subtract two relocatable values, the two values must be defined
in the same assembler segment (e.g., text or data), and the result is an absolute value.

6.3.3 Addressing modes

The SPARC supports two addressing modes: register indirect with index and register indirect with displacement. In the first mode, the effective address is calculated by adding
the contents two integer registers. This addressing mode is commonly used to access an
array element: one of the registers holds the base address of the array, the other holds the
(scaled) index of the element.
In the second mode, the effective address is calculated by adding a 13-bit signed integer
constant to a register. This addressing mode can be used with pointers to structures: the
register holds the address of the structure and the integer constant specifies the offset of
the member (field) being accessed. Register indirect with displacement addressing is also
commonly used when accessing items on the runtime stack (e.g., parameters and local
variables). We consider access to the runtime stack in Lab 12 when we consider standard
procedure calling conventions for the SPARC.
Table 6.3 summarizes the addressing modes supported by SPARC assemblers. In addition to the two basic addressing modes, SPARC assemblers recognize register indirect
addressing and a limited form of direct memory addressing. Direct memory addresses
are limited to values that can be expressed in 13-bits when sign-extended (i.e., very small
addresses and very large addresses).
Table 6.3 Assembler address specifications
Addressing mode
register indirect with index
register indirect with displacement

register indirect
direct memory

Assembler syntax
[r1 +r2 ]
[r1 +siconst13 ]
[siconst13 +r1 ]
[r1 siconst13 ]
[r]
[siconst13 ]

Implementation
basic mode
basic mode

[r+%r0]
[%r0+siconst13 ]

Effective address
reg[r1 ]+reg[r2 ]
reg[r1 ]+siconst13
siconst13 +reg[r1 ]
reg[r1 ] siconst13
reg[r]
siconst13

Addressing modes can only be used with the load and store instructions. Table 6.4
summarizes the load word and store word operations.

6.3.4 The SETHI instruction

We conclude this discussion by introducing another SPARC instruction, sethi, and the %hi
and %lo operators provided by SPARC assemblers. The sethi instruction takes two arguments: a 22-bit constant and a destination register. This instruction sets the most significant
22 bits of the destination register and clears the least significant 10 bits of this register.

42

Lab 6. Assembler Directives, Assembler Expressions, and Addressing Modes

Table 6.4 The SPARC ld and st operations

Operation
load word
store word

Assembler syntax
ld address, rd
st rs, address

Notes:
address
eff addr(x)

Example 6.1

Operation implemented
reg[rd] = memory[eff addr(address)]
memory[eff addr(address)] = reg[rs]

an address specification (see Table 6.3)

the result of the effective address calculation

Show how you would load a 32-bit, bignum, into %r2 without using the set instruction.
.set
sethi
or
ta

bignum, 0x87654321
bignum>>10, %r2
%r2, bignum&0x3ff, %r2
0

Note the use of the expression bignum>>10 to extract the most significant 22 bits
of bignum and the expression bignum&0x3ff to extract the least significant 10 bits of
bignum. To make your code more readable, SPARC assemblers provide two special operators: %hi(x) yields the most significant 22 bits of x, while %lo(x) yields the least significant
10 bits of x. Note that these operators are written using function call notation.
Example 6.2

Rewrite the code fragment given in Example 6.1, using the %hi and %lo operators.
.set
sethi
or
ta

bignum, 0x87654321
%hi(bignum), %r2
%r2, %lo(bignum), %r2
0

Example 6.3 Write an assembly language fragment to sum up the elements in the array. Give directives to
declare an array of 20 words and an additional word to hold the sum.
arr:
sum:

start:

loop:

end:

.data
.skip
.word

20*4
0

! allocate an array of 20 words

! allocate a word to hold the sum

.text
set
mov
mov
set

arr, %r2
%r0, %r3
%r0, %r4
20, %r5

!
!
!
!

%r2
%r3
%r4
%r5

fetch the next element

add it to the running sum
one fewer element
if %r5 > 0 get next element
increment the index (DELAY SLOT)

is
is
is
is

the
the
the
the

base address
index value
running sum
number of elems to add

ld
add
subcc
bne
add

[%r2+%r3], %r6
%r4, %r6, %r4
%r5, 1, %r5
loop
%r3, 4, %r3

!
!
!
!
!

sethi
st
ta

%hi(sum), %r1
%r4, [%r1+%lo(sum)]
0

! store the result in sum

Lab 6. Assembler Directives, Assembler Expressions, and Addressing Modes

43

Note that the code in Example 6.3 stores the result into sum using a sethi instruction
followed by a st instruction. In previous examples we have used a set instruction followed
by a st instruction to accomplish the same task. However, the set instruction is actually a
synthetic instruction and the assembler implements this instruction using a sethi instruction followed by an or instruction (as we showed in Example 6.1). As such, our earlier
code actually requires three instruction for every (load or) store. Using the sethi instruction
directly, we can avoid an unnecessary instruction.

6.4 Summary
6.5 Review Questions
6.6 Exercises
1. The %hi operator yields the most significant 22 bits while %lo operator yields the
least significant 10 bits of a 32-bit value. Considering that the SPARC uses 13-bit
signed integers in lots of contexts, it might seem that it would be better to have the
%hi and %lo operators yield 19 and 13 bits respectively. What problems would this
cause?
2. Write a SPARC assembly language program to count the number of ones in a bit
string. The bit string should be named BitString and the length (in bits) of the bit
string should be named Length. The result should be stored as a word named Count.
Note: there is no limit on the number of bits in the bit string.
3. Given an array, A, and the number of elements in the array, n, write a SPARC program
to sort the array. You may use any method you like to sort the array.

44

Lab 6. Assembler Directives, Assembler Expressions, and Addressing Modes

Laboratory 7
Operand Sizes and Unsigned
Values
7.1 Goal
To complete our coverage of the load and store instructions provided by the SPARC and to
cover a collection of useful synthetic operations.

7.2 Objectives
After completing this lab, you will be able to write assembly language programs that use:






Byte, halfword, word, and double word operands,

The assembler directives .byte, .hword, .quad, and .align,
Signed and unsigned operands, and
The (synthetic) operations clear, negate, increment, and decrement.

7.3 Discussion
In this lab we introduce the assembler directives and operations associated with different sized operands and unsigned operands. We begin by considering the operand sizes
supported by the SPARC. Then we consider assembler directives used to allocate different
amounts of memory. Next, we consider the load and store operations for different sized operations. The we consider the load operations for unsigned operands. Finally, we conclude
by considering a small collection of useful synthetic operations.

7.3.1 Operand sizes

On the SPARC, all data manipulation operations (e.g., integer addition and integer subtraction) manipulate 32-bit values. However, the data transfer operations (e.g., load and store)
can transfer different sized values between the memory and the integer registers. Table 7.1
summarizes the operand sizes supported by the SPARC load and store operations.
Assembler directives
The .word directive introduced in Lab 2 allocates and initializes memory in 32-bit (word)
units. Table 7.2 summarizes the assembler directives for allocating and initializing different
sized units of memory. The name of the directive used to allocate and initialize a 64 bit
doubleword, .quad, is historical. The name dates to a time when words were 16 bits.
Quad is short for quad-word, that is, four 16-bit words.
45

46

Lab 7. Operand Sizes and Unsigned Values

Table 7.1 Operand sizes for the data transfer operations

Name
byte
halfword
word
double word

Size
8 bits
16 bits
32 bits
64 bits

Table 7.2 Directives for allocating and initializing memory

Directive
.byte
.hword
.word
.quad

Size
8 bits
16 bits
32 bits
64 bits

The SPARC load and store operations require that halfword values be aligned on even
addresses (i.e., halfword alignment) and that word and double word values be aligned on
addresses that are a multiple of four (i.e., word aligned). You can use the .align directive to
make sure that your variables are aligned as needed. This directive takes a two arguments,
a number and an optional pad value. When the assembler encounters an align directive,
it makes sure that the next address in the current assembler segment is a multiple of the
first argument. For example, the directive .align 8 will ensure that the next address is
a multiple of 8. To ensure that the next address meets the alignment requirements, the
assembler emits pad bytes. If the second argument is supplied, the assembler uses this
value when it emits pad bytes; otherwise, the assembler emits zeros. Example 7.1 illustrates
the size and alignment directives.
Example 7.1 Consider the following C declarations. Assuming that a character is one byte, a short integer is
two bytes, and an integer is four bytes, give assembler directives to allocate and initialize the memory specified
by these directives.
short int short1 = 22;
char ch1 = a;
short int short2 = 33;
char ch2 = A;
int int1 = 0;
.data
.align 2
short1: .hword 22
ch1:
.byte a
.align 2
short2: .hword 33
ch2:
.byte A
.align 4
int1:
.word 0

!
!
!
!
!
!
!
!

halfword align
allocate and initialize
allocate and initialize
halfword align
allocate and initialize
allocate and initialize
word align
allocate and initialize

a halfword
a byte
a second halfword
a second byte
a word

The load and store operations

The size of the operand for a load or store operation is specified using a suffix: d for doubleword, h for halfword, or b byte. The operation names are summarized in Table 7.3.

Lab 7. Operand Sizes and Unsigned Values

47

Table 7.3 The load and store operations

Size
byte
halfword
word
doubleword

Load
ldb
ldh
ld
ldd

Store
stb
sth
st
std

Notes
the address must be halfword aligned
the address must be word aligned
the address must be word aligned and the
register must be an even number

The load operations take two operands: the (source) memory address followed by the
(destination) register. Similarly, the store operations take two operands: the (source) register followed by the (destination) memory address.
The load byte, halfword, and word operations set all 32 bits of the destination register.
The load byte operation (ldb) fetches an 8-bit value, sign extends this value to 32 bits,
and loads the resulting value into destination register. The load halfword operation (ldh)
fetches a 16-bit value and sign extends this value to 32 bits. The load word operation (load)
fetches a 32-bit value and load this value into the destination register. The load doubleword
operation fetches two 32-bit values and loads them into consecutive registersstarting
with the register specified in the instruction (e.g., %r2 and %r3).
The store byte operation (stb) stores the least significant 8 bits of the source register
to the destination memory location. The store halfword operation (sth) stores the least
significant 16 bits of the source register into the destination memory location. The store
word operation (st) stores the contents of a register to the destination memory address. The
store doubleword operation (stored) stores the contents two consecutive registers (starting
with the register specified in the instruction).
The memory addresses used with the operations that load and store halfwords must
be even (i.e., halfword aligned). The memory addresses used with the operations that
load and store words and doublewords must be multiples of four (i.e., word aligned). The
register used with the operations that load and store doublewords must be even (e.g., %r2
but not %r3). When these operations are used, the most significant 32 bits are stored in the
even register.
Table 7.4 The load and store instructions
Operation
load byte
load halfword
load word
load doubleword

Instruction syntax
ldb address, rd
ldh address, rd
ld
address, rd
ldd address, rd

store byte
store halfword
store word
store doubleword

stb
sth
st
std

rs, address
rs, address
rs, address
rs, address

Operation implemented
reg[rd] = signextend( memory[address]8 )
reg[rd] = signextend( memory[address]16 )
reg[rd] = memory[address]32
reg[rd] = memory[address]32
reg[rd+1] = memory[address+4]32
memory[address]8 = reg[rs]8
memory[address]16 = reg[rs]16
memory[address]32 = reg[rs]32
memory[address]32 = reg[rs]32
memory[address+4]32 = reg[rs+1]32

Example 7.2 Rewrite the code presented in Example 6.3 using an array of 20 bytes (i.e., chars) instead of
words. (You should still store the sum in a word.)
.data

48

Lab 7. Operand Sizes and Unsigned Values

arr:
sum:

start:

loop:

end:

.skip
.word

20
0

! allocate an array of 20 bytes

! allocate a word to hold the sum

.text
set
mov
mov
set

arr, %r2
%r0, %r3
%r0, %r4
20, %r5

!
!
!
!

%r2
%r3
%r4
%r5

fetch the next element

add it to the running sum
one fewer element
if %r5 > 0 get next element
increment the index (DELAY SLOT)

is
is
is
is

the
the
the
the

base address
index value
running sum
number of elems to add

ldb
add
subcc
bne
add

[%r2+%r3], %r6
%r4, %r6, %r4
%r5, 1, %r5
loop
%r3, 4, %r3

!
!
!
!
!

sethi
st
ta

%hi(sum), %r1
%r4, [%r1+%lo(sum)]
0

! store the result in sum

7.3.2 Unsigned operands

The SPARC instruction set also provides a load unsigned byte operation (loadub) and a
load unsigned halfword operation (loaduh). In contrast to the standard load operations
(loadb and loadh), these operations do not sign extend their values. They always set the
most significant bits to zero.
Table 7.5 The unsigned load operations
Operation
load unsigned byte
load unsigned halfword

Instruction syntax
ldub address, rd
lduh address, rd

Operation implemented
reg[rd] = zerofill( memory[address]8 )
reg[rd] = zerofill( memory[address]16 )

7.3.3 Useful (synthetic) operations

We conclude this lab by considering a collection of useful operations. In particular, we consider the (synthetic) operations clear, negate, increment, and decrement. These operations
are summarized in Table 7.6.
Example 7.3

Rewrite the code presented in Example 6.3 using the operations defined in Table 7.6.

arr:
sum:

start:

loop:

.data
.skip
.word

20*4
0

! allocate an array of 20 words

! allocate a word to hold the sum

.text
set
clr
clr
set

arr, %r2
%r3
%r4
20, %r5

!
!
!
!

ld
add

[%r2+%r3], %r6
%r4, %r6, %r4

%r2
%r3
%r4
%r5

is
is
is
is

the
the
the
the

base address
index value
running sum
number of elems to add

! fetch the next element

! add it to the running sum

Lab 7. Operand Sizes and Unsigned Values

49

Table 7.6 Some useful operations

Operation
clear register
clear memory word
clear memory halfword
clear memory byte
negate register
increment register
increment register, set cc
decrement register
decrement register, set cc

end:

Instruction syntax
clr
rd
clr
address
clrh
address
clrb
address
neg
rd
neg
rs,rd
inc
rd
inc
siconst13 , rd
inccc
rd
inccc
siconst13 , rd
dec
rd
dec
siconst13 , rd
deccc rd
deccc siconst13 , rd

Operation implemented
reg[rd = 0
memory[address]32 = 0
memory[address]16 = 0
memory[address]8 = 0
reg[rd] = reg[rd]
reg[rd] = reg[rs]
reg[rd] = reg[rd] + 1
reg[rd] = reg[rd] + siconst13
reg[rd] = reg[rd] + 1
reg[rd] = reg[rd] + siconst13
reg[rd] = reg[rd] 1
reg[rd] = reg[rd] siconst13
reg[rd] = reg[rd] 1
reg[rd] = reg[rd] siconst13

;
;

;
;
;
;

deccc
bne
inc

%r5
loop
4, %r3

! one fewer element

! if %r5 > 0 get next element
! increment the index (DELAY SLOT)

sethi
st
ta

%hi(sum), %r1
%r4, [%r1+%lo(sum)]
0

! store the result in sum

7.4 Summary
7.5 Review Questions
1. Explain why the SPARC does not provide unsigned store operations.

7.6 Exercises
1. Suppose that the SPARC did not have a load unsigned byte operation, explain how
you could implement this operation using the load byte operation (recall, this operation always sign extends the value being loaded). Note, the Intel i860 processor
provides a load byte operation but does not provide a load unsigned byte operation.

50

Lab 7. Operand Sizes and Unsigned Values

Laboratory 8
The ISEM Graphics Accelerator
8.1 Goal
To cover uses of the graphics accelerator device provided by ISEM. (Currently, this device
is only available in the X11 environment.)

8.2 Objectives
After completing this lab, you will be able to write assembly language programs that use:

the graphics accelerator.

8.3 Discussion
In the previous labs, we have focused on assembly language programming, presenting
SPARC instructions and assembler directives. Now, its time for some fun! In this lab
we present the ISEM graphics accelerator device, gx. In addition to showing you how a
simple device works, this lab will give you an opportunity to review the assembly language
constructs covered in the previous labs.
When you open the gx device, it creates a black and white graphics window. By issuing gx commands, you can instruct the gx device to draw lines, fill rectangles, and copy
 pixels. Individual pixels in
rectangles in the window. The visible gx display is
the visible region are addressed by an x; y pair. The pixel in the upper left corner of the
display is addressed by the pair ; . The pixel in the lower right corner is addressed by
the pair
; .
In addition to the visible pixels, the gx device provides a 512x64 rectangle of pixels
that are not displayed. These pixels are commonly used with the blt operation (described
later in this lab). They are addressed using the pixel addresses ;
through
; .
Figure 8.1 illustrates the pixel addresses provided by the gx device.
The gx device is a memory mapped device. This means that the gx device registers are mapped into memory locations and can be accessed using the standard load and
store operations. (The SPARC architecture doesnt provide any special I/O instructions, so
all devices must be memory mapped when they are used with a SPARC.) The gx device
has 256 registers: a status register, a command register, and 254 argument registers. Each
register is one word (four bytes, 32 bits) wide.
The status register is mapped into memory location 0x100000. Storing a value into this
location has no affect; however, when you load a register using this memory location, you
will actually read the status register of the gx device. The value of the status register is 0
when the gx device hasnt been (opened and) displayed. This register has the value 1 when
the device has been opened and mapped onto the display. The gx command register is

(511 511)

(0 0)

( )

512 512

(0 512)

51

(511 575)

52

Lab 8. The ISEM Graphics Accelerator


;
;
(0,0)


512

512
Visible
(displayed)

;

;
(0,512)

Hidden
(not displayed)

?6
64
?

; ;

(511,511)

; ;

(511,575)

Figure 8.1 The gx display

mapped to memory location 0x100004 while the argument registers are mapped to memory
locations 0x1000080x1007fe. Figure 8.2 illustrates the mapping of gx registers into the
ISEM memory.
The gx device provides commands to draw lines, fill rectangles, and copy rectangles on
the display. Table 8.1 summarizes the commands provided by the gx device.
Before you issue any other gx commands, you must first open the device. When you
issue the open command, the gx device creates an X11 window and initializes the display
memory. The display is initially is solid white, and the drawing color is black.
After you are done using the gx device, you can explicitly close the device using the
close command. If you dont issue a close command, the display window will be destroyed
when you exit isem, as such, closing the gx device is not critical.
You can use the color command to set the color used to draw lines and fill rectangles.
This command takes a single argument, the color. If the argument is 0, the gx device draws
lines an fills rectangles in black; otherwise, if the argument is 1, the gx device draws lines
and fills rectangles in white.
The gx op command has a single argument. This command sets the drawing function
to the value of the argument. When the gx device is initialized, the drawing function is
copy. That is, the gx device simply copies the drawing color (when drawing a line or

Lab 8. The ISEM Graphics Accelerator

53

ISEM
memory

0x100000
0x100004
0x100008
0x10000c
0x100010

status
- command
- arg 1
- arg 2
- arg 3

gx
device

..
.

0x1007fe

arg 254

Figure 8.2 The gx register/memory map

Table 8.1 Commands provided by the graphics accelerator
Command
value
0
1
2

Command
name
open
close
color

Arguments

3
4
5

gx op
line
fill

n
x1 y1 x2 y2
xywh

blt

x1 y1 w h x2 y2

Operation performed
open the device
close the device
set the drawing color (n
for black, n
for white)
set the drawing function (see Table 8.2)
draw a line from pixel x1 ; y1 to x2 ; y2
fill the rectangle with upper left corner
x; y , width w, and height h
copy the rectangle with upper left corner
x1 ; y1 , width w, and height h to the rectangle with upper left corner x2 ; y2

=0

( )
( )

) (

filling a rectangle) or the source pixel (when copying a rectangle) to the destination pixel.
Table 8.2 summarizes the other drawing functions provided by the gx device.
The line function draws a line from one pixel to another based on the current drawing
color and function. This command takes four arguments: two arguments to specify the the
coordinates for each pixel.
The fill command fills a rectangle based on the current drawing color and function.
This command takes four arguments: the first two arguments specify the coordinates of the
upper left corner of the rectangle, the third argument specifies the width of the rectangle,
and the fourth argument specifies the height of the rectangle.
The blt command copies (subject to the drawing function) a rectangle from one part of
the gx memory to another. This command takes six arguments: the first four arguments
specify the source rectangle, the last two specify the upper left corner of the destination
rectangle.
To issue a gx command, you first store the arguments in the gx argument registers and
then store the command into the gx command register. The order in which you perform

54

Lab 8. The ISEM Graphics Accelerator

Table 8.2 Drawing operation supported by the graphics accelerator

Name
clear
and
andReverse
copy
andInverted
noop
xor
or
nor
equiv
invert
orReverse
copyInverted
orInverted
nand
set

Value
0x0
0x1
0x2
0x3
0x4
0x5
0x6
0x7
0x8
0x9
0xa
0xb
0xc
0xd
0xe
0xf

Drawing operation
0
source AND destination
source AND NOT destination
source
NOT source AND destination
destination
source XOR destination
source OR destination
NOT source AND NOT destination
NOT source XOR destination
NOT destination
source OR NOT destination
NOT source
NOT source OR destination
NOT source OR NOT destination
1

these stores is critical. You must load the command register after you have loaded the
argument registers. The gx device reads its argument registers as soon as you store a value
in the command register.
Symbolic constant that make devices, like the gx device, easier to use are commonly
defined in header files. Figure 8.3 presents a header file for the gx device. Example 8.1
illustrates the gx device and a use of the gx header file.
Example 8.1 Write an ISEM program that opens the gx device and draws a line from the upper left corner
to lower right corner.
.include "gx.h"

main:

wait:

.text
set

BX_BUFFER, %r1

! %r1 points to gx registers

st
ld
cmp
be
ld

%r0, [%r1+GX_CMD]
[%r1+GX_STATUS], %r2
%r2, 0
wait
[%r1], %r2

! open display
! load status word

! set up the
st
%r0,
st
%r0,
mov
511,
st
%r2,
st
%r2,

command arguments
[%r1+GX_LINE_X1]
[%r1+GX_LINE_Y1]
%r2
[%r1+GX_LINE_X2]
[%r1+GX_LINE_Y2]

! now, issue the command

mov
GX_LINE, %r2
st
%r2, [%r1+GX_CMD]

! wait until window is mapped

! load status word

! x1 = 0
! y1 = 0
! x2 = 511
! y2 = 511

Lab 8. The ISEM Graphics Accelerator

55

!
! gx.h -- symbolic constants for the gx device
!
.set GX_BUFFER,
0x100000 ! start address for GX registers
.set
.set
.set
.set
.set
.set
.set

GX_OPEN,
GX_CLOSE,
GX_COLOR,
GX_OP,
GX_LINE,
GX_FILL,
GX_BLIT,

.set GX_STATUS,
.set GX_CMD,
.set GX_ARG,

0
1
2
3
4
5
6

! command numbers

0
4
8

! symbolic buffer offsets

.set
.set
.set
.set

GX_FILL_X,
GX_FILL_Y,
GX_FILL_W,
GX_FILL_H,

8
12
16
20

for fill

.set
.set
.set
.set

GX_LINE_X1,
GX_LINE_Y1,
GX_LINE_X2,
GX_LINE_Y2,

8
12
16
20

for line

.set
.set
.set
.set
.set
.set

GX_BLIT_X1,
GX_BLIT_Y1,
GX_BLIT_W,
GX_BLIT_H,
GX_BLIT_X2,
GX_BLIT_Y2,

8
12
16
20
24
28

for blit

! drawing functions for the GX_OP command

.set GX_CLEAR,
0x0
! 0
.set GX_AND,
0x1
! source AND destination
.set GX_AND_REVERSE,
0x2
! source AND NOT destination
.set GX_COPY,
0x3
! source
.set GX_AND_INVERTED, 0x4
! NOT source AND destination
.set GX_NOOP,
0x5
! destination
.set GX_XOR,
0x6
! source XOR destination
.set GX_OR,
0x7
! source OR destination
.set GX_NOR,
0x8
! NOT source AND NOT destination
.set GX_EQUIV,
0x9
! NOT source XOR destination
.set GX_INVERT,
0xa
! NOT destination
.set GX_OR_REVERSE,
0xb
! source OR NOT destination
.set GX_COPY_INVERTED, 0xc
! NOT source
.set GX_OR_INVERTED,
0xd
! NOT source OR destination
.set GX_NAND,
0xe
! NOT source OR NOT destination

Figure 8.3 The gx header file. gx.h

56

Lab 8. The ISEM Graphics Accelerator

end:

! all done -- the display will remain until you exit isem
ta
0

8.4 Summary
8.5 Review Questions
8.6 Exercises
1. A bitmap is rectangle of black and white pixels. In X11, bitmaps are stored row-byrow in an array of bytes, i.e., consecutive memory locations. Each bit in this array
represents a pixel on the screen1 for black and 0 for white (the inverse of the color
convention used for the gx device). Each row of the bitmap is stored in an integral
number of bytes. If the number of columns is not a multiple of 8, the last byte is
padded with zeros. The first byte of the array represents the leftmost 8 pixels in the
top row. The next byte represents the next 8 pixels in the top row or the first 8 pixels
of the next row if there are fewer than 9 columns in the bitmap. Within a byte, the
least significant bit represents the leftmost pixel.
As an example, Figure 8.4 illustrates a simple bitmap and assembly language declarations for the X11 representation of this bitmap.

height:
width:
bits:

.word
.word
.byte
.byte

8
8
0x01, 0x02, 0x04, 0x08
0x10, 0x20, 0x40, 0x80

Figure 8.4 A simple bitmap

You can use the bitmap program under X11 to create bitmaps and save their representation in a file. When you save a bitmap, the bitmap program generates C declarations for the bitmap. You can easily convert these declarations into assembly
language declarations.
Write a program to draw a bitmap on the gx display. The gx device does not provide
a simple way to copy bytes from standard memory to the device memory. You will
need to draw the bitmap into the display memory, using GX LINE commands.
Your program should draw the bitmap with its upper left corner at position (0,0) of
the display. You can test your program with the bitmap shown in Figure 8.4, but
make sure you can also display bitmaps with widths that are not a multiple of 8.
Figure 8.5 gives you another bitmap to display.

Lab 8. The ISEM Graphics Accelerator

width:
height:
bits:

.word
.word
.byte
.byte
.byte
.byte
.byte
.byte

31
13
0x00,
0xbf,
0xc1,
0x18,
0x7e,
0x00,

0x00,
0xdf,
0x1f,
0x18,
0xbf,
0x00,

0x00,
0x1d,
0x18,
0xb0,
0xdf,
0x00,

57

0x00,
0x18,
0xbf,
0xc1,
0x18,
0x00,

0x7e,
0x83,
0xc7,
0x18,
0x7e,
0x00,

Figure 8.5 A simple bitmap

0xbf,
0xc1,
0x1a,
0x18,
0xbf,
0x00,

0xdf,
0x1f,
0x18,
0xb0,
0xdf,
0x00

0x18,
0x18,
0xbf,
0xc1,
0x18,

0x7e
0x83
0xc7
0x18
0x00

58

Lab 8. The ISEM Graphics Accelerator

Laboratory 9
The SPARC Instruction Formats
9.1 Goal
To cover the instruction encoding and decoding for the SPARC.

9.2 Objectives
After completing this lab, you will be able to:




Hand assemble SPARC assembly language instructions, and

Hand disassemble SPARC machine language instructions.

9.3 Discussion
In this lab we consider instruction encoding and decoding for the operations that we have
introduced in previous labs. In particular, we will consider encodings for instructions that
use the data manipulation and branching operations. After we introduce instruction encoding, we consider the translation of synthetic operations. Finally, we conclude this lab
by considering instruction decoding on the SPARC.
All SPARC instructions are encoded in a single 32-bit instruction word, there are no
extension words.

9.3.1 Encoding load and store instructions

The SPARC machine language uses two different formats for load and store instructions.
These formats are shown in Figure 9.1. The first format is used for instructions that use one
or two registers in the effective address. The second format is used for instructions that use
an integer constant in the effective address.
In the first format the 32-bit instruction is divided into seven fields. The first field (reading from the left) holds the 2-bit value 11, while the fifth field (bit 13) holds the 1-bit value 0.
These bits are the same for all load and store instructions that use two source registers. The
sixth field (bits 5 through 12) holds the address space indicator, asi. For the present, we will
always set the asi field to zero. The remaining fields, rd, op3 , rs1 , and rs2 , hold encodings
for the destination register, the operation, and the two source registers, respectively.
Registers are encoded using the 5-bit binary representation of the register number. Table 9.1 summarizes the operation encodings for the load and store operations.
Example 9.1
ldd

Hand assemble the instruction:

[%r4+%r7], %r11

59

60

Lab 9. The SPARC Instruction Formats

A. Instructions of the form: op
op

31 30 29

25 24

11

19 18

op3

rd

11

25 24

[rs1 +siconst13 ], rd
rd, [rs1 +siconst13 ]

19 18

op3

rd

(load instructions) or
(store instructions)
5

14 13 12

rs1

B. Instructions of the form: op

op
31 30 29

[rs1 +rs2 ], rd
rd, [rs1 +rs2 ]

rs2

(load instructions) or
(store instructions)
0

14 13 12

rs1

asi

siconst13

Figure 9.1 Instruction formats for load and store instructions

Table 9.1 Operation encodings for the load and store operations
Operation
ld
ldub
lduh
ldd
ldsb
ldsh

op3
000000
000001
000010
000011
001001
001010

Operation
st
stb
sth
std

op3
000100
000101
000110
000111

Because this instruction uses two registers in the address specification, it is encoded using the first
format shown in Figure 9.1. As such, we must determine the values for the rd, op3 , rs1 , and rs2 fields.
The following table summarizes these encodings:
Field Symbolic value Encoded value
rd
%r11
01011
op3
ldd
000011
rs1
%r4
00100
%r7
00111
rs2
These encodings lead to the following machine instruction:
31 30 29

25 24

19 18

14 13 12

5 4

01011
000011
00100
0
00000000
11
That is, 1101 0110 0001 1001 0000 0000 0000 0111 in binary, or 0xD6190007.

00111

If the assembly language instruction only uses a single register in the address specification (e.g., register indirect addressing), the register is encoded in one of the source register
fields (i.e., sr1 or sr2 ) while %r0 is encoded in the other. It doesnt matter which field holds
the register specified in the assembly language instruction and which field holds the encoding for %r0. However, isem-as encodes %r0 in sr2 .
Example 9.2
ldub

Hand assemble the instruction:

[%r23], %r19

Because this instruction uses registers in the address specification, it is encoded using the first format
shown in Figure 9.1. As such, we must determine the values for the rd, op3 , rs1 , and rs2 fields. The
following table summarizes these encodings:

Lab 9. The SPARC Instruction Formats

61

Field Symbolic value Encoded value

rd
%r19
10011
ldub
000001
op3
rs1
%r23
10111
rs2
%r0
00000
These encodings lead to the following machine instruction:
31 30 29

25 24

19 18

5 4

14 13 12

10011
000001
10111
0
00000000
00000
11
That is, 1110 0110 0000 1101 1100 0000 0000 0000 in binary, or 0xE60DC000.

In the second format the 32-bit instruction is divided into six fields. As in the previous
format, the first field holds the 2-bit value 11. However, unlike the previous format, the
fifth field holds the 1-bit value 1. The remaining fields, rd, op3 , rs1 , and siconst13 , hold
encodings for the destination register, the operation, the source register, and the constant
value, respectively. When this format is used, the integer constant is encoded using the
13-bit 2s complement representation and stored in the siconst13 field of the instruction.

9.3.2 Encoding sethi instructions

The format used to encode sethi instructions is shown in Figure 9.2. Sethi instructions are
encoded in four fields. The first field holds the 2-bit value 00. The next field, rd, holds the
5-bit encoding of the destination register. The third field holds the 3-bit value 100. The final
filed holds the 22-bit binary encoding of the value specified in the instruction.
31 30 29

00

25 24

22 21

100

rd

const22

Figure 9.2 Instruction format for sethi instructions

Example 9.3 Hand assemble the instruction:

sethi
%hi(0x87654321), %r2
This instruction is encoded using the format shown in Figure 9.2. As such, we need to determined
the values for the rd and const22 fields. The following table summarizes these encodings:
Field
Symbolic value
Encoded value
rd
%r2
00010
const22 %hi(0x87654321) 1000 0111 0110 0101 0100 00
These encodings lead to the following machine instruction:
31 30 29

25 24

22 21

00
00010
100
1000 0111 0110 0101 0100 00
That is, 0000 0101 0010 0001 1101 1001 0101 0000 in binary, or 0x0521D950.

9.3.3 Encoding integer data manipulation instructions

Data manipulation instructions are encoded using two formats: one for instructions that
use two source registers and another for instructions that use a source register and a small
integer constant. The formats used for integer data manipulation instructions are shown
in Figure 9.3
In the first format the 32-bit instruction is divided into seven fields. The first field (reading from the left) holds the 2-bit value 10, while the fifth field (bit 13) holds the 1-bit value 0.

62

Lab 9. The SPARC Instruction Formats

A. Instructions of the form: op rs1 , rs2 , rd

31 30 29

10

25 24

19 18

op3

rd

14 13 12

rs1

rs2

unused(zero)

B. Instructions of the form: op rs1 , siconst13 , rd

31 30 29

10

25 24

19 18

op3

rd

14 13 12

rs1

siconst13

Figure 9.3 Instruction formats for data manipulation instructions

These bits are the same for all data manipulation instructions that use two source registers.
The sixth field (bits 5 through 12) is unused-the bits in this field must be zero. The remaining fields, rd, op3 , rs1 , and rs2 , hold encodings for the destination register, the operation,
and the two source registers, respectively
In the second format the 32-bit instruction is divided into six fields. As in the previous
format, the first field holds the first field holds the 2-bit value 01. However, unlike the
previous format, the fifth field holds the 1-bit value 1. The remaining fields, rd, op3 , rs1 ,
and siconst13 , hold encodings for the destination register, the operation, the source register,
and the constant value, respectively. When this format is used, the integer constant is
encoded using the 13-bit 2s complement representation and stored in the siconst13 field of
the instruction.
Recall that a SPARC assembly language instruction begins with the name of the operation, followed by the two source operands, followed by the destination operand. In
considering the translation from an assembly language instruction into machine language,
there are a few points to keep in mind:






The operation is encoded in the op3 field.

The first source operand must be a register and it is encoded in the rs1 field.
The second source operand can be a register or a constant value. If it is a register, it is
encoded in the rs2 field; otherwise, it is encoded in the siconst13 field.
The destination register is encoded in the rd field.

Table 9.2 summarizes the operation encodings for the data manipulation operations
that we have covered in the previous labs. When an instruction using one of these operations is encoded, the operator encoding is placed in the op3 field of the machine instruction.
Example 9.4
sub

Hand Assemble the following SPARC instructions.

%r27, %r16, %r26

Because this instruction uses two source registers, it is encoded using the first format shown in Figure 9.3. As such, we must determine the values for the op3 , rd, rs1 , and rs2 fields. The following table
summarizes these encodings:
Field Symbolic value Encoded value
rd
%r27
11011
op3
sub
000100
%r16
10000
rs1
rs2
%r26
11010
These encodings lead to the following machine instruction:
31 30 29

25 24

19 18

14 13 12

5 4

10
11011
000100
10000
0
00000000
11010
That is, 1011 0110 0010 0100 0000 0000 0001 1010 in binary, or 0xB624001A.

Lab 9. The SPARC Instruction Formats

63

Table 9.2 Operation encodings for the data manipulation operations

Operation
add
and
andn
or
orn
udiv
umul
smul
sdiv
sub
xor
xnor
sll
rsl
rsa

op3
000000
000001
000101
000010
000110
001110
001010
001011
001111
000100
000011
000111
100101
100101
100111

Operation
addcc
andcc
andncc
orcc
orncc
udivcc
umulcc
smulcc
sdivcc
subcc
xorcc
xnorcc

op3
010000
010001
010101
010010
010110
011110
011010
011011
011111
010100
010011
010111

Example 9.5 Hand Assemble the following SPARC instructions.

smulcc %r19, -23, %r29
Because this instruction uses one source register and a signed integer constant, it is encoded using
the second format shown in Figure 9.3. As such, we must determine the values for the op3 , rd, rs1 ,
and siconst13 fields. The following table summarizes these encodings:
Field
Symbolic value Encoded value
rd
%r19
10011
op3
smulcc
011011
%r29
11101
rs1
siconst13
23
1111 1111 0100 1
These encodings lead to the following machine instruction:

31 30 29

25 24

19 18

14 13 12

10
10011
011011
11101
1111 1111 0100 1
1
That is, 1010 0110 1101 1111 0111 1111 1110 1001 in binary, or 0xA6DF7FE9.

9.3.4 Encoding conditional branching instructions

The machine language format for the conditional branching operations on the SPARC is
shown in Figure 9.4. This format divides the machine instruction into five fields. The first
and fourth fields hold the fixed values 102 and 0102. The remaining fields, a, cond, and
disp22 , hold the encoded values for the annul bit, the branching condition, and program
counter displacement.
31 30 29 28

00

25 24

cond

22 21

010

disp22

Figure 9.4 Instruction format for conditional branch instructions

64

Lab 9. The SPARC Instruction Formats

The a field of a machine instruction is set (i.e., 1) for instructions that use the annul suffix
(,a). This field is clear (i.e, 0) for conditional branching instructions that do not nullify
the results of the next instruction. The cond field of a machine instruction encodes the
condition under which the branch is taken. Table 9.3 summarizes the operation encodings
for the branching operations supported by the SPARC.
Table 9.3 Operation encodings for the conditional branching operations
Operation
ba
bne (bnz)
bg
bge
bgu
bcc (bgeu)
bpos
bvc

cond
1000
1001
1010
1011
1100
1101
1110
1111

Operation
bn
be (bz)
ble
bl
bleu
bcs (blu)
bneg
bvs

cond
0000
0001
0010
0011
0100
0101
0110
0111

To complete the encoding of an assembly language instruction that uses conditional

branching, you need to determine the value of the disp22 field. We address this issue by
considering how a processor uses this value. When the processor determines that the
branching condition is satisfied, it multiplies the value in the disp22 field by 4 and adds
it to the program counter (PC). To be more precise, the processor sign extends the 22-bit
value stored in the disp22 field to 30 bits and concatenates two zeros to construct a 32-bit
which which it adds to the PC. In effect, the disp22 field holds the distance from the target
to the destination measured in instructions.
Example 9.6

Hand Assemble the branch instruction in the following SPARC code fragment.
cmp
%r2, 8
bne
l1
nop
inc
%r3

l1:
In this case, the target is 3 instructions from the branch instruction, so the disp22 field will be the 22-bit
binary encoding of 3.
Field
Symbolic value Encoded value
a
0
cond
bne
1001
disp2
l1
0000 0000 0000 0000 0000 11
These encodings lead to the following machine instruction:

31 30 29 28

25 24

22 21

00 0 1001
010
0000 0000 0000 0000 0000 11
That is, 0001 0010 1000 0000 0000 0000 0000 0011 in binary, or 0x12800002.

Example 9.7
top:

Hand Assemble the branch instruction in the following SPARC code fragment.
add
%r2, %r3, %r2
deccc
%r4
bne
top

In this case, the target is 2 instructions (back) from the branch instruction, so the disp22 field will be
the 22-bit binary encoding of
.

;2

Lab 9. The SPARC Instruction Formats

65

Field
Symbolic value Encoded value
a
0
cond
bne
1001
disp2
l1
1111 1111 1111 1111 1111 10
These encodings lead to the following machine instruction:

31 30 29 28

25 24

22 21

00 0 1001
010
1111 1111 1111 1111 1111 10
That is, 0001 0010 1011 1111 1111 1111 1111 1110 in binary, or 0x12BFFFFE.

9.3.5 Synthetic Instructions

In most cases, an assembly language instruction is simply a symbolic representation of a
machine language instruction. The SPARC architecture also defines a number of assembly language instructions that do not correspond directly to SPARC machine language
instructions. These are called synthetic instructions. The assembler translates synthetic
instructions one or more machine language instructions. Using synthetic instructions can
frequently make your programs easier to read. Table 9.4 summarizes the translation provided by the assembler for most of the synthetic instructions on the SPARC.
Most of the translations shown in Table 9.4 are straightforward. However, the implementation of the set instruction merits further discussion. The assembler will always try to
use one of the first two translations if it can. That is, if the constant value can be represented
in 13 bits, the assembler will select the first translation. If the least significant 10 bits of the
constant value are 0, it will used the second translation. Otherwise, the assembler will use
the third translation. Note, if the constant value is relocatable, the assembler will always
select the third translation.

9.3.6 The read and write instructions

The Y register, introduced in Lab 4 is one of the SPARC state registers. As shown in Table 9.4, when you use a state register as the destination in a mov instruction, it is translated
to a wr (write) instruction. Similarly, when you use a state register as the source register in
a mov instruction it is translated to a rd (read) instruction.
Write instructions are encoded using the formats shown in Figure 9.3. When the destination register is the Y register, the rd field is set to the 5-bit value 00000 and the op3 field is
set to the 6-bit value 110000.
Read instructions are encoded using the second format shown in Figure 9.3. When the
source register is the Y register, the op3 field is set to the 6-bit value 101000 and the rs1 field
is set to the 5-bit value 00000.

9.3.7 Relocatable expressions

In this lab, we have limited our discussion to the translation of instructions that use absolute expressions. We will consider the translation of relocatable expressions when we
consider linking and loading in Lab 15.

9.3.8 Decoding SPARC instructions

We conclude our discussion of instruction formats by considering instruction decoding.
That is, the process by which a SPARC processor determines the instruction it is executing.
The SPARC uses a distributed opcode. The two most significant bits in an instruction
represent the primary opcode. If the primary opcode is 00, bits 2224 of the instruction

66

Lab 9. The SPARC Instruction Formats

Table 9.4 The synthetic instructions

Synthetic instruction
bclr
rs, rd
bclr
rs, siconst13
bset
rs, rd
bset
siconst13 , rd
btst
rs1 , rs2
btst
rs, siconst13
btog
rs, rd
btog
rs, siconst13
clr
rd
clrb
[address]
clrh
[address]
clr
[address]
cmp
rs1 , rs2
cmp
rs, siconst13
dec
rd
dec
siconst13 , rd
deccc rd
deccc siconst13 , rd
inc
rd
inc
siconst13 , rd
inccc
rd
inccc
siconst13 , rd
mov
rs, rd
mov
siconst13 , rd
mov
statereg, rd
mov
rs, statereg
mov
siconst13 , statereg
neg
rs, rd
neg
rd
not
rd
not
rs, rd
set
iconst, rd

tst

rs

Implementation
andn
rd, rs, rd
andn
rs, siconst13 , rd
or
rd, rs, rd
or
rd, siconst13 , rd
andcc rs1 , rs2 , %g0
andcc rs, siconst13 , %g0
xor
rd, rs, rd
xor
rs, siconst13 , rd
or
%g0, %g0, rd
stb
%g0, [address]
sth
%g0, [address]
st
%g0, [address]
subcc rs1 , rs2 , %g0
subcc rs, siconst13 , %g0
sub
rd, 1, rd
sub
rd, siconst13 , rd
subcc rd, 1, rd
subcc rd, siconst13 , rd
add
rd, 1, rd
add
rd, siconst13 , rd
addcc rd, 1, rd
addcc rd, siconst13 , rd
or
%g0, rs, rd
or
%g0, siconst13 , rd
rd
statereg, rd
wr
%g0, rs, statereg
wr
%g0, siconst13 , statereg
sub
%g0, rs, rd
sub
%g0, rd, rd
xnor
rd, %g0, rd
xnor
rs, %g0, rd
or
%g0, iconst, rd
or
sethi
%hi(iconst), rd
or
sethi
%hi(iconst), rd
or
rd, %lo(iconst), rd
orcc
%g0, rs, %g0

provide the secondary opcode. If the primary opcode is 01, the instruction is a call instruction and the remaining bits (bits 029) are a displacement for the program counter (we will
discuss the call instruction at greater length in Lab 10). Otherwise, if the primary opcode
is either 10 or 11, bits 1924 of the instruction provide the secondary opcode. Figure 9.5
illustrates the positions of the secondary opcodes based on the primary opcode.
Once you have determined the primary and secondary opcodes, youll be able to to
determined the instruction and, knowing the instruction, decode the remaining fields of
the instruction. If the primary opcode is 01, the instruction is a call instruction and you can
easily complete the decoding of the instruction.
If the primary opcode is 00, the instruction is an unimplemented instruction, a condi-

Lab 9. The SPARC Instruction Formats

31 30 29

25 24

67
0

22 21

op2

00
31 30 29

01
31 30 29

25 24

31 30 29

19 18

19 18

op3

10
25 24

op3

11

Figure 9.5 The primary opcode ina SPARC instruction

tional branch instruction, or a sethi instruction. Table 9.5 summarizes how the 3-bit value
in op2 is used to identify the instruction.
Table 9.5 Decoding the op2 field
Value
000
001
010
011
100
101
110
111

Instruction
The unimplemented instruction
illegal
Conditional branchinteger unit
illegal
SETHI
illegal
Conditional branchfloating point unit
Conditional branchcoprocessor

The data manipulation instructions are encoded with a primary opcode of 10. Table 9.6
shows how the 6-bit value in the op3 field is used to determine the instruction when the
primary opcode is 10.
Table 9.6 Decoding the op3 field when the primary opcode is 10
xxx000
xxx001
xxx010
xxx011
xxx100
xxx101
xxx110
xxx111

000xxx
add
and
or
xor
sub
andn
orn
xnor

001xxx
addx

umul
smul
subx

udiv
sdiv

010xxx
addcc
andcc
orcc
xorcc
subcc
andncc
orncc
xnorcc

011xxx
addxcc

umulcc
smulcc
subxcc

udivcc
sdivcc

100xxx
taddcc
tsubcc
taddcctv
tsubcctv
mulscc
sll
srl
sra

101xxx
rd
rd
rd
rd

110xxx
wr
wr
wr
wr
FPU op
FPU op
CP op
CP op

111xxx
jmpl
rett
trap
flush
save
restore

Instructions that access memory are encoded with a primary opcode of 11. Table 9.7
shows how the 6-bit value in the op3 field is used to determine the instruction when the
primary opcode is 11.
When you decode an instrcution that has a primary opcode of 10 or 11, you will need to
examine bit 13 to determine whether bits 012 of the instruction hold an immediate value

68

Lab 9. The SPARC Instruction Formats

Table 9.7 Decoding the op3 field when the primary opcode is 11
xxx000
xxx001
xxx010
xxx011
xxx100
xxx101
xxx110
xxx111

000xxx
ld
ldub
lduh
ldd
st
stb
sth
std

001xxx

ldsb
ldsh

ldstub

swap

010xxx
lda
lduba
lduha
ldda
sta
stba
stha
stda

011xxx

ldsba
ldsha

ldstuba

swapa

100xxx
ldf
ldfsr

lddf
stf
stfsr
stdfq
stdf

101xxx

110xxx
ldc
ldcsr

lddc
stc
stcsr
scdfq
scdf

111xxx

or a register. If bit 13 is 1, bits 012 hold an immediate value.

Example 9.8

Give an instruction that will assemble to the value 0x09012345.

In binary, this instruction is 00 00100 100 000100. . . . That is, the primary opcode is 00 and op2 is 100.
From Table 9.5, this is a sethi instruction. Using the sethi format to partition the bits yields:
31 30 29

25 24

22 21

iconst22 = 01 0010 0011 0100 0101

00 rd = 00100
100
Thus, the destination register is %r4, and the integer constant is 0x12345. The following instruction
will be assembled as 0x09012345.
sethi
%hi(0x12345<<10), %r4

Example 9.9

Give an instruction that will assemble to the value 0x10800006.

In binary, this instruction is 00 01000 010 000000. . . . That is, the primary opcode is 00 and op2 is
010. From Table 9.5, this is a conditional branch instruction. Using the conditional branch format to
partition the bits yields:
31 30 29 28

25 24

22 21

disp=0000 0000 0000 0000 0001 10

00 0 cd=1000
010
Thus, the operator is ba and the displacement is +6 words. The following instruction will be assembled as 0x10800006.
ba
.+(6*4)
(When you use isem-as, . is the address of the current instruction.

Example 9.10

Give an instruction that will assemble to the value 0x8601600E

In binary, the instruction is 10 00011 000000 0001. . . . That is, the primary opcode is 10 and op3 is
000000. From Table 9.6, this is an add instruction. Because bit 13 is 1, we use the second format in
Figure 9.3 to decode this instruction.
31 30 29

25 24

19 18

14 13 12

rs1 =00101 1
siconst13 =0 0000 0000 1110
10 rd=00011
000000
Thus, the destination is %r3, the source register is %r5, and the constant is 0xE. The following instruction will be assembled as 0x8601600E.
add
%r5, 14, %r3

Lab 9. The SPARC Instruction Formats

9.4 Summary
9.5 Review Questions
9.6 Exercises

69

70

Lab 9. The SPARC Instruction Formats

Laboratory 10
Leaf Procedures on the SPARC
10.1 Goal
To introduce the calling conventions associated with leaf procedures on the SPARC.

10.2 Objectives
After completing this lab, you will be able to write assembly language programs that:





Use leaf procedures,

Use the call and link operation, and
Use the (synthetic) return from leaf operation.

10.3 Discussion
This lab is the first of three labs that cover procedure calling conventions on the SPARC.
In this lab we consider the conventions associated with leaf procedures: procedures that
do not make calls to other procedures. In Lab 11 we consider the use of register windows
on the SPARC. In Lab 12 we complete our coverage of procedure calling conventions by
considering the standard calling conventions used by compilers.

10.3.1 Register usage

The SPARC Architecture Manual describes a class of procedures called optimized leaf
procedures. As we have noted, a leaf procedure is a procedure that does not call any
other procedures. Optimized refers to restrictions placed on the procedures use of the
registers.
Table 10.1 summarizes register uses for optimized leaf procedures. This table specifies
which registers can be modified by a leaf procedure. Registers %r8%r13 are used for
parameters passed to the leaf procedure. The first parameter (i.e., the leftmost parameter)
should be placed in %r8, the next in %r9, and so forth. Note that %r8 is used for the first
parameter and the return value.
Beyond the registers used to pass parameters to the leaf procedure, there are a few
other conventions in Table 10.1 that are worth noting. Register %r1 can always be used as a
temporary register and the caller cannot assume that it will retain its value across a call to a
leaf procedure. We have frequently used this register in the sethi instructions used with the
load and store instructions, and you can continue to do this in leaf procedures. The stack
pointer is stored in %r14. We will discuss the use of the stack pointer when we consider
stack based calling conventions in Lab 12. For the moment, note that a leaf procedure
should not alter the value stored in register %r14. Finally, note that the return address
71

72

Lab 10. Leaf Procedures on the SPARC

Table 10.1 Register usage for optimized leaf procedures

Register(s)
%r0
%r1
%r2%r7
%r8
%r8%r13
%r14
%r15
%r16%r31

Use
zero
temporary
callers variables
return value
parameters
stack pointer
return address
callers variable

Changed by leaf procedure

No
Yes
No
Yes
Yes
No
Yes
No

(actually, the address of the call instruction used to call the leaf procedure) is stored in
register %r15. A leaf procedure can alter this register; however, it will be difficult to return
to the point of the call if you alter the value in%r15.
An optimized leaf procedure should only alter the values stored in registers %r1 and
%r8%r13. If the leaf procedure requires more local storage than these registers provide,
or if the parameters do not fit in these registers, the leaf procedure cannot be implemented
as an optimized leaf procedure. We will discuss the techniques use to implement other
types of procedures in the next two labs.

10.3.2 Calling sequence

In assembly language, a procedure is a block of instructions. The first instruction in this
block is labeled by the name of the procedure.
Table 10.2 summarizes the operations used to call and return from optimized leaf procedures. Like the branching operations introduced in Lab 3, these operations have a branch
delay slot. The call and link operation saves the current value of the PC in %r15, updates
the PC, and sets the nPC to the address specified in the call. The retl operation updates
the PC and sets to nPC to the contents of %r15 plus eight. The plus eight is needed to
skip over the call instruction and the instruction in the delay slot of the call. Examples 10.1
and 10.2 illustrate the use of these operations.
Table 10.2 The call and retl operations
Operation
call and link

Assembler syntax
call label

return from leaf

retl

Operation implemented
%r15 = PC
PC = nPC
nPC = label
PC = nPC
nPC = %r15 + 8

Example 10.1 Write SPARC procedure that prints a NULL terminated string. The address of the string to
print will be passed as the first parameter (i.e., in %r8).
.text
! pr_str - print a null terminated string
!
! Parameters:
%r8 - pointer to string (initially)
!

Lab 10. Leaf Procedures on the SPARC

73

! Temporaries:
!
!

%r8 - the character to be printed

%r9 - pointer to string

pr_str: mov
pr_lp: ldub
cmp
be
nop
ta
ba
inc

%r8, %r9
[%r9], %r8
%r8, 0
pr_dn

! we need %r8 for the "ta 1"

! load character
! check for null

1
pr_lp
%r9

! print character

pr_dn:

Example 10.2
ple 10.1.
str:

main:

end:

! increment the pointer

retl
nop

! return from leaf procedure

!
(branch delay)

Write a SPARC assembly language fragment that calls the procedure presented in Exam.data
.asciz
.text
set
call
nop
ta

"Hello, World! n"

str, %r8
pr_str
0

! setup the first argument

! call print string
!
(branch delay)
! exit gracefully

10.3.3 Instruction Encodings

We have introduced two new instructions in this lab: call and retl. Figure 10.1 illustrates
the format used to encode call instructions. This instruction encoding uses two fields. The
first field holds the 2-bit value 01, the second holds a 30-bit displacement for the program
counter. This field is encoded in the same fashion as the displacement field for the conditional branch instructions (see Lab 9 for more details).
31 30 29

disp30

01

Figure 10.1 Instruction format for call instructions

Example 10.3
main:

Show how the call instruction in the following SPARC assembly code fragment is encoded.
.text
set
str, %r8
! setup the first argument
call
pr_string
! call print string
nop
!
(branch delay)
ta
0
! exit gracefully

pr_string:
mov

%r8, %r9

! we need %r8 for the "ta 1"

74

Lab 10. Leaf Procedures on the SPARC

In this case, the target is 3 instructions from the call instruction, so the disp30 field is set to the 30-bit
binary encoding of 3.
This leads to the following machine instruction:
31 30 29

01
0000 0000 0000 0000 0000 0000 0000 11
That is, 0100 0000 0000 0000 0000 0000 0000 0011 in binary, or 0x40000003.

The retl instruction is actually a synthetic instruction that is translated to a jmpl (jump
and link) instruction. The jmpl instruction has two operands: an address, and a destination
register. The address is similar to the addresses used in the load and store instructions;
however, the brackets surrounding the address in the load and store instructions are omitted in the jmpl instruction. When a SPARC processor executes a jmpl instruction, it saves
the address of the jmpl instruction in the destination register and sets the next program
counter to the address specified in the instruction. Figure 10.2 illustrates the formats used
to encode jmpl instructions.
A. Instructions of the form: jmpl rs1 +rs2 , rd
31 30 29

10

25 24

19 18

14 13 12

rs1

111000

rd

rs2

unused(zero)

B. Instructions of the form: jmpl rs1 +siconst13 , rd

31 30 29

10

25 24

19 18

111000

rd

14 13 12

rs1

siconst13

Figure 10.2 Instruction formats for jmpl instructions

The retl instruction is translated to a jmpl instruction with the address set to %r15+8
and the destination register set to %r0. That is, the target of the branch is instruction after
the branch delay slot of the call instruction (remember, call instruction saved its address in
%r15), and the address of the retl instruction is discarded (saved in %r0).
Example 10.4

Show how a retl instruction is encoded.

A retl instruction is translated to the instruction jmpl %r15+8, %r0. This instruction is encoded using
the second format shown in Figure 10.2. Figure 9.3. As such, we must determine the values for the
rd, rs1 , and siconst13 fields. The following table summarizes these encodings:
Field
Symbolic value Encoded value
rd
%r0
00000
rs1
%r15
01111
siconst13 8
0000 0000 0100 0
These encodings lead to the following machine instruction:
31 30 29

25 24

19 18

14 13 12

10
00000
111000
01111
0000 0000 0100 0
1
That is, 1000 0001 1100 0011 1110 0000 0000 1000 in binary, or 0x81C3E008.

10.4 Summary
A leaf procedure is a procedure that never calls any other procedure. In this lab we have
introduced the SPARC instructions used to write leaf procedures: call and retl. In the next
two labs, we will examine more general procedure calling conventions.

Lab 10. Leaf Procedures on the SPARC

75

10.5 Review Questions

1. xxx

10.6 Exercises
1. Write a SPARC assembly language program consisting of of a main program and a
procedure, pr octal, that prints an unsigned integer in octal notation.
2. Write a SPARC assembly language program consisting of of a main program and a
procedure, pr hex, that prints an unsigned integer in hexadecimal notation.
3. Write a SPARC procedure, called strcmp, that compares two strings. Your procedure should accept two parameters, s1 and s2, both pointers to NULL terminated
strings. Your procedure should return an integer based on the comparison. In particular,





if (s1 < s1) return ;1;

if (s1 > s2) return 1;
if (s1 == s2) return 0;

You should also provide a small driver to test your procedure.

4. Write a SPARC procedure, called strchr, that returns a pointer to the first occurrence of a character within a string. The first parameter should be a pointer to a
string. The second parameter should be the character search for. If the character is
not present in the string, the procedure should return NULL (i.e., 0).
You should also provide a small driver to test your procedure.
5. Write a procedure to draw a black pixel at an arbitrary (x, y) location on the GX
device. The C declaration for the procedure would be:
void draw pixel(int x, int y)

76

Lab 10. Leaf Procedures on the SPARC

Laboratory 11
Register Windows
11.1 Goal
To introduce register windows.

11.2 Objectives
After completing this lab, you will be able to use:





Register windows,
The save and restore operations,
The (synthetic) return operation.

11.3 Discussion
In this lab we introduce a more general procedure calling mechanism that uses register
windows. We introduce the save and restore instructions and another synthetic instruction,
ret, for returning from procedures.

11.3.1 Register Windows

To this point, we have used the %r names for the integer registers. From this point on,
we will use the alternate names for these registers. The alternate names are shown in
Table 11.1.
Table 11.1 Names for the integer registers
Integer registers
%r0%r7
%r8%r15
%r16%r23
%r24%r31

Alternate names
%g0%g7
%o0%o7
%l0%l7
%i0%i7

Group name
Global registers
Output registers
Local registers
Input registers

The alternate names reflect the uses of the registers when procedures use register windows. The global registers (%g0%g7) are shared by all procedures. The output registers
are used for parameters when calling another procedure. That is, the output registers are
outputs from the caller to the called procedure. The local registers (%l0%l7) are used to
store local values used by a procedure. The input registers (%i0%i7) are used for the parameters passed into the procedure. That is, the input registers are inputs passed from the
caller to the called procedure.
77

78

Lab 11. Register Windows

When you consider the relationship between the output and input registers, the trick is
to make the callers output registers the same as the called procedures input registers. On
the SPARC, this is done using overlapping register windows.
All procedures share the global registers (%g0%g7). The remaining registers, %o0
%o7, %l0%l7, and %i0%i7, are called a register window. When a procedure starts its execution, it allocates a set of 16 registers (using the save instruction as described in the following section). The new register set provides the procedure with its own output and local
registers (%o0%o7 and %l0%l7). The procedures input registers (%i0%i7) are overlapped with the callers output registers. Figure 11.1 illustrates the overlapping of register
windows between the caller and the callee.
Global registers

%g7 (%r7)
%g6 (%r6)
%g5 (%r5)
%g4 (%r4)
%g3 (%r3)
%g2 (%r2)
%g1 (%r1)
%g0 (%r0)

Callees
register
window

6 6
Overlap

? ?

Callees
Names
%o0 (%r8)
%o1 (%r9)
%o2 (%r10)
%o3 (%r11)
%o4 (%r12)
%o5 (%r13)
%o6 (%r14), %sp
%o7 (%r15)
%l0 (%r16)
%l1 (%r17)
%l2 (%r18)
%l3 (%r19)
%l4 (%r20)
%l5 (%r21)
%l6 (%r22)
%l7 (%r23)
%i0 (%r24)
%i1 (%r25)
%i2 (%r26)
%i3 (%r27)
%i4 (%r28)
%i5 (%r29)
%i6 (%r30), %fp
%i7 (%r31)

Callers
register
window

?
Figure 11.1 Overlapping register windows

Callers
Names
%o0 (%r8)
%o1 (%r9)
%o2 (%r10)
%o3 (%r11)
%o4 (%r12)
%o5 (%r13)
%o6 (%r14), %sp
%o7 (%r15)
%l0 (%r16)
%l1 (%r17)
%l2 (%r18)
%l3 (%r19)
%l4 (%r20)
%l5 (%r21)
%l6 (%r22)
%l7 (%r23)
%i0 (%r24)
%i1 (%r25)
%i2 (%r26)
%i3 (%r27)
%i4 (%r28)
%i5 (%r29)
%i6 (%r30), %fp
%i7 (%r31)

Lab 11. Register Windows

79

In addition to the register names we have discussed, Figure 11.1 introduces two new
names: %sp and %fp. The first of these, %sp, denotes the stack pointer. In an assembly
language program, %sp is simply another name for %o6. Similarly, %fp denotes the frame
pointer and is simply another name for %i6. We will discuss the special uses of these
registers (and hence the additional names) in the next lab when we consider stack frame
organization.
An implementation of the SPARC integer unit may have between 40 and 520 integer
registers. Every SPARC has 8 global registers, plus a circular stack of 2 to 32 register sets.
Each register set has 16 registers. The number of registers sets is implementation dependent. Number of register sets provided a particular implementation of the SPARC architecture has been given the name NWINDOWS. ISEM provides 32 register sets, the maximum
number supported by the SPARC Architecture. Most hardware implementations provide
7 or 8 register sets.

11.3.2 Save and Restore

The current window in the integer registers is given by the current window pointer (CWP).
The CWP is stored in the lower five bits of the processor status register (PSR). The save and
restore instructions let the programmer to manipulate the CWP.
Table 11.2 summarizes the SPARC save and restore operations. SPARC assemblers provide two instruction formats for save instructions and three formats for restore instructions.
Both formats for the save instruction and the first two formats for the restore instruction
use three explicit operandstwo source operands and a destination operand. In the first
format, both source operands are in registers. In the second format, one source operand
is in a register, the other is specified using a small constant value. This constant may be
positive or negative; however its 2s complement representation must fit in 13 bits. Its important to note, for both the save and restore operations, that the destination register is in
the register window after the CWP has been modified. The restore instruction also has a
format that doesnt have any operands.
Table 11.2 Saving and restoring register windows
Operation
save callers
register window

Syntax
save rs1 , rs2 , rd

save rs1 , siconst13 , rd

restore callers
register window

restore rs1 , rs2 , rd

restore rs, siconst13 , rd

restore

Operation implemented
res = reg[rs1 ]+reg[rs2 ]
CWP = (CWP 1) % NWINDOWS
reg[rd] = res
res = reg[rs1 ]+siconst13
CWP = (CWP 1) % NWINDOWS
reg[rd] = res
res = reg[sr1 ]+reg[sr2 ]
CWP = (CWP+1) % NWINDOWS
reg[rd] = res
res = reg[rs]+siconst13
CWP = (CWP+1) % NWINDOWS
reg[rd] = res
CWP = (CWP+1) % NWINDOWS

80

Lab 11. Register Windows

11.3.3 Stack management

The operands of the save instruction are commonly used to allocate space for a stack frame.
Well discuss the stack and stack frames at greater length in the next lab. However, to the
save and restore instructions correctly, you must allocate and maintain a runtime stack.
The runtime stack grows from higher to lower addresses. The stack pointer (%r14, %o6)
should be aligned on an 8 byte address at all times. For now, well use the operands of the
save instruction to subtract 96 from the stack pointer (%sp). The reason for this will be
clearer in the next lab.
In most cases, you will simply use the restore instruction with no arguments to restore
the callers register window. However, you can occasionally use the other versions of this
instruction to return a value from a procedure.

11.3.4 Procedure calling conventions

From the callers perspective using register windows does not change anything about how
it interacts with the called procedure. The caller still puts the outgoing parameters in registers %o0%o5 (%r8%r13) and expect the result in %o0 (%r8). Moreover, the caller may
assume that registers %g0, %g2%g7, %o6 (%sp), %l0%l7, and %i0%i7 will not be altered by the called procedure. Finally, the caller uses the call instruction introduced in the
previous lab to transfer control to the called procedure.
From the perspective of the called procedure, things have changed quite a bit. As soon
as control is transferred to the procedure, it needs to save to the callers register window.
In the body of the procedure, you can modify %g1, %i0%i5, %l0%l7, and %o0%o6. You
could also modify %i7; however, %i7 holds the return address, so its not a good idea to
change it. Finally, just before returning control to the caller, the called procedure needs to
restore the callers register window.
The instruction used to restore the callers register window is usually put in the branch
delay slot of the instruction used to return control to the caller. Because the callers register
window has not been restored when the called procedure issues the return instruction, the
return instruction needs to use %i7 in calculating the return address instead of %o7 (%r15).
SPARC assemblers provide the ret instruction for this purpose. Table 11.3 summarizes the
call and return instructions.
Table 11.3 The SPARC call and ret operations
Operation
call and link

Syntax
call label

return from procedure

ret

Operation implemented
%o7 = PC
PC = nPC
nPC = label
PC = nPC
nPC = %i7 + 8

Example 11.1 Write a SPARC assembly language procedure, pr str, that will print a NULL terminated
string. Your procedure should take a single argument, the address of the string to print. In writing this
procedure, you should assume that the procedure pr ch is available for printing a character.
.text
! pr_str - print a null terminated string
!
! Temporaries: %i0 - pointer to string

Lab 11. Register Windows

!
!
pr_str: save

pr_lp:

pr_dn:

ldub
cmp
be
nop
call
nop
ba
inc

81

%o0 - character to be printed

%sp, -96, %sp

! PROLOGUE - save the current window and

!
allocate the minimum stack frame

[%i0], %o0
%o0, 0
pr_dn

! load character
! check for null

pr_char

! print character

pr_lp
%i0

! increment the pointer (branch delay)

ret
restore

! EPILOGUE return from procedure and

!
restore old window; no return value

Example 11.2 Write a main SPARC assembly language fragment that allocates space for the stack and
calls the pr str procedure in the previous example.
str:

.data
.asciz

"Hello, World! n"

.align 8
stack_top:
. = . + 2048
stack_bot:
.text
start:
set

end:

set
call
nop
ta

stack_bot-96, %sp ! initialize the stack and allocate

! the minimum stack frame
str, %o0
! initialize pointer to str
pr_str
! call print string
!
(branch delay)
0

Example 11.3 Write a procedure that recursively calculates the Nth Fibonacci number. You may assume
that N is non-negative and will be small enough that register overflow will not occur.
! fib - calculate the Nth Fibonacci number
!
!
fib(N) = fib(N-1) + fib(N-2)
!
fib(0) = fib(1) = 1
fib:

save

%sp, -96, %sp

! PROLOGUE

cmp
bg
nop

%i0, 1
fib_call

! call recursively

ret
restore %g0, 1, %o0

! EPILOGUE
! return 1

82

Lab 11. Register Windows

fib_call:
call
sub
mov

fib
%i0, 1, %o0
%o0, %l0

! call with N-1

! (branch delay)
! %l0 = fib(N-1)

call
sub

fib
%i0, 2, %o0

! call with N-2

! (branch delay)

ret
restore %l0, %o0, %o0

! EPILOGUE
! return fib(N-1) + fib(N-2)

11.3.5 Exceptions
Both the save and restore operations can generate exceptions (or traps). Before the CWP is
modified, the bit in the WIM cooresponding to the new value for the CWP is tested. If the
bit in the WIM is 1, an exception is generated. For a save instruction, this causes a window
overflow trap. For a restore instruction, this causes a window underflow trap.
These traps are normally handled by the operating system and are transparent to the
application programmer. In tkisem these traps are handled by the rom code. We will discuss
the code used to handle these traps in Lab 17.

11.3.6 Instruction encoding

We have introduced three new instructions in this lab: save, restore, and ret. The save and
restore instructions are encoded as data manipulation instructions (the instruction formats
are shown in Figure 9.3). The restore instruction with no operands is actually a synthetic
instruction in which all of the operands are %g0. Table 11.4 summarizes the encodings of
the op3 field for the save and restore instructions.
Table 11.4 Encoding op3 in save and restore instructions
Instruction
save
restore

op3
111100
111101

Like the retl instruction, the ret instruction is a synthetic instruction, based on the jmpl
instruction. The ret instruction is translated to jmpl %i7+8, %g0.

11.4 Summary
This lab presents a more general mechanism for procedures on the SPARC. Register windows provide easy access to a large collection of registers and can reduce the need to save
registers in memory. While this mechanism has many advantages there are several disadvantages to keep in mind. The mechanism only provides six registers for procedure
parameters. If you write a procedure with more than six parameters, you will need to to
use the stack for any parameters beyond six. Secondly, most implementations only have 7
or 8 register sets. So, if your call sequence gets deeper than NWINDOWS (as it probably
will in most recursive procedures), you are again forced to use the stack.

Lab 11. Register Windows

83

11.5 Review Questions

11.6 Exercises
1. Write a procedure which will draw a bitmap at an arbitrary (x, y) location on the GX
device. The bitmap is described by an array of chars, a width, and a height (i.e., the
X Windows bitmap format) The C declaration for the procedure would be:
void draw bitmap(char* bits, int w, int h, int x, int y)
The draw bitmap procedure should make use of the draw pixel procedure (from the
previous lab).
2. Write a procedure that tiles a bitmap to the GX display. The procedure should call
draw bitmap for the first tile. But, for all of the subsequent tiles, you should use the
GX BLIT operation. The C declaration for the procedure would be:
void tile bitmap(char* bits, int w, int h)
Dont forget to clip the border tiles appropriately.

84

Lab 11. Register Windows

Laboratory 12
Standard Calling Conventions
12.1 Goal
To cover the standard procedure calling conventions for the SPARC.

12.2 Objectives
After completing this lab, you will be able to write assembly language procedures that:





Follow the standard calling conventions for the SPARC,

Call C functions, and
Can be called from C.

12.3 Discussion
In most cases, you will not want to write entire programs in assembly language. Instead,
you will want to write most of the program in a high-level language (like C) and only write
a few procedures in assembly languagethe procedures that cannot be easily optimized
in the high-level language or that need to take advantage of special features provided by
the machine.
In this lab, we complete our presentation of the SPARC application binary interface
(ABI). The SPARC ABI is a set of conventions that are expected to be folowed by all compilers and assembly language programmers. These conventions cover the uses of registers
and the structure of the stack frame. If you follow the conventions specified by the SPARC
ABI in your assembly language procedures, it will be possible to call your proceudres from
procedures written in high-level languages. You will also be able to call procedures written
in high-level languages from your assembly language procedures.
In Lab 10 we covered the portion of the SPARC ABI that deals with optimized leaf
procedures. In Lab 11 we covered the conventions related to register usage for procedures
that are not implementd as optimized leaf procedures. In this lab we cover the conventions
related to the allocation and structure of stack frames. Throughout this lab we will assume
that we are not implementing an optimized leaf procedure.

12.3.1 The allocation of stack frames

The stack pointer is stored in register %o6. In assembly language, this register can be
referenced using the alias %sp. Due to the overlap of register windows, the stack pointer
for the calling procedure is always available in register %i6. In SPARC terminology, the
previous stack pointer is called the frame pointer and can be accessed using the alias %fp.
The stack grows from addresses with larger numbers to addresses with smaller numbers.
85

86

Lab 12. Standard Calling Conventions

As such, allocation of a stack frame is implemented by subtracting a value from the current
stack pointer (actually, this usually done by adding a negative number to the stack pointer).

%sp (=%o6)

smaller
addresses

Current
stack
frame

%fp (=%i6)
(previous %sp)

6
Stack
growth

Callers
stack
frame

larger
addresses

Figure 12.1 Stack frame allocation in the SPARC ABI

12.3.2 Parameters
As we noted in the previous two labs, registers %o0%o5 are used for the first six parameters passed to a procedure. If a procedure has more than 6 parameters, the remaining
parameters are passed on the stack. SPARC procedures do not push parameters (beyond
the sixth parameter) onto the procedure call stack. Instead, they allocate space in their stack
frame for the parameters and copy parameters into this space. This means that the called
procedure will find its parameters (beyond the sixth) in the callers stack frame. The called
procedure can access these parameters using the frame pointer (%fp) with positive offsets.

12.3.3 Stack frame organization

As a minimum, every procedure that is not implemented as an optimized leaf procedure
(i.e., any procudure that executes a save instruction) must allocate a stack frame or 64
bytes. This space will be used to store the input and local registers (%i0%i7 and %l0%l7)
allocated by this procedure should you run out of register windows in a later procedure
call.
Every nonleaf procedure must allocate an additional 7 words (28 bytes) in its stack
frame. The first word of this space is used to store a hidden parameter. The hidden
parameter is used for procedures that return structured values. Procedures that return
simple values use %i0 (the callers %o0) to return the result. However, if a procedure

Lab 12. Standard Calling Conventions

87

returns a structured value, the result may not fit in a register. In this case, the ccalling
procedure must allocate space for the return value (probably in its stack frame). The calling
procedure then puts the address of this space into the hidden parameter before making the
call.
The remaining 6 words can be used by the called procedure to store the first six arguments (the ones passed in %o0%o5). In most cases, the called procedure will be able
to access these parameters in the registers %i0%i5 and will not need to store them in the
callers stack frame. However, if the called procedure needs to take the address of a parameter, it needs to store the parameter into memory (you cant take the address of a register).
In addition to the regions that we have discussed, a procedure may allocate additional
stack space for: alignment (the stack pointer should always be a multiple of 8), outgoing
parameters (beyond the sixth parameter), automatic local arrays and other automatic local variables that dont fit in the local registers %l0%l7, temporaries, and floating point
registers. Figure 12.2 illustrates the organization of a SPARC stack frame.

%sp

6
space for
%i0%i7
and %l0%l7

%sp
%sp

%sp

+ 64
+ 68
+ 92

hidden parameter

required
for all
procedures
required
for nonleaf
procedures

space for
the first six
parameters

outgoing
parameters
past the sixth
%sp

+ (92 + 4
p)

(4
l)

%fp ;

temporaries
and pad
automatic
local
variables

%fp

Figure 12.2 Stack frame organization

Example 12.1

Translate the following C function into a SPARC procedure.

88

Lab 12. Standard Calling Conventions

int add7( int p1, int p2, int p3, int p4, int p5, int p6, int p7 )

f
g

return p1 + p2 + p3 + p4 + p5 + p6 + p7;

In this case, the parameters p1p6 will be in registers %i0%i5. The parameter p7 will be in the
callers stack frame at offset 92 (that is, %fp + 92).
.text
add7:
save
%sp, -64, %sp
! this is a leaf procedure
ld

[%fp+92], %l0

! well eventually need p7

add
add
add
add
add

%i0,
%i0,
%i0,
%i0,
%i0,

!
!
!
!
!

%i1,
%i2,
%i3,
%i4,
%i5,

%i0
%i0
%i0
%i0
%i0

ret
restore %i0, %l0, %o0

add
add
add
add
add

in
in
in
in
in

p2
p3
p4
p5
p6

! add in p7

Example 12.2 Translate the following C function which includes a call to the function defined in example 12.1
int test( int x1, int x2 )

int l1, l2;

l1 = x1 + x2;
l2 = x2 - x1;
return add7( x1, x2, l1, l2, l1+l2, l2-l1, l2+l2 );

test:

.text
save

%sp, -(92+4), %sp

! allocate the minimum stack frame

! (includes a 4 byte alignment

add
sub

%i0, %i1, %l0

%i1, %i0, %l1

! l1 = x1 + x2;
! l2 = x2 - x1;

mov
mov
mov
mov
add
sub

%i0,
%i1,
%l0,
%l1,
%l0,
%l1,

!
!
!
!
!
!

add
call
st

%l1, %l1, %l2

add7
%l2, [%sp+92]

pad)

%o0
%o1
%o2
%o3
%l1, %o4
%l0, %o5

ret
restore %g0, %o0, %o0

first parameter
second parameter
third parameter
fourth parameter
fifth parameter
sixth parameter

! temp = l2+l2
! seventh parameter (delay slot)

! return the result to callers %o0

Lab 12. Standard Calling Conventions

89

It is also common to access local variables stored in the stack using negative offsets from
the frame pointer.
Example 12.3 Translate the following C function into a SPARC procedure. You should assume that the
procedures read int and write int are defined elsewhere.
void read10( )

int i;
int a[10];
for( i = 0 ; i < 10 ; i++ )
a[i] = read_int();

for( i = 9 ; i >= 0 ; i-- )

write_int( a[i] );

In this case, we will use %l0 for i (scaled by 4) and the array a will be stored in the local space starting
at %fp 40.
.text
add7:
save
%sp, -(92+4*10+4), %sp ! we need 40 words for the array

sub

%fp, 40, %l1

! l1 points to the start of the

clr

%l0

! i = 0

call
nop
st

read_int
%o0, [%l1+%l0]

! a[i] = read_int();

inc
cmp
bl
nop

%l0, 4
%l0, 40
top1

! increment i += 4
! i < 10*4

array

top1:

top2:

mov
ld
call
nop

36, %l0
[%l1+%l0], %o0
write_int

! i = 9*4
! write_int( a[i] )

deccc
bge
nop

%l0, 4
top2

! i -= 4
! i >= 0

ret
restore

90

12.4 Summary
12.5 Review Questions
12.6 Exercises

Lab 12. Standard Calling Conventions

Laboratory 13
Integer Arithmetic on the SPARC
13.1 Goal
To cover

13.2 Objectives
After completing this lab, you will be able to write SPARC programs that:

Implement multiple precision arithmetic.

13.3 Discussion
13.4 Summary
13.5 Review Questions
13.6 Exercises

91

92

Lab 13. Integer Arithmetic on the SPARC

Laboratory 14
The Floating Point Coprocessor
14.1 Goal
To cover the floating point coprocessor on the SPARC.

14.2 Objectives
After completing this lab, you will be able to write SPARC programs that:

Use floating point operations.

14.3 Discussion
14.4 Summary
14.5 Review Questions
14.6 Exercises

93

94

Lab 14. The Floating Point Coprocessor

Laboratory 15
Linking and Loading
15.1 Goal
To cover the translation process implemented by the ISEM tools.

15.2 Objectives
After completing this lab, you will:

coff

15.3 Discussion
15.4 Summary
15.5 Review Questions
15.6 Exercises

95

96

Lab 15. Linking and Loading

Laboratory 16
Traps
16.1 Goal
To cover the basic SPARC trap mechanism and trap instructions.

16.2 Objectives
After completing this lab, you will be able to write trap handlers:




trap always
conditional traps

16.3 Discussion
16.3.1 The Processor Status Register (PSR)
Figure 16.1 presents the fields in the processor status register.

Figure 16.1 The processor status register

16.3.2 Address Spaces

As noted in Lab 1, the SPARC uses separate address spaces for data and text (code). In
fact, the SPARC provides (at least) four address spaces: the user instruction space, the
supervisor instruction space, the user data space, and the supervisor data space. When
the processor is in user state, instructions are fetched from the user instruction space while
data values are loaded from and stored to user data memory. Similarly, when the processor
is in supervisor mode, instructions are fetched from supervisor instruction space and data
values are, by default, loaded from and stored to supervisor data memory.
When the processor is in supervisor state, you can use special load and store instructions to access data values in alternate memory spaces. For examples, you can load a
value from the user data space, or store a value into the user instruction space. These instructions require an explicit address space indicator (ASI). Table 16.1 summarizes the ASI
values used for these instructions.

97

98

Lab 16. Traps

Table 16.1 Address Spaces on the SPARC

Address space
User Instructions
Supervisor Instructions
User Data
Supervisor Data

ASI
7
8
9
10

16.3.3 The ROM Code

16.4 Review Questions

16.5 Exercises
1. Write a trap handler that will print a NULL terminated string in the user data space.
(Because the string is in the user data space and not in supervisor data space, you
cannot use puts function in rom.s to implement this trap.) The starting address of
the string will be available in
The functionality provided by this trap is not absolutely necessary. Application programs could attain equivalent functionality using multiple invocations of trap 1 (putc).
What is the advantage of providing this as an separate trap?

Laboratory 17
Exceptions and Exception
Handling
17.1 Goal
To cover exception handling on the SPARC.

17.2 Objectives
After completing this lab, you will:

exception handlers.

17.3 Discussion
17.4 Summary
17.5 Review Questions
17.6 Exercises

99

100

Lab 17. Exceptions and Exception Handling

Laboratory 18
Interrupts and Interrupt Handling
18.1 Goal
To cover interrupts and interrupt handling on the SPARC.

18.2 Objectives
After completing this lab, you will:

interrupt handlers.

18.3 Discussion
18.4 Summary
18.5 Review Questions
18.6 Exercises

101

102

Lab 18. Interrupts and Interrupt Handling

Laboratory 19
Context Switching
19.1 Goal
To cover context switching on the SPARC.

19.2 Objectives
After completing this lab, you will:

multitasking.

19.3 Discussion
19.4 Summary
19.5 Review Questions
19.6 Exercises

103

104

Lab 19. Context Switching