ZeePedia

Comparison: Motorolla 68K Processors, Sun SPARC Processor

<< Interfacing with High Level Languages: Calling Conventions, Calling C from Assembly
img
17
Comparison with Other
Processors
We emphasized that assembly language has to be learned once and every
processor can be programmed by that person. To give a flavour of two
different widely popular processors we introduce the Motorolla 68K series
and the Sun SPARC processors. The Motorolla 68K processors are very
popular in high performance embedded applications while the Sun SPARC
processors are popular in very high end enterprise servers. We will compare
them with the Intel x86 series which is known for its success in the desktop
market.
17.1. MOTOROLLA 68K PROCESSORS
Motorolla 68K processors are very similar to Intel x86 series in their
architecture and instruction set. The both are of the same era and added
various features at the same time. The instructions are very similar however
the difference in architecture evident from a programmer's point of view must
be understood.
68K processors have 16 23bit general purpose registers named from A0-A7
and D0-D7. A0-A7 can hold addresses in indirect memory accesses. These
can also be used as software stack pointers. Stack in 68K is not as rigit a
structure as it is in x86. There is a 32bit program counter (PC) that holds the
address of currently executing instruction. The 8bit condition code register
(CCR) holds the X (Extend) N (Negative) Z (Zero) V (Overflow) C (Carry) flags.
X is set to C for extended operations (addition, subtraction, or shifting).
Motrolla processors allow bit addressing, that is a specific bit in a byte or a
bit field, i.e. a number of bits can be directly accessed. This is a very useful
feature especially in control applications. Other data types include byte,
word, long word, and quad word. A special MOVE16 instruction also accepts
a 16byte block.
68K allows indirect memory access using any A register. A special memory
access allows post increment or predecrement as part of memory access.
These forms are written as (An), (An)+, and ­(An). Other forms allow
addressing with another regiser as index and with constant displacement.
Using one of the A registers as the stack pointer and using the post
increment and pre decrement forms of addressing, stack is implemented.
Immediates can also be given as arguments and are preceded with a hash
sign (#). Addressing is indicated with parenthesis instead of brackets.
68K has no segmentation; it however has a paged memory model. It used
the big endian format in contrast to the little endian used by the Intel
processors. It has varying instruction lengths from 1-11 words. It has a
decrementing stack just like the Intel one. The format of instructions is
"operation source, destination" which is different from the Intel order of
operands. Some instructions from various instruction groups are given
below.
Data Movement
EXG D0, D2
MOVE.B (A1), (A2)
img
Computer Architecture & Assembly Language Programming
Course Code: CS401
CS401@vu.edu.pk
MOVEA (2222).L, A4
MOVEQ #12, D7
Arithmetic
ADD D7, (A4)
CLR (A3)
(set to zero)
CMP (A2), D1
ASL, ASR, LSL, LSR, ROR, ROL, ROXL, ROXR (shift operations)
Program Control
BRA label
JMP (A3)
BSR label
(CALL)
JSR (A2)
(indirect call)
RTD #4
(RET N)
Conditional Branch
BCC
(branch if carry clear)
BLS
(branch if Lower or Same)
BLT
(branch if Less Than)
BEQ
(branch if Equal)
BVC
(branch if Overflow clear)
17.2. SUN SPARC PROCESSOR
The Sun SPARC is a very popular processing belonging to the RISC
(reduced instruction set computer) family of processors. RISC processors
originally named because of the very few rudimentary instructions they
provided, are now providing almost as many instruction as CISC (complex
instruction set computer). However some properties like a fixed instruction
size and single clock execution for most instructions are there.
SPARC stands for Scalable Processor ARChitecture. SPARC is a 64bit
processor. It byte order is user settable and even on a per program basis. So
one program may be using little endian byte order and another may be using
big endian at the same time. Data types include byte, Halfword, Word (32bit),
and Double Word (64bits) and Quadword. It has a fixed 32bit instruction
size. It has a concept of ASI (Address Space Identifier); an 8bit number that
works similar to a segment.
There are 8 global registers and 8 alternate global registers. One of them is
active at a time and accessible as g0-g7. Apart from that it has 8 in registers
(i0-i7), 8 local registers (l0-l7), and 8 out registers (o0-o7). All registers are
64bit in size. The global registers can also be called r0-r7, in registers as r8-
r15, local registers as r16-r23, and out registers as r24-r31.
SPARC introduces a concept of register window. One window is 24
registers and the active window is pointed to by a special register called
Current Window Pointer (CWP). The actual number of registers in the
processor is in hundreds not restricted by the architecture definition. Two
instruction SAVE and RESTORE move this register window forward and
backward by 16 registers. Therefore one SAVE instruction makes the out
register the in registers and brings in new local and out registers. A
RESTORE instruction makes the in registers out registers and restores the
old local and in registers. This way parameters passing and returning can be
totally done in registers and there is no need to save and restore registers
inside subroutines.
The register o6 is conventionally used as the stack pointer. Return address
is stored in o7 by the CALL instruction. The register g0 (r0) is always 0 so
loading 0 in a register is made easy. SPARC is a totally register based
architecture, or it is called a load-store architecture where memory access is
only allowed in data movement instruction. Rest of the operations must be
done on registers.
184
img
Computer Architecture & Assembly Language Programming
Course Code: CS401
CS401@vu.edu.pk
SPARC instructions have two sources and a distinct destination. This
allows more flexibility in writing programs. Some examples of instructions of
this processor follow.
Data Movement
LDSB [rn], rn
(load signed byte)
LDUW [rn], rn
(load unsigned word)
STH [rn], rn
(store half word)
Arithmetic
source1 = rn
source2 = rn or simm13
dest = rn
ADD r2, r3, r4
SUB r2, 4000, r5
SLL, SRA, SRL
(shifting)
AND, OR, XOR
(logical)
Program Control
CALL
(direct call)
JMPL
(register indirect)
RET
SAVE
RESTORE
BA label
(Branch Always)
BE label
(branch if equal)
BCC label
(branch if carry clear)
BLE label
(branch if less or equal)
BVS label
(branch if overflow set)
185