OpenCores
URL https://opencores.org/ocsvn/zipcpu/zipcpu/trunk

Subversion Repositories zipcpu

[/] [zipcpu/] [trunk/] [doc/] [src/] [spec.tex] - Blame information for rev 37

Go to most recent revision | Details | Compare with Previous | View Log

Line No. Rev Author Line
1 21 dgisselq
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2
%%
3
%% Filename:    spec.tex
4
%%
5
%% Project:     Zip CPU -- a small, lightweight, RISC CPU soft core
6
%%
7
%% Purpose:     This LaTeX file contains all of the documentation/description
8 33 dgisselq
%%              currently provided with this Zip CPU soft core.  It supersedes
9 21 dgisselq
%%              any information about the instruction set or CPUs found
10
%%              elsewhere.  It's not nearly as interesting, though, as the PDF
11
%%              file it creates, so I'd recommend reading that before diving
12
%%              into this file.  You should be able to find the PDF file in
13
%%              the SVN distribution together with this PDF file and a copy of
14
%%              the GPL-3.0 license this file is distributed under.  If not,
15
%%              just type 'make' in the doc directory and it (should) build
16
%%              without a problem.
17
%%
18
%%
19
%% Creator:     Dan Gisselquist
20
%%              Gisselquist Technology, LLC
21
%%
22
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
23
%%
24
%% Copyright (C) 2015, Gisselquist Technology, LLC
25
%%
26
%% This program is free software (firmware): you can redistribute it and/or
27
%% modify it under the terms of  the GNU General Public License as published
28
%% by the Free Software Foundation, either version 3 of the License, or (at
29
%% your option) any later version.
30
%%
31
%% This program is distributed in the hope that it will be useful, but WITHOUT
32
%% ANY WARRANTY; without even the implied warranty of MERCHANTIBILITY or
33
%% FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
34
%% for more details.
35
%%
36
%% You should have received a copy of the GNU General Public License along
37
%% with this program.  (It's in the $(ROOT)/doc directory, run make with no
38
%% target there if the PDF file isn't present.)  If not, see
39
%% <http://www.gnu.org/licenses/> for a copy.
40
%%
41
%% License:     GPL, v3, as defined and found on www.gnu.org,
42
%%              http://www.gnu.org/licenses/gpl.html
43
%%
44
%%
45
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
46
\documentclass{gqtekspec}
47
\project{Zip CPU}
48
\title{Specification}
49
\author{Dan Gisselquist, Ph.D.}
50
\email{dgisselq (at) opencores.org}
51 36 dgisselq
\revision{Rev.~0.4}
52
\definecolor{webred}{rgb}{0.2,0,0}
53
\definecolor{webgreen}{rgb}{0,0.2,0}
54
\usepackage[dvips,ps2pdf,colorlinks=true,
55
        anchorcolor=black,pagecolor=webgreen,pdfpagelabels,hypertexnames,
56
        pdfauthor={Dan Gisselquist},
57
        pdfsubject={Zip CPU}]{hyperref}
58 21 dgisselq
\begin{document}
59
\pagestyle{gqtekspecplain}
60
\titlepage
61
\begin{license}
62
Copyright (C) \theyear\today, Gisselquist Technology, LLC
63
 
64
This project is free software (firmware): you can redistribute it and/or
65
modify it under the terms of  the GNU General Public License as published
66
by the Free Software Foundation, either version 3 of the License, or (at
67
your option) any later version.
68
 
69
This program is distributed in the hope that it will be useful, but WITHOUT
70
ANY WARRANTY; without even the implied warranty of MERCHANTIBILITY or
71
FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
72
for more details.
73
 
74
You should have received a copy of the GNU General Public License along
75
with this program.  If not, see \hbox{<http://www.gnu.org/licenses/>} for a
76
copy.
77
\end{license}
78
\begin{revisionhistory}
79 36 dgisselq
0.4 & 9/19/2015 & Gisselquist & Added DMA controller, improved stall information, and self--assessment info.\\\hline
80 33 dgisselq
0.3 & 8/22/2015 & Gisselquist & First completed draft\\\hline
81 24 dgisselq
0.2 & 8/19/2015 & Gisselquist & Still Draft, more complete \\\hline
82 21 dgisselq
0.1 & 8/17/2015 & Gisselquist & Incomplete First Draft \\\hline
83
\end{revisionhistory}
84
% Revision History
85
% Table of Contents, named Contents
86
\tableofcontents
87 24 dgisselq
\listoffigures
88 21 dgisselq
\listoftables
89
\begin{preface}
90
Many people have asked me why I am building the Zip CPU. ARM processors are
91
good and effective. Xilinx makes and markets Microblaze, Altera Nios, and both
92
have better toolsets than the Zip CPU will ever have. OpenRISC is also
93 24 dgisselq
available, RISC--V may be replacing it. Why build a new processor?
94 21 dgisselq
 
95
The easiest, most obvious answer is the simple one: Because I can.
96
 
97
There's more to it, though. There's a lot that I would like to do with a
98
processor, and I want to be able to do it in a vendor independent fashion.
99 36 dgisselq
First, I would like to be able to place this processor inside an FPGA.  Without
100
paying royalties, ARM is out of the question.  I would then like to be able to
101
generate Verilog code, both for the processor and the system it sits within,
102
that can run equivalently on both Xilinx and Altera chips, and that can be
103
easily ported from one manufacturer's chipsets to another. Even more, before
104
purchasing a chip or a board, I would like to know that my soft core works. I
105
would like to build a test bench to test components with, and Verilator is my
106
chosen test bench. This forces me to use all Verilog, and it prevents me from
107
using any proprietary cores. For this reason, Microblaze and Nios are out of
108
the question.
109 21 dgisselq
 
110
Why not OpenRISC? That's a hard question. The OpenRISC team has done some
111
wonderful work on an amazing processor, and I'll have to admit that I am
112
envious of what they've accomplished. I would like to port binutils to the
113
Zip CPU, as I would like to port GCC and GDB. They are way ahead of me. The
114
OpenRISC processor, however, is complex and hefty at about 4,500 LUTs. It has
115
a lot of features of modern CPUs within it that ... well, let's just say it's
116
not the little guy on the block. The Zip CPU is lighter weight, costing only
117 32 dgisselq
about 2,300 LUTs with no peripherals, and 3,200 LUTs with some very basic
118 21 dgisselq
peripherals.
119
 
120
My final reason is that I'm building the Zip CPU as a learning experience. The
121
Zip CPU has allowed me to learn a lot about how CPUs work on a very micro
122
level. For the first time, I am beginning to understand many of the Computer
123
Architecture lessons from years ago.
124
 
125
To summarize: Because I can, because it is open source, because it is light
126
weight, and as an exercise in learning.
127
 
128
\end{preface}
129
 
130
\chapter{Introduction}
131
\pagenumbering{arabic}
132
\setcounter{page}{1}
133
 
134
 
135 36 dgisselq
The original goal of the Zip CPU was to be a very simple CPU.   You might
136 21 dgisselq
think of it as a poor man's alternative to the OpenRISC architecture.
137
For this reason, all instructions have been designed to be as simple as
138
possible, and are all designed to be executed in one instruction cycle per
139
instruction, barring pipeline stalls.  Indeed, even the bus has been simplified
140
to a constant 32-bit width, with no option for more or less.  This has
141
resulted in the choice to drop push and pop instructions, pre-increment and
142
post-decrement addressing modes, and more.
143
 
144
For those who like buzz words, the Zip CPU is:
145
\begin{itemize}
146
\item A 32-bit CPU: All registers are 32-bits, addresses are 32-bits,
147
                instructions are 32-bits wide, etc.
148 24 dgisselq
\item A RISC CPU.  There is no microcode for executing instructions.  All
149
        instructions are designed to be completed in one clock cycle.
150 21 dgisselq
\item A Load/Store architecture.  (Only load and store instructions
151
                can access memory.)
152
\item Wishbone compliant.  All peripherals are accessed just like
153
                memory across this bus.
154
\item A Von-Neumann architecture.  (The instructions and data share a
155
                common bus.)
156
\item A pipelined architecture, having stages for {\bf Prefetch},
157
                {\bf Decode}, {\bf Read-Operand}, the {\bf ALU/Memory}
158 24 dgisselq
                unit, and {\bf Write-back}.  See Fig.~\ref{fig:cpu}
159
\begin{figure}\begin{center}
160
\includegraphics[width=3.5in]{../gfx/cpu.eps}
161
\caption{Zip CPU internal pipeline architecture}\label{fig:cpu}
162
\end{center}\end{figure}
163
                for a diagram of this structure.
164 21 dgisselq
\item Completely open source, licensed under the GPL.\footnote{Should you
165
        need a copy of the Zip CPU licensed under other terms, please
166
        contact me.}
167
\end{itemize}
168
 
169
Now, however, that I've worked on the Zip CPU for a while, it is not nearly
170
as simple as I originally hoped.  Worse, I've had to adjust to create
171
capabilities that I was never expecting to need.  These include:
172
\begin{itemize}
173 33 dgisselq
\item {\bf External Debug:} Once placed upon an FPGA, some external means is
174 21 dgisselq
        still necessary to debug this CPU.  That means that there needs to be
175
        an external register that can control the CPU: reset it, halt it, step
176 24 dgisselq
        it, and tell whether it is running or not.  My chosen interface
177
        includes a second register similar to this control register.  This
178
        second register allows the external controller or debugger to examine
179 21 dgisselq
        registers internal to the CPU.
180
 
181
\item {\bf Internal Debug:} Being able to run a debugger from within
182
        a user process requires an ability to step a user process from
183
        within a debugger.  It also requires a break instruction that can
184
        be substituted for any other instruction, and substituted back.
185
        The break is actually difficult: the break instruction cannot be
186
        allowed to execute.  That way, upon a break, the debugger should
187
        be able to jump back into the user process to step the instruction
188
        that would've been at the break point initially, and then to
189
        replace the break after passing it.
190
 
191 24 dgisselq
        Incidentally, this break messes with the prefetch cache and the
192
        pipeline: if you change an instruction partially through the pipeline,
193
        the whole pipeline needs to be cleansed.  Likewise if you change
194
        an instruction in memory, you need to make sure the cache is reloaded
195
        with the new instruction.
196
 
197 21 dgisselq
\item {\bf Prefetch Cache:} My original implementation had a very
198
        simple prefetch stage.  Any time the PC changed the prefetch would go
199
        and fetch the new instruction.  While this was perhaps this simplest
200
        approach, it cost roughly five clocks for every instruction.  This
201
        was deemed unacceptable, as I wanted a CPU that could execute
202
        instructions in one cycle.  I therefore have a prefetch cache that
203
        issues pipelined wishbone accesses to memory and then pushes
204
        instructions at the CPU.  Sadly, this accounts for about 20\% of the
205
        logic in the entire CPU, or 15\% of the logic in the entire system.
206
 
207
 
208
\item {\bf Operating System:} In order to support an operating system,
209
        interrupts and so forth, the CPU needs to support supervisor and
210
        user modes, as well as a means of switching between them.  For example,
211
        the user needs a means of executing a system call.  This is the
212
        purpose of the {\bf `trap'} instruction.  This instruction needs to
213
        place the CPU into supervisor mode (here equivalent to disabling
214
        interrupts), as well as handing it a parameter such as identifying
215
        which O/S function was called.
216
 
217 24 dgisselq
My initial approach to building a trap instruction was to create an external
218
peripheral which, when written to, would generate an interrupt and could
219
return the last value written to it.  In practice, this approach didn't work
220
at all: the CPU executed two instructions while waiting for the
221
trap interrupt to take place.  Since then, I've decided to keep the rest of
222
the CC register for that purpose so that a write to the CC register, with the
223
GIE bit cleared, could be used to execute a trap.  This has other problems,
224
though, primarily in the limitation of the uses of the CC register.  In
225
particular, the CC register is the best place to put CPU state information and
226
to ``announce'' special CPU features (floating point, etc).  So the trap
227
instruction still switches to interrupt mode, but the CC register is not
228
nearly as useful for telling the supervisor mode processor what trap is being
229
executed.
230 21 dgisselq
 
231
Modern timesharing systems also depend upon a {\bf Timer} interrupt
232 24 dgisselq
to handle task swapping.  For the Zip CPU, this interrupt is handled
233
external to the CPU as part of the CPU System, found in {\tt zipsystem.v}.
234
The timer module itself is found in {\tt ziptimer.v}.
235 21 dgisselq
 
236
\item {\bf Pipeline Stalls:} My original plan was to not support pipeline
237
        stalls at all, but rather to require the compiler to properly schedule
238 24 dgisselq
        all instructions so that stalls would never be necessary.  After trying
239 21 dgisselq
        to build such an architecture, I gave up, having learned some things:
240
 
241
        For example, in  order to facilitate interrupt handling and debug
242
        stepping, the CPU needs to know what instructions have finished, and
243
        which have not.  In other words, it needs to know where it can restart
244
        the pipeline from.  Once restarted, it must act as though it had
245 24 dgisselq
        never stopped.  This killed my idea of delayed branching, since what
246
        would be the appropriate program counter to restart at?  The one the
247
        CPU was going to branch to, or the ones in the delay slots?  This
248
        also makes the idea of compressed instruction codes difficult, since,
249
        again, where do you restart on interrupt?
250 21 dgisselq
 
251
        So I switched to a model of discrete execution: Once an instruction
252
        enters into either the ALU or memory unit, the instruction is
253
        guaranteed to complete.  If the logic recognizes a branch or a
254
        condition that would render the instruction entering into this stage
255 33 dgisselq
        possibly inappropriate (i.e. a conditional branch preceding a store
256 21 dgisselq
        instruction for example), then the pipeline stalls for one cycle
257
        until the conditional branch completes.  Then, if it generates a new
258 33 dgisselq
        PC address, the stages preceding are all wiped clean.
259 21 dgisselq
 
260
        The discrete execution model allows such things as sleeping: if the
261 24 dgisselq
        CPU is put to ``sleep,'' the ALU and memory stages stall and back up
262 21 dgisselq
        everything before them.  Likewise, anything that has entered the ALU
263
        or memory stage when the CPU is placed to sleep continues to completion.
264
        To handle this logic, each pipeline stage has three control signals:
265
        a valid signal, a stall signal, and a clock enable signal.  In
266
        general, a stage stalls if it's contents are valid and the next step
267
        is stalled.  This allows the pipeline to fill any time a later stage
268
        stalls.
269
 
270 24 dgisselq
        This approach is also different from other pipeline approaches.  Instead
271
        of keeping the entire pipeline filled, each stage is treated
272
        independently.  Therefore, individual stages may move forward as long
273
        as the subsequent stage is available, regardless of whether the stage
274
        behind it is filled.
275
 
276 21 dgisselq
\item {\bf Verilog Modules:} When examining how other processors worked
277
        here on open cores, many of them had one separate module per pipeline
278
        stage.  While this appeared to me to be a fascinating and commendable
279
        idea, my own implementation didn't work out quite so nicely.
280
 
281
        As an example, the decode module produces a {\em lot} of
282
        control wires and registers.  Creating a module out of this, with
283
        only the simplest of logic within it, seemed to be more a lesson
284
        in passing wires around, rather than encapsulating logic.
285
 
286
        Another example was the register writeback section.  I would love
287
        this section to be a module in its own right, and many have made them
288
        such.  However, other modules depend upon writeback results other
289
        than just what's placed in the register (i.e., the control wires).
290
        For these reasons, I didn't manage to fit this section into it's
291
        own module.
292
 
293
        The result is that the majority of the CPU code can be found in
294
        the {\tt zipcpu.v} file.
295
\end{itemize}
296
 
297
With that introduction out of the way, let's move on to the instruction
298
set.
299
 
300
\chapter{CPU Architecture}\label{chap:arch}
301
 
302 24 dgisselq
The Zip CPU supports a set of two operand instructions, where the second operand
303 21 dgisselq
(always a register) is the result.  The only exception is the store instruction,
304
where the first operand (always a register) is the source of the data to be
305
stored.
306
 
307 24 dgisselq
\section{Simplified Bus}
308
The bus architecture of the Zip CPU is that of a simplified WISHBONE bus.
309
It has been simplified in this fashion: all operations are 32--bit operations.
310 36 dgisselq
The bus is neither little endian nor big endian.  For this reason, all words
311 24 dgisselq
are 32--bits.  All instructions are also 32--bits wide.  Everything has been
312
built around the 32--bit word.
313
 
314 21 dgisselq
\section{Register Set}
315
The Zip CPU supports two sets of sixteen 32-bit registers, a supervisor
316 24 dgisselq
and a user set as shown in Fig.~\ref{fig:regset}.
317
\begin{figure}\begin{center}
318
\includegraphics[width=3.5in]{../gfx/regset.eps}
319
\caption{Zip CPU Register File}\label{fig:regset}
320
\end{center}\end{figure}
321
The supervisor set is used in interrupt mode when interrupts are disabled,
322
whereas the user set is used otherwise.  Of this register set, the Program
323
Counter (PC) is register 15, whereas the status register (SR) or condition
324
code register
325 21 dgisselq
(CC) is register 14.  By convention, the stack pointer will be register 13 and
326 24 dgisselq
noted as (SP)--although there is nothing special about this register other
327
than this convention.
328 21 dgisselq
The CPU can access both register sets via move instructions from the
329
supervisor state, whereas the user state can only access the user registers.
330
 
331 36 dgisselq
The status register is special, and bears further mention.  As shown in
332
Fig.~\ref{tbl:cc-register},
333
\begin{table}\begin{center}
334
\begin{bitlist}
335
31\ldots 11 & R/W & Reserved for future uses\\\hline
336
10 & R & (Reserved for) Bus-Error Flag\\\hline
337
9 & R & Trap, or user interrupt, Flag.  Cleared on return to userspace.\\\hline
338
8 & R & (Reserved for) Illegal Instruction Flag\\\hline
339
7 & R/W & Break--Enable\\\hline
340
6 & R/W & Step\\\hline
341
5 & R/W & Global Interrupt Enable (GIE)\\\hline
342
4 & R/W & Sleep.  When GIE is also set, the CPU waits for an interrupt.\\\hline
343
3 & R/W & Overflow\\\hline
344
2 & R/W & Negative.  The sign bit was set as a result of the last ALU instruction.\\\hline
345
1 & R/W & Carry\\\hline
346
 
347
\end{bitlist}
348
\caption{Condition Code Register Bit Assignment}\label{tbl:cc-register}
349
\end{center}\end{table}
350
the lower 11~bits of the status register form
351
a set of CPU state and condition codes.  Writes to other bits of this register
352
are preserved.
353 21 dgisselq
 
354 33 dgisselq
Of the condition codes, the bottom four bits are the current flags:
355 21 dgisselq
                Zero (Z),
356
                Carry (C),
357
                Negative (N),
358
                and Overflow (V).
359
 
360
The next bit is a clock enable (0 to enable) or sleep bit (1 to put
361
        the CPU to sleep).  Setting this bit will cause the CPU to
362
        wait for an interrupt (if interrupts are enabled), or to
363
        completely halt (if interrupts are disabled).
364 33 dgisselq
 
365 21 dgisselq
The sixth bit is a global interrupt enable bit (GIE).  When this
366 32 dgisselq
        sixth bit is a `1' interrupts will be enabled, else disabled.  When
367 21 dgisselq
        interrupts are disabled, the CPU will be in supervisor mode, otherwise
368
        it is in user mode.  Thus, to execute a context switch, one only
369
        need enable or disable interrupts.  (When an interrupt line goes
370
        high, interrupts will automatically be disabled, as the CPU goes
371 32 dgisselq
        and deals with its context switch.)  Special logic has been added to
372
        keep the user mode from setting the sleep register and clearing the
373
        GIE register at the same time, with clearing the GIE register taking
374
        precedence.
375 21 dgisselq
 
376
The seventh bit is a step bit.  This bit can be
377
        set from supervisor mode only.  After setting this bit, should
378
        the supervisor mode process switch to user mode, it would then
379
        accomplish one instruction in user mode before returning to supervisor
380
        mode.  Then, upon return to supervisor mode, this bit will
381
        be automatically cleared.  This bit has no effect on the CPU while in
382
        supervisor mode.
383
 
384
        This functionality was added to enable a userspace debugger
385
        functionality on a user process, working through supervisor mode
386
        of course.
387
 
388
 
389 24 dgisselq
The eighth bit is a break enable bit.  This controls whether a break
390
instruction in user mode will halt the processor for an external debugger
391
(break enabled), or whether the break instruction will simply send send the
392
CPU into interrupt mode.  Encountering a break in supervisor mode will
393
halt the CPU independent of the break enable bit.  This bit can only be set
394
within supervisor mode.
395 21 dgisselq
 
396 32 dgisselq
% Should break enable be a supervisor mode bit, while the break enable bit
397
% in user mode is a break has taken place bit?
398
%
399
 
400 21 dgisselq
This functionality was added to enable an external debugger to
401
        set and manage breakpoints.
402
 
403 36 dgisselq
The ninth bit is reserved for an illegal instruction bit.  When the CPU
404
tries to execute either a non-existant instruction, or an instruction from
405
an address that produces a bus error, the CPU will (once implemented) switch
406
to supervisor mode while setting this bit.  The bit will automatically be
407
cleared upon any return to user mode.
408 21 dgisselq
 
409
The tenth bit is a trap bit.  It is set whenever the user requests a soft
410
interrupt, and cleared on any return to userspace command.  This allows the
411
supervisor, in supervisor mode, to determine whether it got to supervisor
412
mode from a trap or from an external interrupt or both.
413
 
414
\section{Conditional Instructions}
415 36 dgisselq
Most, although not quite all, instructions may be conditionally executed.  From
416 21 dgisselq
the four condition code flags, eight conditions are defined.  These are shown
417
in Tbl.~\ref{tbl:conditions}.
418
\begin{table}
419
\begin{center}
420
\begin{tabular}{l|l|l}
421
Code & Mneumonic & Condition \\\hline
422
3'h0 & None & Always execute the instruction \\
423
3'h1 & {\tt .Z} & Only execute when 'Z' is set \\
424
3'h2 & {\tt .NE} & Only execute when 'Z' is not set \\
425
3'h3 & {\tt .GE} & Greater than or equal ('N' not set, 'Z' irrelevant) \\
426
3'h4 & {\tt .GT} & Greater than ('N' not set, 'Z' not set) \\
427 24 dgisselq
3'h5 & {\tt .LT} & Less than ('N' set) \\
428 21 dgisselq
3'h6 & {\tt .C} & Carry set\\
429
3'h7 & {\tt .V} & Overflow set\\
430
\end{tabular}
431
\caption{Conditions for conditional operand execution}\label{tbl:conditions}
432
\end{center}
433
\end{table}
434 24 dgisselq
There is no condition code for less than or equal, not C or not V.  Sorry,
435 36 dgisselq
I ran out of space in 3--bits.  Conditioning on a non--supported condition
436
is still possible, but it will take an extra instruction and a pipeline stall.  (Ex: \hbox{\em (Stall)}; \hbox{\tt TST \$4,CC;} \hbox{\tt STO.NZ R0,(R1)})
437 21 dgisselq
 
438 36 dgisselq
Conditionally executed ALU instructions will not further adjust the
439
condition codes.
440
 
441 21 dgisselq
\section{Operand B}
442 24 dgisselq
Many instruction forms have a 21-bit source ``Operand B'' associated with them.
443 21 dgisselq
This Operand B is either equal to a register plus a signed immediate offset,
444
or an immediate offset by itself.  This value is encoded as shown in
445
Tbl.~\ref{tbl:opb}.
446
\begin{table}\begin{center}
447
\begin{tabular}{|l|l|l|}\hline
448
Bit 20 & 19 \ldots 16 & 15 \ldots 0 \\\hline
449 24 dgisselq
1'b0 & \multicolumn{2}{l|}{20--bit Signed Immediate value} \\\hline
450
1'b1 & 4-bit Register & 16--bit Signed immediate offset \\\hline
451 21 dgisselq
\end{tabular}
452
\caption{Bit allocation for Operand B}\label{tbl:opb}
453
\end{center}\end{table}
454 24 dgisselq
 
455 33 dgisselq
Sixteen and twenty bit immediate values don't make sense for all instructions.
456
For example, what is the point of a 20--bit immediate when executing a 16--bit
457 24 dgisselq
multiply?  Likewise, why have a 16--bit immediate when adding to a logical
458
or arithmetic shift?  In these cases, the extra bits are reserved for future
459
instruction possibilities.
460
 
461 21 dgisselq
\section{Address Modes}
462 36 dgisselq
The Zip CPU supports two addressing modes: register plus immediate, and
463 21 dgisselq
immediate address.  Addresses are therefore encoded in the same fashion as
464
Operand B's, shown above.
465
 
466
A lot of long hard thought was put into whether to allow pre/post increment
467
and decrement addressing modes.  Finding no way to use these operators without
468 32 dgisselq
taking two or more clocks per instruction,\footnote{The two clocks figure
469
comes from the design of the register set, allowing only one write per clock.
470
That write is either from the memory unit or the ALU, but never both.} these
471
addressing modes have been
472 21 dgisselq
removed from the realm of possibilities.  This means that the Zip CPU has no
473
native way of executing push, pop, return, or jump to subroutine operations.
474 24 dgisselq
Each of these instructions can be emulated with a set of instructions from the
475
existing set.
476 21 dgisselq
 
477
\section{Move Operands}
478
The previous set of operands would be perfect and complete, save only that
479 24 dgisselq
the CPU needs access to non--supervisory registers while in supervisory mode.
480
Therefore, the MOV instruction is special and offers access to these registers
481
\ldots when in supervisory mode.  To keep the compiler simple, the extra bits
482
are ignored in non-supervisory mode (as though they didn't exist), rather than
483
being mapped to new instructions or additional capabilities.  The bits
484
indicating which register set each register lies within are the A-Usr and
485
B-Usr bits.  When set to a one, these refer to a user mode register.  When set
486
to a zero, these refer to a register in the current mode, whether user or
487
supervisor.  Further, because a load immediate instruction exists, there is no
488
move capability between an immediate and a register: all moves come from either
489
a register or a register plus an offset.
490 21 dgisselq
 
491 24 dgisselq
This actually leads to a bit of a problem: since the MOV instruction encodes
492
which register set each register is coming from or moving to, how shall a
493
compiler or assembler know how to compile a MOV instruction without knowing
494
the mode of the CPU at the time?  For this reason, the compiler will assume
495
all MOV registers are supervisor registers, and display them as normal.
496
Anything with the user bit set will be treated as a user register.  The CPU
497
will quietly ignore the supervisor bits while in user mode, and anything
498 36 dgisselq
marked as a user register will always be valid.
499 21 dgisselq
 
500
\section{Multiply Operations}
501 36 dgisselq
The Zip CPU supports two Multiply operations, a 16x16 bit signed multiply
502
({\tt MPYS}) and a 16x16 bit unsigned multiply ({\tt MPYU}).  In both
503 21 dgisselq
cases, the operand is a register plus a 16-bit immediate, subject to the
504
rule that the register cannot be the PC or CC registers.  The PC register
505
field has been stolen to create a multiply by immediate instruction.  The
506
CC register field is reserved.
507
 
508
\section{Floating Point}
509 36 dgisselq
The Zip CPU does not (yet) support floating point operations.  However, the
510 32 dgisselq
instruction set reserves two possibilities for future floating point
511
operations.
512 21 dgisselq
 
513 32 dgisselq
The first floating point operation hole in the instruction set involves
514 36 dgisselq
setting a proposed (but non-existent) floating point bit in the CC register.
515
The next instruction
516
would then simply interpret its operands as floating point instructions.
517 32 dgisselq
Not all instructions, however, have floating point equivalents.  Further, the
518
immediate fields do not apply in floating point mode, and must be set to
519
zero.  Not all instructions make sense as floating point operations.
520
Therefore, only the CMP, SUB, ADD, and MPY instructions may be issued as
521
floating point instructions.  Other instructions allow the examining of the
522
floating point bit in the CC register.  In all cases, the floating point bit
523
is cleared one instruction after it is set.
524 21 dgisselq
 
525 32 dgisselq
The other possibility for floating point operations involves exploiting the
526
hole in the instruction set that the NOOP and BREAK instructions reside within.
527 36 dgisselq
These two instructions use 24--bits of address space, when only a single bit
528
is necessary.  A simple adjustment to this space could create instructions
529
with 4--bit register addresses for each register, a 3--bit field for
530
conditional execution, and a 2--bit field for which operation.
531
In this fashion, such a floating point capability would only fill 13--bits of
532
the 24--bit field, still leaving lots of room for expansion.
533 32 dgisselq
 
534
In both cases, the Zip CPU would support 32--bit single precision floats
535 36 dgisselq
only, since other choices would complicate the pipeline.
536 32 dgisselq
 
537
The current architecture does not support a floating point not-implemented
538
interrupt.  Any soft floating point emulation must be done deliberately.
539
 
540 21 dgisselq
\section{Native Instructions}
541
The instruction set for the Zip CPU is summarized in
542
Tbl.~\ref{tbl:zip-instructions}.
543
\begin{table}\begin{center}
544
\begin{tabular}{|l|l|l|l|l|l|l|l|l|l|l|l|l|l|l|l|l|l|l|l|l|l|l|l|l|l|l|l|l|l|l|l|l|c|}\hline
545 36 dgisselq
\rowcolor[gray]{0.85}
546 21 dgisselq
Op Code & \multicolumn{8}{c|}{31\ldots24} & \multicolumn{8}{c|}{23\ldots 16}
547
        & \multicolumn{8}{c|}{15\ldots 8} & \multicolumn{8}{c|}{7\ldots 0}
548 36 dgisselq
        & Sets CC? \\\hline\hline
549 21 dgisselq
CMP(Sub) & \multicolumn{4}{l|}{4'h0}
550
                & \multicolumn{4}{l|}{D. Reg}
551
                & \multicolumn{3}{l|}{Cond.}
552
                & \multicolumn{21}{l|}{Operand B}
553
                & Yes \\\hline
554 24 dgisselq
TST(And) & \multicolumn{4}{l|}{4'h1}
555 21 dgisselq
                & \multicolumn{4}{l|}{D. Reg}
556
                & \multicolumn{3}{l|}{Cond.}
557
                & \multicolumn{21}{l|}{Operand B}
558
        & Yes \\\hline
559
MOV & \multicolumn{4}{l|}{4'h2}
560
                & \multicolumn{4}{l|}{D. Reg}
561
                & \multicolumn{3}{l|}{Cond.}
562
                & A-Usr
563
                & \multicolumn{4}{l|}{B-Reg}
564
                & B-Usr
565
                & \multicolumn{15}{l|}{15'bit signed offset}
566
                & \\\hline
567
LODI & \multicolumn{4}{l|}{4'h3}
568
                & \multicolumn{4}{l|}{R. Reg}
569
                & \multicolumn{24}{l|}{24'bit Signed Immediate}
570
                & \\\hline
571
NOOP & \multicolumn{4}{l|}{4'h4}
572
                & \multicolumn{4}{l|}{4'he}
573
                & \multicolumn{24}{l|}{24'h00}
574
                & \\\hline
575
BREAK & \multicolumn{4}{l|}{4'h4}
576
                & \multicolumn{4}{l|}{4'he}
577
                & \multicolumn{24}{l|}{24'h01}
578
                & \\\hline
579 36 dgisselq
{\em Reserved} & \multicolumn{4}{l|}{4'h4}
580 21 dgisselq
                & \multicolumn{4}{l|}{4'he}
581
                & \multicolumn{24}{l|}{24'bits, but not 0 or 1.}
582
                & \\\hline
583
LODIHI & \multicolumn{4}{l|}{4'h4}
584
                & \multicolumn{4}{l|}{4'hf}
585
                & \multicolumn{3}{l|}{Cond.}
586
                & 1'b1
587
                & \multicolumn{4}{l|}{R. Reg}
588
                & \multicolumn{16}{l|}{16-bit Immediate}
589
                & \\\hline
590
LODILO & \multicolumn{4}{l|}{4'h4}
591
                & \multicolumn{4}{l|}{4'hf}
592
                & \multicolumn{3}{l|}{Cond.}
593
                & 1'b0
594
                & \multicolumn{4}{l|}{R. Reg}
595
                & \multicolumn{16}{l|}{16-bit Immediate}
596
                & \\\hline
597
16-b MPYU & \multicolumn{4}{l|}{4'h4}
598
                & \multicolumn{4}{l|}{R. Reg}
599
                & \multicolumn{3}{l|}{Cond.}
600
                & 1'b0 & \multicolumn{4}{l|}{Reg}
601
                & \multicolumn{16}{l|}{16-bit Offset}
602
                & Yes \\\hline
603
16-b MPYU(I) & \multicolumn{4}{l|}{4'h4}
604
                & \multicolumn{4}{l|}{R. Reg}
605
                & \multicolumn{3}{l|}{Cond.}
606
                & 1'b0 & \multicolumn{4}{l|}{4'hf}
607
                & \multicolumn{16}{l|}{16-bit Offset}
608
                & Yes \\\hline
609
16-b MPYS & \multicolumn{4}{l|}{4'h4}
610
                & \multicolumn{4}{l|}{R. Reg}
611
                & \multicolumn{3}{l|}{Cond.}
612
                & 1'b1 & \multicolumn{4}{l|}{Reg}
613
                & \multicolumn{16}{l|}{16-bit Offset}
614
                & Yes \\\hline
615
16-b MPYS(I) & \multicolumn{4}{l|}{4'h4}
616
                & \multicolumn{4}{l|}{R. Reg}
617
                & \multicolumn{3}{l|}{Cond.}
618
                & 1'b1 & \multicolumn{4}{l|}{4'hf}
619
                & \multicolumn{16}{l|}{16-bit Offset}
620
                & Yes \\\hline
621
ROL & \multicolumn{4}{l|}{4'h5}
622
                & \multicolumn{4}{l|}{R. Reg}
623
                & \multicolumn{3}{l|}{Cond.}
624
                & \multicolumn{21}{l|}{Operand B, truncated to low order 5 bits}
625
                & \\\hline
626
LOD & \multicolumn{4}{l|}{4'h6}
627
                & \multicolumn{4}{l|}{R. Reg}
628
                & \multicolumn{3}{l|}{Cond.}
629
                & \multicolumn{21}{l|}{Operand B address}
630
                & \\\hline
631
STO & \multicolumn{4}{l|}{4'h7}
632
                & \multicolumn{4}{l|}{D. Reg}
633
                & \multicolumn{3}{l|}{Cond.}
634
                & \multicolumn{21}{l|}{Operand B address}
635
                & \\\hline
636
SUB & \multicolumn{4}{l|}{4'h8}
637
        &       \multicolumn{4}{l|}{R. Reg}
638
        &       \multicolumn{3}{l|}{Cond.}
639 32 dgisselq
        &       \multicolumn{21}{l|}{Operand B}
640 21 dgisselq
        & Yes \\\hline
641
AND & \multicolumn{4}{l|}{4'h9}
642
        &       \multicolumn{4}{l|}{R. Reg}
643
        &       \multicolumn{3}{l|}{Cond.}
644
        &       \multicolumn{21}{l|}{Operand B}
645
        & Yes \\\hline
646
ADD & \multicolumn{4}{l|}{4'ha}
647
        &       \multicolumn{4}{l|}{R. Reg}
648
        &       \multicolumn{3}{l|}{Cond.}
649
        &       \multicolumn{21}{l|}{Operand B}
650
        & Yes \\\hline
651
OR & \multicolumn{4}{l|}{4'hb}
652
        &       \multicolumn{4}{l|}{R. Reg}
653
        &       \multicolumn{3}{l|}{Cond.}
654
        &       \multicolumn{21}{l|}{Operand B}
655
        & Yes \\\hline
656
XOR & \multicolumn{4}{l|}{4'hc}
657
        &       \multicolumn{4}{l|}{R. Reg}
658
        &       \multicolumn{3}{l|}{Cond.}
659
        &       \multicolumn{21}{l|}{Operand B}
660
        & Yes \\\hline
661
LSL/ASL & \multicolumn{4}{l|}{4'hd}
662
        &       \multicolumn{4}{l|}{R. Reg}
663
        &       \multicolumn{3}{l|}{Cond.}
664 33 dgisselq
        &       \multicolumn{21}{l|}{Operand B, imm. truncated to 6 bits}
665 21 dgisselq
        & Yes \\\hline
666
ASR & \multicolumn{4}{l|}{4'he}
667
        &       \multicolumn{4}{l|}{R. Reg}
668
        &       \multicolumn{3}{l|}{Cond.}
669 33 dgisselq
        &       \multicolumn{21}{l|}{Operand B, imm. truncated to 6 bits}
670 21 dgisselq
        & Yes \\\hline
671
LSR & \multicolumn{4}{l|}{4'hf}
672
        &       \multicolumn{4}{l|}{R. Reg}
673
        &       \multicolumn{3}{l|}{Cond.}
674 33 dgisselq
        &       \multicolumn{21}{l|}{Operand B, imm. truncated to 6 bits}
675 21 dgisselq
        & Yes \\\hline
676
\end{tabular}
677
\caption{Zip CPU Instruction Set}\label{tbl:zip-instructions}
678
\end{center}\end{table}
679
 
680
As you can see, there's lots of room for instruction set expansion.  The
681 24 dgisselq
NOOP and BREAK instructions are the only instructions within one particular
682 36 dgisselq
24--bit hole.  The rest of this space is reserved for future enhancements.
683 21 dgisselq
 
684
\section{Derived Instructions}
685 36 dgisselq
The Zip CPU supports many other common instructions, but not all of them
686 24 dgisselq
are single cycle instructions.  The derived instruction tables,
687 36 dgisselq
Tbls.~\ref{tbl:derived-1}, \ref{tbl:derived-2}, \ref{tbl:derived-3}
688
and~\ref{tbl:derived-4},
689 21 dgisselq
help to capture some of how these other instructions may be implemented on
690 36 dgisselq
the Zip CPU.  Many of these instructions will have assembly equivalents,
691 21 dgisselq
such as the branch instructions, to facilitate working with the CPU.
692
\begin{table}\begin{center}
693
\begin{tabular}{p{1.4in}p{1.5in}p{3in}}\\\hline
694
Mapped & Actual  & Notes \\\hline
695 36 dgisselq
ABS Rx
696
        & \parbox[t]{1.5in}{TST -1,Rx\\NEG.LT Rx}
697
        & Absolute value, depends upon derived NEG.\\\hline
698 21 dgisselq
\parbox[t]{1.4in}{ADD Ra,Rx\\ADDC Rb,Ry}
699
        & \parbox[t]{1.5in}{Add Ra,Rx\\ADD.C \$1,Ry\\Add Rb,Ry}
700
        & Add with carry \\\hline
701
BRA.Cond +/-\$Addr
702 33 dgisselq
        & \hbox{MOV.cond \$Addr+PC,PC}
703 24 dgisselq
        & Branch or jump on condition.  Works for 15--bit
704
                signed address offsets.\\\hline
705 21 dgisselq
BRA.Cond +/-\$Addr
706
        & \parbox[t]{1.5in}{LDI \$Addr,Rx \\ ADD.cond Rx,PC}
707
        & Branch/jump on condition.  Works for
708
        23 bit address offsets, but costs a register, an extra instruction,
709 33 dgisselq
        and sets the flags. \\\hline
710 21 dgisselq
BNC PC+\$Addr
711
        & \parbox[t]{1.5in}{Test \$Carry,CC \\ MOV.Z PC+\$Addr,PC}
712
        & Example of a branch on an unsupported
713
                condition, in this case a branch on not carry \\\hline
714
BUSY & MOV \$-1(PC),PC & Execute an infinite loop \\\hline
715
CLRF.NZ Rx
716
        & XOR.NZ Rx,Rx
717
        & Clear Rx, and flags, if the Z-bit is not set \\\hline
718
CLR Rx
719
        & LDI \$0,Rx
720
        & Clears Rx, leaves flags untouched.  This instruction cannot be
721
                conditional. \\\hline
722
EXCH.W Rx
723
        & ROL \$16,Rx
724
        & Exchanges the top and bottom 16'bit words of Rx \\\hline
725
HALT
726
        & Or \$SLEEP,CC
727
        & Executed while in interrupt mode.  In user mode this is simply a
728 33 dgisselq
        wait until interrupt instruction. \\\hline
729 21 dgisselq
INT & LDI \$0,CC
730
        & Since we're using the CC register as a trap vector as well, this
731
        executes TRAP \#0. \\\hline
732
IRET
733
        & OR \$GIE,CC
734
        & Also an RTU instruction (Return to Userspace) \\\hline
735
JMP R6+\$Addr
736
        & MOV \$Addr(R6),PC
737
        & \\\hline
738
JSR PC+\$Addr
739
        & \parbox[t]{1.5in}{SUB \$1,SP \\\
740
        MOV \$3+PC,R0 \\
741
        STO R0,1(SP) \\
742
        MOV \$Addr+PC,PC \\
743
        ADD \$1,SP}
744 24 dgisselq
        & Jump to Subroutine. Note the required cleanup instruction after
745 36 dgisselq
        returning.  This could easily be turned into a three instruction
746
        operand, removing the preliminary stack instruction before and
747
        the cleanup after, by adjusting how any stack frame was built for
748
        this routine to include space at the top of the stack for the PC.
749
        \\\hline
750 21 dgisselq
JSR PC+\$Addr
751
        & \parbox[t]{1.5in}{MOV \$3+PC,R12 \\ MOV \$addr+PC,PC}
752
        &This is the high speed
753
        version of a subroutine call, necessitating a register to hold the
754
        last PC address.  In its favor, this method doesn't suffer the
755
        mandatory memory access of the other approach. \\\hline
756
LDI.l \$val,Rx
757
        & \parbox[t]{1.5in}{LDIHI (\$val$>>$16)\&0x0ffff, Rx \\
758
                        LDILO (\$val \& 0x0ffff)}
759
        & Sadly, there's not enough instruction
760
                space to load a complete immediate value into any register.
761
                Therefore, fully loading any register takes two cycles.
762
                The LDIHI (load immediate high) and LDILO (load immediate low)
763
                instructions have been created to facilitate this. \\\hline
764
\end{tabular}
765
\caption{Derived Instructions}\label{tbl:derived-1}
766
\end{center}\end{table}
767
\begin{table}\begin{center}
768
\begin{tabular}{p{1.4in}p{1.5in}p{3in}}\\\hline
769
Mapped & Actual  & Notes \\\hline
770
LOD.b \$addr,Rx
771
        & \parbox[t]{1.5in}{%
772
        LDI     \$addr,Ra \\
773
        LDI     \$addr,Rb \\
774
        LSR     \$2,Ra \\
775
        AND     \$3,Rb \\
776
        LOD     (Ra),Rx \\
777
        LSL     \$3,Rb \\
778
        SUB     \$32,Rb \\
779
        ROL     Rb,Rx \\
780
        AND \$0ffh,Rx}
781
        & \parbox[t]{3in}{This CPU is designed for 32'bit word
782
        length instructions.  Byte addressing is not supported by the CPU or
783
        the bus, so it therefore takes more work to do.
784
 
785
        Note also that in this example, \$Addr is a byte-wise address, where
786 24 dgisselq
        all other addresses in this document are 32-bit wordlength addresses.
787
        For this reason,
788 21 dgisselq
        we needed to drop the bottom two bits.  This also limits the address
789
        space of character accesses using this method from 16 MB down to 4MB.}
790
                \\\hline
791
\parbox[t]{1.5in}{LSL \$1,Rx\\ LSLC \$1,Ry}
792
        & \parbox[t]{1.5in}{LSL \$1,Ry \\
793
        LSL \$1,Rx \\
794
        OR.C \$1,Ry}
795
        & Logical shift left with carry.  Note that the
796
        instruction order is now backwards, to keep the conditions valid.
797 33 dgisselq
        That is, LSL sets the carry flag, so if we did this the other way
798 21 dgisselq
        with Rx before Ry, then the condition flag wouldn't have been right
799
        for an OR correction at the end. \\\hline
800
\parbox[t]{1.5in}{LSR \$1,Rx \\ LSRC \$1,Ry}
801
        & \parbox[t]{1.5in}{CLR Rz \\
802
        LSR \$1,Ry \\
803
        LDIHI.C \$8000h,Rz \\
804
        LSR \$1,Rx \\
805
        OR Rz,Rx}
806
        & Logical shift right with carry \\\hline
807
NEG Rx & \parbox[t]{1.5in}{XOR \$-1,Rx \\ ADD \$1,Rx} & \\\hline
808 36 dgisselq
NEG.C Rx & \parbox[t]{1.5in}{MOV.C \$-1+Rx,Rx\\XOR.C \$-1,Rx} & \\\hline
809 21 dgisselq
NOOP & NOOP & While there are many
810
        operations that do nothing, such as MOV Rx,Rx, or OR \$0,Rx, these
811
        operations have consequences in that they might stall the bus if
812
        Rx isn't ready yet.  For this reason, we have a dedicated NOOP
813
        instruction. \\\hline
814
NOT Rx & XOR \$-1,Rx & \\\hline
815
POP Rx
816 37 dgisselq
        & \parbox[t]{1.5in}{LOD \$1(SP),Rx \\ ADD \$1,SP}
817 21 dgisselq
        & Note
818
        that for interrupt purposes, one can never depend upon the value at
819
        (SP).  Hence you read from it, then increment it, lest having
820 33 dgisselq
        incremented it first something then comes along and writes to that
821 21 dgisselq
        value before you can read the result. \\\hline
822 36 dgisselq
\end{tabular}
823
\caption{Derived Instructions, continued}\label{tbl:derived-2}
824
\end{center}\end{table}
825
\begin{table}\begin{center}
826
\begin{tabular}{p{1.4in}p{1.5in}p{3in}}\\\hline
827 21 dgisselq
PUSH Rx
828 33 dgisselq
        & \parbox[t]{1.5in}{SUB \$1,SP \\
829 21 dgisselq
        STO Rx,\$1(SP)}
830
        & \\\hline
831 36 dgisselq
PUSH Rx-Ry
832
        & \parbox[t]{1.5in}{SUB \$n,SP \\
833
        STO Rx,\$n(SP)
834
        \ldots \\
835
        STO Ry,\$1(SP)}
836
        & Multiple pushes at once only need the single subtract from the
837
        stack pointer.  This derived instruction is analogous to a similar one
838
        on the Motoroloa 68k architecture, although the Zip Assembler
839
        does not support this instruction (yet).\\\hline
840 21 dgisselq
RESET
841
        & \parbox[t]{1in}{STO \$1,\$watchdog(R12)\\NOOP\\NOOP}
842
        & \parbox[t]{3in}{This depends upon the peripheral base address being
843
        in R12.
844
 
845
        Another opportunity might be to jump to the reset address from within
846
        supervisor mode.}\\\hline
847 36 dgisselq
RET & \parbox[t]{1.5in}{LOD \$1(SP),PC}
848 24 dgisselq
        & Note that this depends upon the calling context to clean up the
849
        stack, as outlined for the JSR instruction.  \\\hline
850 21 dgisselq
RET & MOV R12,PC
851
        & This is the high(er) speed version, that doesn't touch the stack.
852
        As such, it doesn't suffer a stall on memory read/write to the stack.
853
        \\\hline
854
STEP Rr,Rt
855
        & \parbox[t]{1.5in}{LSR \$1,Rr \\ XOR.C Rt,Rr}
856
        & Step a Galois implementation of a Linear Feedback Shift Register, Rr,
857
                using taps Rt \\\hline
858
STO.b Rx,\$addr
859
        & \parbox[t]{1.5in}{%
860
        LDI \$addr,Ra \\
861
        LDI \$addr,Rb \\
862
        LSR \$2,Ra \\
863
        AND \$3,Rb \\
864
        SUB \$32,Rb \\
865
        LOD (Ra),Ry \\
866
        AND \$0ffh,Rx \\
867
        AND \$-0ffh,Ry \\
868
        ROL Rb,Rx \\
869
        OR Rx,Ry \\
870
        STO Ry,(Ra) }
871
        & \parbox[t]{3in}{This CPU and it's bus are {\em not} optimized
872
        for byte-wise operations.
873
 
874
        Note that in this example, \$addr is a
875
        byte-wise address, whereas in all of our other examples it is a
876
        32-bit word address. This also limits the address space
877
        of character accesses from 16 MB down to 4MB.F
878
        Further, this instruction implies a byte ordering,
879
        such as big or little endian.} \\\hline
880
SWAP Rx,Ry
881
        & \parbox[t]{1.5in}{
882
        XOR Ry,Rx \\
883
        XOR Rx,Ry \\
884
        XOR Ry,Rx}
885
        & While no extra registers are needed, this example
886
        does take 3-clocks. \\\hline
887
TRAP \#X
888 36 dgisselq
        & \parbox[t]{1.5in}{LDI \$x,R0 \\ AND ~\$GIE,CC }
889
        & This works because whenever a user lowers the \$GIE flag, it sets
890
        a TRAP bit within the CC register.  Therefore, upon entering the
891
        supervisor state, the CPU only need check this bit to know that it
892
        got there via a TRAP.  The trap could be made conditional by making
893
        the LDI and the AND conditional.  In that case, the assembler would
894
        quietly turn the LDI instruction into an LDILO and LDIHI pair,
895 37 dgisselq
        but the effect would be the same. \\\hline
896 36 dgisselq
\end{tabular}
897
\caption{Derived Instructions, continued}\label{tbl:derived-3}
898
\end{center}\end{table}
899
\begin{table}\begin{center}
900
\begin{tabular}{p{1.4in}p{1.5in}p{3in}}\\\hline
901 21 dgisselq
TST Rx
902
        & TST \$-1,Rx
903
        & Set the condition codes based upon Rx.  Could also do a CMP \$0,Rx,
904
        ADD \$0,Rx, SUB \$0,Rx, etc, AND \$-1,Rx, etc.  The TST and CMP
905
        approaches won't stall future pipeline stages looking for the value
906
        of Rx. \\\hline
907
WAIT
908
        & Or \$SLEEP,CC
909
        & Wait 'til interrupt.  In an interrupts disabled context, this
910
        becomes a HALT instruction.
911
\end{tabular}
912 36 dgisselq
\caption{Derived Instructions, continued}\label{tbl:derived-4}
913 21 dgisselq
\end{center}\end{table}
914
\section{Pipeline Stages}
915 32 dgisselq
As mentioned in the introduction, and highlighted in Fig.~\ref{fig:cpu},
916
the Zip CPU supports a five stage pipeline.
917 21 dgisselq
\begin{enumerate}
918 36 dgisselq
\item {\bf Prefetch}: Reads instruction from memory and into a cache, if so
919
        configured.  This
920 21 dgisselq
        stage is actually pipelined itself, and so it will stall if the PC
921
        ever changes.  Stalls are also created here if the instruction isn't
922
        in the prefetch cache.
923 36 dgisselq
 
924
        The Zip CPU supports one of two prefetch methods, depending upon a flag
925
        set at build time within the {\tt zipcpu.v} file.  The simplest is a
926
        non--cached implementation of a prefetch.  This implementation is
927
        fairly small, and ideal for
928
        users of the Zip CPU who need the extra space on the FPGA fabric.
929
        However, because this non--cached version has no cache, the maximum
930
        number of instructions per clock is limited to about one per five.
931
 
932
        The second prefetch module is a pipelined prefetch with a cache.  This
933
        module tries to keep the instruction address within a window of valid
934
        instruction addresses.  While effective, it is not a traditional
935
        cache implementation.  One unique feature of this cache implementation,
936
        however, is that it can be cleared in a single clock.  A disappointing
937
        feature, though, was that it needs an extra internal pipeline stage
938
        to be implemented.
939
 
940
\item {\bf Decode}: Decodes an instruction into op code, register(s) to read,
941
        and immediate offset.  This stage also determines whether the flags will
942 32 dgisselq
        be set or whether the result will be written back.
943 21 dgisselq
\item {\bf Read Operands}: Read registers and apply any immediate values to
944 24 dgisselq
        them.  There is no means of detecting or flagging arithmetic overflow
945
        or carry when adding the immediate to the operand.  This stage will
946
        stall if any source operand is pending.
947 21 dgisselq
\item Split into two tracks: An {\bf ALU} which will accomplish a simple
948 36 dgisselq
        instruction, and the {\bf MemOps} stage which handles {\tt LOD} (load)
949
        and {\tt STO} (store) instructions.
950 21 dgisselq
        \begin{itemize}
951 36 dgisselq
        \item Loads will stall the entire pipeline until complete.
952
        \item Condition codes are available upon completion of the ALU stage
953
        \item Issuing an instruction to the memory unit while the memory unit
954
                is busy will stall the entire pipeline.  If the bus deadlocks,
955
                only a reset will release the CPU.  (Watchdog timer, anyone?)
956 24 dgisselq
        \item The Zip CPU currently has no means of reading and acting on any
957
        error conditions on the bus.
958 21 dgisselq
        \end{itemize}
959 32 dgisselq
\item {\bf Write-Back}: Conditionally write back the result to the register
960 36 dgisselq
        set, applying the condition.  This routine is bi-entrant: either the
961 21 dgisselq
        memory or the simple instruction may request a register write.
962
\end{enumerate}
963
 
964 24 dgisselq
The Zip CPU does not support out of order execution.  Therefore, if the memory
965
unit stalls, every other instruction stalls.  Memory stores, however, can take
966 36 dgisselq
place concurrently with ALU operations, although memory reads (loads) cannot.
967 24 dgisselq
 
968 32 dgisselq
\section{Pipeline Stalls}
969
The processing pipeline can and will stall for a variety of reasons.  Some of
970
these are obvious, some less so.  These reasons are listed below:
971
\begin{itemize}
972
\item When the prefetch cache is exhausted
973 21 dgisselq
 
974 36 dgisselq
This reason should be obvious.  If the prefetch cache doesn't have the
975
instruction in memory, the entire pipeline must stall until enough of the
976
prefetch cache is loaded to support the next instruction.
977 21 dgisselq
 
978 32 dgisselq
\item While waiting for the pipeline to load following any taken branch, jump,
979 36 dgisselq
        return from interrupt or switch to interrupt context (5 stall cycles)
980 32 dgisselq
 
981
If the PC suddenly changes, the pipeline is subsequently cleared and needs to
982
be reloaded.  Given that there are five stages to the pipeline, that accounts
983 36 dgisselq
for four of the five stalls.  The stall cycle is lost in the pipelined prefetch
984 32 dgisselq
stage which needs at least one clock with a valid PC before it can produce
985 36 dgisselq
a new output.
986 32 dgisselq
 
987 36 dgisselq
The Zip CPU handles {\tt MOV \$X(PC),PC}, {\tt ADD \$X,PC}, and
988
{\tt LDI \$X,PC} instructions specially, however.  These instructions, when
989
not conditioned on the flags, can execute with only 3~stall cycles.
990
 
991 32 dgisselq
\item When reading from a prior register while also adding an immediate offset
992
\begin{enumerate}
993
\item\ {\tt OPCODE ?,RA}
994
\item\ {\em (stall)}
995
\item\ {\tt OPCODE I+RA,RB}
996
\end{enumerate}
997
 
998
Since the addition of the immediate register within OpB decoding gets applied
999
during the read operand stage so that it can be nicely settled before the ALU,
1000
any instruction that will write back an operand must be separated from the
1001
opcode that will read and apply an immediate offset by one instruction.  The
1002
good news is that this stall can easily be mitigated by proper scheduling.
1003 36 dgisselq
That is, any instruction that does not add an immediate to {\tt RA} may be
1004
scheduled into the stall slot.
1005 32 dgisselq
 
1006 36 dgisselq
\item When any write to either the CC or PC Register is followed by a memory
1007
        operation
1008 32 dgisselq
\begin{enumerate}
1009
\item\ {\tt OPCODE RA,PC} {\em Ex: a branch opcode}
1010
\item\ {\em (stall, even if jump not taken)}
1011 36 dgisselq
\item\ {\tt LOD \$X(RA),RB}
1012 32 dgisselq
\end{enumerate}
1013
Since branches take place in the writeback stage, the Zip CPU will stall the
1014
pipeline for one clock anytime there may be a possible jump.  This prevents
1015
an instruction from executing a memory access after the jump but before the
1016
jump is recognized.
1017
 
1018 36 dgisselq
This stall may be mitigated by shuffling the operations immediately following
1019
a potential branch so that an ALU operation follows the branch instead of a
1020
memory operation.
1021 33 dgisselq
 
1022 32 dgisselq
\item When reading from the CC register after setting the flags
1023
\begin{enumerate}
1024 36 dgisselq
\item\ {\tt ALUOP RA,RB} {\em Ex: a compare opcode}
1025
\item\ {\em (stall)}
1026 32 dgisselq
\item\ {\tt TST sys.ccv,CC}
1027
\item\ {\tt BZ somewhere}
1028
\end{enumerate}
1029
 
1030
The reason for this stall is simply performance.  Many of the flags are
1031
determined via combinatorial logic after the writeback instruction is
1032
determined.  Trying to then place these into the input for one of the operands
1033
created a time delay loop that would no longer execute in a single 100~MHz
1034
clock cycle.  (The time delay of the multiply within the ALU wasn't helping
1035
either \ldots).
1036
 
1037 33 dgisselq
This stall may be eliminated via proper scheduling, by placing an instruction
1038
that does not set flags in between the ALU operation and the instruction
1039
that references the CC register.  For example, {\tt MOV \$addr+PC,uPC}
1040
followed by an {\tt RTU} ({\tt OR \$GIE,CC}) instruction will not incur
1041
this stall, whereas an {\tt OR \$BREAKEN,CC} followed by an {\tt OR \$STEP,CC}
1042 36 dgisselq
will incur the stall, while a {\tt LDI \$BREAKEN|\$STEP,CC} will not.
1043 33 dgisselq
 
1044 32 dgisselq
\item When waiting for a memory read operation to complete
1045
\begin{enumerate}
1046
\item\ {\tt LOD address,RA}
1047 36 dgisselq
\item\ {\em (multiple stalls, bus dependent, 4 clocks best)}
1048 32 dgisselq
\item\ {\tt OPCODE I+RA,RB}
1049
\end{enumerate}
1050
 
1051 36 dgisselq
Remember, the Zip CPU does not support out of order execution.  Therefore,
1052 32 dgisselq
anytime the memory unit becomes busy both the memory unit and the ALU must
1053
stall until the memory unit is cleared.  This is especially true of a load
1054 33 dgisselq
instruction, which must still write its operand back to the register file.
1055
Store instructions are different, since they can be busy with no impact on
1056
later ALU write back operations.  Hence, only loads stall the pipeline.
1057 32 dgisselq
 
1058
This also assumes that the memory being accessed is a single cycle memory.
1059
Slower memories, such as the Quad SPI flash, will take longer--perhaps even
1060 33 dgisselq
as long as forty clocks.   During this time the CPU and the external bus
1061 32 dgisselq
will be busy, and unable to do anything else.
1062
 
1063
\item Memory operation followed by a memory operation
1064
\begin{enumerate}
1065
\item\ {\tt STO address,RA}
1066 36 dgisselq
\item\ {\em (multiple stalls, bus dependent, 4 clocks best)}
1067 32 dgisselq
\item\ {\tt LOD address,RB}
1068 36 dgisselq
\item\ {\em (multiple stalls, bus dependent, 4 clocks best)}
1069 32 dgisselq
\end{enumerate}
1070
 
1071 36 dgisselq
In this case, the LOD instruction cannot start until the STO is finished.
1072 32 dgisselq
With proper scheduling, it is possible to do something in the ALU while the
1073 36 dgisselq
memory unit is busy with the STO instruction, but otherwise this pipeline will
1074
stall waiting for it to complete.
1075 32 dgisselq
 
1076
Note that even though the Wishbone bus can support pipelined accesses at
1077
one access per clock, only the prefetch stage can take advantage of this.
1078
Load and Store instructions are stuck at one wishbone cycle per instruction.
1079 36 dgisselq
 
1080
\item When waiting for a conditional memory read operation to complete
1081
\begin{enumerate}
1082
\item\ {\tt LOD.Z address,RA}
1083
\item\ {\em (multiple stalls, bus dependent, 7 clocks best)}
1084
\item\ {\tt OPCODE I+RA,RB}
1085
\end{enumerate}
1086
 
1087
In this case, the Zip CPU doesn't warn the prefetch cache to get off the bus
1088
two cycles before using the bus, so there's a potential for an extra three
1089
cycle cost due to bus contention between the prefetch and the CPU.
1090
 
1091
This is true for both the LOD and the STO instructions, with the exception that
1092
the STO instruction will continue in parallel with any ALU instructions that
1093
follow it.
1094
 
1095 32 dgisselq
\end{itemize}
1096
 
1097
 
1098 21 dgisselq
\chapter{Peripherals}\label{chap:periph}
1099 24 dgisselq
 
1100
While the previous chapter describes a CPU in isolation, the Zip System
1101
includes a minimum set of peripherals as well.  These peripherals are shown
1102
in Fig.~\ref{fig:zipsystem}
1103
\begin{figure}\begin{center}
1104
\includegraphics[width=3.5in]{../gfx/system.eps}
1105
\caption{Zip System Peripherals}\label{fig:zipsystem}
1106
\end{center}\end{figure}
1107
and described here.  They are designed to make
1108
the Zip CPU more useful in an Embedded Operating System environment.
1109
 
1110 21 dgisselq
\section{Interrupt Controller}
1111 24 dgisselq
 
1112
Perhaps the most important peripheral within the Zip System is the interrupt
1113
controller.  While the Zip CPU itself can only handle one interrupt, and has
1114
only the one interrupt state: disabled or enabled, the interrupt controller
1115
can make things more interesting.
1116
 
1117
The Zip System interrupt controller module supports up to 15 interrupts, all
1118
controlled from one register.  Bit~31 of the interrupt controller controls
1119
overall whether interrupts are enabled (1'b1) or disabled (1'b0).  Bits~16--30
1120
control whether individual interrupts are enabled (1'b0) or disabled (1'b0).
1121
Bit~15 is an indicator showing whether or not any interrupt is active, and
1122
bits~0--15 indicate whether or not an individual interrupt is active.
1123
 
1124
The interrupt controller has been designed so that bits can be controlled
1125
individually without having any knowledge of the rest of the controller
1126
setting.  To enable an interrupt, write to the register with the high order
1127
global enable bit set and the respective interrupt enable bit set.  No other
1128
bits will be affected.  To disable an interrupt, write to the register with
1129
the high order global enable bit cleared and the respective interrupt enable
1130
bit set.  To clear an interrupt, write a `1' to that interrupts status pin.
1131
Zero's written to the register have no affect, save that a zero written to the
1132
master enable will disable all interrupts.
1133
 
1134
As an example, suppose you wished to enable interrupt \#4.  You would then
1135
write to the register a {\tt 0x80100010} to enable interrupt \#4 and to clear
1136
any past active state.  When you later wish to disable this interrupt, you would
1137
write a {\tt 0x00100010} to the register.  As before, this both disables the
1138
interrupt and clears the active indicator.  This also has the side effect of
1139
disabling all interrupts, so a second write of {\tt 0x80000000} may be necessary
1140
to re-enable any other interrupts.
1141
 
1142
The Zip System currently hosts two interrupt controllers, a primary and a
1143
secondary.  The primary interrupt controller has one interrupt line which may
1144
come from an external interrupt controller, and one interrupt line from the
1145
secondary controller.  Other primary interrupts include the system timers,
1146
the jiffies interrupt, and the manual cache interrupt.  The secondary interrupt
1147
controller maintains an interrupt state for all of the processor accounting
1148
counters.
1149
 
1150 21 dgisselq
\section{Counter}
1151
 
1152
The Zip Counter is a very simple counter: it just counts.  It cannot be
1153
halted.  When it rolls over, it issues an interrupt.  Writing a value to the
1154
counter just sets the current value, and it starts counting again from that
1155
value.
1156
 
1157
Eight counters are implemented in the Zip System for process accounting.
1158
This may change in the future, as nothing as yet uses these counters.
1159
 
1160
\section{Timer}
1161
 
1162
The Zip Timer is also very simple: it simply counts down to zero.  When it
1163
transitions from a one to a zero it creates an interrupt.
1164
 
1165
Writing any non-zero value to the timer starts the timer.  If the high order
1166
bit is set when writing to the timer, the timer becomes an interval timer and
1167
reloads its last start time on any interrupt.  Hence, to mark seconds, one
1168
might set the timer to 100~million (the number of clocks per second), and
1169
set the high bit.  Ever after, the timer will interrupt the CPU once per
1170 24 dgisselq
second (assuming a 100~MHz clock).  This reload capability also limits the
1171
maximum timer value to $2^{31}-1$, rather than $2^{32}-1$.
1172 21 dgisselq
 
1173
\section{Watchdog Timer}
1174
 
1175
The watchdog timer is no different from any of the other timers, save for one
1176
critical difference: the interrupt line from the watchdog
1177
timer is tied to the reset line of the CPU.  Hence writing a `1' to the
1178
watchdog timer will always reset the CPU.
1179 32 dgisselq
To stop the Watchdog timer, write a `0' to it.  To start it,
1180 21 dgisselq
write any other number to it---as with the other timers.
1181
 
1182
While the watchdog timer supports interval mode, it doesn't make as much sense
1183
as it did with the other timers.
1184
 
1185
\section{Jiffies}
1186
 
1187
This peripheral is motivated by the Linux use of `jiffies' whereby a process
1188
can request to be put to sleep until a certain number of `jiffies' have
1189
elapsed.  Using this interface, the CPU can read the number of `jiffies'
1190
from the peripheral (it only has the one location in address space), add the
1191 24 dgisselq
sleep length to it, and write the result back to the peripheral.  The zipjiffies
1192 21 dgisselq
peripheral will record the value written to it only if it is nearer the current
1193
counter value than the last current waiting interrupt time.  If no other
1194
interrupts are waiting, and this time is in the future, it will be enabled.
1195
(There is currently no way to disable a jiffie interrupt once set, other
1196 24 dgisselq
than to disable the interrupt line in the interrupt controller.)  The processor
1197 21 dgisselq
may then place this sleep request into a list among other sleep requests.
1198
Once the timer expires, it would write the next Jiffy request to the peripheral
1199
and wake up the process whose timer had expired.
1200
 
1201
Indeed, the Jiffies register is nothing more than a glorified counter with
1202
an interrupt.  Unlike the other counters, the Jiffies register cannot be set.
1203
Writes to the jiffies register create an interrupt time.  When the Jiffies
1204
register later equals the value written to it, an interrupt will be asserted
1205
and the register then continues counting as though no interrupt had taken
1206
place.
1207
 
1208
The purpose of this register is to support alarm times within a CPU.  To
1209
set an alarm for a particular process $N$ clocks in advance, read the current
1210
Jiffies value, and $N$, and write it back to the Jiffies register.  The
1211
O/S must also keep track of values written to the Jiffies register.  Thus,
1212 32 dgisselq
when an `alarm' trips, it should be removed from the list of alarms, the list
1213 21 dgisselq
should be sorted, and the next alarm in terms of Jiffies should be written
1214
to the register.
1215
 
1216 36 dgisselq
\section{Direct Memory Access Controller}
1217 24 dgisselq
 
1218 36 dgisselq
The Direct Memory Access (DMA) controller can be used to either move memory
1219
from one location to another, to read from a peripheral into memory, or to
1220
write from a peripheral into memory all without CPU intervention.  Further,
1221
since the DMA controller can issue (and does issue) pipeline wishbone accesses,
1222
any DMA memory move will by nature be faster than a corresponding program
1223
accomplishing the same move.  To put this to numbers, it may take a program
1224
18~clocks per word transferred, whereas this DMA controller can move one
1225
word in two clocks--provided it has bus access.  (The CPU gets priority over the
1226
bus.)
1227 24 dgisselq
 
1228 36 dgisselq
When copying memory from one location to another, the DMA controller will
1229
copy in units of a given transfer length--up to 1024 words at a time.  It will
1230
read that transfer length into its internal buffer, and then write to the
1231
destination address from that buffer.  If the CPU interrupts a DMA transfer,
1232
it will release the bus, let the CPU complete whatever it needs to do, and then
1233
restart its transfer by writing the contents of its internal buffer and then
1234
re-entering its read cycle again.
1235 24 dgisselq
 
1236 36 dgisselq
When coupled with a peripheral, the DMA controller can be configured to start
1237
a memory copy on an interrupt line going high.  Further, the controller can be
1238
configured to issue reads from (or two) the same address instead of incrementing
1239
the address at each clock.  The DMA completes once the total number of items
1240
specified (not the transfer length) have been transferred.
1241
 
1242
In each case, once the transfer is complete and the DMA unit returns to
1243
idle, the DMA will issue an interrupt.
1244
 
1245
 
1246 21 dgisselq
\chapter{Operation}\label{chap:ops}
1247
 
1248 33 dgisselq
The Zip CPU, and even the Zip System, is not a System on a Chip (SoC).  It
1249
needs to be connected to its operational environment in order to be used.
1250
Specifically, some per system adjustments need to be made:
1251
\begin{enumerate}
1252
\item The Zip System depends upon an external 32-bit Wishbone bus.  This
1253
        must exist, and must be connected to the Zip CPU for it to work.
1254
\item The Zip System needs to be told of its {\tt RESET\_ADDRESS}.  This is
1255
        the program counter of the first instruction following a reset.
1256
\item If you want the Zip System to start up on its own, you will need to
1257
        set the {\tt START\_HALTED} parameter to zero.  Otherwise, if you
1258
        wish to manually start the CPU, that is if upon reset you want the
1259
        CPU start start in its halted, reset state, then set this parameter to
1260
        one.
1261
\item The third parameter to set is the number of interrupts you will be
1262
        providing from external to the CPU.  This can be anything from one
1263
        to nine, but it cannot be zero.  (Wire this line to a 1'b0 if you
1264
        do not wish to support any external interrupts.)
1265
\item Finally, you need to place into some wishbone accessible address, whether
1266
        RAM or (more likely) ROM, the initial instructions for the CPU.
1267
\end{enumerate}
1268
If you have enabled your CPU to start automatically, then upon power up the
1269
CPU will immediately start executing your instructions.
1270
 
1271
This is, however, not how I have used the Zip CPU.  I have instead used the
1272 36 dgisselq
Zip CPU in a more controlled environment.  For me, the CPU starts in a
1273 33 dgisselq
halted state, and waits to be told to start.  Further, the RESET address is a
1274
location in RAM.  After bringing up the board I am using, and further the
1275
bus that is on it, the RAM memory is then loaded externally with the program
1276
I wish the Zip System to run.  Once the RAM is loaded, I release the CPU.
1277
The CPU then runs until its halt condition, at which point its task is
1278
complete.
1279
 
1280
Eventually, I intend to place an operating system onto the ZipSystem, I'm
1281
just not there yet.
1282
 
1283 36 dgisselq
The rest of this chapter examines some common programming constructs, and
1284
how they might be applied to the Zip System.
1285 33 dgisselq
 
1286 36 dgisselq
\section{Example: Idle Task}
1287
One task every operating system needs is the idle task, the task that takes
1288
place when nothing else can run.  On the Zip CPU, this task is quite simple,
1289
and it is shown in assemble in Tbl.~\ref{tbl:idle-asm}.
1290
\begin{table}\begin{center}
1291
\begin{tabular}{ll}
1292
{\tt idle\_task:} \\
1293
&        {\em ; Wait for the next interrupt, then switch to supervisor task} \\
1294
&        {\tt WAIT} \\
1295
&        {\em ; When we come back, it's because the supervisor wishes to} \\
1296
&        {\em ; wait for an interrupt again, so go back to the top.} \\
1297
&        {\tt BRA idle\_task} \\
1298
\end{tabular}
1299
\caption{Example Idle Loop}\label{tbl:idle-asm}
1300
\end{center}\end{table}
1301
When this task runs, the CPU will fill up all of the pipeline stages up the
1302
ALU.  The {\tt WAIT} instruction, upon leaving the ALU, places the CPU into
1303
a sleep state where nothing more moves.  Sure, there may be some more settling,
1304
the pipe cache continue to read until full, other instructions may issue until
1305
the pipeline fills, but then everything will stall.  Then, once an interrupt
1306
takes place, control passes to the supervisor task to handle the interrupt.
1307
When control passes back to this task, it will be on the next instruction.
1308
Since that next instruction sends us back to the top of the task, the idle
1309
task thus does nothing but wait for an interrupt.
1310
 
1311
This should be the lowest priority task, the task that runs when nothing else
1312
can.  It will help lower the FPGA power usage overall---at least its dynamic
1313
power usage.
1314
 
1315
\section{Example: Memory Copy}
1316
One common operation is that of a memory move or copy.  Consider the C code
1317
shown in Tbl.~\ref{tbl:memcp-c}.
1318
\begin{table}\begin{center}
1319
\parbox{4in}{\begin{tabbing}
1320
{\tt void} \= {\tt memcp(void *dest, void *src, int len) \{} \\
1321
        \> {\tt for(int i=0; i<len; i++)} \\
1322
        \> \hspace{0.2in} {\tt *dest++ = *src++;} \\
1323
\}
1324
\end{tabbing}}
1325
\caption{Example Memory Copy code in C}\label{tbl:memcp-c}
1326
\end{center}\end{table}
1327
This same code can be translated in Zip Assembly as shown in
1328
Tbl.~\ref{tbl:memcp-asm}.
1329
\begin{table}\begin{center}
1330
\begin{tabular}{ll}
1331
memcp: \\
1332
&        {\em ; R0 = *dest, R1 = *src, R2 = LEN} \\
1333
&        {\em ; The following will operate in 17 clocks per word minus one clock} \\
1334
&        {\tt CMP 0,R2} \\
1335
&        {\tt LOD.Z -1(SP),PC} {\em ; A conditional return }\\
1336
&        {\em ; (One stall on potentially writing to PC)} \\
1337
&        {\tt LOD (R1),R3} \\
1338
&        {\em ; (4 stalls, cannot be scheduled away)} \\
1339
&        {\tt STO R3,(R2)} {\em ; (4 schedulable stalls, has no impact now)} \\
1340
&        {\tt ADD 1,R1} \\
1341
&        {\tt SUB 1,R2} \\
1342
&        {\tt BNZ loop} \\
1343
&        {\em ; (5 stalls, if branch taken, to clear and refill the pipeline)} \\
1344
&        {\tt RET} \\
1345
\end{tabular}
1346
\caption{Example Memory Copy code in Zip Assembly}\label{tbl:memcp-asm}
1347
\end{center}\end{table}
1348
This example points out several things associated with the Zip CPU.  First,
1349
a straightforward implementation of a for loop is not the fastest loop
1350
structure.  For this reason, we have placed the test to continue at the
1351
end.  Second, all pointers are {\tt void} pointers to arbitrary 32--bit
1352
data types.  The Zip CPU does not have explicit support for smaller or larger
1353
data types, and so this memory copy cannot be applied at a byte level.
1354
Third, we've optimized the conditional jump to a return instruction into a
1355
conditional return instruction.
1356
 
1357
\section{Context Switch}
1358
 
1359
Fundamental to any multiprocessing system is the ability to switch from one
1360
task to the next.  In the ZipSystem, this is accomplished in one of a couple
1361
ways.  The first step is that an interrupt happens.  Anytime an interrupt
1362
happens, the CPU needs to execute the following tasks in supervisor mode:
1363
\begin{enumerate}
1364
\item Check for a trap instruction.  That  is, if the user task requested a
1365
        trap, we may not wish to adjust the context, check interrupts, or call
1366
        the scheduler.  Tbl.~\ref{tbl:trap-check}
1367
\begin{table}\begin{center}
1368
\begin{tabular}{ll}
1369
{\tt return\_to\_user:} \\
1370
&       {\em; The instruction before the context switch processing must} \\
1371
&       {\em; be the RTU instruction that enacted user mode in the first} \\
1372
&       {\em; place.  We show it here just for reference.} \\
1373
&       {\tt RTU} \\
1374
{\tt trap\_check:} \\
1375
&       {\tt MOV uCC,R0} \\
1376
&       {\tt TST \$TRAP,R0} \\
1377
&       {\tt BNZ swap\_out} \\
1378
&       {; \em Do something here to execute the trap} \\
1379
&       {; \em Don't need to call the scheduler, so we can just return} \\
1380
&       {\tt BRA return\_to\_user} \\
1381
\end{tabular}
1382
\caption{Checking for whether the user issued a TRAP instruction}\label{tbl:trap-check}
1383
\end{center}\end{table}
1384
        shows the rudiments of this code, while showing nothing of how the
1385
        actual trap would be implemented.
1386
 
1387
You may also wish to note that the instruction before the first instruction
1388
in our context swap {\em must be} a return to userspace instruction.
1389
Remember, the supervisor process is re--entered where it left off.  This is
1390
different from many other processors that enter interrupt mode at some vector
1391
or other.  In this case, we always enter supervisor mode right where we last
1392
left.\footnote{The one exception to this rule is upon reset where supervisor
1393
mode is entered at a pre--programmed wishbone memory address.}
1394
 
1395
\item Capture user counters.  If the operating system is keeping track of
1396
        system usage via the accounting counters, those counters need to be
1397
        copied and accumulated into some master counter at this point.
1398
 
1399
\item Preserve the old context.  This involves pushing all the user registers
1400
        onto the user stack and then copying the resulting stack address
1401
        into the tasks task structure, as shown in Tbl.~\ref{tbl:context-out}.
1402
\begin{table}\begin{center}
1403
\begin{tabular}{ll}
1404
{\tt swap\_out:} \\
1405
&        {\tt MOV -15(uSP),R1} \\
1406
&        {\tt STO R1,stack(R12)} \\
1407
&        {\tt MOV uPC,R0} \\
1408
&        {\tt STO R0,15(R1)} \\
1409
&        {\tt MOV uCC,R0} \\
1410
&        {\tt STO R0,14(R1)} \\
1411
&       {\em ; We can skip storing the stack, uSP, since it'll be stored}\\
1412
&       {\em ; elsewhere (in the task structure) }\\
1413
&        {\tt MOV uR13,R0} \\
1414
&        {\tt STO R0,13(R1)} \\
1415
        & \ldots {\em ; Need to repeat for all user registers} \\
1416
&        {\tt MOV uR0,R0} \\
1417
&        {\tt STO R0,1(R1)} \\
1418
\end{tabular}
1419
\caption{Example Storing User Task Context}\label{tbl:context-out}
1420
\end{center}\end{table}
1421
For the sake of discussion, we assume the supervisor maintains a
1422
pointer to the current task's structure in supervisor register
1423
{\tt R12}, and that {\tt stack} is an offset to the beginning of this
1424
structure indicating where the stack pointer is to be kept within it.
1425
 
1426
        For those who are still interested, the full code for this context
1427
        save can be found as an assembler macro within the assembler
1428
        include file, {\tt sys.i}.
1429
 
1430
\item Reset the watchdog timer.  If you are using the watchdog timer, it should
1431
        be reset on a context swap, to know that things are still working.
1432
        Example code for this is shown in Tbl.~\ref{tbl:reset-watchdog}.
1433
\begin{table}\begin{center}
1434
\begin{tabular}{ll}
1435
\multicolumn{2}{l}{{\tt `define WATCHDOG\_ADDRESS 32'hc000\_0002}}\\
1436
\multicolumn{2}{l}{{\tt `define WATCHDOG\_TICKS 32'd1\_000\_000} {; \em = 10 ms}}\\
1437
&       {\tt LDI WATCHDOG\_ADDRESS,R0} \\
1438
&       {\tt LDI WATCHDOG\_TICKS,R1} \\
1439
&       {\tt STO R1,(R0)}
1440
\end{tabular}
1441
\caption{Example Watchdog Reset}\label{tbl:reset-watchdog}
1442
\end{center}\end{table}
1443
 
1444
\item Interrupt handling.  An interrupt handler within the Zip System is nothing
1445
        more than a task.  At context swap time, the supervisor needs to
1446
        disable all of the interrupts that have tripped, and then enable
1447
        all of the tasks that would deal with each of these interrupts.
1448
        These can be user tasks, run at higher priority than any other user
1449
        tasks.  Either way, they will need to re--enable their own interrupt
1450
        themselves, if the interrupt is still relevant.
1451
 
1452
        An example of this master interrut handling is shown in
1453
        Tbl.~\ref{tbl:pre-handler}.
1454
\begin{table}\begin{center}
1455
\begin{tabular}{ll}
1456
{\tt pre\_handler:} \\
1457
&       {\tt LDI PIC\_ADDRESS,R0 } \\
1458
&       {\em ; Start by grabbing the interrupt state from the interrupt}\\
1459
&       {\em ; controller.  We'll store this into the register R7 so that }\\
1460
&       {\em ; we can keep and preserve this information for the scheduler}\\
1461
&       {\em ; to use later. }\\
1462
&       {\tt LOD (R0),R1} \\
1463
&       {\tt MOV R1,R7 } \\
1464
&       {\em ; As a next step, we need to acknowledge and disable all active}\\
1465
&       {\em ; interrupts. We'll start by calculating all of our active}\\
1466
&       {\em ; interrupts.}\\
1467
&       {\tt AND 0x07fff,R1 } \\
1468
&       {\em ; Put the active interrupts into the upper half of R1} \\
1469
&       {\tt ROL 16,R1 } \\
1470
&       {\tt LDILO 0x0ffff,R1   } \\
1471
&       {\tt AND R7,R1}\\
1472
&       {\em ; Acknowledge and disable active interrupts}\\
1473
&       {\em ; This also disables all interrupts from the controller, so}\\
1474
&       {\em ; we'll need to re-enable interrupts in general shortly } \\
1475
&       {\tt STO R1,(R0) } \\
1476
&       {\em ; We leave our active interrupt mask in R7 so the scheduler can}\\
1477
&       {\em ; release any tasks that depended upon them. } \\
1478
\end{tabular}
1479
\caption{Example checking for active interrupts}\label{tbl:pre-handler}
1480
\end{center}\end{table}
1481
 
1482
\item Calling the scheduler.  This needs to be done to pick the next task
1483
        to switch to.  It may be an interrupt handler, or it may  be a normal
1484
        user task.  From a priority standpoint, it would make sense that the
1485
        interrupt handlers all have a higher priority than the user tasks,
1486
        and that once they have been called the user tasks may then be called
1487
        again.  If no task is ready to run, run the idle task to wait for an
1488
        interrupt.
1489
 
1490
        This suggests a minimum of four task priorities:
1491
        \begin{enumerate}
1492
        \item Interrupt handlers, executed with their interrupts disabled
1493
        \item Device drivers, executed with interrupts re-enabled
1494
        \item User tasks
1495
        \item The idle task, executed when nothing else is able to execute
1496
        \end{enumerate}
1497
 
1498
        For our purposes here, we'll just assume that a pointer to the current
1499
        task is maintained in {\tt R12}, that a {\tt JSR scheduler} is
1500
        called, and that the next current task is likewise placed into
1501
        {\tt R12}.
1502
 
1503
\item Restore the new tasks context.  Given that the scheduler has returned a
1504
        task that can be run at this time, the stack pointer needs to be
1505
        pulled out of the tasks task structure, placed into the user
1506
        register, and then the rest of the user registers need to be popped
1507
        back off of the stack to run this task.  An example of this is
1508
        shown in Tbl.~\ref{tbl:context-in},
1509
\begin{table}\begin{center}
1510
\begin{tabular}{ll}
1511
{\tt swap\_in:} \\
1512
&       {\tt LOD stack(R12),R1} \\
1513
&       {\tt MOV 15(R1),uSP} \\
1514
&       {\tt LOD 15(R1),R0} \\
1515
&       {\tt MOV R0,uPC} \\
1516
&       {\tt LOD 14(R1),R0} \\
1517
&       {\tt MOV R0,uCC} \\
1518
&       {\tt LOD 13(R1),R0} \\
1519
&       {\tt MOV R0,uR12} \\
1520
        & \ldots {\em ; Need to repeat for all user registers} \\
1521
&       {\tt LOD 1(R1),R0} \\
1522
&       {\tt MOV R0,uR0} \\
1523
&       {\tt BRA return\_to\_user} \\
1524
\end{tabular}
1525
\caption{Example Restoring User Task Context}\label{tbl:context-in}
1526
\end{center}\end{table}
1527
        assuming as before that the task
1528
        pointer is found in supervisor register {\tt R12}.
1529
        As with storing the user context, the full code associated with
1530
        restoring the user context can be found in the assembler include
1531
        file, {\tt sys.i}.
1532
 
1533
\item Clear the userspace accounting registers.  In order to keep track of
1534
        per process system usage, these registers need to be cleared before
1535
        reactivating the userspace process.  That way, upon the next
1536
        interrupt, we'll know how many clocks the userspace program has
1537
        encountered, and how many instructions it was able to issue in
1538
        those many clocks.
1539
 
1540
\item Jump back to the instruction just before saving the last tasks context,
1541
        because that location in memory contains the return from interrupt
1542
        command that we are going to need to execute, in order to guarantee
1543
        that we return back here again.
1544
\end{enumerate}
1545
 
1546 21 dgisselq
\chapter{Registers}\label{chap:regs}
1547
 
1548 24 dgisselq
The ZipSystem registers fall into two categories, ZipSystem internal registers
1549
accessed via the ZipCPU shown in Tbl.~\ref{tbl:zpregs},
1550
\begin{table}[htbp]
1551
\begin{center}\begin{reglist}
1552 32 dgisselq
PIC   & \scalebox{0.8}{\tt 0xc0000000} & 32 & R/W & Primary Interrupt Controller \\\hline
1553
WDT   & \scalebox{0.8}{\tt 0xc0000001} & 32 & R/W & Watchdog Timer \\\hline
1554 36 dgisselq
  & \scalebox{0.8}{\tt 0xc0000002} & 32 & R/W & {\em (Reserved for future use)} \\\hline
1555 32 dgisselq
CTRIC & \scalebox{0.8}{\tt 0xc0000003} & 32 & R/W & Secondary Interrupt Controller \\\hline
1556
TMRA  & \scalebox{0.8}{\tt 0xc0000004} & 32 & R/W & Timer A\\\hline
1557
TMRB  & \scalebox{0.8}{\tt 0xc0000005} & 32 & R/W & Timer B\\\hline
1558
TMRC  & \scalebox{0.8}{\tt 0xc0000006} & 32 & R/W & Timer C\\\hline
1559
JIFF  & \scalebox{0.8}{\tt 0xc0000007} & 32 & R/W & Jiffies \\\hline
1560
MTASK  & \scalebox{0.8}{\tt 0xc0000008} & 32 & R/W & Master Task Clock Counter \\\hline
1561
MMSTL  & \scalebox{0.8}{\tt 0xc0000009} & 32 & R/W & Master Stall Counter \\\hline
1562
MPSTL  & \scalebox{0.8}{\tt 0xc000000a} & 32 & R/W & Master Pre--Fetch Stall Counter \\\hline
1563
MICNT  & \scalebox{0.8}{\tt 0xc000000b} & 32 & R/W & Master Instruction Counter\\\hline
1564
UTASK  & \scalebox{0.8}{\tt 0xc000000c} & 32 & R/W & User Task Clock Counter \\\hline
1565
UMSTL  & \scalebox{0.8}{\tt 0xc000000d} & 32 & R/W & User Stall Counter \\\hline
1566
UPSTL  & \scalebox{0.8}{\tt 0xc000000e} & 32 & R/W & User Pre--Fetch Stall Counter \\\hline
1567
UICNT  & \scalebox{0.8}{\tt 0xc000000f} & 32 & R/W & User Instruction Counter\\\hline
1568 36 dgisselq
DMACTRL  & \scalebox{0.8}{\tt 0xc0000010} & 32 & R/W & DMA Control Register\\\hline
1569
DMALEN  & \scalebox{0.8}{\tt 0xc0000011} & 32 & R/W & DMA total transfer length\\\hline
1570
DMASRC  & \scalebox{0.8}{\tt 0xc0000012} & 32 & R/W & DMA source address\\\hline
1571
DMADST  & \scalebox{0.8}{\tt 0xc0000013} & 32 & R/W & DMA destination address\\\hline
1572 32 dgisselq
% Cache  & \scalebox{0.8}{\tt 0xc0100000} & & & Base address of the Cache memory\\\hline
1573 24 dgisselq
\end{reglist}
1574
\caption{Zip System Internal/Peripheral Registers}\label{tbl:zpregs}
1575
\end{center}\end{table}
1576 33 dgisselq
and the two debug registers shown in Tbl.~\ref{tbl:dbgregs}.
1577 24 dgisselq
\begin{table}[htbp]
1578
\begin{center}\begin{reglist}
1579
ZIPCTRL & 0 & 32 & R/W & Debug Control Register \\\hline
1580
ZIPDATA & 1 & 32 & R/W & Debug Data Register \\\hline
1581
\end{reglist}
1582
\caption{Zip System Debug Registers}\label{tbl:dbgregs}
1583
\end{center}\end{table}
1584
 
1585 33 dgisselq
\section{Peripheral Registers}
1586
The peripheral registers, listed in Tbl.~\ref{tbl:zpregs}, are shown in the
1587
CPU's address space.  These may be accessed by the CPU at these addresses,
1588
and when so accessed will respond as described in Chapt.~\ref{chap:periph}.
1589
These registers will be discussed briefly again here.
1590 24 dgisselq
 
1591 33 dgisselq
The Zip CPU Interrupt controller has four different types of bits, as shown in
1592
Tbl.~\ref{tbl:picbits}.
1593
\begin{table}\begin{center}
1594
\begin{bitlist}
1595
31 & R/W & Master Interrupt Enable\\\hline
1596
30\ldots 16 & R/W & Interrupt Enables, write '1' to change\\\hline
1597
15 & R & Current Master Interrupt State\\\hline
1598
15\ldots 0 & R/W & Input Interrupt states, write '1' to clear\\\hline
1599
\end{bitlist}
1600
\caption{Interrupt Controller Register Bits}\label{tbl:picbits}
1601
\end{center}\end{table}
1602
The high order bit, or bit--31, is the master interrupt enable bit.  When this
1603
bit is set, then any time an interrupt occurs the CPU will be interrupted and
1604
will switch to supervisor mode, etc.
1605
 
1606
Bits 30~\ldots 16 are interrupt enable bits.  Should the interrupt line go
1607
ghile while enabled, an interrupt will be generated.  To set an interrupt enable
1608
bit, one needs to write the master interrupt enable while writing a `1' to this
1609
the bit.  To clear, one need only write a `0' to the master interrupt enable,
1610
while leaving this line high.
1611
 
1612
Bits 15\ldots 0 are the current state of the interrupt vector.  Interrupt lines
1613
trip when they go high, and remain tripped until they are acknowledged.  If
1614
the interrupt goes high for longer than one pulse, it may be high when a clear
1615
is requested.  If so, the interrupt will not clear.  The line must go low
1616
again before the status bit can be cleared.
1617
 
1618
As an example, consider the following scenario where the Zip CPU supports four
1619
interrupts, 3\ldots0.
1620
\begin{enumerate}
1621
\item The Supervisor will first, while in the interrupts disabled mode,
1622
        write a {\tt 32'h800f000f} to the controller.  The supervisor may then
1623
        switch to the user state with interrupts enabled.
1624
\item When an interrupt occurs, the supervisor will switch to the interrupt
1625
        state.  It will then cycle through the interrupt bits to learn which
1626
        interrupt handler to call.
1627
\item If the interrupt handler expects more interrupts, it will clear its
1628
        current interrupt when it is done handling the interrupt in question.
1629
        To do this, it will write a '1' to the low order interrupt mask,
1630
        such as writing a {\tt 32'h80000001}.
1631
\item If the interrupt handler does not expect any more interrupts, it will
1632
        instead clear the interrupt from the controller by writing a
1633
        {\tt 32'h00010001} to the controller.
1634
\item Once all interrupts have been handled, the supervisor will write a
1635
        {\tt 32'h80000000} to the interrupt register to re-enable interrupt
1636
        generation.
1637
\item The supervisor should also check the user trap bit, and possible soft
1638
        interrupt bits here, but this action has nothing to do with the
1639
        interrupt control register.
1640
\item The supervisor will then leave interrupt mode, possibly adjusting
1641
        whichever task is running, by executing a return from interrupt
1642
        command.
1643
\end{enumerate}
1644
 
1645
Leaving the interrupt controller, we show the timer registers bit definitions
1646
in Tbl.~\ref{tbl:tmrbits}.
1647
\begin{table}\begin{center}
1648
\begin{bitlist}
1649
31 & R/W & Auto-Reload\\\hline
1650
30\ldots 0 & R/W & Current timer value\\\hline
1651
\end{bitlist}
1652
\caption{Timer Register Bits}\label{tbl:tmrbits}
1653
\end{center}\end{table}
1654
As you may recall, the timer just counts down to zero and then trips an
1655
interrupt.  Writing to the current timer value sets that value, and reading
1656
from it returns that value.  Writing to the current timer value while also
1657
setting the auto--reload bit will send the timer into an auto--reload mode.
1658
In this mode, upon setting its interrupt bit for one cycle, the timer will
1659
also reset itself back to the value of the timer that was written to it when
1660
the auto--reload option was written to it.  To clear and stop the timer,
1661
just simply write a `32'h00' to this register.
1662
 
1663
The Jiffies register is somewhat similar in that the register always changes.
1664
In this case, the register counts up, whereas the timer always counted down.
1665
Reads from this register, as shown in Tbl.~\ref{tbl:jiffybits},
1666
\begin{table}\begin{center}
1667
\begin{bitlist}
1668
31\ldots 0 & R & Current jiffy value\\\hline
1669
31\ldots 0 & W & Value/time of next interrupt\\\hline
1670
\end{bitlist}
1671
\caption{Jiffies Register Bits}\label{tbl:jiffybits}
1672
\end{center}\end{table}
1673
always return the time value contained in the register.  Writes greater than
1674
the current Jiffy value, that is where the new value minus the old value is
1675
greater than zero while ignoring truncation, will set a new Jiffy interrupt
1676
time.  At that time, the Jiffy vector will clear, and another interrupt time
1677
may either be written to it, or it will just continue counting without
1678
activating any more interrupts.
1679
 
1680
The Zip CPU also supports several counter peripherals, mostly in the way of
1681
process accounting.  This peripherals have a single register associated with
1682
them, shown in Tbl.~\ref{tbl:ctrbits}.
1683
\begin{table}\begin{center}
1684
\begin{bitlist}
1685
31\ldots 0 & R/W & Current counter value\\\hline
1686
\end{bitlist}
1687
\caption{Counter Register Bits}\label{tbl:ctrbits}
1688
\end{center}\end{table}
1689
Writes to this register set the new counter value.  Reads read the current
1690
counter value.
1691
 
1692
The current design operation of these counters is that of performance counting.
1693
Two sets of four registers are available for keeping track of performance.
1694
The first is a task counter.  This just counts clock ticks.  The second
1695
counter is a prefetch stall counter, then an master stall counter.  These
1696
allow the CPU to be evaluated as to how efficient it is.  The fourth and
1697
final counter is an instruction counter, which counts how many instructions the
1698
CPU has issued.
1699
 
1700
It is envisioned that these counters will be used as follows: First, every time
1701
a master counter rolls over, the supervisor (Operating System) will record
1702
the fact.  Second, whenever activating a user task, the Operating System will
1703
set the four user counters to zero.  When the user task has completed, the
1704
Operating System will read the timers back off, to determine how much of the
1705
CPU the process had consumed.
1706
 
1707 36 dgisselq
The final peripheral to discuss is the DMA controller.  This controller
1708
has four registers.  Of these four, the length, source and destination address
1709
registers should need no further explanation.  They are full 32--bit registers
1710
specifying the entire transfer length, the starting address to read from, and
1711
the starting address to write to.  The registers can be written to when the
1712
DMA is idle, and read at any time.  The control register, however, will need
1713
some more explanation.
1714
 
1715
The bit allocation of the control register is shown in Tbl.~\ref{tbl:dmacbits}.
1716
\begin{table}\begin{center}
1717
\begin{bitlist}
1718
31 & R & DMA Active\\\hline
1719
30 & R & Wishbone error, transaction aborted (cleared on any write)\\\hline
1720
29 & R/W & Set to '1' to prevent the controller from incrementing the source address, '0' for normal memory copy. \\\hline
1721
28 & R/W & Set to '0' to prevent the controller from incrementing the
1722
        destination address, '0' for normal memory copy. \\\hline
1723
27 \ldots 16 & W & The DMA Key.  Write a 12'hfed to these bits to start the
1724
        activate any DMA transfer.  \\\hline
1725
27 & R & Always reads '0', to force the deliberate writing of the key. \\\hline
1726
26 \ldots 16 & R & Indicates the number of items in the transfer buffer that
1727
        have yet to be written. \\\hline
1728
15 & R/W & Set to '1' to trigger on an interrupt, or '0' to start immediately
1729
        upon receiving a valid key.\\\hline
1730
14\ldots 10 & R/W & Select among one of 32~possible interrupt lines.\\\hline
1731
9\ldots 0 & R/W & Intermediate transfer length minus one.  Thus, to transfer
1732
        one item at a time set this value to 0. To transfer 1024 at a time,
1733
        set it to 1024.\\\hline
1734
\end{bitlist}
1735
\caption{DMA Control Register Bits}\label{tbl:dmacbits}
1736
\end{center}\end{table}
1737
This control register has been designed so that the common case of memory
1738
access need only set the key and the transfer length.  Hence, writing a
1739
\hbox{32'h0fed03ff} to the control register will start any memory transfer.
1740
On the other hand, if you wished to read from a serial port (constant address)
1741
and put the result into a buffer every time a word was available, you
1742
might wish to write \hbox{32'h2fed8000}--this assumes, of course, that you
1743
have a serial port wired to the zero bit of this interrupt control.  (The
1744
DMA controller does not use the interrupt controller, and cannot clear
1745
interrupts.)  As a third example, if you wished to write to an external
1746
FIFO anytime it was less than half full (had fewer than 512 items), and
1747
interrupt line 2 indicated this condition, you might wish to issue a
1748
\hbox{32'h1fed8dff} to this port.
1749
 
1750 33 dgisselq
\section{Debug Port Registers}
1751
Accessing the Zip System via the debug port isn't as straight forward as
1752
accessing the system via the wishbone bus.  The debug port itself has been
1753
reduced to two addresses, as outlined earlier in Tbl.~\ref{tbl:dbgregs}.
1754
Access to the Zip System begins with the Debug Control register, shown in
1755
Tbl.~\ref{tbl:dbgctrl}.
1756
\begin{table}\begin{center}
1757
\begin{bitlist}
1758
31\ldots 14 & R & Reserved\\\hline
1759
13 & R & CPU GIE setting\\\hline
1760
12 & R & CPU is sleeping\\\hline
1761
11 & W & Command clear PF cache\\\hline
1762
10 & R/W & Command HALT, Set to '1' to halt the CPU\\\hline
1763
9 & R & Stall Status, '1' if CPU is busy\\\hline
1764 36 dgisselq
8 & R/W & Step Command, set to '1' to step the CPU, also sets the halt bit\\\hline
1765 33 dgisselq
7 & R & Interrupt Request \\\hline
1766
6 & R/W & Command RESET \\\hline
1767
5\ldots 0 & R/W & Debug Register Address \\\hline
1768
\end{bitlist}
1769
\caption{Debug Control Register Bits}\label{tbl:dbgctrl}
1770
\end{center}\end{table}
1771
 
1772
The first step in debugging access is to determine whether or not the CPU
1773
is halted, and to halt it if not.  To do this, first write a '1' to the
1774
Command HALT bit.  This will halt the CPU and place it into debug mode.
1775
Once the CPU is halted, the stall status bit will drop to zero.  Thus,
1776
if bit 10 is high and bit 9 low, the debug port is open to examine the
1777
internal state of the CPU.
1778
 
1779
At this point, the external debugger may examine internal state information
1780
from within the CPU.  To do this, first write again to the command register
1781
a value (with command halt still high) containing the address of an internal
1782
register of interest in the bottom 6~bits.  Internal registers that may be
1783
accessed this way are listed in Tbl.~\ref{tbl:dbgaddrs}.
1784
\begin{table}\begin{center}
1785
\begin{reglist}
1786
sR0 & 0 & 32 & R/W & Supervisor Register R0 \\\hline
1787
sR1 & 0 & 32 & R/W & Supervisor Register R1 \\\hline
1788
sSP & 13 & 32 & R/W & Supervisor Stack Pointer\\\hline
1789
sCC & 14 & 32 & R/W & Supervisor Condition Code Register \\\hline
1790
sPC & 15 & 32 & R/W & Supervisor Program Counter\\\hline
1791
uR0 & 16 & 32 & R/W & User Register R0 \\\hline
1792
uR1 & 17 & 32 & R/W & User Register R1 \\\hline
1793
uSP & 29 & 32 & R/W & User Stack Pointer\\\hline
1794
uCC & 30 & 32 & R/W & User Condition Code Register \\\hline
1795
uPC & 31 & 32 & R/W & User Program Counter\\\hline
1796
PIC & 32 & 32 & R/W & Primary Interrupt Controller \\\hline
1797
WDT & 33 & 32 & R/W & Watchdog Timer\\\hline
1798
CCHE & 34 & 32 & R/W & Manual Cache Controller\\\hline
1799
CTRIC & 35 & 32 & R/W & Secondary Interrupt Controller\\\hline
1800
TMRA & 36 & 32 & R/W & Timer A\\\hline
1801
TMRB & 37 & 32 & R/W & Timer B\\\hline
1802
TMRC & 38 & 32 & R/W & Timer C\\\hline
1803
JIFF & 39 & 32 & R/W & Jiffies peripheral\\\hline
1804
MTASK & 40 & 32 & R/W & Master task clock counter\\\hline
1805
MMSTL & 41 & 32 & R/W & Master memory stall counter\\\hline
1806
MPSTL & 42 & 32 & R/W & Master Pre-Fetch Stall counter\\\hline
1807
MICNT & 43 & 32 & R/W & Master instruction counter\\\hline
1808
UTASK & 44 & 32 & R/W & User task clock counter\\\hline
1809
UMSTL & 45 & 32 & R/W & User memory stall counter\\\hline
1810
UPSTL & 46 & 32 & R/W & User Pre-Fetch Stall counter\\\hline
1811
UICNT & 47 & 32 & R/W & User instruction counter\\\hline
1812
\end{reglist}
1813
\caption{Debug Register Addresses}\label{tbl:dbgaddrs}
1814
\end{center}\end{table}
1815
Primarily, these ``registers'' include access to the entire CPU register
1816 36 dgisselq
set, as well as the internal peripherals.  To read one of these registers
1817 33 dgisselq
once the address is set, simply issue a read from the data port.  To write
1818
one of these registers or peripheral ports, simply write to the data port
1819
after setting the proper address.
1820
 
1821
In this manner, all of the CPU's internal state may be read and adjusted.
1822
 
1823
As an example of how to use this, consider what would happen in the case
1824
of an external break point.  If and when the CPU hits a break point that
1825
causes it to halt, the Command HALT bit will activate on its own, the CPU
1826
will then raise an external interrupt line and wait for a debugger to examine
1827
its state.  After examining the state, the debugger will need to remove
1828
the breakpoint by writing a different instruction into memory and by writing
1829
to the command register while holding the clear cache, command halt, and
1830
step CPU bits high, (32'hd00).  The debugger may then replace the breakpoint
1831
now that the CPU has gone beyond it, and clear the cache again (32'h500).
1832
 
1833
To leave this debug mode, simply write a `32'h0' value to the command register.
1834
 
1835
\chapter{Wishbone Datasheets}\label{chap:wishbone}
1836 32 dgisselq
The Zip System supports two wishbone ports, a slave debug port and a master
1837 21 dgisselq
port for the system itself.  These are shown in Tbl.~\ref{tbl:wishbone-slave}
1838
\begin{table}[htbp]
1839
\begin{center}
1840
\begin{wishboneds}
1841
Revision level of wishbone & WB B4 spec \\\hline
1842
Type of interface & Slave, Read/Write, single words only \\\hline
1843 24 dgisselq
Address Width & 1--bit \\\hline
1844 21 dgisselq
Port size & 32--bit \\\hline
1845
Port granularity & 32--bit \\\hline
1846
Maximum Operand Size & 32--bit \\\hline
1847
Data transfer ordering & (Irrelevant) \\\hline
1848
Clock constraints & Works at 100~MHz on a Basys--3 board\\\hline
1849
Signal Names & \begin{tabular}{ll}
1850
                Signal Name & Wishbone Equivalent \\\hline
1851
                {\tt i\_clk} & {\tt CLK\_I} \\
1852
                {\tt i\_dbg\_cyc} & {\tt CYC\_I} \\
1853
                {\tt i\_dbg\_stb} & {\tt STB\_I} \\
1854
                {\tt i\_dbg\_we} & {\tt WE\_I} \\
1855
                {\tt i\_dbg\_addr} & {\tt ADR\_I} \\
1856
                {\tt i\_dbg\_data} & {\tt DAT\_I} \\
1857
                {\tt o\_dbg\_ack} & {\tt ACK\_O} \\
1858
                {\tt o\_dbg\_stall} & {\tt STALL\_O} \\
1859
                {\tt o\_dbg\_data} & {\tt DAT\_O}
1860
                \end{tabular}\\\hline
1861
\end{wishboneds}
1862 22 dgisselq
\caption{Wishbone Datasheet for the Debug Interface}\label{tbl:wishbone-slave}
1863 21 dgisselq
\end{center}\end{table}
1864
and Tbl.~\ref{tbl:wishbone-master} respectively.
1865
\begin{table}[htbp]
1866
\begin{center}
1867
\begin{wishboneds}
1868
Revision level of wishbone & WB B4 spec \\\hline
1869 24 dgisselq
Type of interface & Master, Read/Write, single cycle or pipelined\\\hline
1870
Address Width & 32--bit bits \\\hline
1871 21 dgisselq
Port size & 32--bit \\\hline
1872
Port granularity & 32--bit \\\hline
1873
Maximum Operand Size & 32--bit \\\hline
1874
Data transfer ordering & (Irrelevant) \\\hline
1875
Clock constraints & Works at 100~MHz on a Basys--3 board\\\hline
1876
Signal Names & \begin{tabular}{ll}
1877
                Signal Name & Wishbone Equivalent \\\hline
1878
                {\tt i\_clk} & {\tt CLK\_O} \\
1879
                {\tt o\_wb\_cyc} & {\tt CYC\_O} \\
1880
                {\tt o\_wb\_stb} & {\tt STB\_O} \\
1881
                {\tt o\_wb\_we} & {\tt WE\_O} \\
1882
                {\tt o\_wb\_addr} & {\tt ADR\_O} \\
1883
                {\tt o\_wb\_data} & {\tt DAT\_O} \\
1884
                {\tt i\_wb\_ack} & {\tt ACK\_I} \\
1885
                {\tt i\_wb\_stall} & {\tt STALL\_I} \\
1886
                {\tt i\_wb\_data} & {\tt DAT\_I}
1887
                \end{tabular}\\\hline
1888
\end{wishboneds}
1889 22 dgisselq
\caption{Wishbone Datasheet for the CPU as Master}\label{tbl:wishbone-master}
1890 21 dgisselq
\end{center}\end{table}
1891
I do not recommend that you connect these together through the interconnect.
1892 24 dgisselq
Rather, the debug port of the CPU should be accessible regardless of the state
1893
of the master bus.
1894 21 dgisselq
 
1895 24 dgisselq
You may wish to notice that neither the {\tt ERR} nor the {\tt RETRY} wires
1896
have been implemented.  What this means is that the CPU is currently unable
1897
to detect a bus error condition, and so may stall indefinitely (hang) should
1898
it choose to access a value not on the bus, or a peripheral that is not
1899
yet properly configured.
1900 21 dgisselq
 
1901
\chapter{Clocks}\label{chap:clocks}
1902
 
1903 32 dgisselq
This core is based upon the Basys--3 development board sold by Digilent.
1904
The Basys--3 development board contains one external 100~MHz clock, which is
1905 36 dgisselq
sufficient to run the Zip CPU core.
1906 21 dgisselq
\begin{table}[htbp]
1907
\begin{center}
1908
\begin{clocklist}
1909
i\_clk & External & 100~MHz & 100~MHz & System clock.\\\hline
1910
\end{clocklist}
1911
\caption{List of Clocks}\label{tbl:clocks}
1912
\end{center}\end{table}
1913
I hesitate to suggest that the core can run faster than 100~MHz, since I have
1914
had struggled with various timing violations to keep it at 100~MHz.  So, for
1915
now, I will only state that it can run at 100~MHz.
1916
 
1917
 
1918
\chapter{I/O Ports}\label{chap:ioports}
1919 33 dgisselq
The I/O ports to the Zip CPU may be grouped into three categories.  The first
1920
is that of the master wishbone used by the CPU, then the slave wishbone used
1921
to command the CPU via a debugger, and then the rest.  The first two of these
1922
were already discussed in the wishbone chapter.  They are listed here
1923
for completeness in Tbl.~\ref{tbl:iowb-master}
1924
\begin{table}
1925
\begin{center}\begin{portlist}
1926
{\tt o\_wb\_cyc}   &  1 & Output & Indicates an active Wishbone cycle\\\hline
1927
{\tt o\_wb\_stb}   &  1 & Output & WB Strobe signal\\\hline
1928
{\tt o\_wb\_we}    &  1 & Output & Write enable\\\hline
1929
{\tt o\_wb\_addr}  & 32 & Output & Bus address \\\hline
1930
{\tt o\_wb\_data}  & 32 & Output & Data on WB write\\\hline
1931
{\tt i\_wb\_ack}   &  1 & Input  & Slave has completed a R/W cycle\\\hline
1932
{\tt i\_wb\_stall} &  1 & Input  & WB bus slave not ready\\\hline
1933
{\tt i\_wb\_data}  & 32 & Input  & Incoming bus data\\\hline
1934
\end{portlist}\caption{CPU Master Wishbone I/O Ports}\label{tbl:iowb-master}\end{center}\end{table}
1935
and~\ref{tbl:iowb-slave} respectively.
1936
\begin{table}
1937
\begin{center}\begin{portlist}
1938
{\tt i\_wb\_cyc}   &  1 & Input & Indicates an active Wishbone cycle\\\hline
1939
{\tt i\_wb\_stb}   &  1 & Input & WB Strobe signal\\\hline
1940
{\tt i\_wb\_we}    &  1 & Input & Write enable\\\hline
1941
{\tt i\_wb\_addr}  &  1 & Input & Bus address, command or data port \\\hline
1942
{\tt i\_wb\_data}  & 32 & Input & Data on WB write\\\hline
1943
{\tt o\_wb\_ack}   &  1 & Output  & Slave has completed a R/W cycle\\\hline
1944
{\tt o\_wb\_stall} &  1 & Output  & WB bus slave not ready\\\hline
1945
{\tt o\_wb\_data}  & 32 & Output  & Incoming bus data\\\hline
1946
\end{portlist}\caption{CPU Debug Wishbone I/O Ports}\label{tbl:iowb-slave}\end{center}\end{table}
1947 21 dgisselq
 
1948 33 dgisselq
There are only four other lines to the CPU: the external clock, external
1949
reset, incoming external interrupt line(s), and the outgoing debug interrupt
1950
line.  These are shown in Tbl.~\ref{tbl:ioports}.
1951
\begin{table}
1952
\begin{center}\begin{portlist}
1953
{\tt i\_clk} & 1 & Input & The master CPU clock \\\hline
1954
{\tt i\_rst} & 1 & Input &  Active high reset line \\\hline
1955
{\tt i\_ext\_int} & 1\ldots 6 & Input &  Incoming external interrupts \\\hline
1956
{\tt o\_ext\_int} & 1 & Output & CPU Halted interrupt \\\hline
1957
\end{portlist}\caption{I/O Ports}\label{tbl:ioports}\end{center}\end{table}
1958
The clock line was discussed briefly in Chapt.~\ref{chap:clocks}.  We
1959
typically run it at 100~MHz.  The reset line is an active high reset.  When
1960
asserted, the CPU will start running again from its reset address in
1961
memory.  Further, depending upon how the CPU is configured and specifically on
1962
the {\tt START\_HALTED} parameter, it may or may not start running
1963
automatically.  The {\tt i\_ext\_int} line is for an external interrupt.  This
1964
line may be as wide as 6~external interrupts, depending upon the setting of
1965
the {\tt EXTERNAL\_INTERRUPTS} line.  As currently configured, the ZipSystem
1966
only supports one such interrupt line by default.  For us, this line is the
1967
output of another interrupt controller, but that's a board specific setup
1968
detail.  Finally, the Zip System produces one external interrupt whenever
1969
the CPU halts to wait for the debugger.
1970
 
1971 36 dgisselq
\chapter{Initial Assessment}\label{chap:assessment}
1972
 
1973
Having now worked with the Zip CPU for a while, it is worth offering an
1974
honest assessment of how well it works and how well it was designed. At the
1975
end of this assessment, I will propose some changes that may take place in a
1976
later version of this Zip CPU to make it better.
1977
 
1978
\section{The Good}
1979
\begin{itemize}
1980
\item The Zip CPU is light weight and fully featured as it exists today. For
1981
        anyone who wishes to build a general purpose CPU and then to
1982
        experiment with building and adding particular features, the Zip CPU
1983
        makes a good starting point--it is fairly simple. Modifications should
1984
        be simple enough.
1985
\item As an estimate of the ``weight'' of this implementation, the CPU has
1986
        cost me less than 150 hours to implement from its inception.
1987
\item The Zip CPU was designed to be an implementable soft core that could be
1988
        placed within an FPGA, controlling actions internal to the FPGA. It
1989
        fits this role rather nicely. It does not fit the role of a system on
1990
        a chip very well, but then it was never intended to be a system on a
1991
        chip but rather a system within a chip.
1992
\item The extremely simplified instruction set of the Zip CPU was a good
1993
        choice. Although it does not have many of the commonly used
1994
        instructions, PUSH, POP, JSR, and RET among them, the simplified
1995
        instruction set has demonstrated an amazing versatility. I will contend
1996
        therefore and for anyone who will listen, that this instruction set
1997
        offers a full and complete capability for whatever a user might wish
1998
        to do with two exceptions: bytewise character access and accelerated
1999
        floating-point support.
2000
\item This simplified instruction set is easy to decode.
2001
\item The simplified bus transactions (32-bit words only) were also very easy
2002
        to implement.
2003
\item The novel approach of having a single interrupt vector, which just
2004
        brings the CPU back to the instruction it left off at within the last
2005
        interrupt context doesn't appear to have been that much of a problem.
2006
        If most modern systems handle interrupt vectoring in software anyway,
2007
        why maintain hardware support for it?
2008
\item My goal of a high rate of instructions per clock may not be the proper
2009
        measure. For example, if instructions are being read from a SPI flash
2010
        device, such as is common among FPGA implementations, these same
2011
        instructions may suffer stalls of between 64 and 128 cycles per
2012
        instruction just to read the instruction from the flash. Executing the
2013
        instruction in a single clock cycle is no longer the appropriate
2014
        measure. At the same time, it should be possible to use the DMA
2015
        peripheral to copy instructions from the FLASH to a temporary memory
2016
        location, after which they may be executed at a single instruction
2017
        cycle per access again.
2018
\end{itemize}
2019
 
2020
\section{The Not so Good}
2021
\begin{itemize}
2022
\item While one of the stated goals was to use a small amount of logic,
2023
        3k~LUTs isn't that impressively small. Indeed, it's really much
2024
        too expensive when compared against other 8 and 16-bit CPUs that have
2025
        less than 1k LUTs.
2026
 
2027
        Still, \ldots it's not bad, it's just not astonishingly good.
2028
 
2029
\item The fact that the instruction width equals the bus width means that the
2030
        instruction fetch cycle will always be interfering with any load or
2031
        store memory operation, with the only exception being if the
2032
        instruction is already in the cache.  {\em This has become the
2033
        fundamental limit on the speed and performance of the CPU!}
2034
        Those familiar with the Von--Neumann approach of sharing a bus
2035
        between data and instructions will not be surprised by this assessment.
2036
 
2037
        This could be fixed in one of three ways: the instruction set
2038
        architecture could be modified to handle Very Long Instruction Words
2039
        (VLIW) so that each 32--bit word would encode two or more instructions,
2040
        the instruction fetch bus width could be increased from 32--bits to
2041
        64--bits or more, or the instruction bus could be separated from the
2042
        data bus.  Any and all of these approaches would increase the overall
2043
        LUT count.
2044
 
2045
\item The (non-existant) floating point unit was an after-thought, isn't even
2046
        built as a potential option, and most likely won't support the full
2047
        IEEE standard set of FPU instructions--even for single point precision.
2048
        This (non-existant) capability would benefit the most from an
2049
        out-of-order execution capability, which the Zip CPU does not have.
2050
 
2051
        Still, sharing FPU registers with the main register set was a good
2052
        idea and worth preserving, as it simplifies context swapping.
2053
 
2054
        Perhaps this really isn't a problem, but rather a feature.  By not
2055
        implementing FPU instructions, the Zip CPU maintains a lower LUT count
2056
        than it would have if it did implement these instructions.
2057
 
2058
\item The CPU has no character support. This is both good and bad.
2059
        Realistically, the CPU works just fine without it. Characters can be
2060
        supported as subsets of 32-bit words without any problem. Practically,
2061
        though, it will make compiling non-Zip CPU code difficult--especially
2062
        anything that assumes sizeof(int)=4*sizeof(char), or that tries to
2063
        create unions with characters and integers and then attempts to
2064
        reference the address of the characters within that union.
2065
 
2066
\item The Zip CPU does not support a data cache. One can still be built
2067
        externally, but this is a limitation of the CPU proper as built.
2068
        Further, under the theory of the Zip CPU design (that of an embedded
2069
        soft-core processor within an FPGA, where any ``address'' may reference
2070
        either memory or a peripheral that may have side-effects), any data
2071
        cache would need to be based upon an initial knowledge of whether or
2072
        not it is supporting memory (cachable) or peripherals. This knowledge
2073
        must exist somewhere, and that somewhere is currently (and by design)
2074
        external to the CPU.
2075
 
2076
        This may also be written off as a ``feature'' of the Zip CPU, since
2077
        the addition of a data cache can greatly increase the LUT count of
2078
        a soft core.
2079
 
2080
\item Many other instruction sets offer three operand instructions, whereas
2081
        the Zip CPU only offers two operand instructions. This means that it
2082
        takes the Zip CPU more instructions to do many of the same operations.
2083
        The good part of this is that it gives the Zip CPU a greater amount of
2084
        flexibility in its immediate operand mode, although that increased
2085
        flexibility isn't necessarily as valuable as one might like.
2086
 
2087
\item The Zip CPU does not currently detect and trap on either illegal
2088
        instructions or bus errors.  Attempts to access non--existent
2089
        memory quietly return erroneous results, rather than halting the
2090
        process (user mode) or halting or resetting the CPU (supervisor mode).
2091
 
2092
\item The Zip CPU doesn't support out of order execution. I suppose it could
2093
        be modified to do so, but then it would no longer be the ``simple''
2094
        and low LUT count CPU it was designed to be. The two primary results
2095
        are that 1) loads may unnecessarily stall the CPU, even if other
2096
        things could be done while waiting for the load to complete, 2)
2097
        bus errors on stores will never be caught at the point of the error,
2098
        and 3) branch prediction becomes more difficult.
2099
 
2100
\item Although switching to an interrupt context in the Zip CPU design doesn't
2101
        require a tremendous swapping of registers, in reality it still
2102
        does--since any task swap still requires saving and restoring all
2103
        16~user registers. That's a lot of memory movement just to service
2104
        an interrupt.
2105
 
2106
\item The Zip CPU is by no means generic: it will never handle addresses
2107
        larger than 32-bits (16GB) without a complete and total redesign.
2108
        This may limit its utility as a generic CPU in the future, although
2109
        as an embedded CPU within an FPGA this isn't really much of a limit
2110
        or restriction.
2111
 
2112
\item While the Zip CPU has its own assembler, it has no linker and does not
2113
        (yet) support a compiler. The standard C library is an even longer
2114
        shot. My dream of having binutils and gcc support has not been
2115
        realized and at this rate may not be realized. (I've been intimidated
2116
        by the challenge everytime I've looked through those codes.)
2117
 
2118
\item While the Wishbone Bus (B4) supports a pipelined mode with single cycle
2119
        execution, the Zip CPU is unable to exploit this parallelism. Instead,
2120
        apart from the DMA and the pipelined prefetch, all loads and stores
2121
        are single wishbone bus operations requiring a minimum of 3 clocks.
2122
        (In practice, this has turned into 7-clocks.)
2123
 
2124
\iffalse
2125
\item There is no control over whether or not an instruction sets the
2126
        condition codes--certain instructions always set the condition codes,
2127
        other instructions never set them. This effectively limits conditional
2128
        instructions to a single instruction only (with two or more
2129
        instructions as an exception), as the first instruction that sets
2130
        condition codes will break the condition code chain.
2131
 
2132
        {\em (A proposed change below address this.)}
2133
 
2134
\item Using the CC register as a trap address was a bad idea--it limits the CC
2135
        registers ability to be used in future expansion, such as by adding
2136
        exception indication flags: bus error, floating point exception, etc.
2137
 
2138
        {\em (This can be changed by a different O/S implementation of the trap
2139
        instruction.)}
2140
\item The current implementation suffers from too many stalls on any
2141
        branch--even if the branch is known early on.
2142
 
2143
        {\em (This is addressed in proposals below.)}
2144
        % Addressed, 20150918
2145
 
2146
\item In a similar fashion, a switch to interrupt context forces the pipeline
2147
        to be cleared, whereas it might make more sense to just continue
2148
        executing the instructions already in the pipeline while the prefetch
2149
        stage is working on switching to the interrupt context.
2150
 
2151
        {\em (Also addressed in proposals below.)}
2152
        % This should happen so rarely that it is not really a problem
2153
\fi
2154
 
2155
\end{itemize}
2156
 
2157
\section{The Next Generation}
2158
This section could also be labeled as my ``To do'' list.
2159
 
2160
Given the feedback listed above, perhaps its time to consider what changes could be made to improve the Zip CPU in the future. I offer the following as proposals:
2161
 
2162
\begin{itemize}
2163
\item {\bf Remove the low LUT goal.} It wasn't really achieved, and the
2164
        proposals below will only increase the amount of logic the Zip CPU
2165
        requires.  While I expect that the Zip CPU will always be somewhat
2166
        of a light weight, it will never be the smallest kid on the block.
2167
 
2168
        I'm actually struggling with this idea.  The whole goal of the Zip
2169
        CPU was to be light weight.  Wouldn't it make more sense to create and
2170
        maintain options whereby it would remain lightweight?  For example, if
2171
        the process accounting registers are anything but light weight, why
2172
        keep them?  Why not instead make some compile flags that just turn them
2173
        off, keeping the CPU lightweight?  The same holds for the prefetch
2174
        cache.
2175
 
2176
\iffalse
2177
\item {\bf Adjust the Zip CPU so that conditional instructions do not set
2178
        flags}, although they may explicitly set condition codes if writing
2179
        to the CC register.
2180
 
2181
        This is a simple change to the core, and may show up in new releases.
2182
        % Fixed, 20150918
2183
\fi
2184
 
2185
\item The `{\tt .V}' condition was never used in any code other than my test
2186
        code.  Suggest changing it to a `{\tt .LE}' condition, which seems
2187
        to be more useful.
2188
 
2189
\iffalse
2190
\item Add in an {\bf unpredictable branch delay slot}, so that on any branch
2191
        the delay slot may or may not be executed before the branch.
2192
        Instructions that do not depend upon the branch, and that should be
2193
        executed were the branch not taken, could be placed into the delay
2194
        slot. Thus, if the branch isn't taken, we wouldn't suffer the stall,
2195
        whereas it wouldn't affect the timing of the branch if taken. It would
2196
        just do something irrelevant.
2197
 
2198
        % Changes made, 20150918, make this option no longer relevant
2199
 
2200
\item {\bf Re-engineer Branch Processing.}  There's no reason why a {\tt BRA}
2201
        instruction should create five stall cycles.  The decode stage, plus
2202
        the prefetch engine, should be able to drop this number of stalls via
2203
        better branch handling.
2204
 
2205
        Indeed, this could turn into a simple means of branch prediction:
2206
        if {\tt BRA} suffered a single stall only, whereas {\tt BRA.C}
2207
        suffered five stalls, then {\tt BRA.!C} followed by {\tt BRA} would
2208
        be faster than a {\tt BRA.C} instruction.  This would then allow a
2209
        compiler to do explicit branch optimizations.
2210
 
2211
        Of course, longer branches using {\tt ADD X,PC} would still not be
2212
        optimized.
2213
 
2214
        % DONE: 20150918 -- to include the ADD X,PC instructions
2215
 
2216
\item {\bf Request bus access for Load/Store two cycles earlier.}  The problem
2217
        here is the contention for the bus between the memory unit and the
2218
        prefetch unit.  Currently, the memory unit must ask the prefetch
2219
        unit to release the bus if it is in the middle of a bus cycle.  At this
2220
        point, the prefetch drops the {\tt STB} line on the next clock and must
2221
        then wait for the last {\tt ACK} before releasing the bus.  If the
2222
        request takes one clock, dropping the strobe line the next, waiting
2223
        for an acknowledgement takes another, and then the bus must be idle
2224
        for one cycle before starting again, these extra cycles add up.
2225
        It should be possible to tell the prefetch stage to give up the bus
2226
        as soon as the decoder knows the instruction will need the bus.
2227
        Indeed, if done in the decode stage, this might drop the seven cycle
2228
        access down by two cycles.
2229
 
2230
        % FIXED: 20150918
2231
\fi
2232
 
2233
\item {\bf Consider a more traditional Instruction Cache.}  The current
2234
        pipelined instruction cache just reads a window of memory into
2235
        its cache.  If the CPU leaves that window, the entire cache is
2236
        invalidated.  A more traditional cache, however, might allow
2237
        common subroutines to stay within the cache without invalidating the
2238
        entire cache structure.
2239
 
2240
\iffalse
2241
\item {\bf Very Long Instruction Word (VLIW).}  Now, to speed up operation, I
2242
        propose that the Zip CPU instruction set be modified towards a Very
2243
        Long Instruction Word (VLIW) implementation. In this implementation,
2244
        an instruction word may contain either one or two separate
2245
        instructions. The first instruction would take up the high order bits,
2246
        the second optional instruction the lower 16-bits. Further, I propose
2247
        that any of the ALU instructions (SUB through LSR) automatically have
2248
        a second instruction whenever their operand `B' is a register, and use
2249
        the full 20-bit immediate if not. This will effectively eliminate the
2250
        register plus immediate mode for all of these instructions.
2251
 
2252
        This is the minimal required change to increase the number of
2253
        instructions per clock cycle. Other changes would need to take place
2254
        as well to support this. These include:
2255
        \begin{itemize}
2256
        \item Instruction words containing two instructions would take two
2257
                clocks to complete, while requiring only a single cycle
2258
                instruction fetch.
2259
        \item Instructions preceded by a label in the asseembler must always
2260
                start in the high order word.
2261
        \item VLIW's, once started, must always execute to completion. The
2262
                upper word may set the PC, the lower word may not. Regardless
2263
                of whether the upper word sets the PC, the lower word must
2264
                still be guaranteed to complete before the PC changes. On any
2265
                switch to (or from) interrupt context, both instructions must
2266
                complete or none of the instructions in the word shall
2267
                complete prior to the switch.
2268
        \item STEP commands and BREAK instructions will only take place after
2269
                the entire word is executed.
2270
        \end{itemize}
2271
 
2272
        If done well, the assembler should be able to handle these changes
2273
        with the biggest impacts to the user being increased performance and
2274
        a loss of the register plus immediate ALU modes. (These weren't really
2275
        relevant for the XOR, OR, AND, etc. operations anyway.) Machine code
2276
        compatibility will not be maintained.
2277
 
2278
        A proposed secondary instruction set might consist of: a four bit
2279
        operand (any of the prior instructions would be supported, with some
2280
        exceptions such as moves to and from user registers while in
2281
        supervisor mode not being supported), a 4-bit register result (PC not
2282
        allowed), a 3-bit conditional (identical to the conditional for the
2283
        upper word), a single bit for whether or not an immediate is present
2284
        or not, followed by either a 4-bit register or a 4-bit signed
2285
        immediate. The multiply instruction would steal the immediate flag to
2286
        be used as a sign indication, forcing both operands to be registers
2287
        without any immediate offsets.
2288
 
2289
        {\em Initial conversion of several library functions to this secondary
2290
        instruction set has demonstrated little to no gain.   The problem was
2291
        that the new instruction set was made by joining a rarely used
2292
        instruction (ALU with register and not immediate) with a more common
2293
        instruction.  The utility was then limited by the utility of the rare
2294
        instrction, which limited the impact of the entire approach.  }
2295
\else
2296
\item {\bf Very Long Instruction Word (VLIW).}  The goal here would be to
2297
        create a new instruction set whereby two instructions would be encoded
2298
        in each 32--bit word.  While this may speed up
2299
        CPU operation, it would necessitate an instruction redesign.
2300
\fi
2301
 
2302
\end{itemize}
2303
 
2304
 
2305 21 dgisselq
% Appendices
2306
% Index
2307
\end{document}
2308
 
2309
 

powered by: WebSVN 2.1.0

© copyright 1999-2022 OpenCores.org, equivalent to Oliscience, all rights reserved. OpenCores®, registered trademark.