File: TODO

package info (click to toggle)
simulavr 1.0.0+git20160221.e53413b-1
  • links: PTS
  • area: main
  • in suites: bullseye, buster, sid
  • size: 5,748 kB
  • sloc: cpp: 35,491; python: 6,991; ansic: 3,567; makefile: 1,072; sh: 653; asm: 414; tcl: 320
file content (117 lines) | stat: -rw-r--r-- 5,110 bytes parent folder | download
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
#
#  $Id$
#

+ Host specific issues
  - Cygwin: Basic Simulator Functionality Works.  Need to check Tcl/X.
  - Mingw: Uses BSD sockets.  Need to have WinSock as alternative
  - check for makeinfo/texi2html program, else a full build will fail!

+ Feature parity with avrtest.  Known issues:
  - Verify gcc test suite can be run with avrtest

+ Feature parity with simulavr.  Known issues:
  - avr-libc mostly tests with AT90S8515 and ATmega128 so most is OK
  - Models not yet supported:
      + AT90S2313 -
      + AT90S4414 - essentially AT90S8515 w/less memory so easy
      + AT90S8515 -
      + ATmega8 - could be built based on the existing AT90S4433 code
      + ATmega16 -
      + ATmega32 -
      + AT90S1200 - different, can be ignored as obsolete
      + ATmega103 - can be ignored as obsolete
      + range of AT43USBxxx controllers - can be ignored as replaced (sorta)
        by the AT90USB* / ATmega???U? devices.
   - simulavr has a core dump feature

+ Configure issues
  - Joel's Cygwin has texi2pdf but not a valid TeX install.  It configures
    OK but fails when building documentation.
  - libiconv.a and libintl.a are not needed on Fedora 10 but required on
    Cygwin.  They are forced in on Cygwin now.  Can we know dynamically
    if these are needed?

+ Known limitations of the M128 as the most comprehensive currently simulated CPU
  - TWI Serial Interface missing
  - Analog to Digital Converter Subsystem need to be validated
  - Self-Programming not supported
  - Boot Loader Support (fuses, different reset vector, switching interrupt
    vector table, ...) not supported
  - Real Time Clock missing
  - Watchdog Timer status unclear
  - Sleep-command not working
  - Reset-pin is not available. Also different reset reasons are not supported
  - With activating the Tx-Pin of an UART the DDR-Register is not set properly
    to output. Workaround: Set the Pin's default value to PULLUP. While the
    Pin behaves as Open Colletor (pulls down only) the pull-up "resistor" lets
    the system run as it should.
  - There are always 64kByte of external memory automatically attached to the
    Mega128. Different memory sizes are not supported.
  - Segments dfferent than .text, .data, .bss are not supported

+ Known limitation of other CPU-models
  - While Atmel changed some function details of the EEPROM, Watchdog Timer,
    Timer Subsystem, ADC, and USART / USI these subsystems have identical
    names but different functions.  Therefore adding a new CPU needs a detailed
    analysis of the functions of each IO-subsystem
  - CPUs with 24bit program counter are not supported
  - USI-subsystem is missing
  - CAN-subsystem is missing
  - USB subsystem is missing
  - X-Mega specifics are missing

+ Known issues of the simulator
  - LCD subsystem is simlified to
    - only a 4*20 character type is available
    - the timing is fixed as described for the 5V-versions
    - only the 4 bit interface is supported. At start-up the commands are
      interpreted as if an eight bit interface is available (one write cycle
      per command). After finishing the initialization switching to the four
      bit interface is permitted at any time.
    - the graphic representation is character based. Display of of characters
       follows the rules of your display, not of the LCD character generator.
    - loadable characters are not supported.
    - reading of display is not supported.
    - reading of busy flag does not give the current address in the lower bits.
    - scrolling not supported.
    - shift right / left of the display content is not supported.
    - only one character set is supported - based on your diplay font.

  - Scope subsystem is an empty frame

  - Serial Rx/Tx is simplfied to
    - only 8 bit data transmission is supported
    - 9 bit handling (MPM) is not supported

  - Waiting at a breakpoint for a GDB user input: the active wait-loop should
    be replaced by something using less CPU power.

+ Testing
  - All tests need to be integrated into a nice regression test suite (Maybe
    python unittest?)
    Right now, the tests are split up and there is a custom made regression
    test suite for testing the handling of instructions in the core.

+ Proper teardown
  Currently, there are various points in the simulator where a simple
  exit(x) is performed. This may hurt additional instrumentation software
  which may need to close files or it may hurt any external program
  which is controlling simulavr.

+ State saving support
  Saving/restoring the complete state of the simulation (this needs
  a lot of infrastructure)

+ Export additional hardware (LCDs etc.) to verilog

+ Clean up the private/protected/public declaration scopes
  for the various classes. This is quite inconsistent.

+ Tracing/Dumping:
  - Make the device names configurable and not only a simple index into a list.
    Do not update bxxxx values in trace output.
  - Implement PCb back.
  - Bug, nothing in SREG? Why SREG twice?
  - Have an optional constraint on the number of bits in trace_direct (to e.g.
    only trace the necessary stack bits).