Scott, take a look at the interrupt-driven input of my posting to the TI e2e
site. Using that technique, you can collect the input string, saving it in an
input buffer via interrupt, let the interrupt routine notify main() when the
string terminator arrives, then process the string in main() (outside the
interrupt), with something like:
sscanf( Inbuf, fmt, ...);.
That eliminates the input wait-loop that you mentioned.
In my example, the input interrupt routine is not putting the incoming
characters into Inbuf, but is only sending a flag to main(), and main() is
printing out a simple response, just to show that the technique works.
Hope this helps,
Jim Smith
--- In m..., "swhitney14" wrote: >
> Hi Jim,
>
> Thank you for the link and the fast answer. Your example looks great, and I
have been able to implement printf, putc, and puts on uartA1 on my board. I
cannot imagine how much longer it would have taken without your help and that of
the others in this group. The CCSv4 documentation on device drivers is simply
dismal, and not having an out-of-the-box way to connect the standard I/O to an
on-chip resource is nearly inexcusable. Good for IAR users, I guess.
>
> Regarding the second part of my question, has anyone come up with an
equivalent solution for scanf/gets/getc? Formatted input is equally important
to us, and would greatly help with generation of a more elaborate command parser
for development testing and field use.
>
> I assume that similar low-level routines are available for the innards of
scanf, and there may be routines that we can replace for something like _inc()
and _ins(), similar to _outc() and _outs().
>
> Also, has anyone implemented a buffering scheme to provide interrupt driven
transmission of output data to prevent the "busy-wait" required for something
like:
>
> while(!IFG2 & UCA0TXIFG) ; // wait for room in Tx register
>
> Thanks again, Jim and Hardy!
>
Thank you for the link and the fast answer. Your example looks great, and I
have been able to implement printf, putc, and puts on uartA1 on my board. I
cannot imagine how much longer it would have taken without your help and that of
the others in this group. The CCSv4 documentation on device drivers is simply
dismal, and not having an out-of-the-box way to connect the standard I/O to an
on-chip resource is nearly inexcusable. Good for IAR users, I guess.
Regarding the second part of my question, has anyone come up with an equivalent
solution for scanf/gets/getc? Formatted input is equally important to us, and
would greatly help with generation of a more elaborate command parser for
development testing and field use.
I assume that similar low-level routines are available for the innards of scanf,
and there may be routines that we can replace for something like _inc() and
_ins(), similar to _outc() and _outs().
Also, has anyone implemented a buffering scheme to provide interrupt driven
transmission of output data to prevent the "busy-wait" required for something
like:
while(!IFG2 & UCA0TXIFG) ; // wait for room in Tx register
Jim Smith
PS: I had tried (and failed, miserably) to come up with a device driver, as you
mentioned below. That's when I started this thread and started getting
helpful suggestions right away!
--- In m..., "swhitney14" wrote: >
> Jim,
>
> This looks like terrific work, and thanks to you and to Hardy. I have been
trying to find out how to actually install a device driver using CCSv4, and the
documentation is pretty useless. The manual only talks about writing open(),
close(), read(), write(), rename(), unlink(), and then installing the new device
driver using add_device(). Sounds great, but I cannot find any decent examples
of connecting a serial port in this manner. At least your method provides a
reasonable alternative.
>
> On the flip side, have you or anyone else in the group written some code to
implement an equivalent of scanf() for UART input? Formatted I/O in both
directions is essential for the project I am working on, so any pointers are
greatly appreciated!
>
> Also, it would be helpful to have the transmit interrupt handler installed, so
that busy-wait loops for the transmit buffer to become empty could be avoided.
In many embedded "super-loop" applications, having to sit in a busy-wait for any
extended period of time is going to negatively impact overall system
performance.
>
> I plan to try your posted solution on the printf() side, and also am looking
forward to hearing about possible scanf() implementations for a UART. I have
asked this question on the TI E2E board, but no luck so far...
>
> Scott
>
Reply by swhitney14●August 4, 20112011-08-04
Jim,
This looks like terrific work, and thanks to you and to Hardy. I have been
trying to find out how to actually install a device driver using CCSv4, and the
documentation is pretty useless. The manual only talks about writing open(),
close(), read(), write(), rename(), unlink(), and then installing the new device
driver using add_device(). Sounds great, but I cannot find any decent examples
of connecting a serial port in this manner. At least your method provides a
reasonable alternative.
On the flip side, have you or anyone else in the group written some code to
implement an equivalent of scanf() for UART input? Formatted I/O in both
directions is essential for the project I am working on, so any pointers are
greatly appreciated!
Also, it would be helpful to have the transmit interrupt handler installed, so
that busy-wait loops for the transmit buffer to become empty could be avoided.
In many embedded "super-loop" applications, having to sit in a busy-wait for any
extended period of time is going to negatively impact overall system
performance.
I plan to try your posted solution on the printf() side, and also am looking
forward to hearing about possible scanf() implementations for a UART. I have
asked this question on the TI E2E board, but no luck so far...
Scott
--- In m..., "Jim" wrote: >
> Well, attaching a zip file to an email reply didn't work, so here is my
little test program (source) for uart_printf(). If you are reading this on the
http://tech.groups.yahoo.com/group/msp430/messages site, on the upper-right,
under 'Show Message Info', select 'Use Fixed Width Font' to
get proper formatting of the code below. Now, you can cut and paste the code
into your editor, if you like.
>
> Jim Smith
>
> //***** Begin uart_printf_3.c ************************************
>
> //************************************************************************
> // MSP430x26x Demo - USCI_A0, Ultra-Low Pwr UART 1200 Echo ISR, 32kHz ACLK
> //
> // Simple program to get low-level character and string output to UART
> // working, and interfaced to printf() (or its replacement), using TI's
> // CCSv4 Compiler/IDE.
> // ACLK = BRCLK = LFXT1 = 32768Hz, MCLK = SMCLK = DCO ~8.0MHz
> // Baud rate divider with 32768Hz XTAL @1200 = 32768Hz/1200 = 27.31
> // ***** An external watch crystal is required on XIN XOUT for ACLK *****
> //
> // MSP430F261x/241x
> // -----------------
> // /|\| XIN|-
> // | | | 32kHz
> // --|RST XOUT|-
> // | |
> // | P3.4/UCA0TXD|------------>
> // | | 1200 - 7E1
> // | P3.5/UCA0RXD|<------------
> //
> // Started with, and modified TI example MSP430x261x_uscia0_uart_03.c
> // Jim Smith
> // Aug 2, 2011
> // Built with CCS Version: 4.2.4
> //************************************************************************
> #include
> #include
> #include
> #include
> #include "msp430x26x.h"
>
> char ch = 'A';
> int i = 325;
> float j = 234.5678;
> //----- Code from Hardy Griech, pretty-much verbatim ---------------------
> // js used _printfi() instead of _printfi_nofloat(), ie full floating
> // point. This requires '--printf_support=full' when linking. You
may
> // select this in the CCS IDE by: r-click on your project name ->
> // C/C++ Build -> Library Function Assumptions -> choose full as level
> // of printf support. Do it for both Debug & Release Builds.
>
> //extern int _printfi_nofloat(char const **_format, va_list _ap,
> extern int _printfi(char const **_format, va_list _ap, // js changed
> void *_op, int (*_outc)(char c, void *op),
> int (*_outs)(char *s, void *op));
>
> void uart_putc( uint8_t c )
> {
> // send character directly to UART or send it to a buffer
> // to decouple character output
> while(!(IFG2 & UCA0TXIFG)); // js added
> UCA0TXBUF = c; // js added
> }
> void uart_puts( const char *s )
> {
> while ( *s != '\0') {
> uart_putc( (uint8_t)*(s++) );
> }
> } // uart_puts
> static int _outc( char c, void *_op )
> {
> uart_putc( (uint8_t)c );
> return 1;
> } // _outc
> static int _outs( char *s, void *_op )
> {
> size_t len = strlen( s );
> uart_puts( s );
> return (int)len;
> } // _outs
> void uart_printf( const char *fmt, ... ) {
> va_list _ap;
>
> va_start( _ap, fmt );
> // _printfi_nofloat( &fmt, _ap, NULL, _outc, _outs);
> _printfi( &fmt, _ap, NULL, _outc, _outs); // js changed
> va_end( _ap );
> } // uart_printf
> //----- End code from Hardy Griech ---
>
> //----- MAIN fuction -----------------
> void main(void)
> {
> WDTCTL = WDTPW + WDTHOLD; // Stop WDT
> //----- Setup the UART -----------
> P3OUT &= ~(BIT4+BIT5); // From TI example, is this necessary?
> P3SEL = 0x30; // P3.4,5 = USCI_A0 TXD/RXD
> UCA0CTL1 |= (UCSSEL_1 + UCBRKIE); // CLK = ACLK, BRK int enabled
> UCA0BR0 = 27; // 1200 baud: 32768/1200 = 27.31
> UCA0BR1 = 0;
> UCA0MCTL = UCBRS_2; // Modulation UCBRSx = 2 (UG pg 15-30)
> UCA0CTL0 |= (UCPEN | UCPAR | UC7BIT );// Parity Enable, Even, 7-bit
> UCA0CTL1 &= ~UCSWRST; // **Initialize USCI state machine**
> IE2 |= UCA0RXIE; // Enable USCI_A0 RX interrupt
>
> //----- Sign-on ------------------
> // Insert some delay for everything to stabilize
> for (i = 0; i < 15; i++) { // 15 rollovers @ 1.1 mHz, about 1 sec
delay
> __delay_cycles(65535);
> } // This delay cleared up garbage when output "Hello, World!", below.
> // i should be 15, because of its use in for-loop, above.
> uart_puts("\r\nHello, World!\r\n");
> uart_printf("\
> This is just to prove uart_putc(), uart_puts() and uart_printf()
work.\r\n\n\
> char: %c\r\n\
> Integer: %d\r\n\
> float: %1.4f\r\n", ch, i, j);
>
> //----- MAIN's 'forever' loop ----
> // The UART0 interrupt echos incoming characters
> while (1)
> {
> __bis_SR_register( LPM3_bits + GIE); // Enter LPMx, ints enabled
> __no_operation(); // For using a BreakPoint
> __no_operation();
> } // End of main's "forever" loop
> } // End of main()
>
> //----- UCIA0 RX Interrupt -----------
> #pragma vector=USCIAB0RX_VECTOR
> __interrupt void USCI0RX_ISR(void)
> {
> char temp;
> temp = ( UCA0RXBUF & 0x7F ); // Strip off parity bit in temp
> while(!(IFG2 & UCA0TXIFG));
> UCA0TXBUF = temp; // Echo the character
> } // End of USCIAB0RX_isr
>
> //***** End uart_printf_3.c **************************************
>
> --- In m..., "Jim Smith" wrote:
> >
> > I have followed the suggestion of Hardy Griech and used his prototypes to
> >
> > come up with a printf replacement: uart_printf(). As promised, I am
> > including
> >
> > my working code as an attachment. Maybe that will preserve the
formatting.
>
Reply by Neil Cherry●August 3, 20112011-08-03
On 08/02/2011 08:14 PM, Jim wrote:
> and complexity of development software requires it!
On the
> other hand, development has gotten much easier! I started
> with a ROM assembler running on a 6502 (AIM-65) with two
I still have the AIM65 (and a Z80STK & a Z80 hardware emulator
sitting on my desk to demostrate the old days :-) ). While things
have gotten complex, things are also easier and we have new methods
of programming complex things (OOP, multi-core/CPU/computer, SAS,
etc.). Of course programming in 2k is a lot easier now. :-)
Well, attaching a zip file to an email reply didn't work, so here is my
little test program (source) for uart_printf(). If you are reading this on the
http://tech.groups.yahoo.com/group/msp430/messages site, on the upper-right,
under 'Show Message Info', select 'Use Fixed Width Font' to
get proper formatting of the code below. Now, you can cut and paste the code
into your editor, if you like.
Jim Smith
//***** Begin uart_printf_3.c ************************************
//************************************************************************
// MSP430x26x Demo - USCI_A0, Ultra-Low Pwr UART 1200 Echo ISR, 32kHz ACLK
//
// Simple program to get low-level character and string output to UART
// working, and interfaced to printf() (or its replacement), using TI's
// CCSv4 Compiler/IDE.
// ACLK = BRCLK = LFXT1 = 32768Hz, MCLK = SMCLK = DCO ~8.0MHz
// Baud rate divider with 32768Hz XTAL @1200 = 32768Hz/1200 = 27.31
// ***** An external watch crystal is required on XIN XOUT for ACLK *****
//
// MSP430F261x/241x
// -----------------
// /|\| XIN|-
// | | | 32kHz
// --|RST XOUT|-
// | |
// | P3.4/UCA0TXD|------------>
// | | 1200 - 7E1
// | P3.5/UCA0RXD|<------------
//
// Started with, and modified TI example MSP430x261x_uscia0_uart_03.c
// Jim Smith
// Aug 2, 2011
// Built with CCS Version: 4.2.4
//************************************************************************
#include
#include
#include
#include
#include "msp430x26x.h"
char ch = 'A';
int i = 325;
float j = 234.5678;
//----- Code from Hardy Griech, pretty-much verbatim ---------------------
// js used _printfi() instead of _printfi_nofloat(), ie full floating
// point. This requires '--printf_support=full' when linking. You may
// select this in the CCS IDE by: r-click on your project name ->
// C/C++ Build -> Library Function Assumptions -> choose full as level
// of printf support. Do it for both Debug & Release Builds.
//extern int _printfi_nofloat(char const **_format, va_list _ap,
extern int _printfi(char const **_format, va_list _ap, // js changed
void *_op, int (*_outc)(char c, void *op),
int (*_outs)(char *s, void *op));
void uart_putc( uint8_t c )
{
// send character directly to UART or send it to a buffer
// to decouple character output
while(!(IFG2 & UCA0TXIFG)); // js added
UCA0TXBUF = c; // js added
}
void uart_puts( const char *s )
{
while ( *s != '\0') {
uart_putc( (uint8_t)*(s++) );
}
} // uart_puts
static int _outc( char c, void *_op )
{
uart_putc( (uint8_t)c );
return 1;
} // _outc
static int _outs( char *s, void *_op )
{
size_t len = strlen( s );
uart_puts( s );
return (int)len;
} // _outs
void uart_printf( const char *fmt, ... ) {
va_list _ap;
//----- Sign-on ------------------
// Insert some delay for everything to stabilize
for (i = 0; i < 15; i++) { // 15 rollovers @ 1.1 mHz, about 1 sec delay
__delay_cycles(65535);
} // This delay cleared up garbage when output "Hello, World!", below.
// i should be 15, because of its use in for-loop, above.
uart_puts("\r\nHello, World!\r\n");
uart_printf("\
This is just to prove uart_putc(), uart_puts() and uart_printf() work.\r\n\n\
char: %c\r\n\
Integer: %d\r\n\
float: %1.4f\r\n", ch, i, j);
//----- MAIN's 'forever' loop ----
// The UART0 interrupt echos incoming characters
while (1)
{
__bis_SR_register( LPM3_bits + GIE); // Enter LPMx, ints enabled
__no_operation(); // For using a BreakPoint
__no_operation();
} // End of main's "forever" loop
} // End of main()
//----- UCIA0 RX Interrupt -----------
#pragma vector=USCIAB0RX_VECTOR
__interrupt void USCI0RX_ISR(void)
{
char temp;
temp = ( UCA0RXBUF & 0x7F ); // Strip off parity bit in temp
while(!(IFG2 & UCA0TXIFG));
UCA0TXBUF = temp; // Echo the character
} // End of USCIAB0RX_isr
//***** End uart_printf_3.c **************************************
--- In m..., "Jim Smith" wrote: >
> I have followed the suggestion of Hardy Griech and used his prototypes to
>
> come up with a printf replacement: uart_printf(). As promised, I am
> including
>
> my working code as an attachment. Maybe that will preserve the formatting.
Reply by Jim Smith●August 3, 20112011-08-03
I have followed the suggestion of Hardy Griech and used his prototypes to
come up with a printf replacement: uart_printf(). As promised, I am
including
my working code as an attachment. Maybe that will preserve the formatting.
I liked Hardy's approach because it allowed getting rid of a declared
RAM
buffer, and the possibility of buffer overrun.
I made only minor changes/additions to Hardy's prototypes. Instead of
Debug_printf(), I used uart_printf() because I need to output formatted text
and numbers, including floating point, over a UART link, both for data
transmission and for user interface. The Debug name just didn't seem
appropriate. I also used _printfi() instead of _printfi_nofloat(),
because I wanted the full floating point capability.
This working code was compiled by CCSv4 and downloaded to hardware using
TI's
MSP430 USB Debug-Interface. The hardware is a MSP430F2618, socket-mounted
On a 64-bit MSP-TS430PM64 board, to which I have added a RS232 Level
Shifter, to interface to a terminal.
Thanks, Hardy. And thanks to all the others who have made helpful
suggestions.
I have learned a lot from being a member of and monitoring this MSP430 Group
for several years. I always like to see the code: how others did things! I
hope that others may benefit from this, as well.
Jim Smith
Reply by Jim●August 2, 20112011-08-02
--- In m..., Hardy Griech wrote:
: >
> in rtssrc.zip/SHARED/_printfi.c the functions are declared as follows:
>
> #if !defined(NOFLOAT) && !defined(MINIMAL)
> extern int _printfi(char **_format, va_list _ap, void *_op,
> int (*_outc)(char, void *), int (*_outs)(char *,
> void *));
> #elif defined(NOFLOAT)
> extern int _printfi_nofloat(char **_format, va_list _ap, void *_op,
> int (*_outc)(char, void *), int (*_outs)(char *,
> void *));
> #elif defined(MINIMAL)
> extern int _printfi_minimal(char **_format, va_list _ap, void *_op,
> int (*_outc)(char, void *), int (*_outs)(char *,
> void *));
> #endif
>
> Unfortunately I cannot remember how I stumbled into the prototypes some
> years ago. Nevertheless it is working for me (with buffering and some
> other features...).
>
> Hardy
> :
Yes, you are correct, I don't know how I came up with
"_printfi_nf", and the functions ARE declared for access.
Thanks, I'm still learning! Over the years, I've moved from
assembler to compiler to full IDE for development. Thank
goodness we have fast machines for development, now! The size
and complexity of development software requires it! On the
other hand, development has gotten much easier! I started
with a ROM assembler running on a 6502 (AIM-65) with two
audio tape machines to handle the multi-pass assembler. Each
pass of the assembler required playing the source in on one
tape machine and writing results on the other. You can guess
what happens if any glitch comes along! I can remember
patching HEX code in the wee hours to avoid a new assembly.
I graduated from that to a 6502 cross-assembler running on
a floppy-based CPM machine. At the time, a real joy! I
could never have imagined or dreamed of the stuff we now
have to work with, all the peripherals, memory, etc..
built-in to the CPU chip, along with wonderful debugging!
Enough reminiscing. Thanks for your help.
Jim Smith
Reply by Hardy Griech●August 2, 20112011-08-02
On 02.08.2011 23:21, Jim Smith wrote:
: > back and have another go at this. I think the CCS
library terminology is
> "_printfi_nf", and :
in rtssrc.zip/SHARED/_printfi.c the functions are declared as follows:
#if !defined(NOFLOAT) && !defined(MINIMAL)
extern int _printfi(char **_format, va_list _ap, void *_op,
int (*_outc)(char, void *), int (*_outs)(char *,
void *));
#elif defined(NOFLOAT)
extern int _printfi_nofloat(char **_format, va_list _ap, void *_op,
int (*_outc)(char, void *), int (*_outs)(char *,
void *));
#elif defined(MINIMAL)
extern int _printfi_minimal(char **_format, va_list _ap, void *_op,
int (*_outc)(char, void *), int (*_outs)(char *,
void *));
#endif
Unfortunately I cannot remember how I stumbled into the prototypes some
years ago. Nevertheless it is working for me (with buffering and some
other features...).
Hardy
Reply by Jim Smith●August 2, 20112011-08-02
Thanks for the input. I tried implementing this approach, early on, and ran
into all kinds
of problems. Your code looks to be very straightforward and complete,
however, so I may go
back and have another go at this. I think the CCS library terminology is
"_printfi_nf", and
I'm not sure it is accessible outside the library, so may need to redo
the
library (which
I'd most certainly like to avoid). I wanted the full floating point
capability, so I tried
"_printfi", instead, but never got it all fully working. Now that I've
reviewed, and
hopefully, have a better understanding of the use of pointers ( '*' )
and
address ( '&' ),
as well as your code (below) to guide me, I might have a better chance of
success. I will
post results if I attempt and succeed with this approach.
Jim Smith
-----Original Message-----
From: Hardy Griech [mailto:n...@mardys.de]
Sent: Tuesday, August 02, 2011 3:07 PM
To: m...
Cc: Jim
Subject: Re: [msp430] Re: printf to UART - Revisited - CCSv4
There is a way to implement your own printf() without an extra buffer:
extern int _printfi_nofloat(char const **_format, va_list _ap,
void *_op, int (*_outc)(char c, void *op),
int (*_outs)(char *s, void *op));
void Debug_putc( uint8_t c )
{
// send character directly to UART or send it to a buffer
// to decouple character output
}
