V6M6 Software Release Notes V1.4.1 / V1.4.2

V6M6 Software Release Notes
Version 1.4.1 and 1.4.2

Version 1.4.1 and 1.4.2 of the V6M6 host software and FPGA configurations include some significant changes:

Enhancements to the IM2E1 and IM2T1 modules,
Completed support for the DM4C51 module,
ADPCM bit-reverse mode for the IMG726 module,
Beta release of the TDM Configurator.

General additions and changes in Versions 1.4.1 and 1.4.2

Note:

pciflashup

Completed support for the IMG726 module, including a bug fix in its FPGA configuration, a mode to reverse the bits in 32K ADPCM data and the new pcig726 diagnostic program which is integrated into pciburn.
Modifications were made in the host library and the pciinit program to support the 512KB version of the DM2C31 module.
A new FPGA configuration for the PM4600/4700 module fixes problems with requesting the PCI bus when DM4C51 modules are present and performing burst reads from I/O space.
A problem with transmitting message channel data was fixed in the FPGA configuration for the IMSCSA module.

IM2E1 and IM2T1 Enhancements in Version 1.4.1

Problems were reported using certain TDM configurations with the IM2T1 module. The IM2T1 module only has one phased-locked loop for converting between 1.544 MHz and 2.048, 4.096 or 8.192 MHz clocks. Previous versions could not simultaneously generate TDM clocks from one of its received T1 clocks and derive T1 transmit clock from TDM clock.

The new FPGA configuration (imtelco13.mcs) has added logic to avoid the problem. If the module is generating TDM clock from an RCLK and is told to derive T1 transmit clock from TDM clock, it will automatically revert to using the RCLK selected as the source of TDM clock as the source for the T1 transmit clock.

In some application situations, the timing of multiple incoming E1 or T1 lines may be derived from different time bases; their clocks may be running at slightly different frequencies. Depending on which clock is faster, data from one of the incoming lines could be either lost or repeated as the elastic storage buffer in the framer chip underflows or overflows. The solution to this problem has two parts:

First, the new FPGA configuration for the IM2E1 and IM2T1 includes logic to compare the receive clocks (RCLK) extracted from the incoming lines to detect which is the faster of the two. The library functions, im2e1_rcvclk and im2t1_rcvclk, now accept a value of 3 for their select argument. When select is 3 the module will choose the faster of the two RCLKs to derive the TDM clock. This ensures that any frame slips are due to elastic store underflows and data will be repeated rather than lost.

Second, logic in the new FPGA along with a hardware modification (Revision 2 of the IM2E1 and IM2T1 module) allows the module to flag data from a repeated frame as invalid on the TDM Subsystem. When used in conjunction with conditional TDM destinations for MIPS modules, the MIPS processor receiving E1 or T1 data via TDM will not receive data from repeated frames.

Functions were added to the host library to enable or disable the slipped frame invalidation mode:

    int
    im2e1_slipinval (pci, select)

    int
    im2t1_slipinval (pci, select)

        PCI_MOD *pci    - pointer to an open IM2E1 or IM2T1.
        int      select - controls slip frame invalidation
                          mode for lines 0 and 1.

Bit 0 of select controls slip frame invalidation mode for line 0: a 1 in this bit enables and a 0 disables. Bit 1 enables or disables this mode for line 1.

Note 1: Using this mode requires that the Framer chips be configured to generate an interrupt. For both the IM2E1 and IM2T1, this means turning on bit 4 of Interrupt Mask Register 1 in their respective framer chips. For the im2e1_outreg function, the register address is IM2E1REG_MASK1. For the im2t1_outreg function, the register address is IM2T1REG_MASK1.

Note 2: The interrupt must be serviced by the host system or a processor on the V6M6 in time for the next frame slip to be recognized. The time between slips will, of course, depend on the frequency offset between the two incoming RCLKs. For example, with a frequency offset of 0.01%, a slip can be expected every 1.25 seconds.

For applications using two more IM2E1 or IM2T1 modules, logic was added to enable a PCI interrupt when one of the module's RCLKs is detected faster then the frequency of the TDM clock. This is enabled using the im2e1_intenb or im2t1_intenb function, whose intmask argument has been extended to include:

      Bit 8 - RCLK faster than TDMCLK interrupts to PCI INT A
      Bit 9 - RCLK faster than TDMCLK interrupts to PCI INT B

The status of this detection is available using the im2e1_intstat or im2t1_intstat functions. The value returned by these functions now includes:

      Bit 4 - RCLK detected faster than TDMCLK

When this interrupt is serviced, the application running on the host or a MIPS processor may decide to switch which IM2E1 or IM2T1 module is sourcing the TDM clock. Note that the interrupt condition is latched and will remain active until clear by calling a new function:

    int
    im2e1_clrfastclk (pci)

    int
    im2t1_clrfastclk (pci)

        PCI_MOD *pci - pointer to an open IM2E1 or IM2T1.

DM4C51 Support in Version 1.4.1

An in-depth user manual for the DM4C51 is under development. The major features of the host application library are described in the following notes. Unless otherwise specified, functions return values of:

C51 Processor Access
Accessing a C51 Processor unit for control and data transfer operations requires accessing the Unix device and initializing a PCI_MOD structure. This is handled by the dm4c51_open function.
```
    PCI_MOD *
    dm4c51_open(devname)
  
        char  *devname  - name of PCI device and C51 DSP
```
The devname argument specifies the name of the V6M6 board, the module location and the C51 DSP on the module. The format of the name is pciNMD.
The PCI_MOD pointer returned by dm4c51_open is used to pass information about the module and DSP to other library functions.
The dm4c51_init function is used to initialize the C51 DSP referenced by the pci pointer. Note that because a given DM4C51 module is accessed through a single Unix device, the four C51 DSP cannot be accessed by different host programs.
```
    int
    dm4c51_init(pci)
  
        PCI_MOD *pci - pointer to an open DM4C51
```

Host Data Transfers:

Most data transfers are handled using the general purpose V6M6 data transfer functions:

    pci_dl_i8b   pci_up_i8b   pci_dl_a8b   pci_up_a8b
    pci_dl_i16b  pci_up_i16b  pci_dl_a16b  pci_up_a16b

Due to the way the compiler for the C51 arranges long integers in memory, the standard transfer functions for 32-bit values and arrays will not work. The library includes the following functions for 32-bit data transfers:

dm4c51_dl_i32b downloads a 32-bit value to memory of the C51 referenced by pci.

    int
    dm4c51_dl_i32b(pci, pciadr, value)
  
        PCI_MOD *pci     - pointer to an open DM4C51
        u_long   pciadr  - PCI memory address on the DM4C51,
                           must be modulo-2 aligned
        u_long   value   - Data value to be downloaded

dm4c51_up_i32b uploads a 32-bit value from memory of the C51 referenced by pci. The value returned is the data read from the C51's memory.

    long
    dm4c51_up_i32b(pci, pciadr)
  
        PCI_MOD *pci     - pointer to an open DM4C51
        u_long   pciadr  - PCI memory address on the DM4C51,
                           must be modulo-2 aligned

dm4c51_dl_a32b and dm4c51_up_a32b download and upload arrays of 32-bit values between the host and the C51 DSP referenced by pci.

    int
    dm4c51_dl_a32b (pci, pciadr, nlwords, buf)
  
    int
    dm4c51_up_a32b (pci, pciadr, nlwords, buf)
  
        PCI_MOD *pci     - pointer to an open DM4C51
        u_long   pciadr  - DM4C51 PCI memory address,
                           must be modulo-2 aligned
        u_long   nlwords - number of 32-bit words to transfer
        u_long  *buf     - pointer to source data

Loading and executing C51 Programs:
C51 executable code is loaded in C51 memory on a DM4C51 using the pci_load_code function. When pci_load_code returns a 0 value, it indicates an error reading the file or downloading the program data. For example:
```
     char *prog_name = "c51program";

     if (!pci_load_code(pci, prog_name, MM_COFF))
     {
         perror(prog_name);
         exit(1);
     }
```
The dm4c51_run and dm4c51_halt functions are used to control execution of programs in a C51 DSP.
```
    int
    dm4c51_run(pci)
  
    int
    dm4c51_halt(pci)
  
        PCI_MOD *pci     - pointer to an open DM4C51
```
DM4C51 TDM Interface:
The C51 DSPs communicate with the V6M6 TDM subsystem through an SC4000 SCSA interface. Each C51 DSP is allocated 32 time slots on the DM4C51 side of the SC4000. The 32 time slots may be mapped to any TDM time slot and TDM bus on the TDM subsystem side of the SC4000.
The C51 DSPs read and write TDM data in their local memory using double buffers. Data is transferred between the TDM data memory and the SC4000 by a common DMA controller on the DM4C51 module. The TDM data buffers are fixed at 32 bytes in size and are at fixed locations in the DSP's memory:
```
    DSP mem addr     Buffer Use
    ------------     ----------
      0x1ff80        TDM Out Buffer 1  (to TDM)
      0x1ffa0        TDM Out Buffer 2  (to TDM)
      0x1ffc0        TDM In  Buffer 1  (from TDM)
      0x1ffe0        TDM In  Buffer 2  (from TDM)
```
A register bit for each C51 DSP indicating which incoming TDM buffer has data available. The same flag indicates which outgoing TDM buffer may be written to by the C51 DSP.
The standard TDM mapping functions, pci_tdm_src_add, pci_tdm_dst_add, etc. are not used for the DM4C51. Instead, use the functions described below to establish TDM connectivity for the DM4C51. Note that these functions assume the TDM subsystem is configured to use 8-bit TDM slots.
The dm4c51_sc4000_init function resets and initializes the SC4000 and disables its DMA controller. The tdmframe argument specifies the number of 8-bit TDM time slots per TDM frame. In cases where the TDM time slots are larger than 8 bits, this value should specify the equivalent number of 8-bit slots.
```
    int
    dm4c51_sc4000_init(pci, tdmframe)

        PCI_MOD *pci - pointer to an open DM4C51
        int tdmframe - slots per TDM frame:
                       32, 64 or 128
```
For adding TDM source and destination connections use the dm4c51_tdm_src_add and dm4c51_tdm_dst_add functions. These functions configure the next available SC4000 slot of the SC4000 to source the TDM bus or receive from the TDM bus during the specified TDM time slot. They also enable the TDM data DMA controller and configure the DMA data count.
```
    int
    dm4c51_tdm_src_add(pci, TDMslot, TDMbus)

    int
    dm4c51_tdm_dst_add(pci, TDMslot, TDMbus)
  
        PCI_MOD *pci  - pointer to an open DM4C51
        u_int TDMbus  - TDM bus: TDM_BUSA, TDM_BUSB,
                                 TDM_BUSC or TDM_BUSD
        u_int TDMslot - time slot number (1 - 128).
```
TDM connections are deleted using the dm4c51_tdm_src_del and dm4c51_tdm_dst_del functions. These functions modify the SC4000 configuration to stop driving the TDM bus or stop receiving from the TDM bus during the specified TDM time slot.
```
    int
    dm4c51_tdm_src_del(pci, TDMslot, TDMbus)

    int
    dm4c51_tdm_dst_del(pci, TDMslot, TDMbus)

        PCI_MOD *pci  - pointer to an open DM4C51
        u_int TDMbus  - TDM bus: TDM_BUSA, TDM_BUSB,
                                 TDM_BUSC or TDM_BUSD
        u_int TDMslot - time slot number (1 - 128).
```

ADPCM 32K Bit-Reverse Mode for IMG726 in Version 1.4.2

A mode was added to reverse the 4 bits of 32K ADPCM data. This mode is selected using a new function, img726_adpcm_32k_reverse:

    int
    img726_adpcm_32k_reverse (pci, reverse)

        PCI_MOD *pci     - pointer to an open IM2E1 or IM2T1.
        int      reverse - 1 enables reverse mode
                           0 disables reverse mode.

img726_init

Beta Release of TDM Configurator in Version 1.4.1

The process of generating C code to configure the TDM subsystem using the various library functions can be tedious. Although using the library functions to select timing sources, set parameters and build the TDM connection map provides the most flexible approach, the code tends to be hard to read and modify. The TDM Configurator provides a more integrated approach. It may be used to set up a static or initial TDM configuration using a TDM description file for the configuration of each board in the system.

The TDM Configurator is available as both a host library function call, pci_tdmreadcfg, and as a stand-alone program, pcitdmconfig. Both forms use an input source file; a TDM description file. This file, created by the user, contains a description of the TDM components, operational parameters and connection map for a V6M6 board and its modules. It provides the ability to specify and configure the three following aspects of the serial data flow in your system:

Clock and timing source.
Data routing and connection maps.
T1 and E1 Framer parameters for IM2T1 and IM2E1 modules.

The TDM map compiler parses a TDM description file and makes the appropriate function calls to configure the TDM controller, TDM map RAM, SCSA interfaces, T1 and E1 modules, and any other modules involved in generating timing for the TDM subsystem.

A complete manual for the TDM Configurator and the TDM description file is being written. It will be available on our Internet site or by request soon. In the meantime, the initial version of the software is provided in this release to allow customers to use example applications developed at CAC and WMI.