The initialization processing for the user program is configured by the following elements.
In the sample program, these elements are located in cstart.asm.
The stack area is used for code generated by the compiler. The CC-RH generates code based on the assumption that the stack pointer (sp) is allocated at the 4-byte boundary and the stack area is made to be grown in the direction of address 0x0. Therefore, an address aligned at the 4-byte boundary on the 0xffffffff address side of the .stack.bss section needs to be specified for the stack pointer (sp).
The entry point is the label (address) to which the hardware initialization routine branches at a reset.
The stack pointer, gp register, and ep register are all initialized.
mov #_stacktop, sp ; set sp register mov #__gp_data, gp ; set gp register mov #__ep_data, ep ; set ep register |
For details on __gp_data and __ep_data, see "8.4 Symbols".
Variable areas defined in the C source program or assembly source program are initialized. Prepare a table that holds the addresses of sections to be initialized and pass the address of the table to the _INITSCT_RH library function and call it.
The initialization table for initialized data sections should be written in the following format.
The initialization-destination section of an initialized data section can be specified by using the -rom option of the optimizing linker.
In this case, code of the initialization table is as follows:
.section ".INIT_DSEC.const", const .align 4 .dw #__s.data, #__e.data, #__s.data.R .dw #__s.sdata, #__e.sdata, #__s.sdata.R |
The initialization table for uninitialized data sections should be written in the following format.
.section ".INIT_BSEC.const", const .align 4 .dw #__s section-name-1, #__e section-name-1 .dw #__s section-name-2, #__e section-name-2 : .dw #__s section-name-n, #__e section-name-n |
The start address and end address of the initialization table are passed through parameters of the _INITSCT_RH function and initialization is executed.
The start of each initialization table must be aligned at the 4-byte boundary.
mov #__s.INIT_DSEC.const, r6 mov #__e.INIT_DSEC.const, r7 mov #__s.INIT_BSEC.const, r8 mov #__e.INIT_BSEC.const, r9 jarl32 __INITSCT_RH, lp ; initialize RAM area |
For the usage method of _INITSCT_RH, see "7.4.11 RAM section initialization function".
Reference the PID register to confirm whether the FPU is available.
If available, make the initial settings.
Set the PSW.CU0 bit to 1 to enable usage of the FPU.
Set the FPU operating mode for the FPSR register.
The FPEPC register is initialized as a preparation to use the lockstep function.
Delete these codes from programs that do not use the FPU.
Set the PSW.ID bit to 0 to enable occurrence of exceptions.
Set the PSW.UM bit to 1 to enter the user mode.
To reflect the settings as soon as a branch to the main function occurs, write the settings to the FEPSW instead of the PSW; the FEPSW settings are reflected to the PSW when the feret instruction is executed.
stsr 5, r10, 0 ; r10 <- PSW xori 0x0020, r10, r10 ; enable interrupt movhi 0x4000, r0, r11 or r11, r10 ; supervisor mode -> user mode ldsr r10, 3, 0 ; FEPSW <- r10 |
A branch to the main function occurs when the feret instruction is executed with the address of the main function set to the FEPC register and the address of the _exit function to be executed subsequent to the main function set to the r31 register.