What is this macro style definition?

Hi guys,
I am looking an ADC driver and I found a definition that I do not understand how is expanded.

Code:

define NRF_SAADC_BASE 0x40007000UL

Code:

typedef struct { /*!< (@ 0x40007000) SAADC Structure */ __OM uint32_t TASKS_START; /*!< (@ 0x00000000) Starts the SAADC and prepares the result buffer in RAM */ __OM uint32_t TASKS_SAMPLE; /*!< (@ 0x00000004) Takes one SAADC sample */ __OM uint32_t TASKS_STOP; /*!< (@ 0x00000008) Stops the SAADC and terminates all on-going conversions */ __OM uint32_t TASKS_CALIBRATEOFFSET; /*!< (@ 0x0000000C) Starts offset auto-calibration */ __IM uint32_t RESERVED[60]; __IOM uint32_t EVENTS_STARTED; /*!< (@ 0x00000100) The SAADC has started */ __IOM uint32_t EVENTS_END; /*!< (@ 0x00000104) The SAADC has filled up the result buffer */ } NRF_SAADC_Type; /*!< Size = 1592 (0x638)

This is the definition I do not understand. It is like creating a pointer NRF_SAADC to struct NRF_SAADC_Type and then initialize with NRF_SAADC_BASE? Could you re-write the following definition into a non-macro style function to understand how this expands?

Code:

#define NRF_SAADC ((NRF_SAADC_Type*) NRF_SAADC_BASE)

Then he accesses the NRF_SAADC_Type struct to set a specific register like this:

Code:

NRF_SAADC->INTENSET = SAADC_INTENSET_END_Msk;

Thanks Nick

It is literally just a type cast. If you expand it, then this:

Code:

NRF_SAADC->INTENSET = SAADC_INTENSET_END_Msk;

becomes:

Code:

((NRF_SAADC_Type*)0x40007000UL)->INTENSET = SAADC_INTENSET_END_Msk;

The NRF_SAADC_Type struct type doesn't appear to have a member named INTENSET though, so I'd guess it is either a typo or you're omitting something.

Struggling to understand why he is doing like that. Why is required this definition?

Code:

define NRF_SAADC_BASE 0x40007000UL

Why not implement like this?

Code:

NRF_SAADC_Type *NRF_SAADC

Quote:

Originally Posted by Nikosant03

Struggling to understand why he is doing like that. Why is required this definition?

Code:

define NRF_SAADC_BASE 0x40007000UL

Why not implement like this?

Code:

NRF_SAADC_Type *NRF_SAADC

Doing it the way you suggest is probably the "right" way to do it, in terms of producing code that can be tested and debugged.

However, it might produce "worse" code in objective terms. Many small CPUs have a stark performance difference between "immediate" and "indirect" operations. Doing something like "load immediate from address" versus "load indirect" may cost more bytes of code, or cost more instruction cycles, or both. Especially when there is an extra offset involved (for any struct members not at offset 0).

Try writing a dummy function that does things both ways, and see what the generated assembly looks like. (But keep in mind there's a difference between local and global variables.)

Well, I have always programmed for hosted implementations where there's an OS and thus the notion of casting an integer constant (other than zero) to a pointer is weird to me. But you're looking at driver code, so the rules of what's weird and what's not are presumably different. I'd guess the answer is as simple as "the author wanted the struct object to be located in that region of memory". Why? You have to read the documentation, ask the author, etc.

Quote:

Originally Posted by laserlight

"the author wanted the struct object to be located in that region of memory"

That makes sense!!
This is the base address of the ADC peripheral

Code:

define NRF_SAADC_BASE 0x40007000UL

So by locating the struct at the base address provides the offset address for all registers of the ADC peripheral. That's very clever and efficient!!
SAADC — Successive approximation analog-to-digital converter

This is the entire struct

Code:

typedef struct { /*!< (@ 0x40007000) SAADC Structure */ __OM uint32_t TASKS_START; /*!< (@ 0x00000000) Starts the SAADC and prepares the result buffer in RAM */ __OM uint32_t TASKS_SAMPLE; /*!< (@ 0x00000004) Takes one SAADC sample */ __OM uint32_t TASKS_STOP; /*!< (@ 0x00000008) Stops the SAADC and terminates all on-going conversions */ __OM uint32_t TASKS_CALIBRATEOFFSET; /*!< (@ 0x0000000C) Starts offset auto-calibration */ __IM uint32_t RESERVED[60]; __IOM uint32_t EVENTS_STARTED; /*!< (@ 0x00000100) The SAADC has started */ __IOM uint32_t EVENTS_END; /*!< (@ 0x00000104) The SAADC has filled up the result buffer */ __IOM uint32_t EVENTS_DONE; /*!< (@ 0x00000108) A conversion task has been completed. Depending on the configuration, multiple conversions might be needed for a result to be transferred to RAM. */ __IOM uint32_t EVENTS_RESULTDONE; /*!< (@ 0x0000010C) Result ready for transfer to RAM */ __IOM uint32_t EVENTS_CALIBRATEDONE; /*!< (@ 0x00000110) Calibration is complete */ __IOM uint32_t EVENTS_STOPPED; /*!< (@ 0x00000114) The SAADC has stopped */ __IOM SAADC_EVENTS_CH_Type EVENTS_CH[8]; /*!< (@ 0x00000118) Peripheral events. */ __IM uint32_t RESERVED1[106]; __IOM uint32_t INTEN; /*!< (@ 0x00000300) Enable or disable interrupt */ __IOM uint32_t INTENSET; /*!< (@ 0x00000304) Enable interrupt */ __IOM uint32_t INTENCLR; /*!< (@ 0x00000308) Disable interrupt */ __IM uint32_t RESERVED2[61]; __IM uint32_t STATUS; /*!< (@ 0x00000400) Status */ __IM uint32_t RESERVED3[63]; __IOM uint32_t ENABLE; /*!< (@ 0x00000500) Enable or disable SAADC */ __IM uint32_t RESERVED4[3]; __IOM SAADC_CH_Type CH[8]; /*!< (@ 0x00000510) Unspecified */ __IM uint32_t RESERVED5[24]; __IOM uint32_t RESOLUTION; /*!< (@ 0x000005F0) Resolution configuration */ __IOM uint32_t OVERSAMPLE; /*!< (@ 0x000005F4) Oversampling configuration. The RESOLUTION is applied before averaging, thus for high OVERSAMPLE a higher RESOLUTION should be used. */ __IOM uint32_t SAMPLERATE; /*!< (@ 0x000005F8) Controls normal or continuous sample rate */ __IM uint32_t RESERVED6[12]; __IOM SAADC_RESULT_Type RESULT; /*!< (@ 0x0000062C) RESULT EasyDMA channel */ } NRF_SAADC_Type; /*!< Size = 1592 (0x638) */

So it can very easily modify a register

Code:

NRF_SAADC->INTENSET = SAADC_INTENSET_END_Msk;