Posted on May 17, 2011 at 12:54I have an intriguing problem with the STM32:
I have been enhancing the sample code provided with the STM32F10X-128K-EVAL board with the intention of using some of the modules within a new project. Part of the enhancement is to make the code more readable.
I am using gcc as my compiler and Eclipse as the IDE.
Initially I used #defines to access registers, as in the following example:
#define SPI2_BASE (u32*)0x40003800
#define PeriphArrayReg(Address, Index, Element) *(u32*)(Address##Index##_BASE + offsetof(t##Address, Element))
#define SPI_DR(x) PeriphArrayReg(SPI, x, DR)
In the demo board SPI2 is used to access the LCD display, so a sample of the code is as follows:
SPI_DR(2) = Start_Byte;
The machine code created is as follows, with the processor registers shown when stopped at the highlighted line:
0x08004656 : ldr r3, [pc, #16] (0x8004668 )
0x08004658 : str r0, [r3, #0]
r0 0x70
r1 0x0
r2 0x42218188
r3 0x4000380c
This code works fine and correctly writes to the display.
I then tried further enhancing the code to use pointers to structures:
#define SPI2_BASE (u32*)0x40003800
tSPI* pSPI[4] ={[2] = (tSPI*)SPI2_BASE };
#defineSPI_DR(x) pSPI[x]->DR
The same sample code produces the following machine code, with the processor registers again shown when stopped at the highlighted line:
0x08004662 : ldr r3, [pc, #20] (0x8004678 )
0x08004664 : ldr r3, [r3, #8]
0x08004666 : str r0, [r3, #12]
r0 0x70
r1 0x0
r2 0x0
r3 0x40003800
I believe this should perform exactly the same. Instead of the r3 register pointing directly to the SPI_DR register it points to the start of the SPI structure, but the code writes the contents of the r0 register to (SPI + 12).
However, this code fails to write to the LCD display! I am using the same technique to read from flash on SPI1 and that works OK. I have proved conclusively that this is the cause of the problem by including both #defineSPI_DR(x) variants and commenting out one or the other.
Can anyone explain the anomaly?
Regards
Alan Rouse
