‘C’ for Microcontrollers, Just Being Efficient Lloyd Moore, President [email protected] www.CyberData-Robotics.com Seattle Robotics Society 9/15/2012 

Agenda  Microcontroller Resources  Knowing Your Environment  Memory Usage  Code Structure  Optimization  Summary 

Disclaimer  Some microcontroller techniques necessarily need to trade one benefit for another – typically lower resource usage for maintainability  Point of this presentation is to point out various techniques that can be used as needed  Use these suggestions when necessary  Feel free to suggest better solutions as we go along 

Microcontroller Resources  EVERYTHING resides on one die inside one package: RAM, Flash, Processor, I/O  Cost is a MAJOR design consideration  Typical costs are $0.25 to $25 each (1000’s)  RAM: 16 BYTES to 256K Bytes typical  Flash/ROM: 384 BYTES to 1M Byte  Clock Speed: 4MHz to 175MHz typical  Much lower for battery saving modes (32KHz)  Bus is 8, 16, or 32 bits wide  Have dedicated peripherals (MAC, Phys, etc) 

Power Consumption  Microcontrollers typically used in battery operated devices  Power requirements can be EXTREMELY tight  Energy harvesting applications  Long term battery installations (remote controls, hard to reach devices, etc.)  EVERY instruction executed consumes power, even if you have the time and memory! 

Know Your Environment  Traditionally we ignore hardware details  Need to tailor code to hardware available  Specialized hardware MUCH more efficient  Compilers typically have extensions  Interrupt – specifies code as being ISR  Memory model – may handle banked memory and/or simultaneous access banks  Multiple data pointers / address generators  Debugger may use some resources 

Memory Usage  Put constant data into program memory (Flash/ROM)  Alignment / padding issues  Typically NOT an issue, non-aligned access ok  Avoid dynamic memory allocation, even if available  Take extra space and processing time  Memory fragmentation a big issue  Use and reuse static buffers  Reduces variable passing overhead  Allows for smaller / faster code due to reduced indirections  Does bring back over write bugs if not done carefully  More reliable for mission critical systems  Use the appropriate variable type  Don’t use int and double for everything!!  Affects processing time as well as storage 

C99 Datatypes – inttypes.h  int8_t, int16_t, int32_t, int64_t  uint8_t, uint16_t, uint32_t, uint_64_t  Avoids the ambiguity of int and uint when moving code between processors of different native size  Makes code more portable and upgradable over time 

Char vs. Int Increment on 8051 char cX; cX++; 000A 900000 MOV DPTR,#cX 000D E0 MOVX A,@DPTR 000E 04 INC A 000F F0 MOVX @DPTR,A  6 Bytes of Flash  4 Instruction cycles int iX; iX++; 0000 900000 MOV DPTR,#iX 0003 E4 CLR A 0004 75F001 MOV B,#01H 0007 120000 LCALL ?C?IILDX  10 Bytes of Flash + subroutine overhead  Many more than 4 instruction cycles with a LCALL 

Code Structure  Count down instead of up  Saves a subtraction on all processors  Decrement-jump-not-zero style instruction on some processors  Pointers vs. array notation  Generally better using pointers  Bit Shifting  May not always generate what you think  May or may not have barrel shifter hardware  May or may not have logical vs. arithmetic shifts 

Shifting Example on 8051 cX = cX << 3; 0006 33 RLC A 0007 33 RLC A 0008 33 RLC A 0009 54F8 ANL A,#0F8H  Constants turn into seperate statements  Variables turn into loops  Both of these can be one instruction with a barrel shifter cA = 3; cX = cX << cA; 000B 900000 MOV DPTR,#cA 000E E0 MOVX A,@DPTR 000F FE MOV R6,A 0010 EF MOV A,R7 0011 A806 MOV R0,AR6 0013 08 INC R0 0014 8002 SJMP ?C0005 0016 ?C0004: 0016 C3 CLR C 0017 33 RLC A 0018 ?C0005 0018 D8FC DJNZ R0,?C0004 

Indexed Array vs Pointer on M8C ucMode = g_Channels[uc_Channel].ucMode; 01DC 52FC mov A,[X-4] 01DE 5300 mov [__r1],A 01E0 5000 mov A,0 01E2 08 push A 01E3 5100 mov A,[__r1] 01E5 08 push A 01E6 5000 mov A,0 01E8 08 push A 01E9 5007 mov A,7 01EB 08 push A 01EC 7C0000 xcall __mul16 01EF 38FC add SP,-4 01F1 5F0000 mov [__r1],[__rX] 01F4 5F0000 mov [__r0],[__rY] 01F7 060000 add[__r1],<_g_Channels 01FA 0E0000 adc[__r0],>_g_Channels 01FD 3E00 mvi A,[__r1] 01FF 5403 mov [X+3],A ucMode = pChannel->ucMode; 01ED 5201 mov A,[X+1] 01EF 5300 mov [__r1],A 01F1 3E00 mvi A,[__r1] 01F3 5405 mov [X+5],A  Does the same thing  Saves 29 bytes of memory AND a call to a 16 bit multiplication routine!  Pointer version will be at least 4x faster to execute as well, maybe 10x  Most compilers not this bad – but you do find some! 

More Code Structure  Actual parameters typically passed in registers if available  Keep function parameters to less than 3  May also be passed on stack or special parameter area  May be more efficient to pass pointer to struct  Global variables  While generally frowned upon for most code can be very helpful here  Typically ends up being a direct access  Read assembly code for critical areas  Know which optimizations are present  Small compilers do not always have common optimizations  Inline, loop unrolling, loop invariant, pointer conversion 

Switch Statement Implementation  Switch statements can be implemented in various ways  Sequential compares  In line table look up for case block  Special function with look up table  Specific implementation can also vary based case clauses  Clean sequence (1, 2, 3, 4, 5)  Gaps in sequence (1, 10, 30, 255)  Ordering of sequence (5, 4, 1, 2, 3)  Knowing which method gets implemented is critical to optimizing! 

Switch Statement Example switch(cA) { case 0: cX = 4; break; case 1: cX = 10; break; case 2: cX = 30; break; default: cX = 0; break; } 0006 900000 MOV DPTR,#cA 0009 E0 MOVX A,@DPTR 000A FF MOV R7,A 000B EF MOV A,R7 000C 120000 LCALL ?C?CCASE 000F 0000 DW ?C0003 0011 00 DB 00H 0012 0000 DW ?C0002 0014 01 DB 01H 0015 0000 DW ?C0004 0017 02 DB 02H 0018 0000 DW 00H 001A 0000 DW ?C0005 001C ?C0002: 001C 900000 MOV DPTR,#cX 001F 7404 MOV A,#04H 0021 F0 MOVX @DPTR,A 0022 8015 SJMP ?C0006 ...More blocks follow for each case 

Optimization Process  Step 0 – Before coding anything, think about risk points and prototype unknowns!!!  Use available dedicated hardware  Step 1 – Get it working!!  Fast but wrong is of no use to anyone  Optimization will typically reduce readability  Step 2 – Profile to know where to optimize  Usually only one or two routines are critical  You need to have specific performance metrics to target 

Optimization Process  Step 3 – Let the tools do as much as they can  Turn off debugging!  Select the correct memory model  Select the correct optimization level  Step 4 – Do it manually  Read the generated code! Might be able to make a simple code or structure change.  Last – think about assembly coding 

Summary  Microcontrollers are a resource constrained environment  Be familiar with the hardware in your microcontroller  Be familiar with your compiler options and how it translates your code  For time or space critical code look at the assembly listing from time to time 

Questions?