Home

Keil C166 Compiler Report

1. Company X Ltd T AWORD TECHDOC c166 report companyX compiler report doc Revision C Last amended by M Beach on 22 06 00 14 14 Page 47 of 83 2 4 6 The Special Sections C_CLRMEMSEC And C_INITSEC These are special sections created by C166 to hold the initial values of RAM variables When a variable is declared such as int x 0x80 the 0x80 is actually placed in ROM based look up table During the startup a66 this ROM data is transferred to its final resting place in RAM By default L166 will place these near the bottom of memory on the assumption that this must be EPROM On systems where the program is at for example 0x40000 it is up to the user to put these into the correct area This is achieved by using the sections control in L166 They are best kept together the only point to watch is that the size of the two sections will vary according to how much initialised data there is in the program Thus it is quite possible that they will grow such that they will overlap or move into an illegal memory area The linker will issue a warning which must not be ignored Example main obj amp start1l67 obj amp to exec amp VECTAB 0x40000 amp CLASSES NCODE 0x40400 NCONST 0x40400 NDATA 0x48000 amp SECTIONS C_CLRMEMSEC 0x40400 C_INITSEC REGFILE exec reg Company X Ltd T AWORD TECHDOC c166 report companyX compiler report doc Revision C Last amended by M Beach on 22 06 00 14 14 P
2. x x WARNING 137 IN LINE 21 OF MAIN C constant in condition expression 22 2 23 2 svar 1 j Missing y 24 2 x ERROR 25 IN LINE 24 OF MAIN C syntax error near 25 2 26 2 while 1 x ERROR 25 IN LINE 26 OF MAIN C syntax error near 1 27 2 C166 COMPILATION COMPLETE 2 WARNING S 2 ERROR S L166 Linker Diagnostics The L166 linker also issues WARNING and ERROR diagnostic messages These are stored exclusively in the M66 file Linker Syntax ERRORs Syntax errors in the LIN linker control file inhibit the production of an executable OMF66 file and the M66 map file Details of the ERROR are reported in the LER file This is machine readable only and is intended only to be used by the uVISION environment Company X Ltd T AWORD TECHDOC c166 report companyX compiler report doc Revision C Last amended by M Beach on 22 06 00 14 14 Page 18 of 83 2 3 3 4 2 Linker Linking And Location Diagnostics WARNINGs and ERRORs messages are stored by the linker at the end of the M66 file besides being reported in the LER file Example Of WARNING x WARNING 23 NDATA NDATAO OR NCONST MUST FIT IN ONE 16KB PAGE CLASS NDATAO L166 RUN COMPLETE 1 WARNING S 0 ERROR S Such a warning will still produce an executable OMF66 output file However a run time failure may occur due to the misplacement or mismatch of data LINKER WARNINGS MUST NEVER BE IGNORED Linker ERRORs are produced by references
3. Pee E aa The arithmetic diagnosed has unacceptable Serious Defects Potentially fatal FAILURE may have spoiled this program s subsequent diagnoses END OF TEST Tasking C166 v6r3 Results Program failed to compile and no executable could be produced This has been referred to Tasking for comments Company X Ltd T AWORD TECHDOC c166 report companyX compiler report doc Revision C Last amended by M Beach on 22 06 00 14 14 Page 49 of 83 2 5 2 3 Microsoft C v11 00 Results Note To force the compiler to behave in an equivalent mode to the Keil C166 the 386 code generator was used and the floating point emulator library used These measures prevented the Pentium floating point co processor being used The number of FAILURES encountered Go The number of SERIOUS DEFECTs discovered 5 The number of DEFECTs discovered TOR The number of FLAWs discovered PAR The arithmetic diagnosed has unacceptable Serious Defects Potentially fatal FAILURE may have spoiled this program s subsequent diagnoses END OF TEST 2 5 3 Discussion Of Results The Keil compiler was the only one tested that was able to compile and run the test program without major modification The results appear to suggest that the floating point performance is inadequate but this should be viewed in the context that the notionally equivalent Tasking C166 compiler was unable to compile the test piece and that the reported performance of an industry standa
4. To continue press RETURN Diagnosis resumes after milestone Number 0 Page 1 Users are invited to help debug and augment this program so it will cope with unanticipated and newly uncovered arithmetic pathologies Please send suggestions and interesting results to Richard Karpinski Computer Center U 76 University of California San Francisco CA 94143 0704 USA In doing so please include the following information Precision single Version 10 February 1989 Computer Compiler Optimization level Other relevant compiler options To continue press RETURN Diagnosis resumes after milestone Number 1 Page 2 Running this program should reveal these characteristics Radix 1 2 4 8 10 16 100 256 Precision number of significant digits carried U2 Radix Radix Precision One Ulp OneUlpnit in the Last Place of 1 000xxx Ul 1 Radix Precision One Ulp of numbers a little less than 1 0 Adequacy of guard digits for Mult Div and Subt Whether arithmetic is chopped correctly rounded or something else for Mult Div Add Subt and Sqrt Whether a Sticky Bit used correctly for rounding UnderflowThreshold an underflow threshold E0 and PseudoZero tell whether underflow is abrupt gradual or fuzzy V an overflow threshold roughly vO tells roughly whether Infinity is represented Comparisions are checked for consistency with subtraction and for contamination with pseudo zeros Sqrt is tested Y X is not t
5. C166 COMPILATION COMPLETE 0 WARNING S 1 ERROR S 2 3 15 The outcome when the result of an integer arithmetic function abs div labs or Idiv cannot be represented div and div are not implemented in C166 abs and labs do not return unrepresentable values In the cases where the smallest possible numbers 32768 and 2147483648 respectively are passed the functions do not attempt to produce the absolute value and simply return the input parameter unchanged Company X Ltd T AWORD TECHDOC c166 report companyX compiler report doc Revision C Last amended by M Beach on 22 06 00 14 14 Page 30 of 83 2 3 16 The outcome when an value with an incomplete type is used in a context that requires the value of the designated object Incomplete types may be declared and compiled successfully However any attempt to use an object derived from an incomplete type will result in an error message Example struct testl No member list struct testl test2 test3 Create two identical structures void main void test3 test2 Attempt to copy structure fails Result x ERROR 129 IN LINE 45 OF MAIN C test3 uses undefined struct union testl x ERROR 129 IN LINE 45 OF MAIN C test2 uses undefined struct union testl ERROR 52 IN LINE 45 OF MAIN C use of undefined type testl1 2 3 17 Mismatches in type between value and object The outcome when an object has its stored value accessed by an value that
6. Description The atan2 function computes the principal value of the arc tangent of y x using the signs of both arguments to determine the quadrant of the return value A domain error may occur if both arguments are zero Returns The atan2 function returns the arc tangent of y x in the range pi pi radians No domain error value is return if input parameters are both zero Caution The user should check for zero parameters before calling the function 2 3 13 5 Hyperbolic Functions That Are Subject To Domain Or Range Errors 2 3 13 5 1 The cosh Function Synopsis include lt math h gt double cosh double x Description The cosh function computes the hyperbolic cosine of x A range error occurs if the magnitude of x is too large Returns The cosh function returns the hyperbolic cosine value The range error value returned is INF 0x7F8000 Company X Ltd T WORD TECHDOC c166 report companyX compiler report doc Revision C Last amended by M Beach on 22 06 00 14 14 Page 26 of 83 2 3 13 5 2 The sinh Function Synopsis include lt math h gt double sinh double x Description The sinh function computes the hyperbolic sine of x A range error occurs if the magnitude of x is too large Returns The sinh function returns the hyperbolic sine value The range error value returned is INF Ox7F8000 2 3 13 6 Integer Arithmetic Functions 2 3 13 6 1 The abs function Synopsis include lt stdlib
7. Running this program should reveal these characteristics Radix 1 2 4 8 10 16 100 256 Precision number of significant digits carried U2 Radix Radix Precision One Ulp OneUlpnit in the Last Place of 1 000xxx Ul 1 Radix Precision One Ulp of numbers a little less than 1 0 Adequacy of guard digits for Mult Div and Subt Whether arithmetic is chopped correctly rounded or something else for Mult Div Add Subt and Sqrt Whether a Sticky Bit used correctly for rounding UnderflowThreshold an underflow threshold E0 and PseudoZero tell whether underflow is abrupt gradual or fuzzy V an overflow threshold roughly vO tells roughly whether Infinity is represented Comparisions are checked for consistency with subtraction and for contamination with pseudo zeros Sqrt is tested Y X is not tested Extra precise subexpressions are revealed but NOT YET tested Decimal Binary conversion is NOT YET tested for accuracy To continue press RETURN Diagnosis resumes after milestone Number 2 Page 3 The program attempts to discriminate among FLAWs like lack of a sticky bit Serious DEFECTs like lack of a guard digit and FAILUREs like 2 2 5 Failures may confound subsequent diagnoses The diagnostic capabilities of this program go beyond an earlier program called MACHAR which can be found at the end of the book Software Manual for the Elementary Functions 1980 by W J Cody and W Waite Although both p
8. int strcoll const char sl const char s2 size_t strxfrm char sl const char s2 size_t n char strstr const char sl const char s2 char strtok char sl const char s2 char strerror int errnum Keil C166 Memory Copying Functions Such functions are not C166 memory copying functions differ from ISO C versions in that Source pointer does not have the const attribute Where a parameter or return value is of type size_t C166 gives the type as unsigned int void s2 void s2 char val void s2 void s2 char val void sl1 void s1 void s void s1 void sl void s unsigned int n unsigned int n unsigned int n char val unsigned int n unsigned int n unsigned int n extern int memcmp extern void memcpy extern void memchr extern void memccpy extern void memmove extern void memset Additional versions of these functions are provided that cater for those situations where memory regions are to be handled that are larger than the default location qualifier for the current memory model permits These functions are listed below unsigned int n unsigned long n extern int fmemcmp void far sl void far s2 unsigned int n extern void far fmemcpy void far sl void far s2 unsigned int n extern void far fmemchr void far s char val unsigned int n extern void far fmemccpy void far sl void far s2 char val extern void far fmemmove vo
9. Fh Fh Fh H n nothree I echo off em compiler compile source file 1 with include dirs 2 3 em logic you must configure all this script for your compiler em remove old object file in case something abends f exist 1 0BJ del 1 0BJ em start a new clg compiler log file h compile 1 gt S1 cl em choose the compile recipe according to number of args 2 goto notwo 3 goto nothree S4 goto nofour f exist phstatus del phstatus em set INCLUDE set external INCLUDE if compiler uses one em format you need to hand configure the compile commands et SPHINS PHIP S 2 3 4 SPHDST SPHSRC conform c OJ 1 S0BI SCFLAGSS gt gt 1 clg I Q Q oe oX N ole oO ct ole lgo I H ole SPHIP S 2 3 SPHDST SPHSRC conform OJ 1 0BJ SCFLAGS gt gt Sl clg lo I Q oe oe N oe e Ill Q oto done ole et SPH SPHIP 2 SPHDST SPHSRC conform PHCC S 2 1 c OJ 1 0BJ SCFLAGS S gt gt 1 clg oto done notwo r i e e g em using only one arg is allowed only if PHSRC PHDST f PHSRC PHDST goto onetree cho usage 0 pgm name source dir include dir cho The source dir is required if you are compiling remote sources oto done onetree S 2 a T r et PHINS PHIP SPHDST SPHSRC conform PHCC S 1 c OJ 1 0BJ CFLAGS gt gt 1 clg done f errorle
10. 1 exvar R4 R4 exvar testvar R4 exvar testvar R4 exvar testvar R4 exvar testvar R4 exvar testvar R4 exvar testvar R4 exvar testvar CC_UC RO RO T AWORD TECHDOC c166 report companyX compiler report doc Revision C Last amended by M Beach on 22 06 00 14 14 R4 R4 28 Page 34 of 83 2 3 20 Incomplete Types And Tentative Declarations The outcome when an identifier for an object with internal linkage and an incomplete type is declared with a tentative definition The declaration static short arr will result in an error and compilations ends No object file is produced and the build process stops There is thus no hazard resulting from this tentative declaration 2 3 21 Shift Left And Shift Right Operations 2 3 21 1 2 3 21 2 The implementation of the shift left and shift right operators shall be documented Left Shift Signed signedtest 0x0001 for i 0 i lt 16 i signedtest lt lt 1 The object is shifted left with zero shifted into the least significant bit The most significant bit is shifted into the carry flag of the C167 program status word PSW Right Shift Signed signedtest 0x8000 for i 0 i lt 16 i signedtest gt gt 1 If the most significant bit 615 was set ONES are shifted into most significant bit If the most significant bit 015 was clear ZEROS are shifted into most significant bit The least s
11. _BOS EQU _TOS 512 gt gt _STKSZ bottom of system stack PUBLIC C_SYSSTKBOT C_SYSSTKBOT EQU _BOS MOV STKOV _BOS 6 2 INITIALIZE STACK OVFL REGISTER L166 automatically reserves the appropriate on chip memory and so no special actions are required by the user Company X Ltd T AWORD TECHDOC c166 report companyX compiler report doc Revision C Last amended by M Beach on 22 06 00 14 14 Page 55 of 83 4 1 1 3 Setting Up The BUSCONx ADDRSELx Registers The operation of BUSCON2 3 4 and ADDRSEL2 3 4 on the C165 7 is identical to the BUSCON1 and ADDRSEL1 on the C166 The big difference to the C166 is that there is a chip select pin for each ADDRSEL Thus if an address is accessed in a region covered by an ADDRSELx BUSCONx pair the corresponding chip select CSx pin goes low to enable the appropriate memory or other device It is essential that the ADDRSELx registers are initialised before the corresponding BUSCONx This is to some extent common sense the BUSCONx contains a BUSACT bus active bit which activates the bus characteristics over the preset ADDRSEL range Unexpected results can occur if the BUSCONx is configured first for the following reasons i The C167 5 data book warns that no two ADDRSEL registers must describe an overlapping region As all the ADDRSELs are set to zero when coming out of RESET enabling two BUSCONSs will cause an overlapping condition ii The BUSCONx region bus characteristics may diff
12. long y long z SOURCE LINE 20 MOV R5 WORD y MOV R4 WORD x MULU R4 R5 MOV R6 WORD z DIVLU R6 MOV R4 MDL MOV WORD a R4 Company X Ltd T AWORD TECHDOC c166 report companyX compiler report doc Revision C Last amended by M Beach on 22 06 00 14 14 Page 40 of 83 2 4 Compiler System Controls Which Impact Software Integrity And Maintainability 2 4 1 Overview This section is the most important part of the report when assessing the integrity of the C166 executable within the product Assuming that the software has been written in accordance with the standards and recommendations listed in the project coding standard the highest risk of malfunction will be due to CPU configuration errors and incorrect placement of specific ROM and RAM objects The sizing and placement of SYSTEM and USER stack must also be verified A detailed examination of the START167 A66 assembler file should be made to ensure that the CPU bus interface is configured in accordance with the hardware design parameters This topic is outside the scope of this document The L166 linker control file should be assessed to ensure that ROM and RAM objects are located at valid hardware addresses Moreover the limitations on physical placement of certain CLASSES must be borne in mind This topic is considered to be within the scope of this report and is examined in section 2 4 4 3 and 2 4 4 8 Other C167 specific safety critical items to be checked for e Critical in
13. the effect of the data page pointer mechanism under all memory models except HCOMPACT and HLARGE Such programming techniques are not advisable in any commercial application It should be noted that all pointers that have no explicitly stated memory space qualifier under the HLARGE and HCOMPACT models contain 4 bytes and are thus more correctly converted to Long types Cast int type to pointer If the value of the int is lt 0x8000 then an address will be access as expected under the HCOMPACT and HLARGE models Under all other models unexpected results may be obtained Cast pointer to int Under HLARGE and HCOMPACT models the address contained in the pointer is transferred to the int as expected provided it is lt 0x10000 Note It would be more appropriate to cast the pointer to an object of type long Refer to the section on recommend memory models for further information A full review of casting basic types to pointers and the associated risks is given in Appendix B 2 3 7 What constitutes an access to an object that has volatile qualified type short volatile xptr A volatile quantity is one whose value may change without CPU and hence compiler intervention Memory mapped IO devices and C167CR sfrs are regarded as volatile Pointers used to test RAM areas must also be volatile 2 3 8 The maximum number of declarators that may modify an arithmetic structure or union type This item has been referred to KEIL Company
14. 11 of 83 2 3 1 1 Test Conditions The following compiler controls were used during testing WARNINGLEVEL 8 The special WARNINGLEVEL 8 will cause the compiler to generate WARNINGs in a lint like fashion It will indicate most sources of potential run time failure due to poor use of the C language or misuse of pointers It has been used throughout the testing phase OPTIMIZE X In some cases the aggressive optimisation carried out by C166 prevents a proper examination of implementation specific areas addressed in this report In those cases where results would not be invalidated by inhibiting the optimiser the pragma OPTIMIZE 0 control was used ERRORS Compiler errors prevent the production of an object file so therefore program cannot be linked or executed Under such circumstances the result of the test for implementation specific issues is irrelevant MOD167 The target CPU was a C167CR that contains instruction set extensions only accessible through the pragma MOD167 control HLARGE The recommended HLARGE memory model was used A justification for this can be found in section 2 4 Company X Ltd T AWORD TECHDOC c166 report companyX compiler report doc Revision C Last amended by M Beach on 22 06 00 14 14 Page 12 of 83 2 3 2 The layout of storage for parameters 2 3 2 1 Basic Data Types In C166 The data types available in C166 are bit 1 bit 0 to 1 signed char 8 bits 128 tov 127 unsigned char 8
15. 2 4 5 5 2 4 5 6 2 4 5 7 High Speed Data Variable data accessed frequently can benefit by being located in the C167 on chip RAM This can be achieved via the idata location qualifier Example Of idata unsigned short idata fast_variable Put this data on chip Very Large Objects Data objects of above OxFFFF bytes in size should be declared with the xhuge keyword Such objects can be up to 16MB in size Example unsigned short xhuge verybigarray 0x20000 A very big array unsigned short xhuge ram_memtestptr A pointer than will test 256kb RAM Objects declared with xhuge will be placed into the XDATA and XDATAO classes These can be placed at any address Typical Pointer Declarations A generalised pointer declaration is thus lt type gt lt const volatile gt lt typequalifier gt lt dataname gt lt type gt lt const volatile gt lt address gt kI Pointers defined with nothing between the asterisk and the lt dataname gt are located by default in the data class determined by the memory model Itis entirely possible to force the pointer itself into a specific memory space By putting a type and typequalifier after the the pointer can be placed This is most frequently required when is desired to put a pointer into EPROM i e a constant class Jump tables or tables of function pointers fall into this category also Example Put a huge constant pointer to a constant string into EPROM itse
16. EXTS seg off instruction MOD167 Huge Pointer Access A hptr 0x55 SOURCE LINE 19 MOV R6 85 MOV R5 WORD hptr 2 MOV R4 WORD hptr EXTS R5 1 MOV R4 R6 R5 holds the 64kb segment number and R6 holds the offset into the segment In effect this a full 32 bit access being equivalent to MOV R5 R4 R6 Company X Ltd T AWORD TECHDOC c166 report companyX compiler report doc Revision C Last amended by M Beach on 22 06 00 14 14 Page 61 of 83 7 Appendix D 7 1 BSI Definition Of Year 2000 Compliance Keil C166 Tools The following Keil tools of equal or later version number to those listed below are Year 2000 compliant according to the DISC PD2000 1 definition follows C51 Compiler for 8051 V5 50 A51 Assembler V5 50 BL51 Linker V3 70 uVision 51 Windows V1 31 DScope51 for Windows OH51 C166 Compiler V3 11 A166 Assembler V3 12 L166 Linker V3 11 uVision 166 Windows V1 31 DScope166 for Windows V1 02 OH166 C251 Compiler V2 12a A251 Assembler V1 4 L251 Linker V1 29 uVision 251 Windows V1 31 DScope251 for Windows V1 3w OH251 V1 4 Notes i Earlier versions of compilers will continue to function but date related operation may not return the correct values ii Older projects that are in the support and maintenance phase using earlier versions of Keil compilers may suffer from Year 2000 problems if they rely on the _ DATE __ intrinsic macro Other problems may arise and you sh
17. Library Functions Not Implemented In Keil C166 abort fgets ldexp asctime fmod ldiv assert fopne localeconv atexit fprintf localtime bsearch fputc mblen clearerr fputs mbstowcs clock fread mkt ime ctime freopen perror difftime frexp putc div Fscanft qsort exit fseek raise fclose fsetpos remove feof ftell rename ferror fwrite reweind fflush getc setlocale fgetc getenv setvbuf fgetpos gmt ime strcoll Company X Ltd T AWORD TECHDOC c166 report companyX compiler report doc Revision F Last amended by M Beach on 22 06 00 14 14 S S trerror trftime trstr trtod trok trol troul trxfrm system time tmpfile tmpnam ungetc vfprintf wcstombs wctomb Page 7 of 83 2 2 2 2 2 2 1 2 2 2 2 2 2 2 3 Standard ISO C library functions that differ in prototype from what would be expected Type Of size_t size_t is the unsigned integer type result of the sizeof operator It is implemented as a typedef in stddef h Example typedef unsigned int size_t Keil C166 String Handling Function Prototypes The string handling functions whose prototypes are listed in string h differ slightly from the ISO standard in that the source pointer does not have the const attribute const objects in the C166 compiler are defined as being located in a read only area such as ROM The limitation of source strings only being located in ROM would be impracticable in a real C167 system Where a para
18. The LST file emitted by the compiler contains the original C source lines with a line number prefix and the WARNING or ERROR message placed in the file so as to indicated where the problem occurred Example Of A Warning Diagnostic x WARNING 34 IN LINE 7 OF MAIN C incompletevar missing declaration specifiers The WARNING number can be referenced in the Keil C166 User Manual Example Of An Error Diagnostic ERROR 25 IN LINE 24 OF MAIN C syntax error near The ERROR number can be referenced in the Keil C166 User Manual Company X Ltd T WORD TECHDOC c166 report companyX compiler report doc Revision C Last amended by M Beach on 22 06 00 14 14 Page 17 of 83 2 3 3 4 2 3 3 4 1 Example Of A LST File With ERROR And WARNING Messages C166 COMPILER V3 12 MAIN 13 02 99 20 24 00 PAGE 1 DOS C166 COMPILER V3 12 COMPILATION OF MODULE MAIN OBJECT MODULE PLACED IN MAIN OBJ COMPILER INVOKED BY C C166 BIN C166 EXE MAIN C DB M167 WL 3 HLARGE stmt level source pragma WARNINGLEVEL 3 pragma OT 0 Sample Error And Warning Messages test MIAH BRUNE incompletevar Deliberate incomplete definition C166 assumes type Warts xxx WARNING 34 IN LINE 8 OF MAIN C incompletevar missing declaration specifiers 9 10 unsigned short svar 1 12 extern arr 13 14 15 void main void 16 1 17 ac arr 1 svar 18 1 19 1 incompletevar arr 3 20 1 21 1 if svar 1
19. X Ltd T WORD TECHDOC c166 report companyX compiler report doc Revision C Last amended by M Beach on 22 06 00 14 14 Page 21 of 83 2 3 9 Whether the value of a single character character constant in a constant expression that controls conditional inclusion matches the value of the same character constant in the execution character set Whether such a character constant may have a negative value The C166 compiler has common source and execution character sets Thus a single character character constant in a constant expression that controls conditional inclusion will always match the value of the same character constant in the execution character It should be noted that character constants used in conditional include controls must not be simple numerical quantities either positive or negative Furthermore all conditional inclusion constants manifest constants and macros must not begin with a numerical character of any kind 2 3 10 The mapping of source file character sequences The C166 compiler uses identical source and execution character sets There is therefore no mapping of characters between source and execution sets 2 3 11 The null pointer constant to which the macro NULL expands The NULL macro expands to void OL A NULL or uninitialised pointer will return the value found at location 0x0000 In most cases this will be byte value OxFA this being the first byte of the JMPS instruction that is normally located at ad
20. after milestone Number 40 Page 6 Checking rounding on multiply divide and add subtract Multiplication appears to round correctly is neither chopped nor correctly rounded Addition Subtraction neither rounds nor chops Sticky bit used incorrectly or not at all FLAW lack s of guard digits or failure s to correctly round or chop noted above count as one flaw in the final tally below Does Multiplication commute Testing on 20 random pairs No failures found in 20 integer pairs Running test of square root x Testing if sqrt X X X for 20 Integers X Test for sqrt monotonicity sqrt has passed a test for Monotonicity Testing whether sqrt is rounded or chopped Square root is neither chopped nor correctly rounded Observed errors run from 0 0000000e 00 to 5 0000000e 01 ulps To continue press RETURN Diagnosis resumes after milestone Number 90 Page 7 Testing powers Z i for small Integers Z and i WARNING computing 0 00000000000000000e 00 0 00000000000000000e 00 yielded WWYVVWWVVVIWVWYVV which compared unequal to correct 1 00000000000000000e 00 they differ by VWVVVVVWVWVWVIWVWYVY Similar discrepancies have occurred 5 times DEFECT computing 5 00000000000000000e 01 2 60000000000000000e 01 yielded 1 49011400000000000e 08 which compared unequal to correct 1 49011600000000000e 08 they differ by 2 04281000000000000e 14 Errors like this may invalidate financial calculations involving interes
21. are stored in the following arrangement Address 0 1 2 3 Contents MMMM MMMM MMMM MMMM MMMM MMMM SEEE EEEE The mantissa MMM MMMMM is a 24 bit quantity The exponent EEE EEEE is stored as a twos complement value with a 127 offset The sign of the exponent S is a single bit value Company X Ltd T AWORD TECHDOC c166 report companyX compiler report doc Revision C Last amended by M Beach on 22 06 00 14 14 Page 39 of 83 2 3 25 Type promotion char To int Automatic promotion of char to int only occurs in arithmetic operations Therefore no explicit casting from char to int is required in the following example Example resultcharO varcharO varcharl varchar2 Gives the correct result as the implied 16 bit value in the numerator is allowed for by the use of a 16x8 divide instruction Note Promotion from char to int in compare operations does not occur 2 3 26 Type Promotion From int To long As dictated by the ISO C standard there is no automatic promotion from int to Long This must be considered in the following situation unsigned int x y z ZS eR y fz Here there is an implied Long value in the numerator that will be lost unless the appropriate cast to long is used The correct ISO C way to perform this calculation is unsigned int x y z z unsigned long x unsigned long y unsigned long z In tested version of C166 the following efficient code will result i a long x
22. bits O to 255 signed int 16 bits 32768 to S3 2767 signed short 16 bits 32768 to 32767 unsigned int 16 bits 0 to 65535 unsigned short 16 bits 0 to 65535 signed long 32 bits 2147483648 to 2147483647 unsigned long 32 bits O 429496795 float 32 bits 1 176E 38 to 3 4E 38 double 64 bits 1 7E 308 to 1 7E 308 pointer 16 32 bits Variable address Notes i The 16 bit ANSI short type equates exactly to int The latter takes the natural size of the CPU here 16 bits ii The machine size of the C166 is 16 bits hence int short or unsigned int unsigned short should be used when possible to produce the most compact and efficient code The use of char and unsigned char will result in a lot of MOVBZ type instructions which waste time iii char and unsigned char Unless a signed 8 bit number is required always use unsigned char Char is normally reserved for ASCII characters which have no sign 2 3 2 2 Representation Of Data The C166 uses the Intel little endian byte order 2 3 2 2 1 Significant Character Values The C166 compiler uses the following significant character values NEWLINE n LINEFEED r String null terminator 0 Company X Ltd T AWORD TECHDOC c166 report companyX compiler report doc Revision C Last amended by M Beach on 22 06 00 14 14 Page 13 of 83 2 3 2 2 2 2 3 2 2 3 2 3 2 2 4 2 3 2 3 Representation Of 16 Bit Quantities Thus
23. guard digits or failure s to correctly round or chop noted above count as one flaw in the final tally below Does Multiplication commute Testing on 20 random pairs DEFECT X Y Y X trial fails Running test of square root x SERIOUS DEFECT sqrt 1 01330991615836160e 016 1 00663296000000000e 008 00e 000 1 000000000000000 instead of correct value 0 SERIOUS DEFECT sqrt 9 86864969599650250e 017 9 93410775862457740e 009 90e 017 9 868649843050824 instead of correct value 0 SERIOUS DEFECT sqrt 9 86864969599650250e 017 9 93410775862457740e 009 90e 017 9 868649843050824 instead of correct value 0 To continue press RETURN sqrt 1 01330991615836160e 016 1 00663296000000000e 008 00e 000 1 000000000000000 Company X Ltd T AWORD TECHDOC c166 report companyX compiler report doc Revision C Last amended by M Beach on 22 06 00 14 14 Page 71 of 83 instead of correct value 0 SERIOUS DEFECT sqrt 9 86864969599650250e 017 9 93410775862457740e 009 9 868649843050824 90e 017 instead of correct value 0 SERIOUS DEFECT sqrt 9 86864969599650250e 017 9 93410775862457740e 009 90e 017 9 868649843050824 instead of correct value 0 To continue press RETURN Diagnosis resumes after milestone Number 70 Page 7 Testing if sqrt X X X for 20 Integers X DEFECT sqrt 1 01330991615836160e 016 1 00663296000000000e 008 1 000000029802322 40e 0
24. h gt int abs int j Description The abs function computes the absolute value of an integer j If the parameter is the most negative integer the behaviour is undefined Returns The abs function returns the absolute value If the input parameter is the most negative integer the result is the most negative integer no action having been taken by abs Company X Ltd T AWORD TECHDOC c166 report companyX compiler report doc Revision C Last amended by M Beach on 22 06 00 14 14 Page 27 of 83 2 3 13 6 2 The labs function Synopsis include lt stdlib h gt long labs long 4 Description The labs function is equivalent to the abs function except that the argument and the return value each have type long Returns The labs function returns the absolute value If the input parameter is the most negative integer the result is the most negative integer no action having been taken by labs 2 3 14 Compiler Actions With Unusually Terminated Source Files This section determines the outcome when a non empty source file is compiled and 2 3 14 1 Source file does not end in a new line character Result Compilation succeeds without error or warning Company X Lid T AWORD TECHDOC c166 report companyX compiler report doc Revision C Last amended by M Beach on 22 06 00 14 14 Page 28 of 83 2 3 14 2 Source file ends in a new line character immediately preceded by a backslash character Result The compiler
25. not be acceptable This control is therefore better used as a pragma STATIC with only those modules which contain functions which are used non reentrantly such as interrupt routines HOLD location size Instructs the compiler to put any object less than the given size into the stated memory area In the example any object of 4 bytes or less i e long short int char should be placed on chip This is a way of making sure that large objects such as arrays and structures do not eat up valuable fast on chip RAM Here IDATA means the on chip RAM at OxFAO00 OxF600 on the C167 PECDEF a b c To ensure that the linker does not attempt to place data at the PEC pointer addresses the user must list the PEC channels initialised as arguments to the PECDEF control Failure to do so will result in malfunctions of the PEC system with unpredictable results Company X Ltd T AWORD TECHDOC c166 report companyX compiler report doc Revision C Last amended by M Beach on 22 06 00 14 14 Page 43 of 83 2 4 3 Linker Controls The use of the linker via its LIN control file is central to ensuring correct and reliable system operation The user should satisfy himself that the addresses give for the CLASSES and SECTIONS created by the compiler are suitable for the target hardware platform Some guidance on this is given in section 2 4 3 2 4 4 CPU Configuration Controls In START167 A66 The proper configuration of the START 167 A66 file is cru
26. to undefined symbols ERROR messages are stored at the end of the M66 file as well as being reported in the machine readable only LER file The occurance of errors during linking will prevent the production of an executable OMF66 file Any previous OMF66 will be deleted It is therefore not possible to execute a program containing linking errors Example Of ERROR k k ERROR 127 UNRESOLVED EXTERNAL SYMBOL SYMBOL nonexistantfunc MODULE D C166DEV ALARIS PTRCAST DUMMY OBJ DUMMY x k ERROR 128 REFERENCE MADE TO UNRESOLVED EXTERNAL SYMBOL nonexistantfunc MODULE D C166DEV ALARIS PTRCAST DUMMY OBJ DUMMY ADDRESS 01B1H k k ERROR 128 REFERENCE MADE TO UNRESOLVED EXTERNAL SYMBOL nonexistantfunc MODULE D C166DEV ALARIS PTRCAST DUMMY OBJ DUMMY ADDRESS 01B2H L166 RUN COMPLETE 0 WARNING S 3 ERROR S 2 3 4 What constitutes an interactive device For the purposes of a C166 project the definition of an interactive device may sensibly restricted to the on chip serial port or any off chip UART that is utilised by formatted IO functions such as printf scanf and their derivatives This definition may be extended to include on and off chip peripherals Company X Ltd T AWORD TECHDOC c166 report companyX compiler report doc Revision C Last amended by M Beach on 22 06 00 14 14 Page 19 of 83 2 3 5 The number of bits in a character in the execution character set 2 3 5 1 2 3 5 2 The C166 compiler uses the
27. 00 instead of correct value 0 Test for sqrt monotonicity sqrt has passed a test for Monotonicity Testing whether sqrt is rounded or chopped PROGRAM HANGS UP AT THIS POINT PROGRAM MODIFIED TO REMOVE SQRT TEST OUTPUT LOGGING RESUMED Diagnosis resumes after milestone Number 70 Page 7 Testing if sqrt X X X for 20 Integers X DEFECT sqrt 1 01330991615836160e 016 1 00663296000000000e 008 1 000000029802322 40e 000 instead of correct value 0 Test for sqrt monotonicity sqrt has passed a test for Monotonicity To continue press RETURN Diagnosis resumes after milestone Number 90 Page 8 Testing powers Z i for small Integers Z and i DEFECT computing 1 00000001490116120e 001 7 00000000000000000e 000 yielded 1 00000008274037100e 007 which compared unequal to correct 1 00000015379464460e 007 they differ by 7 10542735760100190e 015 Errors like this may invalidate financial calculations involving interest rates Similar discrepancies have occurred 10 times To continue press RETURN calculation may suffer larger Relative error than merely roundoff DEFECT Range is too narrow U1 5 Underflows Since underflow occurs below the threshold UfThold 1 00663296000000000e 008 3 96250009536743160e 000 only underflow should afflict the expression 1 00663296000000000e 008 7 92500019073486330e 000 actually calculating yields 0 00000000000000000e 000 This computed value is O K Te
28. 166 compiler controls must be used if the userstack is moved on chip This is to allow the compiler to correctly produce pointers to data that is on this stack Failure to do this will result in undefined results The System Stack With the user stack taking care of function parameters and local variables the system stack is used for storing return addresses the current PSW and CP plus any general purpose registers in the current register bank used for local register variables This stack is always located on chip and defaults to 256 words in length 80C 166 at Oxfbff down to Oxfa00 The required stack size is set in the START167 A66 file Values of 32 64 128 and 256 words can be selected via SYSCON The CPU register SP has its top 5 bits hard wired to 1 the stack is always in the range Oxf800 to Oxfffe i e on chip Setting The System Stack Size STKSZ Maximum System Stack Size selection SYSCON 13 SYSCON 14 _STKSZ EQU 0 System stack sizes A 0 256 words Reset Value 1 128 words 3 2 64 words 3 32 words The C166 is endowed with two special registers STKOV and STKUN which set the top and bottom limits of the stack The default value of STKOV Stack overflow is Oxfa00 whilst STKUN Stack underflow defaults to Oxf00 in line with the default 256 words The address of the stack is defined by loading the STKOV register is startup a66 Setup Stack Overflow OS EQU OFCOOH top of system stack
29. 24 324 0 0 conform exprtest bandeq out 288 288 0 0 conform exprtest cast out 162 162 0 0 conform exprtest compl out 288 288 0 0 conform exprtest div out 120 120 0 0 conform exprtest diveql out 156 156 0 0 conform exprtest diveq2 out 156 156 0 0 conform exprtest diveq3 out 144 144 0 0 conform exprtest diveq4 out 288 288 0 0 conform exprtest eq out 288 288 0 0 conform exprtest ge out 288 288 0 0 conform exprtest gt out 13 13 0 0 conform exprtest intl out 9 9 0 0 conform exprtest int2 out 9 9 0 0 conform exprtest int3 out 12 12 0 0 conform exprtest int4 out 13 13 0 0 conform exprtest int5 out T2 12 0 0 conform exprtest int6 out 13 13 0 0 conform exprtest int7 out 15 15 0 0 conform exprtest int8 out 13 13 0 0 conform exprtest int9 out 10 10 0 0 conform exprtest int10 out 13 13 0 0 conform exprtest int11l out 14 14 0 0 conform exprtest int12 out 11 1 0 0 conform exprtest int13 out 288 288 0 0 conform exprtest le out 162 162 0 0 conform exprtest 1sh out 324 324 0 0 conform exprtest 1sheq out 288 288 0 0 conform exprtest 1t out T22 122 0 0 conform exprtest mineql out 154 154 0 0 conform exprtest mineq2 out 152 152 0 0 conform exprtest mineq3 out 136 136 0 0 conform exprtest mineq4 out 288 288 0 0 conform exprtest minus out 12 12 0 0 conform exprtest mixl out 9 9 0 0 conform exprtest mix2 out 9 9 0 0 conform exprtest mix3 out 12 12 0 0 conform exprtest mix4 out 9 9 0 0 conform exprtest mix5 out 14 14 0 0 conform exprtest mix6 out 11 T1 0 0 co
30. 3 out 144 144 0 0 conform exprtest diveq4 out 288 288 0 0 conform exprtest eq out 288 288 0 0 conform exprtest ge out 288 288 0 0 conform exprtest gt out 13 13 0 0 conform exprtest intl out 9 9 0 0 conform exprtest int2 out 9 9 0 0 conform exprtest int3 out 12 12 0 0 conform exprtest int4 out 13 13 0 0 conform exprtest int5 out T2 12 0 0 conform exprtest int6 out 13 13 0 0 conform exprtest int7 out 15 15 0 0 conform exprtest int8 out 13 13 0 0 conform exprtest int9 out 10 10 0 0 conform exprtest int10 out 13 13 0 0 conform exprtest int11l out 14 14 0 0 conform exprtest int12 out 11 1 0 0 conform exprtest int13 out 288 288 0 0 conform exprtest le out 162 162 0 0 conform exprtest 1sh out 324 324 0 0 conform exprtest 1sheq out 288 288 0 0 conform exprtest 1t out T22 122 0 0 conform exprtest mineql out 154 154 0 0 conform exprtest mineq2 out 152 152 0 0 conform exprtest mineq3 out 136 136 0 0 conform exprtest mineq4 out 288 288 0 0 conform exprtest minus out 12 12 0 0 conform exprtest mixl out 9 9 0 0 conform exprtest mix2 out 9 9 0 0 conform exprtest mix3 out 12 12 0 0 conform exprtest mix4 out 9 9 0 0 conform exprtest mix5 out 14 14 0 0 conform exprtest mix6 out 11 T1 0 0 conform exprtest mix7 out 17 17 0 0 conform exprtest mix8 out 11 11 0 0 conform exprtest mix9 out T2 T2 0 0 conform exprtest mix10 out 15 15 0 0 conform exprtest mix11l out 16 16 0 0 conform exprtest mix12 out 14 14 0 0 conform exprtest mix13 out 288 288 0 0 conform ex
31. 45 A listing of the results of the test are given in appendix F Note Plum Hall is a registered trademark of Plum Hall Inc USA 2 1 4 Empirical Measures Of Compiler Integrity The Keil C166 compiler is in use at approximately 300 UK locations Based on UK usage on average one minor bug is reported and confirmed once every 12 weeks across approximately 300 active users of C166 v3 05a to 3 12 The average time to fix a bug is three weeks from confirmation of the problem Keil has been represented in the UK by Hitex since 1989 and is a tier one supplier Company X Ltd T AWORD TECHDOC c166 report companyX compiler report doc Revision F Last amended by M Beach on 22 06 00 14 14 Page 6 of 83 2 1 5 Compiler Problem Status By special arrangement with Keil a list of known problems for the tested release can be made available 2 2 Standard Functions 2 2 1 Standard ISO C library functions not provided Such information for C166 v3 12j can be found in appendix G The Keil C166 compiler contains a subset of the ANSI library function set that is appropriate for use in a deeply embedded C167 environment It omits the following specific and classes of library functions Stream IO functions fflush fputc fprintf All file orientated functions fopen fclose All signal functions All error handling functions All time and date functions All environment communication functions Type generic functions List Of ISO C
32. ASCII 7 bit character set Extended characters i e those with a value of greater than Ox7F are not supported Multi byte or wide characters are also not supported Legal Characters The available character set is summarised below The 26 uppercase letters of the Latin alphabet ABCDEFGHIJKLMNOPQRSTUVWXYZ The 26 lowercase letters of the Latin alphabet abcdefghigjgkimnopgqrstuvwxyaz The 10 decimal digits 012345678 9 The following 29 graphic characters fe aS pe Ber 8 OG Jo he Ge ge of PER Saree ee ON EO Millie he plus the SPACE character and control characters representing HORIZONTAL TAB VERTICAL TAB and FORMFEED If any character not listed above is encountered in a source file compilation ends and an ERROR message generated Termination Of Character Strings Strings of characters in the source character set are terminated with a character having numerical value ZERO Company X Ltd T AWORD TECHDOC c166 report companyX compiler report doc Revision C Last amended by M Beach on 22 06 00 14 14 Page 20 of 83 2 3 6 The result of casting a pointer to an integer or vice versa Note This aspect of the C166 compiler implementation is very much dependent on the model used Warning This will result in unpredictable accesses to memory if a pointer is used that was the result of a cast from int THIS MUST NOT BE DONE Only unsigned int types can be cast to pointers but even then the user must take into account
33. C Company X Ltd T AWORD TECHDOC c166 report companyX compiler report doc Revision C Last amended by M Beach on 22 06 00 14 14 Page 80 of 83 PHLI L166 PHSRC plumhall suite PHTMP This is a very long string that sets aside environment space PHTMP2 This is a very long string that sets aside environment space PROCESSOR_ARCHITECTURE x86 PROCESSOR_IDENTIFIER x86 Family 6 Model 5 Stepping 2 GenuineIntel PROCESSOR_LEVEL 6 PROCESSOR_REVISION 0502 PROMPT SPSG STARTUP strt 67mm SYSPATH C WINNT system32 C WINNT C KEIL C166 BIN C UTIL SystemDrive C SystemRoot C WINNT TEMP C TEMP TMP C TEMP USERDOMAIN RK USERNAME Administrator USERPROFILE C WINNT Profiles Administrator windir C WINNT WRKDRV D 9 1 4 Explanation Of Error Reports From The Plum Hall Test 9 1 4 1 9 1 4 2 9 1 4 3 9 1 4 4 The PLUM HALL test reports 4 Errors for the Keil C166 compiler that are characterised below LANG Problem LANG this is a complex macro that is not expanded correctly ERROR in c68 c at line 239 1 2 h NEITHER 1 2 NOR 1 h 2 This problem appears when multiple nested macros are called with text expansions M61 Problem M61 there is not compiler error message for following construct enum A INT_MAX B B overflows int value range for enum M64 Problem M64 there is not compiler error message for switch i case LONG_MAX 4 LONG_MAX 4 overflow
34. ECHDOC c166 report companyX compiler report doc Revision C Last amended by M Beach on 22 06 00 14 14 Page 37 of 83 2 3 24 2 65535 99 00000 01 65536 00 0x10000 Double Precision Double represents 14 15 significant digits However it has been determined that in typical engineering calculations only 13 digits can be relied on var64 65535 5000000001 If the length of the initialisation value exceeds 14 significant digits the entire fractional portion is lost Rounding up of fractional parts only occurs when the value to be converted is within one unit of resolution i e the lowest representable value possible given the precision and size of the value to be converted The absolute value of this lowest representable value is dependent entirely on the magnitude of the object and thus cannot be defined here Example double integer 65535 9999999999 65535 OxFFFF 65535 9999999999 00000 0000000001 65536 0x10000 Company X Ltd T AWORD TECHDOC c166 report companyX compiler report doc Revision C Last amended by M Beach on 22 06 00 14 14 Page 38 of 83 2 3 24 3 Limits Of Single And Double Precision Floating Point Types The C166 compiler has the following maximum and minimum values for floating point quantities Maximum float value 3 40282e38 Minimum float value 1 17549e 38 Maximum double value 1 79769313486231500e 308 Minimum double value 2 22507385850720200e 308 Note Single precision floating point numbers
35. FF Positive Infinity INF 0x7FF0000000000000 Negative Infinity INF OxFFFO000000000000 Note there is no provision for floating point exceptions in C166 and it is up to the user to adequately respond to floating point errors Company X Ltd T WORD TECHDOC c166 report companyX compiler report doc Revision C Last amended by M Beach on 22 06 00 14 14 Page 24 of 83 2 3 13 3 Trigonometric Functions That Are Subject To Domain Or Range Errors 2 3 13 4 The acos Function Synopsis include lt math h gt double acos double x Description The acos function computes the principal value of the arc cosine of x A domain error occurs for arguments not in the range 1 1 Returns The acos function returns the arc cosine in the range 0 pi radians The domain error value returned is NaN OxFFFFFFFF 2 3 13 4 1 The asin Function Synopsis include lt math h gt double asin double x Description The asin function computes the principal value of the arc sine of x A domain error occurs for arguments not in the range 1 1 Returns The asin function returns the arc sine in the range pi 2 radians The domain error value returned is NaN OxFFFFFFFF Company X Lid T AWORD TECHDOC c166 report companyX compiler report doc Revision C Last amended by M Beach on 22 06 00 14 14 Page 25 of 83 2 3 13 4 2 The atan2 Function Synopsis include lt math h gt double atan2 double y double x
36. Federation of Small Businesses IBM ICI National Power Paymaster Agency Prudential Assurance Reuters Tesco Stores While every care has been taken in developing this document the contributing organisations accept no liability for any loss or damage caused arising directly or indirectly in connection with reliance on its contents except to the extent that such liability may not be excluded at law Independent legal advice should be sought by any person or organisation intending to enter into a contractual commitment relating to Year 2000 conformity requirements THE DEFINITION Year 2000 conformity shall mean that neither performance nor functionality is affected by dates prior to during and after the year 2000 In particular Rule 1 No value for current date will cause any interruption in operation Rule 2 Date based functionality must behave consistently for dates prior to during and after year 2000 Rule 3 In all interfaces and data storage the century in any date must be specified either explicitly or by unambiguous algorithms or inferencing rules Rule 4 Year 2000 must be recognised as a leap year Company X Ltd T WORD TECHDOC c166 report companyX compiler report doc Revision C Last amended by M Beach on 22 06 00 14 14 Page 63 of 83 AMPLIFICATION OF THE DEFINITION AND RULES General Explanation Problems can arise from some means of representing dates in computer equipment and products and from date logi
37. KEN Setup SYSCON Register Company X Ltd T AWORD TECHDOC c166 report companyX compiler report doc Revision C Last amended by M Beach on 22 06 00 14 14 Page 56 of 83 It is possible to create pointers to local data i e automatics although it is not really good practice If the user stack has been moved into the IDATA on chip RAM you must use the USERSTACKDPP3 compilation control This will force C166 to use DPP3 as the base for the pointer so that it can happily point to the user stack in IDATA rather than via DPP2 into the NDATA area pragma USERSTACKDPP3 Force compiler to use DPP3 when finding address of local array void funcl char a char b char rx_buffer 0x20 This array will be on user stack char rx_ptr rx_ptr amp rx_buffer a Point to ath object in local array Company X Ltd T AWORD TECHDOC c166 report companyX compiler report doc Revision C Last amended by M Beach on 22 06 00 14 14 Page 57 of 83 5 Appendix B 5 1 Pointer Casting And Conversions It is entirely possible to turn a simple C data type int long into a pointer However the user must be aware of the number of bytes required to form the pointer in each case near pointers One word carrying information on offset from DPP2 page number 0 0x3FFF far huge xhuge pointers Two words carrying information on the page number in the upper word and offset from page base in the lower sdata pointers One wo
38. O is never changed from its power up value unless the user alters it himself Therefore C166 need not stack these two registers on entry to an interrupt routine thereby saving two PUSHes and POPs 0 4us 20MHz NOALIAS The NOALIAS control can cause unexpected results in when global or static objects are modified by pointers In most cases source code that could be upset by this control does not conform to the MISRA C guidelines and should therefore not be present anyway It is recommended that this control is not used Example Taken From C166 User Manual Without NOALIAS struct int il inti2 p_struct int val void funcl int p_val p_struct gt il val Read val p_val 0 Zero val via pointer p_struct gt i2 val Read val again Now it is zero void func2 void funcl amp val Company X Ltd T AWORD TECHDOC c166 report companyX compiler report doc Revision C Last amended by M Beach on 22 06 00 14 14 Page 42 of 83 With NOALIAS Struct int ily intiz pustruct int val void funcl int p_val p_struct gt il val Read val p val 0 Zero val via pointer p_struct gt i2 val Read val again Value from first read carried forward so zeroing by pointer in previous line is not seen void func2 void funcl amp val NOFIXDPP The user stack is by default in the near data area which is addressed by DPP2 The NOFIXDPP control is requi
39. OS SOURCE LINE 166 OOF4 F68E0200 R MOV trigger_offset ZEROS SOURCE LINE 168 OOF8 EOOA MOV R10 00H The use of REGISTERMASKs is controlled by the global Register Optimisation item on the Options Make Miscellaneous tab in uVISION Note1 Although the global register optimisation mechanism has been shown to be reliable it is recommended that it is not selected in a safety critical system Note2 Although the ISO C register location qualifier is compiled it has no effect on the compiler register variable allocation strategy and is thus redundant in C166 With assembler coded functions the user should manually work out the register mask and add it to the assembler function s prototype at the top of the C source file Functions written in assembler are assumed to have the worst case register usage and so any C functions calling assembler will not have any registers available to them It is thus up to the user to manually generate the register mask and attach it to the function prototype that the calling module sees extern void asm_funcO void 0x0010 extern void asm_funcl void 0x0018 extern void asm func2 void 0x0000 extern void asm_func3 void 0x8000 The assembler function uses only R4 The assembler function uses R4 amp R3 The assembler function uses all registers The assembler function no registers Nee eS i s The register mask for the user s own assembler functions by reference to the f
40. PC compilers such as Microsoft C MSC for the 80x86 The calculation of the 166 s page number is made during the cast at run time hence all the extra code produced above In MSC the long must have the segment number already in the right place In the above example rather than containing 0x30010 the long would have to have held 0x3000 0010 This could cause problems when converting MSC programs to the 166 Company X Ltd T AWORD TECHDOC c166 report companyX compiler report doc Revision C Last amended by M Beach on 22 06 00 14 14 Page 59 of 83 6 Appendix C 6 1 Pointers In The C166 Compiler Pointers to data in the near data area unsurprisingly called termed near pointers They may point at any object in the near data area and as might be expected from the 16k pages no single object may be over 16k A near pointer in C166 is itself 16 bits with the top two bits fixed at 02 to indicate DPP2 6 2 The Most Common Pointers In C166 6 2 1 far Pointers far pointers are less restricted in that the object being pointed at can be at any 18 or 24 bit address but the size of the object being pointed at must be limited to 16kb As the value of DPPO is calculated each time the pointer is used far pointers are thus somewhat slower than near pointers The 16kb limit only causes problems the user attempts to increment a pointer over this range as might happen when using a pointer to access a large array The reason for the limit is that w
41. age 48 of 83 2 5 Floating Point Library Test A paranoia check will be performed to validate the adequacy of the floating point arithmetic provided by the compiler 2 5 1 Preliminary Work The supplied PARANOIA program was compiled with the Keil C166 compiler Some modifications were required to the keyboard reading function to allow the required user input The HLARGE memory model was used and the single floating point default mechanism enabled 2 5 2 Preliminary Results 2 5 2 1 2 5 2 2 It was noted that the program source was seriously at variance with the requirements of MISRA C and accepted good C programming practice A large number of set jmp goto and other inappropriate statements were observed The operation of the program was undocumented The program consisted of one enormous monolithic function and if nothing else the ability of C166 to compile it without error or memory exhaustion provides considerable reassurance as to the compiler s robustness The suitability of the program for testing floating point performance of an embedded CPU is questionable As aconfidence check the program was also tested on the alternative Tasking C166 compiler and industry standard Microsoft C v11 00 compiler The full report listings are given in Appendix E Keil C166 3 12j Results The number of FAILURES encountered The number of SERIOUS DEFECTs discovered The number of DEFECTs discovered The number of FLAWs discovered
42. ast one function and a number of other sections typically 1 to 8 depending on the memory types used in variable declarations 2 Number of global public identifiers 32767 All file level data and function definitions not explicitly declared with the storage class st atic are public Number of external identifiers 32767 Number of interrupt procedures 128 This limit is imposed by the C167 architecture Company X Ltd T WORD TECHDOC c166 report companyX compiler report doc Revision C Last amended by M Beach on 22 06 00 14 14 Page 51 of 83 3 Composition Of The Compiler Toolkit The Keil C166 compiler toolkit consists of the following items Name Revision Description C166 3 12 C Compiler A166 3 13 C166 7 Assembler L166 3 13 Linker LIB166 4 01 Librarian Tool OH166 3 20 Object To Hex Converter uVISION 1 32 Windows workbench amp make utility Company X Ltd T WORD TECHDOC c166 report companyX compiler report doc Revision C Last amended by M Beach on 22 06 00 14 14 Page 52 of 83 APPENDICES Company X Ltd T AWORD TECHDOC c166 report companyX compiler report doc Revision C Last amended by M Beach on 22 06 00 14 14 Page 53 of 83 4 Appendix A 4 1 1 Moving The USERSTACK On Chip The user stack is fixed in a section named C_CUSERSTACK part of the NDATA class However it is quite common to place it in the IDATA class so that it can be on chip As the user stack is rare
43. c embedded in purchased goods or services as the year 2000 approaches and during and after that year As a result equipment or products including embedded control logic may fail completely malfunction or cause data to be corrupted To avoid such problems organisations must check and modify if necessary internally produced equipment and products and similarly check externally supplied equipment and products with their suppliers The purpose of this document is to allow such checks to be made on a basis of common understanding Where checks are made with external suppliers care should betaken to distinguish between claims of conformity and the ability to demonstrate conformity Rule 1 1 1 This rule is sometimes known as general integrity 1 2 If this requirement is satisfied roll over between all significant time demarcations e g days months years centuries will be performed correctly 1 3 Current date means today s date as known to the equipment or product Rule 2 2 1 This rule is sometimes known as date integrity 2 2 This rule means that all equipment and products must calculate manipulate and represent dates correctly for the purposes for which they were intended 2 3 The meaning of functionality includes both processes and the results of those processes 2 4 If desired a reference point for date values and calculations may be added by organisations e g as defined by the Gregorian calendar 2 5 No equipment or p
44. ced range UfThold V 0 00000000000000000e 000 is too far from 1 What message and or values does Division by Zero produce This can interupt your program You can skip this part if you wish Do you wish to compute 1 0 REPLY YES This much error may spoil financial calculations involving tiny interest rates Testing powers Z Q at four nearly extreme values DEFECT computing 1 00000000000000000e 001 3 20000000000000000e 001 yielded 1 00000003318135350e 032 which compared unequal to correct 1 02679679275673470e 032 they differ by 2 67967595753811630e 030 Similar discrepancies have occurred 4 times To continue press RETURN Diagnosis resumes after milestone Number 190 Page 10 DEFECT Badly unbalanced range UfThold V 0 00000000000000000e 000 is too far from 1 What message and or values does Division by Zero produce This can interupt your program You can skip this part if you wish Do you wish to compute 1 0 y Trying to compute 1 0 produces 1 INFO00e 000 Do you wish to compute 0 0 REPLY YES 1 00000000000000000e 001 3 20000000000000000e 001 yielded 1 00000003318135350e 032 which compared unequal to correct 1 02679679275673470e 032 they differ by 2 67967595753811630e 030 Similar discrepancies have occurred 4 times To continue press RETURN Diagnosis resumes after milestone Number 190 Page 10 DEFECT Badly unbalanced range UfThold V 0 00000000000000000e 000 is too far fro
45. ch on 22 06 00 14 14 Page 83 of 83
46. cial to correct system performance The user should verify that the settings made are appropriate to the hardware design of the target system This topic is outside the scope of this report but Hitex would be willing to assist in any appraisal of the configuration proposed or in use 2 4 5 Characteristics Of HLARGE Model Programs 2 4 5 1 Defaults Under HLARGE Model Default data memory space for HLARGE model huge Under HLARGE any data declaration that does not contain a location qualifier will be of type huge except under circumstances described in section 2 4 1 3 under the pragma HOLD control Examples unsigned short hvar Variable in HDATAO class unsigned short hptr Pointer to huge default unsigned short object unsigned short const hconstant 2 unsigned short constant in HCONST class in ROM area Company X Ltd T AWORD TECHDOC c166 report companyX compiler report doc Revision C Last amended by M Beach on 22 06 00 14 14 Page 44 of 83 2 4 5 2 2 4 5 3 2 4 5 4 Data Placement And Size Limitations e Total size of all data declared without huge memory space qualifier lt 16MB e Largest single array structure or union lt OxFFFF bytes e No array structure or union must be allowed to straddle a 64k segment boundary Note The L166 linker will not permit this to happen e No pointer must be incremented or decremented across a 64k segment boundary Note Pointers used for RAM tests checksum or CRC mus
47. does not have one of the following types e the declared type of the object e a qualified type of the declared type of the object e the signed or unsigned type corresponding to the declared type of the object e the signed or unsigned type corresponding to a qualified version of the declared type of the object e an aggregate or union type that recursively includes one of the aforementioned types among its members e a character type Company X Ltd T AWORD TECHDOC c166 report companyX compiler report doc Revision C Last amended by M Beach on 22 06 00 14 14 Page 31 of 83 2 3 17 1 Summary Of Results Under Type Mismatch Conditions Conversion from bit to int long float double is carried out with no loss of data Conversion from int to long float double is carried out with no loss of data Conversion from float to double is not required as both float and double are 64 bits when pragma FLOAT64 is in force Conversion from any other type to bit results in one if the source object is non zero It otherwise results in zero Conversion to int from float is only valid if value of source object is lt 32768 00000 2P15 Conversion to int from double under FLOAT64 is only valid if source object is lt 32768 00000 Conversion to long from float is only valid if source object is lt 2147483648 00000 2P31 Conversion to long from double under FLOAT64 is only valid if source object is lt 2147483648 00000 2P31 Note Whe
48. dress 0x0000 Example int test_ptr Pointer will point at NULL and return int value at address 0x0000 2 3 12 CTYPE H Functions 2 3 12 1 Note1 All ctype h functions return value 0x01 for true and value 0x00 for false Such functions operate on and return int values Note 2 All ctype h functions return false 0x00 when a parameter is passed that is not part of the range of characters that it is designed to check for All such functions behave deterministically at all times The Sets Of Characters Tested For By isa num isalnun tests the input parameter for alphanumeric characters in the ranges A to Z a to z and O to 9 Company X Ltd T WORD TECHDOC c166 report companyX compiler report doc Revision C Last amended by M Beach on 22 06 00 14 14 Page 22 of 83 2 3 12 2 The Sets Of Characters Tested For By iscnitrl The iscntr1 function tests for any control character whose value lies from 0x00 NUL through 0x1F US or has a value of 0x7F 2 3 12 3 The Sets Of Characters Tested For By islower The islower function tests for any character that is a lowercase letter in the range of a to z 2 3 12 4 The Sets Of Characters Tested For By isprint The isprint function tests for any printing character in the range 0x20 to Ox7E 2 3 12 5 The Sets Of Characters Tested For By isupper The isupper function tests for any character that is a uppercase letter in the range of A to Z Compa
49. e represented by the organisation s business processes Tests for specifically critical dates may also be added e g for leap years end of year etc Organisations may wish to append additional material in support of local requirements Where the term century is used clear distinction should be made between the value denoting the century e g 20th and its representation in dates e g 19xx similarly 21st and 20xx ISBN 0 580 29746 2 DISC is a part of the British Standards Institution BSI 389 Chiswick High Road London W4 4AL Tel 0181 996 9000 Copyright DISC 1995 1996 1997 Company X Ltd T AWORD TECHDOC c166 report companyX compiler report doc Revision C Last amended by M Beach on 22 06 00 14 14 Page 65 of 83 8 Appendix E 8 1 Results Log Of PARANOIA Test For KEIL C166 v3 12j The following test was captured from Windows Hyperterm terminal via the clipboard It was emitted from the C167 serial port at 9600 baud Lest this program stop prematurely i e before displaying END OF TEST try to persuade the computer NOT to terminate execution when an error like Over Underflow or Division by Zero occurs but rather to persevere with a surrogate value after perhaps displaying some warning If persuasion avails naught don t despair but run this program anyway to see how many milestones it passes and then amend it to make further progress Answer questions with Y y N or n unless otherwise indicated
50. egister Masks And Global Register Optimisation C166 will attempt to place local data into 12 registers to improve execution speed by making best use of the single cycle MOV Rw Rw type instructions Local data that overflows the available registers is placed on the userstack C166 will only allocate locals to registers in a particular function provided that no further functions are called This limitation exists as there is no way that the compiler can know whether the called function will use registers that are already used by the caller In addition the called function will push those registers it requires onto the system stack with a consequent increase in run time The C166 compiler uses the REGISTERMASK concept to overcome this limitation These masks are generated in the list file by the compiler in the form of a code string prefixed with within which is contained coded information on each function s register usage This new information can Company X Lid T AWORD TECHDOC c166 report companyX compiler report doc Revision C Last amended by M Beach on 22 06 00 14 14 Page 15 of 83 then be automatically attached to the function prototypes to tell the compiler in advance about which registers any subsequently called functions use and hence remove the need to push registers onto the system in called functions FUNCTION locate_trigger_point BEGIN RMASK 0x6DFF SOURCE LINE 165 OOFO F68E2C03 R MOV trigger_count ZER
51. er from those of the background SYSCON If the BUSCON x is enabled while the ADDRSELx is set to zero the area currently executing could be changed Note1 The user must verify that the START167 A66 configures the ADDRSEL and BUSCON registers in the order given above Note2 The use must ensure that the BUSCONx and ADDRSELx for the currently executing region are not inadvertently changed The code to initialise the BUSCONx and ADDRSELx must be placed in the START 167 A66 just after the BFLDH and BFLDL instructions that set up BUSCONDO It is not sensible to put the BUSCONx set up in C as any RAM areas described by a BUSCONx will not enabled and hence be zeroed or otherwise initialised by the C_STARTUP code in START167 A66 C_RESET PROC TASK C_STARTUP ININO RESET 0 C_STARTUP SIF WATCHDOG 0 DISWDT Disable watchdog timer SENDIF BCONOL SET _MTTCO lt lt 5 OR _RWDCO lt lt 4 OR NOT _MCTCO AND OFH BCONOL SET BCONOL AND NOT _RDYENO lt lt 2 BCONOL SET BCONOL OR _RWDCO lt lt 4 OR _MTTCO lt lt 5 BCONOH SET _ALECTLO lt lt OR _BUSACTO lt lt 2 OR _RDYENO lt lt 4 BFLDL BUSCONO 3FH BCONOL BFLDH BUSCONO 17H BCONOH Add ADDRSEL and BUSCON setups here MOV ADDRSEL1 421H MOV ADDRSEL2 421H MOV BUSCON1 421H MOV BUSCON2 421H SYS_H SET _STKSZ lt lt 5 OR _ROMS1 lt lt 4 OR _SGTEN lt lt 3 SYS_H SET SYS_H OR _ROMEN lt lt 2 OR _BYTDIS lt lt 1 OR _CL
52. ested Extra precise subexpressions are revealed but NOT YET tested Decimal Binary conversion is NOT YET tested for accuracy To continue press RETURNLest this program stop prematurely i e before display Company X Ltd T AWORD TECHDOC c166 report companyX compiler report doc Revision C Last amended by M Beach on 22 06 00 14 14 ing END OF TEST try to persuade the computer NOT to terminate execution when an error like Over Underflow or Division by Zero occurs but rather to persevere with a surrogate value after perhaps displaying some warning If persuasion avails naught don t despair but run this program anyway to see how many milestones it passes and then amend it to make further progress Answer questions with Y y N or n unless otherwise indicated To continue press RETURN Diagnosis resumes after milestone Number 0 Page 1 Users are invited to help debug and augment this program so it will cope with unanticipated and newly uncovered arithmetic pathologies Please send suggestions and interesting results to Richard Karpinski Computer Center U 76 University of California San Francisco CA 94143 0704 USA In doing so please include the following information Precision single Version 10 February 1989 Computer Compiler Keil C166 v3 12j Optimization level OT 6 Other relevant compiler options HLARGE MOD167 To continue press RETURN Diagnosis resumes after milestone Number 1 Page 2
53. f type int Example 34 incompletevar x WARNING 34 IN LINE 34 OF MAIN C incompletevar missing declaration specifiers Company X Ltd T AWORD TECHDOC c166 report companyX compiler report doc Revision C Last amended by M Beach on 22 06 00 14 14 Page 33 of 83 2 3 19 Volatile non volatile mismatches The outcome when an attempt is made to refer to an object with volatile qualified type by means of an value with non volatile qualified type Example testvar testvar testvar testvar testvar testvar testvar Data is transferred to value at each attempted access to the volatile type exvar exvar exvar exvar exvar exvar exvar Ne Ne Ne Ne Ne ee Example Of Sequential Volatile Accesses CSP 0x0122 E014 MOV CSP 0x0124 F6F40090 MOV 14 testvar exvar CSP 0x0128 F2F40090 MOV CSP 0x012C F6F40290 MOV 16 testvar exvar CSP 0x0130 F2F40090 MOV CSP 0x0134 F6F40290 MOV 18 testvar exvar CSP 0x0138 F2F40090 MOV f CSP 0x013C F6F40290 MOV 20 testvar exvar CSP 0x0140 F2F40090 MOV CSP 0x0144 F6F40290 MOV 22 testvar exvar CSP 0x0148 F2F40090 MOV CSP 0x014C F6F40290 MOV 24 testvar exvar CSP 0x0150 F2F40090 MOV CSP 0x0154 F6F40290 MOV 26 testvar exvar CSP 0x0158 F2F40090 MOV CSP 0x015C F6F40290 MOV 28 while 1 CSP 0x0160 ODFF JMPR C_ENDINIT 0000 ADD Company X Ltd R4
54. guard digits for Mult Div and Subt Whether arithmetic is chopped correctly rounded or something else for Mult Div Add Subt and Sqrt Whether a Sticky Bit used correctly for rounding UnderflowThreshold an underflow threshold E0 and PseudoZero tell whether underflow is abrupt gradual or fuzzy V an overflow threshold roughly vO tells roughly whether Infinity is represented Comparisions are checked for consistency with subtraction and for contamination with pseudo zeros Sqrt is tested Y X is not tested Extra precise subexpressions are revealed but NOT YET tested Decimal Binary conversion is NOT YET tested for accuracy To continue press RETURN Diagnosis resumes after milestone Number 2 Page 3 Company X Ltd T AWORD TECHDOC c166 report companyX compiler report doc Revision C Last amended by M Beach on 22 06 00 14 14 Page 70 of 83 The program attempts to discriminate among FLAWs like lack of a sticky bit Serious DEFECTs like lack of a guard digit and FAILUREs like 2 2 5 Failures may confound subsequent diagnoses The diagnostic capabilities of this program go beyond an earlier program called MACHAR which can be found at the end of the book Software Manual for the Elementary Functions 1980 by W J Cody and W Waite Although both programs try to discover the Radix Precision and range over underflow thresholds of the arithmetic this program tries to cope with a wider variety
55. hen incrementing a far pointer when the offset exceeds Ox3FFF the DPPO is not incremented and the offset simply wraps around to zero again Far pointers occupy 32 bits two words 6 2 2 huge Pointers huge pointers and objects can be up to 64k in size as the overflow into the next page is not catered for However as C166 does not allow an overflow from a 16 to 24 bit offset huge pointers will just wrap around once they have been incremented more than 64kb from their start point 6 2 3 xhuge Pointers xhuge pointers remove the 64kb limitation and allow objects of any size to be addressed without restriction They are however relatively slow unless you are using the C167 5 6 2 4 sdata Pointers The final pointer type sdata is C166 family specific This is a pointer which is always points into the system area indicated by DPP3 between 0xC000 and OxFFFF sdata pointers are 16 bits in size and are best used for pointing at internal RAM objects or IO mapped into the 0xC000 region 6 3 Summary Of Pointer Declarations Declare a near pointer Company X Lid T WORD TECHDOC c166 report companyX compiler report doc Revision C Last amended by M Beach on 22 06 00 14 14 Page 60 of 83 int near near_ptr Declare a far pointer int far far pir 3 Declare a sdata pointer int sdata s_ptr 6 4 Special Note On pragma MOD167 For C167 CPUs The pragma MOD167 is used for huge basic types and pointer accesses by using the
56. hitex mmmn DEVELOPMENT TOOLS Company X SOFTWARE QUALITY MANAGEMENT C COMPILER REPORT KEIL C166 v3 12 APPROVAL NAME SIGNATURE DATE Author Software Engineering Project Manager Validation Manager Engineering Director Please destroy all earlier versions Copyright Hitex UK Ltd 1999 University Of Warwick Science Park Coventry CV4 7EZ Tel 02476 692066 Web www hitex co uk Company X Lid T AWORD TECHDOC c166 report companyX compiler report doc Revision F Last amended by M Beach on 22 06 00 14 14 Page 1 of 83 Contents Whether the value of a single character character constant in a constant expression that controls conditional inclusion matches the value of the same character constant in the execution character set Whether such a character constant may have Company X Lid T AWORD TECHDOC c166 report companyX compiler report doc Revision F Last amended by M Beach on 22 06 00 14 14 Page 2 of 83 Company X Lid T AWORD TECHDOC c166 report companyX compiler report doc Revision F Last amended by M Beach on 22 06 00 14 14 Page 3 of 83 Revision History New Revision Description Of Changes A Preliminary version Preliminary version with register and userstack usage described Typographic corrections made and section on library functions added Retesting with v3 12 Compiler amp OMF66 limits added PARANOIA test repeated with single precision Additi
57. ial port 0 gets is covered in section 2 2 2 8 puts const char s writes a string followed by a NEWLINE character n to the output stream using the putchar function putchar char c transmits the input character c to the C167 serial port zero It is common for this function to be modified to allow characters to be directed to other output devices such as an LCD panel or alternative serial port The source code for the put char incorporated in the Keil runtime library set can be found in C166 LIB PUTCHAR C getchar reads a single character from the input stream using the _getkey function The character read is passed back to the putchar function to be echoed getkey waits for a character to be received by the C167 serial port zero It is common for this function to be modified to allow characters to be received from other input devices such as a keypad or alternative serial port The source code for _getkey incorporated in the Keil runtime library set can be found in C166 LIB GETKEY C 2 3 Implementation Specific Issues The ISO ANSI C standard identifies a large number of items that are specific to the implementation of the compiler This section seeks to characterise those areas Where indicated small example program has been generated to illustrate issues raised Company X Ltd T AWORD TECHDOC c166 report companyX compiler report doc Revision C Last amended by M Beach on 22 06 00 14 14 Page
58. id far sl void far s2 unsigned int n extern void far fmemset void far s char val unsigned int n extern int xmemcemp void xhuge sl void xhuge s2 unsigned long n extern void xhuge xmemcpy void xhuge sl void xhuge s2 unsigned long n extern void xhuge xmemchr void xhuge s char val unsigned long n extern void xhuge xmemccpy void xhuge sl void xhuge s2 char val extern void xhuge xmemmove void xhuge sl void xhuge s2 unsigned long n extern void xhuge xmemset void xhuge s char val unsigned long n Company X Ltd T WORD TECHDOC c166 report companyX compiler report doc Revision C Last amended by M Beach on 22 06 00 14 14 Page 9 of 83 2 2 2 7 ISO C Memory Copying Functions void memcpy void sl const void s2 size_t n void memmove void sl const void s2 size_t n char strcpy char sl const char s2 char strncpy char sl const char s2 Size t tyy int memcmp const void sl const void s2 size_t n void memchr const void s int c size_t n void memset void s int c size_t n 2 2 2 8 Input Output Functions Example Of gets char gets char s The ISO C gets function reads characters from the input stream pointed to by stdin into the array pointed to by s until END OF FILE is encountered or a NEWLINE character is read Any NEWLINE character is discarded and a null character is written immediately after the last character read into the ar
59. ignificant bit is shifted into carry flag Company X Ltd T AWORD TECHDOC c166 report companyX compiler report doc Revision C Last amended by M Beach on 22 06 00 14 14 Page 35 of 83 2 3 21 3 Left Shift Unsigned unsignedtest 0x0001 for i 0 i lt 16 i unsignedtest lt lt 1 The object is shifted left with zeros shifted into the least significant bit The most significant bit is shifted into the carry flag of the C167 program status word PSW 2 3 21 4 Right Shift Unsigned unsignedtest 0x8000 for i 0 i lt 16 i unsignedtest gt gt 1 The object is shifted right with zeros shifted into the most significant bit The least significant bit is shifted into the carry flag of the C167 program status word PSW 2 3 22 Integer Division Behaviour The implementation of integer division shall be determined documented and taken into account Attention shall be paid to the sign of the remainder Positive dividend amp divisor gt Positive result positive remainder Positive dividend Negative divisor gt Negative result positive remainder Negative dividend positive divisor gt Negative result negative remainder Negative dividend negative divisor gt Positive result negative remainder Positive dividend zero divisor gt indeterminate result Negative dividend zero divisor gt indeterminate result 2 3 23 Floating Point Truncation The manner in which the chosen compi
60. int and short quantities are represented as follows Value 0x1234 Address 0 Address 1 0x34 0x12 Representation Of 32 Bit Quantities long quantities are represented thus Value 0x12345678 Address 0 Address 1 Address 2 Address 3 0x78 0x56 0x34 0x12 Representation Of Floating Point Quantities The IEEE754 format is used and data is stored in a little endian format Use Of Registers And Userstack The C166 compiler uses two areas to store local data automatics and function parameters Under the default optimisation level of 6 C166 will attempt to use 12 registers in the current register bank for local data and up to 5 registers for parameter passing Bit parameters are passed exclusively in register R15 Note The RO register is used to create the userstack and its value MUST NEVER BE ALTERED BY THE USER S PROGRAM Company X Ltd T WORD TECHDOC c166 report companyX compiler report doc Revision C Last amended by M Beach on 22 06 00 14 14 Page 14 of 83 2 3 2 4 Example Of Userstack Layout After Function Call Given the following function with three parameters and four local variables at OPTIMIZE level 0 the userstack will be loaded as shown in the illustration pragma OPTIMIZE 0 void func char a long b int c char_xl char_x2 int y 3 float zj 5 r UserStack Layout RO 16 RO 14 RO 10 Q D s RO 9 RO 8 RO 4 RO 2 RO 1 RO 0 R
61. ions involving tiny interest rates Testing powers Z Q at four nearly extreme values no discrepancies found To continue press RETURN Diagnosis resumes after milestone Number 190 Page 10 DEFECT Badly unbalanced range UfThold V 0 00000000000000000e 00 is too far from 1 SERIOUS DEFECT X X differs from 1 when X 0 00000000000000000e 00 instead X X 1 2 1 2 What message and or values does Division by Zero produce This can interupt your program You can skip this part if you wish Do you wish to compute 1 0 Trying to compute 1 0 produces 0 0 00000 00e 00 Do you wish to compute 0 0 Trying to compute 0 0 produces 0 0 0000000e 00 To continue press RETURN Diagnosis resumes after milestone Number 220 Page 11 The number of FAILURES encountered The number of SERIOUS DEFECTs discovered The number of DEFECTs discovered The number of FLAWs discovered Ft Gd Fe The arithmetic diagnosed has unacceptable Serious Defects Potentially fatal FAILURE may have spoiled this program s subsequent diagnoses END OF TEST Company X Ltd T AWORD TECHDOC c166 report companyX compiler report doc Revision C Last amended by M Beach on 22 06 00 14 14 Page 69 of 83 8 2 Results Log Of PARANOIA Test For Microsoft C v11 00 Configured for 386 code generation using Floating point Emulator The following test was captured from a DOS window in Windows 95 via the clipboard Lest this program s
62. issues ERROR 25 description syntax error near lt EOF gt Example C166 COMPILER V3 12 MAINI 13 02 99 18 43 04 PAGE 1 DOS C166 COMPILER V3 12 COMPILATION OF MODULE MAIN1 OBJECT MODULE PLACED IN MAIN1 OBJ COMPILER INVOKED BY C C166 BIN C166 EXE MAIN1 C WL 3 stmt level source 1 pragma OT 0 WL 3 2 3 End of file is backslash newline 4 5 6 int test0 7 int testi 8 9 void main void 10 AL TI T test0 1 12 al testl 2 13 1 14 1 FN ERROR 25 IN LINE 14 OF MAIN1 C syntax error near lt EOF gt ERROR 7 IN LINE 14 OF MAIN1 C compilation aborted C166 COMPILATION COMPLETE 0 WARNING S 2 ERROR S 2 3 14 3 Source file ends in a partial pre processing token or comment Result The compiler issues ERROR 300 description unterminated comment Company X Ltd T AWORD TECHDOC c166 report companyX compiler report doc Revision C Last amended by M Beach on 22 06 00 14 14 Page 29 of 83 Example C166 COMPILER V3 12 MAIN3 13 02 99 18 44 17 PAGE 1 DOS C166 COMPILER V3 12 COMPILATION OF MODULE MAIN3 OBJECT MODULE PLACED IN MAIN3 OBJ COMPILER INVOKED BY C C166 BIN C166 EXE MAIN3 C WL 3 stmt level source ds pragma OT 0 WL 3 2 tf 3 End of file is in comment 4 5 6 int test0O ih int testl 8 9 void main void 10 d T T test0 1 12 al testl 2 13 1 14 1 15 16 This is a test comment ERROR 300 IN LINE 16 OF MAIN3 C unterminated comment
63. ler implements truncation and rounding when a floating point number is converted to a smaller floating point number shall be determined documented and taken into account The Keil C166 supports single precision float and double precision double floating point types to the IEEE754 standard Without the pragma FLOAT64 the double type is implemented as float With pragma FLOAT64 float and double are implemented as double It is therefore not possible to convert a double precision double to float There is therefore no potential for loss of data through truncation Company X Ltd T AWORD TECHDOC c166 report companyX compiler report doc Revision C Last amended by M Beach on 22 06 00 14 14 Page 36 of 83 2 3 24 Float To Integer Conversions 2 3 24 1 The implementation of truncation from floating point quantities to integer types in the chosen compiler shall be determined documented and taken into account Converting from floating point types to integer types is not recommended It should only be done once and only then during some form of user interface function Floating point numbers are only an approximation to true values and the behaviour of conversions of float values near to integer values is not defined It is thus difficult to quantify the exact outcome of conversions and care must be exercised Floating point types must not be used in any control loop or function being reserved only for user interface functions Some ge
64. lf char const huge const fixed_string String in EPROM Company X Ltd T AWORD TECHDOC c166 report companyX compiler report doc Revision C Last amended by M Beach on 22 06 00 14 14 Page 46 of 83 Example Put a huge constant pointer to a ram variable into EPROM char huge ram_variable 1 Variable in far RAM char huge const fixed_string amp ram_variable Set up a pointer in EPROM to ram variable Example Create a pointer suitable for performing a destructive read write RAM test over a 256k RAM that is not itself within the RAM to be tested The pointer to be located in the on chip RAM unsigned short xhuge volatile idata mem_ptr 2 4 5 8 Typical Linker Control File This is intended as an example only and users must satisfy themselves as to suitability of this file to their application Fix classes for HLARGE model CLASSES ICODE 000200H FCODE 000200H OOO07FFFH 018000H O2FFFFH CONST OO4000H O7FFFH IDATA OOF 600H OOFDFFH IDATAO OOF600H OOFDFFH FCONST 018000H HCONST 018000H SDATA OOCOOOH OODFFFH SDATAO OOCOOOH OODFFFH DATA 040000H 043FFFH DATA 040000H 043FFFH FDATA 040000H FDATAO 040000H HDATA 040000H HDATAO 040000 XDATAO 040000H XDATA 040000H F r r SECTIONS C_CLRMEMSEC 0x400 Put initialisation tables into ROM C_INITSEC End of linking process
65. ly very large IDATA should be able to contain it A small modification is required to START167 A66 to move the USERSTACK into IDATA as shown below START167 A66 Modified To Put User Stack In On Chip RAM The modifications required are Upper part of START166 A66 at line 479 C_USERSTACK SECTION DATA PUBLIC NDATA Should be modified to C_USERSTACK SECTION DATA PUBLIC IDATA J and further down the file USTSZ User Stack Size Definition Defines the user stack space available for automatics This stack space is accessed by RO The user stack space must be adjusted according the actual requirements of the application USTSZ EQU 100H set User Stack Size to 40H Bytes and further down the file after the EINIT instruction at line 700 Make RO be loaded with DPP3 for an on chip USER STACK rather than DPP2 as at present EPRE EDEDEDES EINIT IF NOT TINY MOV RO DPP3 C_USERSTKTOP This was DPP2 C_USERSTKTOP SENDIF SIF TINY MOV RO C_USERSTKTOP SENDIF Company X Lid T AWORD TECHDOC c166 report companyX compiler report doc Revision C Last amended by M Beach on 22 06 00 14 14 Page 54 of 83 4 1 1 1 4 1 1 2 If your have forgotten to reduce the size of the user stack to something small enough to fit in the on chip RAM the default 1000H bytes will cause L166 to issue a warning about the IDATA class being out of group range Note The USERSTACKDPP3 and NOFIXDPP C
66. m 1 Company X Ltd T WORD TECHDOC c166 report companyX compiler report doc Revision C Last amended by M Beach on 22 06 00 14 14 Page 73 of 83 What message and or values does Division by Zero produce This can interupt your program You can skip this part if you wish Do you wish to compute 1 0 y Trying to compute 1 0 produces 1 INFO00e 000 Do you wish to compute 0 0 y Trying to compute 0 0 produces 1 INDO00e 000 To continue press RETURN What message and or values does Division by Zero produce This can interupt your program You can skip this part if you wish Do you wish to compute 1 0 y Trying to compute 1 0 produces 1 INFO00e 000 Do you wish to compute 0 0 y Trying to compute 0 0 produces 1 INDOO0e 000 To continue press RETURN Diagnosis resumes after milestone Number 220 Page 11 The number of FAILURES encountered 6 The number of SERIOUS DEFECTs discovered 5 1 2 0 The number of DEFECTs discovered The number of FLAWs discovered The arithmetic diagnosed has unacceptable Serious Defects Potentially fatal FAILURE may have spoiled this program s subsequent diagnoses END OF TEST Company X Ltd T AWORD TECHDOC c166 report companyX compiler report doc Revision C Last amended by M Beach on 22 06 00 14 14 Page 74 of 83 9 Appendix F 9 1 9 1 1 Plu m Hall test summary files Invocation Batch File T ei r H H H B O K H
67. meter or return value is of type size_t C166 gives the type as unsigned int extern int strpos tehar s cbar ji extern char strcat char s1 char s2 extern char strncat char sl char s2 unsigned int n extern int strcmp char sl char s2 extern int strncemp char sl char s2 unsigned int n extern char strcpy char sl char s2 extern char strncpy char sl char s2 unsigned int n extern size_t strlen char extern char strchr char s char cli extern char strrchr char s char c extern int strrpos char s char c extern int strspn char s char set extern int strcspn char s char set extern char strpbrk char s char set extern char strrpbrk char s char set Note The passing of source and destination pointers the above library functions with an attribute other than near will result in a WARNING which must not be ignored Additional Keil C166 String Functions C166 provides additional versions of the ISO C string functions that are intended for those cases where the source and destination pointers are of type far extern char far fstrcat char far sl extern char far fstrncat char far sl extern int fstrcmp char far sl char f extern int fstrncemp char far sl char f extern char fstrcpy char far sl char extern char fstrncpy char far sl char extern size_t fstrlen char far extern char far fstrchr const char far e
68. n converting from float types to integer or long care must be taken that the value is not greater than the maximum value that the target object can contain Further information on conversions between integer and float types is given in section 2 3 23 to 2 3 24 3 Company X Ltd T AWORD TECHDOC c166 report companyX compiler report doc Revision C Last amended by M Beach on 22 06 00 14 14 Page 32 of 83 2 3 18 Incomplete Data Declarations The outcome when an identifier for an object is declared with no linkage and the type of the object is incomplete after its declarator or after its init declarator if it has an initializer Example int complete idata incomplete_idata near incomplete_near arr j int in NDATA int in IDATA int in NDATA object not created int x idata y near z arr_init 1 2 3 4 int in NDATA int in NDATA int array in NDATA with initialised data Remarks Objects declared with an incomplete type take the default type of int Individual objects in lists of object declarations do not assume the type or location qualifier of the preceding object Example int a idata b near c int in NDATA int in IDATA int in NDATA The idata qualifier for object b is not applied to c or any other object declaration that might follow in the list Note At WARNINGLEVEL 2 or above an incomplete declaration will result in a WARNING 34 The object created will be o
69. neral guidelines can be given however If a floating point object has a positive or negative value greater than the highest value of an integer type but less than the highest value of a long type the value transferred is meaningless Caution Before converting any float to an integer type the value must be checked for potential overflow which would give a meaningless result If a floating point object has a positive or negative value greater than the highest value of a long type the value transferred is OxFFFFFFF This value will also be transferred to an integer type Caution Before converting any float to a long type the value must be checked for potential overflow and hence a meaningless result Single Precision Float single precision represents 6 to 7 significant digits However it has been determined that in typical engineering calculations only 5 digits can be relied on var32 65535 99 If the length of the initialisation value exceeds 7 significant digits the entire fractional portion is lost Rounding up of fractional parts only occurs when the value to be converted is within one unit of resolution i e the lowest representable value possible given the precision and size of the value to be converted The absolute value of this lowest representable value is dependent entirely on the magnitude of the object and thus cannot be defined here Example float integer 65535 99 65535 OxFFFF Company X Ltd T AWORD T
70. nform exprtest mix7 out 17 17 0 0 conform exprtest mix8 out 11 11 0 0 conform exprtest mix9 out T2 T2 0 0 conform exprtest mix10 out 15 15 0 0 conform exprtest mix11l out 16 16 0 0 conform exprtest mix12 out 14 14 0 0 conform exprtest mix13 out 288 288 0 0 conform exprtest ne out 288 288 0 0 conform exprtest not out 162 162 0 0 conform exprtest or out 288 288 0 0 conform exprtest orelse out Company X Ltd T AWORD TECHDOC c166 report companyX compiler report doc Revision C Last amended by M Beach on 22 06 00 14 14 Page 79 of 83 324 324 0 0 conform exprtest oreq out 122 122 0 0 conform exprtest plueql out 156 156 0 0 conform exprtest plueq2 out 154 154 0 0 conform exprtest plueq3 out 144 144 0 0 conform exprtest plueq4 out 288 288 0 0 conform exprtest plus out 576 576 0 0 conform exprtest postdec out 576 576 0 0 conform exprtest preinc out 576 576 0 0 conform exprtest quest out 12 12 0 0 conform exprtest reall out 11 TI 0 0 conform exprtest real2 out 13 13 0 0 conform exprtest real3 out TE 11 0 0 conform exprtest real4 out 10 10 0 0 conform exprtest real5 out 412 12 0 0 conform exprtest real6 out 15 15 0 0 conform exprtest real7 out 11 11 0 0 conform exprtest real8 out 8 8 0 0 conform exprtest real9 out 15 15 0 0 conform exprtest reall0 out 10 10 0 0 conform exprtest realll out 16 16 0 0 conform exprtest reall2 out 15 15 0 0 conform exprtest reall13 out 162 162 0 0 conform exprtest remain out 324 324 0 0 conform expr
71. nform exprtest timesop out 288 288 0 0 conform exprtest uminus out 162 162 0 0 conform exprtest xor out 324 324 0 0 conform exprtest xoreq out 16291 23372 4 20 TOTAL SSCPUSS DF __MOD166__ SFLOAT FLOAT64 SSMODELSS LA SSOTSS DF __OPT__ BACKSLASH C166INC C KEIL C166 INC C KEIL C166 INC C KEIL C166 INC C KEIL C166 INC C KEIL C166 INC 7 C KEIL C166 INC C KEIL C166 INC C KEIL C166 INC C KEIL C166 INC C KEIL C166 INC C166LIB C KEIL C166 LIB C251INC C C251 INC C251LIB C C251 LIB C51INC C C51 INC C51LIB C C51 LIB CFLAGS DB SB NOPR LA FLOAT64 DF __MOD166 DFE __OPT COMPUTERNAME RK ComSpec C WINNT system32 cmd exe DIRNAME XXXXXXX XXXXXXXX XXXXXXXXX DPPUSE N EXE 66 EXEHO exe EXENAME XXXXXXXX HOCFLAGS c AL Za HOLFLAGS F 2800 HOMEDRIVE C HOMEPATH LFLAGS CLASSES FCODE 0x10000 0x2FFFF CASE LINES SYMBOLS PRINT NUL LOGONSERVER RK NUMBER_OF_PROCESSORS 1 OBJ 0bj OBJHO obj OS Windows_NT Os2LibPath C WINNT system32 os2 d1l Path plumhall test 166 C WINNT system32 C WINNT C KEIL C166 BIN C UTIL PATHEXT COM EXE BAT CMD PHCC C166 PHDST plumhall test 166 PHDSTDRV D PHHOCC cl PHIN C166INC PHIP D KEIL C166 INC Company X Ltd T AWORD TECHDOC c166 report companyX compiler report doc Revision C Last amended by M Beach on 22 06 00 14 14 Page 77 of 83 PHLI L166 PHSRC plumhall suite PHTMP This is a very long string that sets aside envi
72. ny X Ltd T WORD TECHDOC c166 report companyX compiler report doc Revision C Last amended by M Beach on 22 06 00 14 14 Page 23 of 83 2 3 13 The values returned by the mathematics functions on domain errors Only those mathematical functions that can generate domain or range errors are listed here Their behaviour under such conditions is documented 2 3 13 1 C166 Mathematical Function List Note the double type float unless pragma FLOAT64 is in force int long double double double double double double double double double double double double double double double double double double double abs labs fabs sqrt exp log log10 sin cos tan asin acos atan sinh cosh tanh atan2 ceil floor modf pow unsigned int unsigned int int val long val double val double val double val double val double val double val double val double val double val double val double val double val double val double val double y double val double val double val double x _chkfloat_ _chkdouble_ 2 3 13 2 Floating Point Limits Single Precision Not a number NaN for float types OxFFFFFFF de i double x double n double y float x double x Positive Infinity INF 0x7F80000 Negative Infinity INF OxFF80000 Double Precision Not a number NaN for double types OxFFFFFFFFFFFFFF
73. o f pathologies and to say how well the arithmetic is implemented The program is based upon a conventional radix representation for floating point numbers but also allows logarithmic encoding as used by certain early WANG machines BASIC version of this program C 1983 by Prof W M Kahan see source comments for more history To continue press RETURN Diagnosis resumes after milestone Number 3 Page 4 Program is now RUNNING tests on small integers aTe Oy L 252 Lp 27 By 4p or Oy Sy 32 2A are ORK Searching for Radix and Precision Radix 1073741824 000000 Closest relative separation found is Ul 9 3132257e 010 Recalculating radix and precision gets better closest relative separation U1 9 9341078e 009 MYSTERY recalculated Radix 1 0066330e 008 DEFECT Radix is too big roundoff problems FLAW Radix is not as good as 2 or 10 FAILURE 1 U1 1 2 lt 1 2 is FALSE prog fails The number of significant digits of the Radix is 1 000000 SERIOUS DEFECT Precision worse than 5 decimal figures Some subexpressions appear to be calculated extra precisely with about 0 993608 extra B digits i e roughly 7 95172 extra significant decimals That feature is not tested further by this program To continue press RETURN FAILURE Incomplete carry propagation in Addition Add Subtract appears to be chopped Addition Subtraction neither rounds nor chops Sticky bit used incorrectly or not at all FLAW lack s of
74. ollowing Bit Allocations In Register Mask BPS I A one in any of the various fields in a register mask indicates the following MDX gt user s assembler function uses the multiply divide unit R12 R1 gt user s assembler function uses a general purpose register DPP3 gt user s assembler function modifies DPP3 DPPO gt user s assembler function modifies DPPO Company X Ltd T AWORD TECHDOC c166 report companyX compiler report doc Revision C Last amended by M Beach on 22 06 00 14 14 Page 16 of 83 2 3 3 How a diagnostic is identified 2 3 3 1 2 3 3 2 2 3 3 3 Error and warning messages with example reports A diagnostic is considered to be a WARNING or ERROR message A WARINING will prevent the uVISION Make system from proceeding to the link stage A valid OBJECT file will still be produced however An ERROR will inhibit the production of an OBJECT file and delete any previous one that may already exist Thus linking cannot take place Under the uVISION environment WARNINGs and ERRORs are indicated by a green bar over the C line in which a problem was encountered by the compiler Diagnostic messages are stored in two compiler generated output files the ERR and LST files The C166 Compiler ERR file Contains a list of WARNING and ERROR messages that occurred during the compilation in a machine readable form only It is intended to be used only by the uViSION editor The C166 Compiler LST File
75. on of v3 12j problem list Addition of Plum Hall error descriptions Q n im o JO w Change to divide by zero integer division results to indeterminate Company X Ltd T AWORD TECHDOC c166 report companyX compiler report doc Revision F Last amended by M Beach on 22 06 00 14 14 Page 4 of 83 1 Introduction 1 1 Overview The XXX product range has been designed using devices from the Siemens C16x family of microcontrollers Upon the advice of Hitex UK Ltd the Keil C166 C compiler toolkit was selected to support software development This report has been prepared to address the issues raised in the Company X C Software Coding Requirements Revision X X Its purpose is to e Provide evidence of the adequacy of the compiler toolkit for this purpose e Provide details of how the compiler implements those features that are either not specified by the ISO ANSI C standard or are explicitly identified as specific to the implementation of the compiler This will enable to software development and validation personnel to be aware of items that may not be portable to other platforms Where reference is made to the ISO ANSI C Standard this shall be taken to mean ISO IEC 9899 1990 Programming Languages C 1990 with Technical Corrigendum 1995 as published by ISO 1 2 Disclaimer Whilst every effort has been made to ensure the accuracy and completeness of this report neither Hi
76. ould make a full audit of older projects to see what the implications of this are iti Neither Keil nor Hitex can be held responsible for the consequences arising from the continued use of tool versions of earlier release numbers than those listed above Company X Ltd T WORD TECHDOC c166 report companyX compiler report doc Revision C Last amended by M Beach on 22 06 00 14 14 Page 62 of 83 DISC PD2000 1 A Definition of Year 2000 Conformity Requirements A DEFINITION OF YEAR 2000 CONFORMITY REQUIREMENTS Introduction This document addresses what is commonly known as Year 2000 Conformity also sometimes known as century or millennium compliance It provides a definition of this expression and requirements that must be satisfied in equipment and products which use dates and times It has been prepared by British Standards Institution committee BDD 1 3 in response to demand from UK industry commerce and the public sector It is the result of work from the following bodies whose contributions are gratefully acknowledged BT CapGemini CCTA Coopers amp Lybrand Halberstam Elias ICL National Health Service National Westminster Bank BSI DISC would also like to thank the following organisations for their support and encouragement in the development of this definition taskforce 2000 Barclays Bank British Airways Cambridgeshire County Council Computer Software Services Association Department of Health Ernst amp Young
77. prtest ne out 288 288 0 0 conform exprtest not out 162 162 0 0 conform exprtest or out 288 288 0 0 conform exprtest orelse out Company X Ltd T AWORD TECHDOC c166 report companyX compiler report doc Revision C Last amended by M Beach on 22 06 00 14 14 Page 76 of 83 324 324 0 0 conform exprtest oreq out 122 122 0 0 conform exprtest plueql out 156 156 0 0 conform exprtest plueq2 out 154 154 0 0 conform exprtest plueq3 out 144 144 0 0 conform exprtest plueq4 out 288 288 0 0 conform exprtest plus out 576 576 0 0 conform exprtest postdec out 576 576 0 0 conform exprtest preinc out 576 576 0 0 conform exprtest quest out 12 12 0 0 conform exprtest reall out 11 TI 0 0 conform exprtest real2 out 13 13 0 0 conform exprtest real3 out TE 11 0 0 conform exprtest real4 out 10 10 0 0 conform exprtest real5 out 412 12 0 0 conform exprtest real6 out 15 15 0 0 conform exprtest real7 out 11 11 0 0 conform exprtest real8 out 8 8 0 0 conform exprtest real9 out 15 15 0 0 conform exprtest reall0 out 10 10 0 0 conform exprtest realll out 16 16 0 0 conform exprtest reall2 out 15 15 0 0 conform exprtest reall13 out 162 162 0 0 conform exprtest remain out 324 324 0 0 conform exprtest remeq out 162 162 0 0 conform exprtest rsh out 324 324 0 0 conform exprtest rsheq out 120 120 0 0 conform exprtest timeql out 154 154 0 0 conform exprtest timeq2 out 156 156 0 0 conform exprtest timeq3 out 146 146 0 0 conform exprtest timeq4 out 288 288 0 0 co
78. ray Example Of The Keil C166 gets char gets char unsigned int n The Keil C166 gets function calls the get char function to read a line of characters into a string s The line consists of all characters up to and including the first NEWLINE character n The NEWLINE character is replaced by a null character 0 in the string The parameter n specifies the maximum number of characters that may be read If n characters are read before ENWLINE is encountered the gets function terminates the string with the null character 0 and returns Company X Ltd T AWORD TECHDOC c166 report companyX compiler report doc Revision C Last amended by M Beach on 22 06 00 14 14 Page 10 of 83 2 2 3 Standard ISO C functions behaving in a non standard manner 2 2 3 1 2 2 3 2 Memory Allocation Functions Memory management functions assume that the non ISO C init_mempool function has been previously called with parameter consisting of the address of a RAM area which can be used for the heap plus its overall size malloc calloc realloc can then access blocks in this area Note These functions should not be used in a safety critical system Formatted And Other IO Functions printf directs its formatted output to putchar contained in PUTCHAR C which by default addresses C167 serial port 0 scanf obtains it input characters from _getkey contained in GETHAR C which be default addresses C167 ser
79. rd PC compiler was significantly worse Note The user should examine the results presented here and in Appendx E In the light of these and on the actual usage of floating point quantities in the application the user should decide what action to take Company X Ltd T AWORD TECHDOC c166 report companyX compiler report doc Revision C Last amended by M Beach on 22 06 00 14 14 Page 50 of 83 2 6 Limits Of The Keil C166 Compiler System 2 6 1 Compiler Implementation Limits 1 Maximum of 20 levels of indirection to any standard data type 2 Names can be up to 255 characters long Only the first 32 are significant Names are case sensitive 3 The maximum number of cases in a switch block is not fixed and is limited only by the available memory size and the maximum size of individual functions The maximum number of nested function calls in and invocation parameter list is 10 The maximum number of nested include files is 9 The maximum depth of directives for conditional compilation is 20 Instructions blocks may be nested up to 32 levels Macros may be nested up to 8 levels SOS 5009 NY SO OT ge A maximum of 32 parameters may be passed in a macro or function call 10 The maximum length of a line or macro definition is 8000 characters 2 6 2 Siemens Keil OMF66 Object Module Format Limits 1 Number of sections 32767 Each compilation unit generates one code section it is assumed that it contains at le
80. rd carrying information on offset from page number 3 DPP3 0xCO00 OxFFFF 5 1 1 Casting From Basic Types To Pointers To make a near pointer an unsigned int can be used Example int near near_address 0x4000 int y 0x8000 y unsigned char near near_address You must make very sure though with conversions to near pointers that the address is actually in the near data area To make a far pointers the source data must be of type long as only this type has sufficient bytes to accommodate the pagenumber offset information Example unsigned long address 0x38010 unsigned char test test unsigned char far address Company X Lid T WORD TECHDOC c166 report companyX compiler report doc Revision C Last amended by M Beach on 22 06 00 14 14 Page 58 of 83 It should be noted that this on the fly casting from long to pointer is very inefficient and it is always better to use a proper pointer type as the following shows int var unsigned long faddr 0x30010 int far fptr int far 0x30010 Using a proper pointer type var fptr MOV R5 WORD fptr 2 MOV DPP0 R5 MOV R4 WORD fptr MOV R4 R4 MOV WORD var R4 Cast from long to pointer x var int far faddr MOV R4 WORD faddr MOV R5 WORD faddr 2 ADD R4 R4 ADDC R5 R5 ADD R4 R4 ADDC R5 R5 MOV DPP0 R5 SHR R4 2 MOV R8 R4 MOV DPPO 12 Note The conversion from long to a pointer is not like industry standard
81. red if the DPPUSE control is used at the linking stage It is also used if the USER STACK is moved to the on chip IDATA RAM as required in many single chip applications STATIC Normally C166 will try to put as many local variables into registers R1 R15 as possible as the MOV Rw Rw register to register instructions execute in 100ns 20MHz All the normal ADD SUB CMP type instructions are available in the register to register variety so that any such operation will 100ns at 2OMHz Variables that overflow the available local registers are placed on the User Stack as in a PC type compiler and are addressed via MOV R1 RO displacement type instructions As a RISC CPU there are very few stack relative instructions so that operating on user stack variables usually takes several instructions A significant performance advantage for interrupt functions or those with a large number of local variables can therefore be had by forcing the compiler to put locals that cannot fit into registers R1 R15 into near static RAM segments to create a compiled stack as in the C51 compiler The common ADD SUB and CMP instructions all can operate directly on RAM so that there is little performance loss when compared to register variables Note Any functions within modules compiled with this control will no longer be reentrant thus if enabled here in the C166 Options menu no reentrancy will be possible across the entire program which may
82. roduct shall use particular date values for special meanings e g 99 to signify no end value or end of file or 00 to mean not applicable or beginning of file Rule 3 3 1 This rule is sometimes known as explicit implicit century 3 2 It covers two general approaches a explicit representation of the year in dates e g by using four digits or by including a century indicator In this case a reference may be inserted e g 4 digit years as allowed by ISO standard 8601 1988 and it may be necessary to allow for exceptions where domain specific standards e g standards relating to Electronic Data Interchange Automatic Teller Machines or Bankers Automated Clearing Services should have precedence b the use of inferencing rules e g two digit years with a value greater than 50 imply 19xx those with a value equal to or less than 50 imply 20xx Rules for century inferencing as a whole must Company X Lid T AWORD TECHDOC c166 report companyX compiler report doc Revision C Last amended by M Beach on 22 06 00 14 14 Page 64 of 83 apply to all contexts in which the date is used although different inferencing rules may apply to different datesets General Notes For Rules 1 and 2 in particular organisations may wish to specify allowable ranges for values of current date and dates to be manipulated The ranges may relate to one or more of the feasible life span of equipment or products or the span of dates required to b
83. rograms try to discover the Radix Precision and range over underflow thresholds of the arithmetic this program tries to cope with a wider variety of pathologies and to say how well the arithmetic is implemented Company X Ltd T WORD TECHDOC c166 report companyX compiler report doc Revision C Last amended by M Beach on 22 06 00 14 14 Page 67 of 83 The program is based upon a conventional radix representation for floating point numbers but also allows logarithmic encoding as used by certain early WANG machines BASIC version of this program C 1983 by Prof W M Kahan see source comments for more history To continue press RETURN Diagnosis resumes after milestone Number 3 Page 4 Program is now RUNNING tests on small integers iy 07 2 25 Vy 2 37 49 5 O77 25 32 240 ave OTK Searching for Radix and Precision Radix 2 000000 Closest relative separation found is Ul 5 9604640e 08 Recalculating radix and precision confirms closest relative separation U1 Radix confirmed FAILURE 1 U1 1 2 lt 1 2 is FALSE prog fails FAILURE Comparison is fuzzy X 1 but X 1 2 1 2 0 The number of significant digits of the Radix is 24 000000 To continue press RETURN Diagnosis resumes after milestone Number 30 Page 5 Subtraction appears to be normalized as it should be Checking for guard digit in and x and appear to have guard digits as they should To continue press RETUR Diagnosis resumes
84. ronment space PHTMP2 This is a very long string that sets aside environment space PROCESSOR_ARCHITECTURE x86 PROCESSOR_IDENTIFIER x86 Family 6 Model 5 Stepping 2 GenuineIntel PROCESSOR_LEVEL 6 PROCESSOR_REVISION 0502 PROMPT SP G STARTUP strt66lm SYSPATH C WINNT system32 C WINNT C KEIL C166 BIN C UTIL SystemDrive C SystemRoot C WINNT TEMP C TEMP TMP C TEMP USERDOMAIN RK USERNAME Administrator USERPROFILE C WINNT Profiles Administrator windir C WINNT WRKDRV D Company X Ltd T AWORD TECHDOC c166 report companyX compiler report doc Revision C Last amended by M Beach on 22 06 00 14 14 Page 78 of 83 9 1 3 Summary File 2 EXPECTED ACTUAL ERRORS SKIPPED FILE NAME 5 00 1994 01 03 16 16 0 0 conform environ out 24 24 0 4 conform interp91 out 1224 1223 1 0 conform lang out 123 7208 0 16 conform libfrst out 821 821 0 0 conform precl out 1180 1180 0 0 conform prec2 out 1 1 0 0 conform capacity capacity out 11 Tt 0 0 conform errauto i91 out 151 151 0 0 conform errauto j91 out 32 30 2 0 conform errauto m61 out 0 0 0 0 conform errauto m62 out 75 75 0 0 conform errauto m63 out 5 4 1 0 conform errauto m64 out 77 76 1 0 conform errauto m65 out 24 24 0 0 conform errauto m66 out 17 Ty 0 0 conform errauto m67 out 39 39 0 0 conform errauto m68 out 288 288 0 0 conform exprtest andif out 576 576 0 0 conform exprtest assign out 162 162 0 0 conform exprtest band out 3
85. rrect Warning expr with no effect and code removal much like 30 ANWEISUNGtyp memcpy amp gv0 BUF_AB PROG_LES_BUFinst IX amp gv0 SATZ PROG_LES_BUFinst LOOPCOUNTER sizeof gv0 BUF_AB PROG_LES_BUFinst IX CORR GENEXP C GenExp 3 11 98 Company X Ltd T WORD TECHDOC c166 report companyX compiler report doc Revision C Last amended by M Beach on 22 06 00 14 14 38 Missing Error Meassage on constant modification const int c_val 3 void main void c_val c_val 2 gives correct error meassage c_val 2 is not considered as a constant modification CORR ENODE C DoAdd 25 2 99 39 Unnecessary Warning static unsigned char suc char test void register signed char rsc static signed char ssc if unsigned char rsc gt suc signed unsigned type mismatch ESC if unsigned char ssc gt suc SScrry return ssctrsc CORR CFOLD C RelOps 25 2 99 WARNING 173 IN LINE 9 OF TEST130 C 40 Word access to a odd address when pragma order is used pragma code debug or void test char b char c char const abc 3 void main void Tate char dummy2 hello hello starts at an odd address and is copied word by word char dummy3 hell test dummy2 dummy3 use dummy variables Company X Ltd T AWORD TECHDOC c166 report companyX compiler report doc Revision C Last amended by M Bea
86. s value range M65 Problem M65 is a duplication of the problem M61 int main enum A INT_MAX B Missing Error message return 0 Company X Ltd T WORD TECHDOC c166 report companyX compiler report doc Revision C Last amended by M Beach on 22 06 00 14 14 Page 81 of 83 10 Appendix G 10 1 Known Problems In C166 v3 12j These problems have been fixed in C166 v3 12k 33 INCDIR increased to max 50 path specs CORR BOM H 5 10 98 34 Assembly Errors when SRC is used pragma ot 3 size pragma hold near 2048 extern unsigned char m 1034 struct x int x 5 di int i typedef struct unsigned char S char oj A const static A far a 1 2 3 4 hi const static A near b 1 2 3 4 di void main void struct x m gt x b 0 o 0 struct x m gt x a 0 o 0 struct x m gt x i 0 CORR REG166 C ASMGEN C 9 10 98 35 Gp or hangup CORR CFOLD C PostFold 9 10 98 36 tagName space clash struct test_struct outer test_struct int structivar0 int structivarl int Z void main void struct test_struct inner test_struct int i struct test_struct next error resolves to outer test_struct di struct test_struct ptr array 15 if ptr gt next gt i 25 z 1 gives next undefined member CORR GRM C rsutp 22 10 98 37 Inco
87. sting X X 1 X 1 vs exp 2 7 38905572891235350e 000 as X gt 1 DEFECT Calculated 2 67379508591339170e 002 for 1 2 55000000000000000e 002 1 00784313678741460e 000 differs from correct value by 2 59990447998046870e 002 This much error may spoil financial calculations involving tiny interest rates Testing powers Z Q at four nearly extreme values DEFECT computing 1 00000000000000000e 001 3 20000000000000000e 001 yielded 1 00000003318135350e 032 which compared unequal to correct 1 02679679275673470e 032 Company X Ltd T WORD TECHDOC c166 report companyX compiler report doc Revision C Last amended by M Beach on 22 06 00 14 14 Page 72 of 83 they differ by 2 67967595753811630e 030 Similar discrepancies have occurred 4 times To continue press RETURN DEFECT Calculated 2 67379508591339170e 002 for 1 2 55000000000000000e 002 1 00784313678741460e 000 differs from correct value by 2 59990447998046870e 002 This much error may spoil financial calculations involving tiny interest rates Testing powers Z Q at four nearly extreme values DEFECT computing 1 00000000000000000e 001 3 20000000000000000e 001 yielded 1 00000003318135350e 032 which compared unequal to correct 1 02679679275673470e 032 they differ by 2 67967595753811630e 030 Similar discrepancies have occurred 4 times To continue press RETURN Diagnosis resumes after milestone Number 190 Page 10 DEFECT Badly unbalan
88. t be of type xhuge e Pointers occupy 4 bytes e Executable functions may be located at any memory address HLARGE Model Data CLASS Placement HDATA HDATAO HCONST classes ideally should be located on 64k segment boundaries NDATA NDATAO classes must be located on a page boundary i e an address divisible by 0x4000 IDATA and IDATAO classes must be located at OxF600 SDATA and SDATAO classes must be located in the region OxC000 OxFDFF FCODE class must be locate in the ROM area NCONST class must be locate on a 0x4000 byte boundary in the ROM area The sections C_CLRMEMSEC and C_INITSEC must be located near the base of he FCODE class in ROM Management Of Near Data NDATA Under HLARGE The pragma HOLD near 6 control is applied by default This has the effect of placing all objects of up and including 6 bytes in length into the near data classes Thus all char int short and long types will be placed into a 16k near block NDATA amp NDATAO classes that has the benefit of faster addressing Small arrays of up to 6 bytes will also be placed into the near area The total of all such near objects must not exceed 16kb The NDATA and NDATAO classes must be placed at 0x4000 byte boundaries The end address of these classes must not be more than 0x4000 bytes above the start address and must be stated Company X Ltd T AWORD TECHDOC c166 report companyX compiler report doc Revision C Last amended by M Beach on 22 06 00 14 14 Page 45 of 83
89. t rates Similar discrepancies have occurred 14 times To continue press RETURN Diagnosis resumes after milestone Number 100 Page 8 Company X Ltd T AWORD TECHDOC c166 report companyX compiler report doc Revision C Last amended by M Beach on 22 06 00 14 14 Page 68 of 83 Seeking Underflow thresholds UfThold and EO FAILURE multiplication gets too many last digits wrong To continue press RETURN Diagnosis resumes after milestone Number 110 Page 9 Smallest strictly positive number found is EO 1 17549e 38 Since comparison denies Z 0 evaluating Z Z Z should be safe What the machine gets for Z Z Z is 2 00000000000000000e 00 This is O K provided Over Underflow has NOT just been signaled The Underflow threshold is 1 17549400000000000e 38 below which calculation may suffer larger Relative error than merely roundoff Since underflow occurs below the threshold U o fThold 2 00000000000000000e 00 1 26000000000000000e 02 nly underflow should afflict the expression 2 00000000000000000e 00 2 52000000000000000e 02 actually calculating yields 0 00000000000000000e 00 This computed value is O K Testing X X 1 X 1 vs exp 2 7 38905700000000000e 00 as X gt 1 DEFECT Calculated 1 26641700000000000e 14 for 1 5 96046400000000000e 08 3 35544300000000000e 07 differs from correct value by 7 38905700000000000e 00 This much error may spoil financial calculat
90. terrupt regions and atomic sequences e Read modify write pipeline effects e Data coherency in 32 bit quantities updated by interrupts e Illegal sharing of interrupt priority and group level e Illegal sharing of registerbanks by interrupt functions e Misuse of near and huge pointers e Trapping of runtime exceptions via the CLASS A amp B trap system e Unexpected results with pointer arrays to constant strings Note To satisfactorily audit these aspects specialist knowledge of the C167 architecture and C166 compiler linker relationship is required Company X Ltd T WORD TECHDOC c166 report companyX compiler report doc Revision C Last amended by M Beach on 22 06 00 14 14 Page 41 of 83 2 4 2 Compiler Controls The objectives of portability ease of programming and most efficient use of the C167 architecture can be attained through the correct choice of memory model For all external and internal ROM C167 systems the HLARGE memory model provides the best fit to these objectives Under this model the user need only use the idat a keyword to place data into the on chip RAM area of the C167 CPU and sdata to place data in the XRAM area s The user should be fully acquainted with the consequences of using the following controls if they are applied or check whether the programming techniques employed require their use NODPPSAVE On the C167 DPP3 is always set to 3 to indicate the base of the SYSTEM area at 0xC000 and DPP
91. test remeq out 162 162 0 0 conform exprtest rsh out 324 324 0 0 conform exprtest rsheq out 120 120 0 0 conform exprtest timeql out 154 154 0 0 conform exprtest timeq2 out 156 156 0 0 conform exprtest timeq3 out 146 146 0 0 conform exprtest timeq4 out 288 288 0 0 conform exprtest timesop out 288 288 0 0 conform exprtest uminus out 162 162 0 0 conform exprtest xor out 324 324 0 0 conform exprtest xoreq out 16291 23371 5 20 TOTAL SSCPUSS MOD167 SFLOAT FLOAT64 SSMODELS MD SSOTSS DF __OPT__ BACKSLASH C166INC C KEIL C166 INC C KEIL C166 INC C KEIL C166 INC C KEIL C166 INC C KEIL C166 INC 7 C KEIL C166 INC C KEIL C166 INC C KEIL C166 INC C KEIL C166 INC C KEIL C166 INC C166LIB C KEIL C166 LIB C251INC C C251 INC C251LIB C C251 LIB C51INC C C51 INC C51LIB C C51 LIB CFLAGS DB SB NOPR MD FLOAT64 MOD167 DF __OPT__ COMPUTERNAME RK ComSpec C WINNT system32 cmd exe DIRNAME XXXXXXX XXXXXXXX XXXXXXXXX DPPUSE N EXE 66 EXEHO exe EXENAME XXXXXXXX HOCFLAGS c AL Za HOLFLAGS F 2800 HOMEDRIVE C HOMEPATH LFLAGS CLASSES FCODE 0x10000 0x2FFFF CASE LINES SYMBOLS PRINT NUL LOGONSERVER RK NUMBER_OF_PROCESSORS 1 OBJ 0bj OBJHO obj OS Windows_NT Os2LibPath C WINNT system32 os2 d1l Path plumhall test 166 C WINNT system32 C WINNT C KEIL C166 BIN C UTIL PATHEXT COM EXE BAT CMD PHCC C166 PHDST plumhall test 166 PHDSTDRV D PHHOCC cl PHIN C166INC PHIP D KEIL C166 IN
92. tex UK Ltd or the author can be held liable for any consequences resulting from errors omissions or interpretation of information presented therein Company X Lid T AWORD TECHDOC c166 report companyX compiler report doc Revision F Last amended by M Beach on 22 06 00 14 14 Page 5 of 83 2 2 1 Items Addressed By This Report Quality Control amp Validation Evidence shall be provided that the C compiler has been developed to the requirements of a quality management system that meets the requirements of ISO 9001 1994 and that its performance is satisfactory having passed tests executed using a recognised validation suite e g Plum Hall Details of known problems with the version of the compiler used shall be provided along with details of how such information can be kept up to date as new releases become available 2 1 1 Quality Systems Keil do not have an accredited ISO9000 quality management system as is common in the embedded compiler field Hitex is certified to ISO EN 9001 08 94 certificate number report Q1 01 97 02 27639 001 and Keil Elektronik GmbH as a major supplier since 1989 has been fully assessed as part of this process 2 1 2 Year 2000 Issues The tested 3 12 compiler is Year 2000 compliant as defined by BSI PD2000 appendix D 2 1 3 Compiler Testing Prior To Release The Keil C166 compiler is tested using the Plum Hall C validation suite prior to release The Keil Plum Hall licence number is CVS2
93. top prematurely i e before displaying END OF TEST try to persuade the computer NOT to terminate execution when an error like Over Underflow or Division by Zero occurs but rather to persevere with a surrogate value after perhaps displaying some warning If persuasion avails naught don t despair but run this program anyway to see how many milestones it passes and then amend it to make further progress Answer questions with Y y N or n unless otherwise indicated To continue press RETURN Diagnosis resumes after milestone Number 0 Page 1 Users are invited to help debug and augment this program so it will cope with unanticipated and newly uncovered arithmetic pathologies Please send suggestions and interesting results to Richard Karpinski Computer Center U 76 University of California San Francisco CA 94143 0704 USA In doing so please include the following information Precision single Version 10 February 1989 Computer Compiler MSC v8 00 Optimization level MAX Other relevant compiler options G3 for 386 code with no FPU To continue press RETURN Diagnosis resumes after milestone Number 1 Page 2 Running this program should reveal these characteristics Radix 1 2 4 8 10 16 100 256 Precision number of significant digits carried U2 Radix Radix Precision One Ulp OneUlpnit in the Last Place of 1 000xxx Ul 1 Radix Precision One Ulp of numbers a little less than 1 0 Adequacy of
94. vel 1 echo 1 gt phstatus em Any tests for success or failure of the compile command need em to test PHSTATUS em You could execute the diagnost command here to display your em diagnostic messages rem diagnost 1 clg exit 0 Company X Ltd T AWORD TECHDOC c166 report companyX compiler report doc R evision C Last amended by M Beach on 22 06 00 14 14 Page 75 of 83 9 1 2 Summary File 1 EXPECTED ACTUAL ERRORS SKIPPED FILE NAME 5 00 1994 01 03 16 16 0 0 conform environ out 24 24 0 4 conform interp91 out 1224 1223 1 0 conform lang out 123 7208 0 16 conform libfrst out 821 821 0 0 conform precl out 1180 1180 0 0 conform prec2 out 1 1 0 0 conform capacity capacity out 11 Ti 0 0 conform errauto i91 out 151 154 0 0 conform errauto j91 out 32 31 1 0 conform errauto m61 out 0 0 0 0 conform errauto m62 out 75 75 0 0 conform errauto m63 out 5 4 1 0 conform errauto m64 out 77 76 1 0 conform errauto m65 out 24 24 0 0 conform errauto m66 out 17 Ty 0 0 conform errauto m67 out 39 39 0 0 conform errauto m68 out 288 288 0 0 conform exprtest andif out 576 576 0 0 conform exprtest assign out 162 162 0 0 conform exprtest band out 324 324 0 0 conform exprtest bandeq out 288 288 0 0 conform exprtest cast out 162 162 0 0 conform exprtest compl out 288 288 0 0 conform exprtest div out 120 120 0 0 conform exprtest diveql out 156 156 0 0 conform exprtest diveq2 out 156 156 0 0 conform exprtest diveq
95. xtern char far fstrrchr const char far extern int fstrpos const char far s extern int fstrrpos const char far s Company X Lid T WORD TECHDOC c166 report companyX compiler report doc Revision C Last amended by M Beach on 22 06 00 14 14 char far s2 char far s2 ar s2 i ar s2 far s2 far s2 s s cnar cnar char c char c c c These are unsigned int n unsigned int n unsigned int n Page 8 of 83 2 2 2 4 2 2 2 5 2 2 2 6 extern int fstrspn char far s char far set extern int fstrcspn char far s char far set extern char fstrpbrk char far s char far set extern char fstrrpbrk char far s char far set Note The f prefix does not imply that these functions are file orientated available in C166 ISO C String Handling Function Prototypes const char s2 const char s2 const char s2 const char s2 size_t n const char s2 const char s2 int c int c const char s2 const char s2 const char s2 har strcpy char sl har strncpy char sl har strcat char sl har strncat char sl nt strcemp const char sl1 size_t n c C i int strncmp const char s1 fa s s size_t n har strchr const char s har strrchr const char s ize_t strspn const char s1 ize_t strcspn const char s1 har strpbrk const char s1 Unimplemented String Functions In C166

Keil C166 Compiler Report

Contents

Download Pdf Manuals

Related Search

Related Contents