An Improved Quake C Compiler
The -show-defaults flag instructs the compiler to print out the defaults used related to the standard, optimization, and code generation. When this flag is specified, the compiler will just print the defaults and quit. No compilation is performed.
$ gmqcc -show-defaults
gmqcc | default standard |
fteqcc | fteqcc standard |
qcc | vanila QuakeC standard |
Options | What it does |
-l<path> | Adds <path> to the directories searched by the preprocessor for include file resolution. |
-o <file> | Generates the named executable (progs.src) file (when not specified default is progs.src). |
-O<level> | Specfies the optimization level: highest being 3, lowest being 0 (no optimization). |
-g | Enables generation of debug info for engine backtraces (turns on -flno) |
-E | Instructs the compiler to only preprocess the input, writing the preprocessed output to stdout |
-D | "define" a macro. Optionally you may supply a value to the macro with "=". Implicitally turns on -fftepp |
-Wall | Enables all compiled warnings for the selcted standard |
-Werror | Instruct the compiler to treat all warnings as errors |
-std=<standard> | Selects the standard dialect |
Macro | What it represents |
__STD_GMQCC__ | Specifies the current selected standard is gmqcc. |
__STD_FTEQCC__ | Specifies the current selected standard is fteqcc. |
__STD_QCC__ | Specifies the current selected standard is qcc. |
GMQCC | Defined always regardless of the selected standard |
__STD_VERSION_MINOR__ | Specifies the current selected stanadards minor version number. |
__STD_VERSION_MAJOR__ | Specifies the current selected stanadards major version number. |
__FILE__ | Expands to a const string literal of the current file (requires -fftepp-predef) |
__LINE__ | Expands to the current line number (requires -fftepp-predef) |
__RANDOM__ | Expands to a random number between [0, 256) (requires -fftepp-predef) |
__COUNTER__ | Expands to a unique number, each expansion causes it to increment (requires -fftepp-predef) |
__RANDOM_LAST__ | Expands to the last random number (requires -fftepp-predef) |
__COUNTER_LAST__ | Expands to the last counter number (requires -fftepp-predef) |
GMQCC strives hard for compatability with standard dialects, but not all features of those standards might be implemented. The unsupported features are presented below:
Feature | Standard |
Inline Assembly | FTEQCC |
Macro expansion in strings | FTEQCC |
Option | What it does |
-foverlap-locals | Reduces codesize by overlapping locals where possible |
-fdarkplaces-string-table-bug | Works around a bug in older Darkplaces engine builds where the stringtable size is computed wrong |
-fadjust-vector-fields | corrects assignment of vector field pointers (STORE_V instead of STORE_FLD) |
-fftepp | Enables FTEQ preprocessor |
-fftepp-predef | Enables additional predefined macros for the FTEQ preprocessor |
-frelaxted-switch | Relaxes switch statement semantics |
-fshort-logic | Enables short circut evaluation/logic |
-fperl-logic | Enables perl evalutaion/logic |
-ftranslatable-strings | Enables translatable strings via .po file |
-finitialized-nonconstants | Prevents initializations from becoming constant unless 'const' is specified as a qualifer |
-fassign-function-types | Allows function types to be assignable even if their signature is invariant |
-flno | Enables generation of progs.lno for engine VM backtraces (enabled with -g as well) |
-fcorrect-ternary | Enabling this corrects the ternary percedecce bugs present in fteqcc |
-fsingle-vector-defs | Prevents the creation of progdefs for vectors |
-fcorrect-logic | Use cast vector to real booleans when used in logic expressions |
-ftrue-empty-strings | Always treat empty strings as true |
-ffalse-empty-strings | Opposite of above, empty strings always false |
-futf8 | Recognize utf-8 characters in character constants, and encode codepoint escape sequences in strings as utf-8 |
-fbail-on-werror | When a warning is printed and it is set to be treated as error via -Werror, compilation will be stopped. |
-floop-labels | Allow loops and switches to be labeled and break and continue to take optional label to target it. |
-funtyped-nil | Enable the 'nil' null constant, which has no type. It can be used as the right hand of any assignment. |
-fpermissive | Be "permissive", e.g with -funtyped-nil, this will allow local variables with the name 'nil' |
Option | What it does |
-Wunused-uninitialized | Enables warnings about unused or uninitialized variables |
-Wunknwon-control-sequence | Enables warnings about unknown control sequences |
-Wextension | Enables warnings about the use of (an) extension(s) |
-Wfield-redeclared | Enables warnings about redeclared fields |
-Wmissing-return-values | Enables warnings about missing return values |
-Wtoo-few-paramaters | Enables warnings about missing paramaters for function calls |
-Wlocal-shadows | Enables warnings about locals shadowing paramaters or other locals |
-Wlocal-constants | Enables warnings about constants specified as locals |
-Wvoid-variables | Enables warnings about variables declared as type void |
-Wimplicit-function-pointer | Enables warnings about implicitly declared function pointers |
-Wvariadic-function | Enables warnings for use of varadics for non-builtin functions |
-Wframe-macros | Enables warnings about duplicated frame macros |
-Weffectless-statement | Enables warnings about effectiveless statements |
-Wend-sys-field | Enables warnings of end_sys_fields being declared a field |
-Wassign-function-types | Enables warnings for incompatible function pointer signatures used in assignment |
-Wpreprocessor | Enables warnings about redefined macros |
-Wmultifile-if | Enables warnings about multifile if statements |
-Wdouble-declaration | Enables warnings about double declarations |
-Wconst-var | Enables warnings about 'const var' and 'var const' |
-Wmultibyte-character | Enables warnings about use of multibyte characters |
-Wternary-precedence | Enables warnings about ternary expressions whos precedence may be not what expected |
-Wunknown-pragmas | Enables warnings about unknown pragmas |
-Wunreachable-code | Enables warnings about unreachable code |
-Wcpp | Enables warnings about the preprocessor |
Wunknown-attribute | Enables warnings for when an unknown attribute is encountered. |
-Wreserved-names | Enables warnings for variables or fields that are named with a reserved name |
-Wuninitialized-constant | Enables warnings about const global variables with no initializing value |
-Wuninitialized-global | Enables warnings about non-const gloal variable with no initializing value |
-Wdeprecated | Enables warnings about calls to functions marked deprecated |
Options | What it does |
-Opeephole | Enables peephole optimizations |
-Olocal-temps | Enables local-temp omission optimizations |
-Oglobal-temps | Enables global-temp omission optimizations |
-Otail-recursion | Enables tail recursion optimization |
-Otail-calls | Enables tail-call optimizations |
-Ooverlap-locals | Every function where it's safe to do so will share its local section with others. The critera being that the function must have any possibly uninitialized locals, or local arrays regardless of how they are initialized. |
-Ostrip-constant-names | Strip the names of constants |
-Ooverlap-strings | Agressivly reuse strings in the string-section. |
-Ocall-stores | Have expressions which are used as function parameters evaluate directly into the parameter-globals if possible. This avoids copying. |
-Ovoid-return | Do not create a return instruction at the end of return-type void functions |
Individual warnings may be disabled with -Wno-<warning>
$ gmqcc -Wno-frame-macros # disables frame duplication warning
Option | What it does |
-force-crc=<num> | Forces a specific checsum into the header |
-debug | Turns on compiler debug messages |
-memchk | Turns on compiler memory leak checker |
-dump | Dump IR before optimizing and finalizing |
-dumpfin | Dump IR after optimizing and finalizing |
-Whelp or -W? | Lists all warning options |
-fhelp or -f? | Lists all code generation options |
-redirout=<file> | Redirect stdout to any file. |
-redirerr=<file> | Redirect stderr to any file. |
-nocolor | Turn off colored stdout/stderr. |
-config=<file> | Supply a configuration file to set options. Note: If a file named gmqcc.ini or gmqcc.cfg is found it will be loaded implicitally. |
To compile GMQCC on BSD/NIX the following things are required:
$ git clone git://github.com/graphitemaster/gmqcc.git $ cd gmqccThe Makefile contains a few rules, depending on what you want to compile, the following rules are:
Rule | What it does |
gmqcc | Builds the gmqcc compiler |
qcvm | Builds a standable QuakeC VM |
testsuite | Builds the testsuite for GMQCC |
check, test | Build and executes the testsuite for GMQCC |
pak | Builds the pak utility |
splint | Runs splint (static analysis) on the source |
gource | Runs gource visualization on the source |
gource-record | Runs gource visualization and produces a rendered mp4 |
depend | Builds dependinces into the Makefile |
all | Builds gmqcc, qcvm, pak, and testsuite |
install | Installs gmqcc, qcvm, and documentaion |
uninstall | Uninstalls gmqcc, qcvm, and documentation |
check | Runs the testsuite to verify correctness |
To compile GMQCC on windows the following things are required:
Once obtained you may checkout the development repository with the following msysGit commands from a msysGit shell.$ git clone git://github.com/graphitemaster/gmqcc.gitIncluded is a VS project file.
To run the testsuite you can either use
$ make checkOr if you're on windows or have already compiled the testsuite from source:
$ ./testsuiteOptionally you may provide the testsuite with additional arguments:
Argument | What it does |
-redirout=<file> | Redirect stdout to any file. |
-redirerr=<file> | Redirect stderr to any file. |
-debug | Turn on testsuite debug messages. |
-memchk | Turn on testsuite memleak checker. |
-nocolor | Turn off colored stdout/stderr. |
GMQCC comes with a complete testsuite for verifying semantics and syntatics. The testsuite executes files from the test/ directory, by reading task template files.
templates are rules for a specific test, used to create a "task" that is executed with those set of rules (arguments, and what not). Tests that don't have a template with them cannot become tasks, since without the information for that test there is no way to properly "test" them. Rules for these templates are described in a template file, using a task template language.
The languge is composed entierly of "tags" which describe a string of text for a task. Think of it much like a configuration file. Except it's been designed to allow flexibility and future support for prodecual semantics.
The following "tags" are suported by the language:
Tag | Description of what the tag does |
D: | Used to set a description of the current test, this must be provided, this tag is NOT optional. |
F: | Used to set test-suite specific flags, currently the only supported flag is -no-defs which tells the testsuite to exclude defs.qh. |
T: | Used to set the procedure for the given task, there are four
options for this:
|
C: | Used to set the compilation flags for the given task, this must be provided, this tag is NOT optional. |
E: | Used to set the execution flags for the given task. This tag must be provided if T == -execute, otherwise it's erroneous as compilation only takes place. |
M: | Used to describe a string of text that should be matched from the output of executing the task. If this doesn't match the task fails. This tag must be provided at least once if T == -execute or T == -pp, otherwise it's erroneous as compilation only takes place. Multiple M tags are required for multi-line comparision |
I: | Used to specify the INPUT source file to operate on, this must be provided, this tag is NOT optional |
These tags (with exception to M) have one-time use, using them more than once will result in template compilation errors,
Lines beginning with # or // in the template file are comments and are ignored by the template parser. Whitespace is optional, with exception to the colon ':' between the tag and it's assignment value.
The template compiler will detect erronrous tags (optional tags that need not be set), as well as missing tags, and error accordingly which will result in that task failing.Included with GMQCC is a minimal implementation of the QCVM used in many game engines. It's primarly used for the testsuite, but you may also use it as a standalone runtime, or even embed it with existing appliciations.
To run the standalone application you need to have a compiled progs.dat, with an entry function named main The main function can have any amount of arguments as the standalone executor allows main to be invoked with your choice of arguments. An example of invoking the VM:
$ ./qcvm progs.dat -float 200 #execute passing in 200 as a float to mainIf main doesn't require arguments:
$ ./qcvm progs.dat #call main with no argumentsThe standalone executor supports the following arguments for passing arguments to main 
Argument | What it does |
-string | Passes in a string to main |
-float | Passes in a float to main |
-vector | Passes in a vector to main |
void main(float a, vector b)Then to pass the arguments you'd use the same order:
$ ./qcvm -float 200 -vector '1 2 3'
Argument | What it does |
-h, --help | Print help message |
-trace | Trace the execution call hierarchy. |
-profile | Profile the bytecode to find hotspots. |
-info | Get info of the running bytecode. |
-disasm | Dissasemble the bytecode into assembly. |
-diasm-func | Dissasmble function |
-printdefs | Prints all definitions for the bytecode running. |
-printfields | Prints all fields for the bytecode running. |
-printfuns | Prints all functions for the bytecode running. |
-v | Be verbose |
-v | Be even more verbose |
-version, --version | Print version information |
Builtin | Number |
1 | |
ftos | 2 |
spawn | 3 |
kill | 4 |
vtos | 5 |
error | 6 |
vlen | 7 |
etos | 8 |
stof | 9 |
strcat | 10 |
strcmp | 11 |
normalize | 12 |
sqrt | 13 |
floor | 14 |
Having trouble with GMQCC? Join our IRC channel at #kf-engine on irc.freenode.net or contact Us