Before compiling Ajla, install the packages libgmp-dev, libffi-dev, libnuma-dev,
libz3-dev. If libgmp-dev is not installed, Ajla will use slow built-in version.
If libffi-dev is not installed, the possibility to call C functions from Ajla
will be disabled. If libnuma-dev is not installed, NUMA optimizations will be
disabled, if libz3-dev is not installed, verification of programs using z3 will
be disabled.

You can compile and install Ajla with ./configure && make && make install

If you cloned the git repository, there is no ./configure script. You can
generate it with the ./autogen.sh script. You need autoconf, automake and ed
installed.

It is recommended to use gcc. clang works, but the compilation is very slow, it
may take an hour or more to compile the files ipret.c and ipretc.c.

When running ajla programs, it is recommended to set cpu governor to
'performance'. Ajla starts and stops threads rapidly and the 'ondemand' governor
underclocks the cores, resulting in slower performance.


Ajla arguments:

--chicken=features	disable the specified comma-separated features
	cg		disable the code generator and only use the interpreter
	cg-flag-cache	disable the flag cache
	cg-must-be-flat	disable the optimization for flat variables
	cg-must-be-data	disable the optimization for data variables
	cg-ra		disable the register allocator
	cg-optimize	disable the machine code optimizer
	cg-traps	disable traps on architectures that use them
--compile		compile the whole program and don't run it
--compile-run		compile the whole program and run it
--debug			enable all debugging options
--debug=features	enable specific comma-separated debugging features
	magic		put magic number before all allocations
	redzone		put magic number after all allocations
	fill		fill memory block on allocation and freeing
	leak		identify memory leaks
	memory		enable magic, redzone, fill, leak
	mutex		enable mutex debugging
	mutex-errorcheck use the errorcheck attribute on mutexes
	cond		enable condition variable debugging
	thread		enable thread debugging
	handles		enable handle debugging
	objects		enable mutex, mutex-errorcheck, cond, thread, handles
--dump-code		write the generated machine code to "dump.s"
--dump-opencl		write the generated OpenCL code to stderr
--dump-pcode		write the generated pcode code to stderr
--dump-z3		write the generated z3 assertions to stderr
--help			display help
--noinline		disable automatic inlining
--nosave		do not save and load the compiled program
--numa-nodes=x		override the number of numa nodes to "x"
--opencl-device=x	use the OpenCL device with index "x"
--optimize-fp		optimize for floating point calculations (default)
--optimize-int		optimize for integer calculations
--profile		enable all profilng options
--profile=features	enable specified comma-separated profiling features
	function	print functions that consumed most CPUs
	escape		print positions where the code escaped from interpreter
	memory		print places where most memory was allocated
--ptrcomp		enable pointer compression
--strict-calls		disable automatic parallelization
--system-malloc		use the system malloc instead of Ajla malloc
--thread-tick		use a thread instead of a signal for timer ticks
--threads=x		override the number of CPUs to "x"
--tick=x		tick interval in microseconds (default 10000)
--verify		verify the program using z3
--verify-timeout=x	a timeout to verify a function in microseconds
--version		display Ajla version
--warnings		enable warnings in the OpenCL code

vim: textwidth=80
