2144 commits

This branch

This branch

All branches

Author	SHA1	Message	Date
jjgarcia	71ed18e2db	Inline funcalls and fix problem with multiple-value-prog1 and macrolet + declarations	2008-06-19 15:05:08 +00:00
jjgarcia	90370c96d0	Fixed typo in COMPILED-FUNCTION-NAME.	2008-06-19 15:04:54 +00:00
jjgarcia	9a5f1a7751	Introduce bytecode closure objects to save space in FLET/LABELS forms	2008-06-19 15:04:27 +00:00
jjgarcia	77d53061eb	Revert to the old code for multiple-value-setq	2008-06-19 15:04:12 +00:00
jjgarcia	b2d1a996b3	More compact format for bytecodes OP_FLET/LABELS	2008-06-19 15:03:56 +00:00
jjgarcia	590129c767	No need to check when searching environments	2008-06-19 15:03:40 +00:00
jjgarcia	271d33ee3f	Slightly leaner code for searching environment	2008-06-19 15:03:28 +00:00
jjgarcia	83c35e01c2	SETQS does not check whether argument is constant, just as in SBCL.	2008-06-19 15:03:12 +00:00
jjgarcia	ce724349c5	Split operator MSETQ into VSETQ and VSETQS.	2008-06-19 15:02:58 +00:00
jjgarcia	5aa07e402d	Only use OP_VBIND[S] for values other than 0-th	2008-06-19 15:02:40 +00:00
jjgarcia	0c3af3fbc8	ecl_interpret returns the first value	2008-06-19 15:02:27 +00:00
jjgarcia	311995ae4f	Inline the stack push operation	2008-06-19 15:02:12 +00:00
jjgarcia	9dcff352d7	Save a pointer to the current environment in the interpreter.	2008-06-19 15:01:59 +00:00
jjgarcia	f5b4ff25af	COND and PROG1 now operate using only REG0 and the stack	2008-06-19 15:01:47 +00:00
jjgarcia	8fca58aec5	OP_JT/JNIL need not set NVALUES	2008-06-19 15:01:35 +00:00
jjgarcia	ecd1053c26	Allow COND forms to use REG0 to store computations, instead of forcing use of VALUES.	2008-06-19 15:01:23 +00:00
jjgarcia	37022df5d6	Since OP_JT/JNIL check reg0, we can simplify the bytecodes for WHILE	2008-06-19 15:01:05 +00:00
jjgarcia	64ca376393	Given that reg0 contains always VALUES(0), we can simplify OP_JT and OP_JNIL.	2008-06-19 15:00:50 +00:00
jjgarcia	aa6b20fad1	Simplify bytecodes compiler by rewriting IF as a COND.	2008-06-19 15:00:20 +00:00
jjgarcia	8e2237108e	Use the local IHS frame for storing the lexical environment.	2008-06-19 15:00:01 +00:00
jjgarcia	5668d23f2b	Inline OP_PROGV.	2008-06-19 14:59:49 +00:00
jjgarcia	9d3d2b34fe	The IHS record is not created in apply lambda but in ecl_interpret	2008-06-19 14:59:37 +00:00
jjgarcia	6870a9753f	The lexical environment of a function is stored in the same IHS record as its name.	2008-06-19 14:59:24 +00:00
jjgarcia	c1f08a04e6	Removed global environment field cl_env.lex_env	2008-06-19 14:59:11 +00:00
jjgarcia	8ae45ee30e	Remove useless field frs_lex from the frame stack records.	2008-06-19 14:58:57 +00:00
jjgarcia	200ec27464	Inline OP_MSETQ	2008-06-19 14:58:44 +00:00
jjgarcia	327b6b9640	Inline OP_FLET/LABELS and let them access cl_env.lex_env less often.	2008-06-19 14:58:32 +00:00
jjgarcia	ff083d3511	OP_PROTECT also pushes the environment onto the stack.	2008-06-19 14:58:20 +00:00
jjgarcia	45b24c328e	OP_TAGBODY also pushes the environment to be saved	2008-06-19 14:58:07 +00:00
jjgarcia	2c3a2f33a6	Avoid using field lex_env in frame stack records, at least for blocks	2008-06-19 14:57:45 +00:00
jjgarcia	33a53695cf	ecl_interpret() now takes an explicit lexical environment argument.	2008-06-19 14:57:12 +00:00
jjgarcia	6833a3cf0a	lambda_bind_var() now takes also a generic environment as argument	2008-06-19 14:56:59 +00:00
jjgarcia	734d3edadb	All functions that operate on the lexical environment take a generic environment as argument, not inspecting cl_env.	2008-06-19 14:56:47 +00:00
jjgarcia	980b6d68b4	The interpreter is rewritten using indirect threaded code	2008-06-19 14:56:24 +00:00
jjgarcia	9788340759	Export a condition for detected user interrupts (SIGINT)	2008-06-05 12:45:02 +00:00
jjgarcia	9b050098f5	In methods function's type declarations, the type and the name were swapped	2008-06-05 12:44:31 +00:00
jjgarcia	7201e26e85	Import version 1.118 of ASDF	2008-06-05 12:44:07 +00:00
jjgarcia	376eef6143	We need sys/time.h for struct timeval.	2008-05-31 13:27:50 +00:00
jjgarcia	92152c66ee	When using the generational garbage collector, do not use dlopen/dlclose wrapper ... s, which deactivate the garbage collector. In this case, we need not explicit ca lls to si_gc(Ct) in si_load_binary()	2008-05-31 13:14:05 +00:00
jjgarcia	f42ca59c92	Accessors for low-level socket timeout attributes (by G. Carncross).	2008-05-31 13:04:07 +00:00
jjgarcia	af6c551e80	When using the generational garbage collector, do not use dlopen/dlclose wrappers, which deactivate the garbage collector. In this case, we need not explicit calls to si_gc(Ct) in si_load_binary()	2008-05-31 12:58:01 +00:00
jjgarcia	70c30b8ef9	In compile-file, ensure that :output-file is appended the proper file extension	2008-05-30 23:55:54 +00:00
jjgarcia	a63f6a5bda	Allow depth to be infinite (coded as depth=0) in _hash_equal[p]	2008-05-30 23:52:32 +00:00
jjgarcia	abec6e581f	Stronger hashing for C files initialization names	2008-05-30 23:52:16 +00:00
jjgarcia	7677ee8ce6	When computing the hash key of a pathname, since they are not circular structures, reset the depth parameter.	2008-05-30 23:51:57 +00:00
jjgarcia	75a324fbbe	The last 12 bytes of a string were not used for the hash	2008-05-30 23:51:44 +00:00
jjgarcia	d93be391f9	Seems we need to call the garbage collector in order to stabilize loading of binary files	2008-05-30 21:00:17 +00:00
jjgarcia	263c837285	Mark the date at which the generational garbage collector becomes safe to use	2008-05-30 17:03:59 +00:00
jjgarcia	fe3af8d9a2	The Boehm-Weiser garbage collector needs to use SIGSEGV for the generational garbage collector algorithm.	2008-05-30 17:01:06 +00:00
jjgarcia	eadbb9cf11	Make the incremenal garbage collector optional until we solve why some platforms break with it.	2008-05-30 15:33:07 +00:00