diff options
author | Denys Vlasenko <vda.linux@googlemail.com> | 2018-12-07 12:57:32 +0100 |
---|---|---|
committer | Denys Vlasenko <vda.linux@googlemail.com> | 2018-12-07 12:57:32 +0100 |
commit | 18c6b54f820923549135724fee6cf66c26929b07 (patch) | |
tree | 60030a7234fd7e5d6eed7e1093a263bfe1c55e09 | |
parent | b9c321d6d94fc8bbae5fe657e141cbd9f2397037 (diff) | |
download | busybox-18c6b54f820923549135724fee6cf66c26929b07.tar.gz |
bc: use more compact parsing data structures
function old new delta
dc_lex_token 697 701 +4
bc_parse_next_rel 20 - -20
bc_parse_next_read 20 - -20
bc_parse_next_print 20 - -20
bc_parse_next_param 20 - -20
bc_parse_next_for 20 - -20
bc_parse_next_expr 20 - -20
bc_parse_next_elem 20 - -20
common_parse_expr 62 40 -22
bc_parse_expr 49 24 -25
dc_lex_regs 52 13 -39
bc_parse_name 581 539 -42
bc_parse_expr_empty_ok 2157 2108 -49
dc_parse_insts 332 83 -249
dc_lex_tokens 364 91 -273
bc_parse_stmt 2261 1868 -393
------------------------------------------------------------------------------
(add/remove: 0/7 grow/shrink: 1/8 up/down: 4/-1232) Total: -1228 bytes
text data bss dec hex filename
987037 485 7296 994818 f2e02 busybox_old
985814 485 7296 993595 f293b busybox_unstripped
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
-rw-r--r-- | miscutils/bc.c | 62 |
1 files changed, 35 insertions, 27 deletions
diff --git a/miscutils/bc.c b/miscutils/bc.c index 0330c43e5..685427e58 100644 --- a/miscutils/bc.c +++ b/miscutils/bc.c @@ -632,11 +632,6 @@ typedef struct BcLex { BC_PARSE_FLAG_LOOP | BC_PARSE_FLAG_LOOP_INNER | BC_PARSE_FLAG_IF | \ BC_PARSE_FLAG_ELSE | BC_PARSE_FLAG_IF_END))) -typedef struct BcParseNext { - uint32_t len; - BcLexType tokens[4]; -} BcParseNext; - struct BcParse; struct BcProgram; @@ -834,34 +829,40 @@ static const uint8_t bc_parse_ops[] = { #define bc_parse_op_PREC(i) (bc_parse_ops[i] & 0x0f) #define bc_parse_op_LEFT(i) (bc_parse_ops[i] & 0x10) +// Byte array of up to 4 BC_LEX's, packed into 32-bit word +typedef uint32_t BcParseNext; + // These identify what tokens can come after expressions in certain cases. -#define BC_PARSE_NEXT_TOKENS(...) .tokens = { __VA_ARGS__ } -#define BC_PARSE_NEXT(a, ...) \ - { \ - .len = (a), BC_PARSE_NEXT_TOKENS(__VA_ARGS__) \ - } -static const BcParseNext bc_parse_next_expr = - BC_PARSE_NEXT(4, BC_LEX_NLINE, BC_LEX_SCOLON, BC_LEX_RBRACE, BC_LEX_EOF); -static const BcParseNext bc_parse_next_param = - BC_PARSE_NEXT(2, BC_LEX_RPAREN, BC_LEX_COMMA); -static const BcParseNext bc_parse_next_print = - BC_PARSE_NEXT(4, BC_LEX_COMMA, BC_LEX_NLINE, BC_LEX_SCOLON, BC_LEX_EOF); -static const BcParseNext bc_parse_next_rel = BC_PARSE_NEXT(1, BC_LEX_RPAREN); -static const BcParseNext bc_parse_next_elem = BC_PARSE_NEXT(1, BC_LEX_RBRACKET); -static const BcParseNext bc_parse_next_for = BC_PARSE_NEXT(1, BC_LEX_SCOLON); -static const BcParseNext bc_parse_next_read = - BC_PARSE_NEXT(2, BC_LEX_NLINE, BC_LEX_EOF); +enum { +#define BC_PARSE_NEXT4(a,b,c,d) ( (a) | ((b)<<8) | ((c)<<16) | ((((d)|0x80)<<24)) ) +#define BC_PARSE_NEXT2(a,b) BC_PARSE_NEXT4(a,b,0xff,0xff) +#define BC_PARSE_NEXT1(a) BC_PARSE_NEXT4(a,0xff,0xff,0xff) + bc_parse_next_expr = BC_PARSE_NEXT4(BC_LEX_NLINE, BC_LEX_SCOLON, BC_LEX_RBRACE, BC_LEX_EOF), + bc_parse_next_param = BC_PARSE_NEXT2(BC_LEX_RPAREN, BC_LEX_COMMA), + bc_parse_next_print = BC_PARSE_NEXT4(BC_LEX_COMMA, BC_LEX_NLINE, BC_LEX_SCOLON, BC_LEX_EOF), + bc_parse_next_rel = BC_PARSE_NEXT1(BC_LEX_RPAREN), + bc_parse_next_elem = BC_PARSE_NEXT1(BC_LEX_RBRACKET), + bc_parse_next_for = BC_PARSE_NEXT1(BC_LEX_SCOLON), + bc_parse_next_read = BC_PARSE_NEXT2(BC_LEX_NLINE, BC_LEX_EOF), +#undef BC_PARSE_NEXT4 +#undef BC_PARSE_NEXT2 +#undef BC_PARSE_NEXT1 +}; #endif // ENABLE_BC #if ENABLE_DC -static const BcLexType dc_lex_regs[] = { +static const //BcLexType - should be this type, but narrower type saves size: +uint8_t +dc_lex_regs[] = { BC_LEX_OP_REL_EQ, BC_LEX_OP_REL_LE, BC_LEX_OP_REL_GE, BC_LEX_OP_REL_NE, BC_LEX_OP_REL_LT, BC_LEX_OP_REL_GT, BC_LEX_SCOLON, BC_LEX_COLON, BC_LEX_ELSE, BC_LEX_LOAD, BC_LEX_LOAD_POP, BC_LEX_OP_ASSIGN, BC_LEX_STORE_PUSH, }; -static const BcLexType dc_lex_tokens[] = { +static const //BcLexType - should be this type +uint8_t +dc_lex_tokens[] = { BC_LEX_OP_MODULUS, BC_LEX_INVALID, BC_LEX_INVALID, BC_LEX_LPAREN, BC_LEX_INVALID, BC_LEX_OP_MULTIPLY, BC_LEX_OP_PLUS, BC_LEX_INVALID, BC_LEX_OP_MINUS, BC_LEX_INVALID, BC_LEX_OP_DIVIDE, @@ -889,7 +890,9 @@ static const BcLexType dc_lex_tokens[] = { BC_LEX_INVALID }; -static const BcInst dc_parse_insts[] = { +static const //BcInst - should be this type. Using signed narrow type since BC_INST_INVALID is -1 +int8_t +dc_parse_insts[] = { BC_INST_INVALID, BC_INST_INVALID, BC_INST_INVALID, BC_INST_REL_GE, BC_INST_INVALID, BC_INST_POWER, BC_INST_MULTIPLY, BC_INST_DIVIDE, BC_INST_MODULUS, BC_INST_PLUS, BC_INST_MINUS, @@ -4768,7 +4771,7 @@ static BcStatus bc_parse_expr_empty_ok(BcParse *p, uint8_t flags, BcParseNext ne BcInst prev = BC_INST_PRINT; BcLexType top, t = p->l.t.t; size_t nexprs = 0, ops_bgn = p->ops.len; - uint32_t i, nparens, nrelops; + unsigned nparens, nrelops; bool paren_first, paren_expr, rprn, done, get_token, assign, bin_last; paren_first = p->l.t.t == BC_LEX_LPAREN; @@ -4993,9 +4996,14 @@ static BcStatus bc_parse_expr_empty_ok(BcParse *p, uint8_t flags, BcParseNext ne if (prev == BC_INST_BOOL_NOT || nexprs != 1) return bc_error_bad_expression(); - for (i = 0; i < next.len; ++i) - if (t == next.tokens[i]) + // next is BcParseNext, byte array of up to 4 BC_LEX's, packed into 32-bit word + for (;;) { + if (t == (next & 0x7f)) goto ok; + if (next & 0x80) // last element? + break; + next >>= 8; + } return bc_error_bad_expression(); ok: |