------- Comment #7 from j at uriah dot heep dot sax dot de 2006-05-02 15:37
-------
Forwarding this comment on behalf of Bjoern Haase:
Preliminary analysis of the RTL generated without optimization shows that
the problem is present already directly after expand. Maybe the problem
is triggered by the fact that the avr port does not have move patterns
for DI, however, there seems to be a general problem in the mid-end.:
Excerpt of test.c.01.sibling. Personal comments marked with. #
;; Function main
(note 2 0 5 NOTE_INSN_DELETED)
(note 5 2 3 0 [bb 0] NOTE_INSN_BASIC_BLOCK)
(note 3 5 6 0 NOTE_INSN_FUNCTION_BEG)
(note 6 3 8 1 [bb 1] NOTE_INSN_BASIC_BLOCK)
(insn 8 6 9 1 (set (reg:HI 44)
(const_int 42 [0x2a])) -1 (nil)
(nil))
#push int parameter on stack (last parameter of function call)
(insn 9 8 10 1 (set (mem/i:HI (post_dec:HI (reg/f:HI 32 __SPL__)) [0 S2 A8])
(reg:HI 44)) -1 (nil)
(nil))
#allocate 8 Bytes on the stack for the uint64_t var, i.e. stack -= 8
(insn 10 9 11 1 (set (reg/f:HI 32 __SPL__)
(plus:HI (reg/f:HI 32 __SPL__)
(const_int -8 [0xfffffff8]))) -1 (nil)
(nil))
#load first data byte in reg
(insn 11 10 12 1 (set (reg:QI 45)
(const_int -16 [0xfffffff0])) -1 (nil)
(nil))
#write first data byte to stack + 8. instead of stack + 0 !!!! BUG !!!!
(insn 12 11 13 1 (set (mem/i:QI (plus:HI (reg/f:HI 32 __SPL__)
(const_int 8 [0x8])) [0 S1 A8])
(reg:QI 45)) -1 (nil)
(nil))
#load second data byte in reg
(insn 13 12 14 1 (set (reg:QI 46)
(const_int -34 [0xffffffde])) -1 (nil)
(nil))
#write second data byte to stack + 9. instead of stack + 1 : !!!! BUG !!!!
(insn 14 13 15 1 (set (mem/i:QI (plus:HI (reg/f:HI 32 __SPL__)
(const_int 9 [0x9])) [0 S1 A8])
(reg:QI 46)) -1 (nil)
(nil))
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=27386
