https://gcc.gnu.org/bugzilla/show_bug.cgi?id=70359
--- Comment #15 from Fredrik Hederstierna
---
Created attachment 38185
--> https://gcc.gnu.org/bugzilla/attachment.cgi?id=38185=edit
tok.c
I took another example for CSiBE and stripped it down. I'm
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=70359
--- Comment #14 from kugan at gcc dot gnu.org ---
(In reply to Jeffrey A. Law from comment #13)
> The change to the assignment of p_22 is made by forwprop1.
>
> It does create a situation where p_2 is live outside the loop and hides the
> CSE
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=70359
--- Comment #13 from Jeffrey A. Law ---
The change to the assignment of p_22 is made by forwprop1.
It does create a situation where p_2 is live outside the loop and hides the CSE
opportunity, which may be the cause of the more significant
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=70359
--- Comment #12 from kugan at gcc dot gnu.org ---
However, diff of cfgexand is significantly different:
;; Full RTL generated for this function:
;;
32: NOTE_INSN_DELETED
- 38: NOTE_INSN_BASIC_BLOCK 2
+ 39: NOTE_INSN_BASIC_BLOCK 2
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=70359
--- Comment #11 from kugan at gcc dot gnu.org ---
Optimized gimple diff between 5.3 and trunk is :
-;; Function inttostr (inttostr, funcdef_no=0, decl_uid=5268, cgraph_uid=0,
symbol_order=0)
+;; Function inttostr (inttostr, funcdef_no=0,
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=70359
--- Comment #10 from kugan at gcc dot gnu.org ---
I am looking into it. -mcpu=arm966e-s does not uses the
TARGET_NEW_GENERIC_COSTS. i.e, the rtx costs setup by the back-end might not be
optimal.
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=70359
--- Comment #9 from Segher Boessenkool ---
(In reply to Jakub Jelinek from comment #5)
> CCing authors of the other commits. That said, complaining about size
> regressions generally should be only if it (significantly) increases sizes
> of
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=70359
--- Comment #8 from Jakub Jelinek ---
All I wanted to say is that it is to be expected that on some code a newer GCC
version ends up needing one or two more instructions, even at -Os, and what
matters is not the size of a single function, but
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=70359
--- Comment #7 from Jeffrey A. Law ---
I would suggest that you definitely keep reporting these things and extracting
examples from csibe or other benchmarks to show the codesize increases.
While some folks will prioritize performance, it
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=70359
--- Comment #6 from Fredrik Hederstierna
---
Thanks for your analysis on this. One comment on this 'complaint', it's not
only about size - in my example the compiler uses 2 more regs push and pop, and
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=70359
Jeffrey A. Law changed:
What|Removed |Added
Priority|P3 |P2
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=70359
Jakub Jelinek changed:
What|Removed |Added
CC||jakub at gcc dot gnu.org,
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=70359
Ramana Radhakrishnan changed:
What|Removed |Added
Target|arm-none-eabi |arm-none-eabi,
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=70359
Jeffrey A. Law changed:
What|Removed |Added
CC||law at redhat dot com
--- Comment #3
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=70359
--- Comment #2 from Andrew Pinski ---
(In reply to Richard Biener from comment #1)
> My #1 bet would be FSM threading.
I doubt it as if I read the asm differences correctly, GCC 6 just no longer
does store with post increment and that causes
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=70359
Richard Biener changed:
What|Removed |Added
Keywords||missed-optimization
16 matches
Mail list logo