[Bug rtl-optimization/23523] New: code size regression on x86

[Bug rtl-optimization/23523] New: code size regression on x86

[dann at godzilla dot ics dot uci dot edu]

Compiling the code in the main.i attachment with -march=i686 -O2 the code
generated for the "first_map_occurred" is bigger for 4.1 compared to 4.0
The sdiff between 4.0 and 4.1:

xorl    %eax, %eax                     |   xorl    %ecx, %ecx
movl    %eax, waiting_for_initial_map  |   movl    %ecx, waiting_for_initial_map


Trying to minimize the testcase changes the code generated for 4.1
It is a bit worse. 


typedef struct {
  int max_col;
  int max_row;


} TScreen;


typedef struct _XtermWidgetRec {
    TScreen screen;
} XtermWidgetRec, *XtermWidget;



static int cp_pipe[2];


static int pc_pipe[2];
typedef enum {
    PTY_BAD,
    PTY_FATALERROR,
    PTY_GOOD,
    PTY_NEW,
    PTY_NOMORE,
    UTMP_ADDED,
    UTMP_TTYSLOT,
    PTY_EXEC
} status_t;

typedef struct {
    status_t status;
    int error;
    int fatal_error;
    int tty_slot;
    int rows;
    int cols;
    char buffer[1024];
} handshake_t;

extern XtermWidget term;
extern int waiting_for_initial_map;

__extension__ typedef int __ssize_t;

typedef __ssize_t ssize_t;
typedef unsigned int size_t;


extern ssize_t write (int __fd, __const void *__buf, size_t __n) ;
extern int close (int __fd);

void
first_map_occurred(void)
{
    handshake_t handshake;
    TScreen *screen = &term->screen;

    handshake.status = PTY_EXEC;
    handshake.rows = screen->max_row;
    handshake.cols = screen->max_col;
    write(pc_pipe[1], (char *) &handshake, sizeof(handshake));
    close(cp_pipe[0]);
    close(pc_pipe[1]);
    waiting_for_initial_map = 0;
}

This is one of the reasons for the code size regression in PR23153.

-- 
           Summary: code size regression on x86
           Product: gcc
           Version: 4.1.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: rtl-optimization
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: dann at godzilla dot ics dot uci dot edu
                CC: gcc-bugs at gcc dot gnu dot org
GCC target triplet: i686-pc-linux-gnu


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23523

[Bug rtl-optimization/23523] code size regression on x86

[dann at godzilla dot ics dot uci dot edu]

------- Additional Comments From dann at godzilla dot ics dot uci dot edu  2005-08-23 05:45 -------
Created an attachment (id=9559)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=9559&action=view)
preprocessed source, look for "first_map_occurred"


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23523

[Bug rtl-optimization/23523] code size regression on x86

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-08-23 11:26 -------
This is an ra issue.  There is only one load from pc_pipe[1] which x86 does not like.  There is an 
actually a different on the tree level but the ra should have handled that but it does not.



-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
  BugsThisDependsOn|                            |18427
           Severity|normal                      |minor
             Status|UNCONFIRMED                 |NEW
     Ever Confirmed|                            |1
           Keywords|                            |missed-optimization, ra
   Last reconfirmed|0000-00-00 00:00:00         |2005-08-23 11:26:01
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23523

[Bug rtl-optimization/23523] code size regression on x86

[dann at godzilla dot ics dot uci dot edu]

------- Additional Comments From dann at godzilla dot ics dot uci dot edu  2005-08-25 05:43 -------
The issue is the peephole2 pass in 4.1. Before it the insn looks like:

(insn:HI 36 34 37 0 (set (mem/i:SI (symbol_ref:SI ("waiting_for_initial_map")
[flags 0x40] <var_decl 0xb7b71898 waiting_for_initial_map>) [7
waiting_for_initial_map+0 S4 A32])
        (const_int 0 [0x0])) 34 {*movsi_1} (nil)
    (nil))

and after: 

(insn 58 34 59 0 (parallel [
            (set (reg:SI 2 cx)
                (const_int 0 [0x0]))
            (clobber (reg:CC 17 flags))
        ]) -1 (nil)
    (expr_list:REG_UNUSED (reg:CC 17 flags)
        (nil)))

(insn 59 58 37 0 (set (mem/i:SI (symbol_ref:SI ("waiting_for_initial_map")
[flags 0x40] <var_decl 0xb7b71898 waiting_for_initial_map>) [7
waiting_for_initial_map+0 S4 A32])
        (reg:SI 2 cx)) -1 (nil)
    (expr_list:REG_DEAD (reg:SI 2 cx)
        (nil)))

4.0 uses "ax" instead of "cx". %eax is free at that point, so it is strange that 
it's not used in 4.1


-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
  BugsThisDependsOn|18427                       |
OtherBugsDependingO|                            |23153
              nThis|                            |
           Keywords|ra                          |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23523