[Bug rtl-optimization/17236] New: inefficient code for long long multiply on x86

[Bug rtl-optimization/17236] New: inefficient code for long long multiply on x86

[dann at godzilla dot ics dot uci dot edu]

Compiling:
unsigned long long LLM(unsigned long long arg1, unsigned long long arg2)
{
  return arg1 *  arg2;
}

with gcc -O2 -fomit-frame-pointer generates:

        pushl   %ebp
        pushl   %edi
        pushl   %ebx
        pushl   %edi
        pushl   %edi
        movl    32(%esp), %ecx
        movl    24(%esp), %eax
        movl    36(%esp), %ebx
        movl    24(%esp), %edi
        mull    %ecx
        imull   28(%esp), %ecx
        imull   %ebx, %edi
        movl    %edx, %ebp
        movl    %eax, (%esp)
        addl    %edi, %ebp
        movl    (%esp), %eax
        leal    (%ebp,%ecx), %ecx
        movl    %ecx, 4(%esp)
        movl    4(%esp), %edx
        popl    %ecx
        popl    %ebx
        popl    %ebx
        popl    %edi
        popl    %ebp
        ret

ICC -O2 generates much smaller code:

        movl      8(%esp), %ecx                                 #516.18
        imull     12(%esp), %ecx                                #516.18
        movl      16(%esp), %eax                                #516.18
        imull     4(%esp), %eax                                 #516.18
        addl      %eax, %ecx                                    #516.18
        movl      4(%esp), %eax                                 #516.18
        mull      12(%esp)                                      #516.18
        addl      %ecx, %edx                                    #516.18
        ret                                                     #516.18

-- 
           Summary: inefficient code for long long multiply on x86
           Product: gcc
           Version: 3.5.0
            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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17236

[Bug rtl-optimization/17236] inefficient code for long long multiply on x86

[giovannibajo at libero dot it]

------- Additional Comments From giovannibajo at libero dot it  2004-10-06 23:27 -------
Confirmed. The code is know a little bit better:

LLM:
     pushl      %ebp
     pushl      %edi
     pushl      %ebx
     subl       $8, %esp
     movl       32(%esp), %ecx
     movl       36(%esp), %ebx
     movl       24(%esp), %eax
     mull       %ecx
     movl       %eax, (%esp)
     movl       24(%esp), %edi
     imull      %ebx, %edi
     movl       %edx, %ebp
     addl       %edi, %ebp
     imull      28(%esp), %ecx
     leal       (%ebp,%ecx), %ecx
     movl       %ecx, 4(%esp)
     movl       (%esp), %eax
     movl       4(%esp), %edx
     addl       $8, %esp
     popl       %ebx
     popl       %edi
     popl       %ebp
     ret
     .size      LLM, .-LLM
     .ident     "GCC: (GNU) 4.0.0 20041003 (experimental)"




On 2.95, we used to emit:

LLM:
     pushl %ebp
     pushl %edi
     pushl %esi
     pushl %ebx
     movl 20(%esp),%ecx
     movl 24(%esp),%ebx
     movl 28(%esp),%eax
     movl 32(%esp),%esi
     mull %ecx
     movl %eax,%edi
     movl 28(%esp),%eax
     imull %ecx,%esi
     imull %ebx,%eax
     movl %edx,%ebp
     addl %esi,%ebp
     addl %eax,%ebp
     movl %edi,%eax
     movl %ebp,%edx
     popl %ebx
     popl %esi
     popl %edi
     popl %ebp
     ret
.Lfe1:
     .size       LLM,.Lfe1-LLM
     .ident     "GCC: (GNU) 2.95.3 20010315 (release)"



Not a regression, and ICC does so much better.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
           Severity|normal                      |enhancement
             Status|UNCONFIRMED                 |NEW
     Ever Confirmed|                            |1
           Keywords|                            |missed-optimization
      Known to fail|                            |2.95.3 3.0.4 3.3.3 3.4.1
                   |                            |4.0.0
   Last reconfirmed|0000-00-00 00:00:00         |2004-10-06 23:27:47
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17236

[Bug rtl-optimization/17236] inefficient code for long long multiply on x86

[cvs-commit at gcc dot gnu dot org]

------- Additional Comments From cvs-commit at gcc dot gnu dot org  2005-03-14 18:24 -------
Subject: Bug 17236

CVSROOT:	/cvs/gcc
Module name:	gcc
Changes by:	sayle@gcc.gnu.org	2005-03-14 18:24:15

Modified files:
	gcc            : ChangeLog optabs.c 

Log message:
	PR rtl-optimization/17236
	* optabs.c (expand_doubleword_mult): New helper function split out
	from expand_binop.  Permute the order in which instructions are
	emitted to minimize the number of simultaneously live registers.
	(expand_binop): Call expand_doubleword_mult to synthesize a double
	word multiplication.

Patches:
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/ChangeLog.diff?cvsroot=gcc&r1=2.7846&r2=2.7847
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/optabs.c.diff?cvsroot=gcc&r1=1.260&r2=1.261



-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17236

[Bug rtl-optimization/17236] inefficient code for long long multiply on x86

[giovannibajo at libero dot it]

------- Additional Comments From giovannibajo at libero dot it  2005-03-15 10:04 -------
Roger explains what else needs to be done here:
http://gcc.gnu.org/ml/gcc-patches/2005-03/msg01386.html

Right now, after his patch, mainline generates this code:

        pushl   %edi
        pushl   %esi
        pushl   %ebx
        movl    16(%esp), %eax
        movl    20(%esp), %edx
        movl    24(%esp), %ecx
        movl    28(%esp), %ebx
        movl    %edx, %esi
        imull   %ecx, %esi
        movl    %ebx, %edi
        imull   %eax, %edi
        addl    %edi, %esi
        mull    %ecx
        leal    (%esi,%edx), %edx
        popl    %ebx
        popl    %esi
        popl    %edi
        ret


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17236

[Bug rtl-optimization/17236] inefficient code for long long multiply on x86

[giovannibajo at libero dot it]

------- Additional Comments From giovannibajo at libero dot it  2005-03-15 10:07 -------
Uros did some additional comments:
http://gcc.gnu.org/ml/gcc-patches/2005-03/msg01427.html

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
  BugsThisDependsOn|                            |19398


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17236