Strange regression in egcs-current/x86

[Jeffrey A Law]

  In message <19980704150630.38682@kali.lrz-muenchen.de>you write:
  > Hello,
  > 
  > There is something strange in the x86 code generated by egcs-current. 
I don't think so.

  > f(int i)
  > {
  > 	float b[i + 1];
  > 
  > 	for (int c = 0; c <= i; c++) { 
  > 		b[c] = sin(c - 1); /* nonsense */ 
  > 	} 
  > }


  >         movl 8(%ebp),%ecx
  >         leal 1(%ecx),%eax        <-------
  >         sall $5,%eax             <------- shift to the left
  > 	    shrl $3,%eax             <------- shift back
  >         subl %eax,%esp
  > [...]
  > 
  > g++ 2.7.2 did it with:
  > 
  > [..]
  >         movl %esp,-8(%ebp)
  >         leal 4(,%esi,4),%eax
  >         subl %eax,%esp
  > 
  > I know that often more primitive operations are faster on later x86 CPUs,
  > but I doubt that first shifting to the left, and the backshifting to the 
  > right is the fastest way to do it. It seems some patterns got lost or the
  > combiner has some problems. 
Combining shifts like that is not safe.

Consider the value 0x20000000

If you shift it left 5 times, then shift it right 3 times you get the value
zero.

If you combine the shifts and just shift it left 2 times you get 0x80000000.

jeff