[PATCH] Fix PR c/8281 (Sparc)

[David S. Miller]

   From: Eric Botcazou <ebotcazou@libertysurf.fr>
   Date: Wed, 26 Mar 2003 01:05:59 +0100
   
   So the first part of the fix is to disable the second splitter the same way 
   David disabled the first one. Then there is a new problem: the 
   "movdi_insn_sp32_v9" pattern above contains a "f-f" alternative which 
   expects a splitter, which we have just disabled because we use < 32 instead 
   of < 64. But this "f-f" alternative is actually unnecessary since it is 
   superseded by the "e-e" alternative (the 64 FP registers are equivalent in 
   DFmode), so we can remove it. This is sufficient to let the testcase pass.

This patch and testcase looks fine.

Thank you for looking into this.