Re: "Fix" recent povray failure with -mfpmath=sse -ffast-math

[Uros Bizjak <uros at kss-loka dot si>]

Richard Henderson wrote:

> I assume this fixes it.  Though it would be a Good Thing if we could
> get a reduced testcase affected by this.
>
> I'm not sure what real benefit we get from claiming that we have GE
> as a subset of the real UNGE comparison.  Seems to me that we'd only
> get an observable benefit from matching e.g. LT like so:
>  
:(( With your patch I'm not able to compile PovRay. It looks that it 
chokes on

(insn:HI 384 383 385 (set (reg:CCFP 17 flags)
       (compare:CCFP (reg/v:DF 21 xmm0 [orig:69 v ] [69])
           (mem:DF (plus:SI (mult:SI (reg/v:SI 2 cx [orig:61 i ] [61])
                       (const_int 8 [0x8]))
                   (reg:SI 1 dx [orig:76 POVFPU_Consts ] [76])) [17 S8 
A64]))) 24 {*cmpfp_i_sse} (insn_list:REG_DEP_TRUE 375 (nil))
   (nil))

(jump_insn:HI 385 384 603 (set (pc)
       (if_then_else (ltgt (reg:CCFP 17 flags)
               (const_int 0 [0x0]))
           (label_ref 38)
           (pc))) 338 {*jcc_1} (insn_list:REG_DEP_TRUE 384 (nil))
   (expr_list:REG_DEAD (reg:CCFP 17 flags)
       (expr_list:REG_BR_PROB (const_int 8900 [0x22c4])
           (nil))))

Again, ltgt triggers

c++ -DPREFIX=\"/usr/local\" 
-DPOV_LIB_DIR=\"/usr/local/share/povray-3.5\" 
-DCOMPILER_VER=\".Linux.gcc\" -DSYSCONFDIR=\"/usr/local/etc\" 
-DUSE_IO_RESTRICTIONS=\"\"     `if [ "Xgcc" = "Xgcc" ]; then echo 
"-Wno-multichar"; fi ` -O3 -march=pentium4 -mfpmath=sse -ffast-math 
-D__NO_MATH_INLINES -finline-functions -fomit-frame-pointer 
-funroll-loops -fexpensive-optimizations -malign-double 
-foptimize-sibling-calls -minline-all-stringops -Wno-multichar -c 
fnpovfpu.cpp
fnpovfpu.cpp: In function âdouble POVFPU_RunDefault(FUNCTION)â:
fnpovfpu.cpp:1461: internal compiler error: in ix86_expand_fp_movcc, at 
config/i386/i386.c:9813
Please submit a full bug report,

Uros.