llvm-svn: 31122

llvm-svn: 31121

This will allow (after llvm-gcc linkage fix) more dummy tests using
libstdc++ to pass on mingw :)

llvm-svn: 31120

llvm-svn: 31119

optimization opportunity pointed out by Chris Lattner.

llvm-svn: 31118

llvm-svn: 31117

opportunity pointed out by Andrew Lewycky.

llvm-svn: 31115

transformation.  This speeds up a C++ app 2.25x.

llvm-svn: 31113

has a different prototype than the one #included from <setjmp.h>. This
patch fixes siod and a number of other test cases on Linux that were
failing the CBE because of this _setjmp issue.

llvm-svn: 31112

llvm-svn: 31111

llvm-svn: 31110

llvm-svn: 31109

llvm-svn: 31108

llvm-svn: 31107

1. Better document what is going on here.
2. Only hack on one branch per iteration, making the results less conservative.
3. Handle the problematic case by marking edges executable instead of by
   playing with value lattice states.  This is far less pessimistic, and fixes
   SCCP/ipsccp-gvar.ll.

llvm-svn: 31106

llvm-svn: 31103

llvm-svn: 31102

This provides stuff like:

        cmpw cr0, r15, r29
        mr r14, r15
-       bge cr0, LBB3_111       ;bb656
-       b LBB3_90       ;bb501
+       blt cr0, LBB3_90        ;bb501
 LBB3_111:      ;bb656
        lwz r18, 68(r1)

which is particularly good for dispatch group formation.

llvm-svn: 31101

        call L_strcmp$stub
        testl %eax, %eax
-       jne LBB26_208   #cond_true6020
-       jmp LBB26_227   #bb7119
+       je LBB26_227    #bb7119
 LBB26_208:     #cond_true6020
        movl $l31_str14, 4(%esp)

        testl %eax, %eax
-       jne LBB26_704   #cond_true13042
-       jmp LBB26_713   #bb13151
+       je LBB26_713    #bb13151
 LBB26_704:     #cond_true13042
        movl $_str52, 4(%esp)

        cmpl 76(%ecx), %eax
-       jge LBB26_1628  #cond_false63.i.i
-       jmp LBB26_1769  #_Z8makeGridP13mrSurfaceListidiidd.exit.i
+       jl LBB26_1769   #_Z8makeGridP13mrSurfaceListidiidd.exit.i
 LBB26_1628:    #cond_false63.i.i
        movl $0, 48964(%esp)

llvm-svn: 31100

llvm-svn: 31099

  jle FOO
  jmp BAR
BAR:

into:

  jle FOO
BAR:

... whoa!

llvm-svn: 31098

llvm-svn: 31097

llvm-svn: 31096

llvm-svn: 31095

1. Remove a bunch of ifdef'd code.
2. When a block just contains an uncond branch, change all blocks branching
   to it to jump to the destination instead.
3. If branch analysis tells us some edges in the machinecfg are not actually
   possible, remove them.

#2 triggers a suprisingly large number of times.

llvm-svn: 31094

llvm-svn: 31093

This is currently disabled by default and limited in several ways, but does
have a positive effect.

llvm-svn: 31090

llvm-svn: 31089

llvm-svn: 31088

llvm-svn: 31087

llvm-svn: 31086

llvm-svn: 31074

llvm-svn: 31073

but apparently didn't make it into the final patch.

llvm-svn: 31070

llvm-svn: 31069

http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20061016/038780.html

llvm-svn: 31068

output.

llvm-svn: 31067

llvm-svn: 31066

llvm-svn: 31065

llvm-svn: 31064