re PR tree-optimization/33244 (Missed opportunities for vectorization)

	PR tree-optimization/33244
	* tree-ssa-sink.c (statement_sink_location): Don't sink into
	empty loop latches.

testsuite/
	PR tree-optimization/33244
	* gfortran.dg/vect/fast-math-vect-8.f90: New test.

From-SVN: r163998

4 files changed, 110 insertions, 0 deletions

diff --git a/gcc/ChangeLog b/gcc/ChangeLog
index 6da2bc4..ff56d4b 100644
--- a/gcc/ChangeLog
+++ b/gcc/ChangeLog
@@ -1,3 +1,9 @@
+2010-09-08  Michael Matz  <matz@suse.de>
+
+	PR tree-optimization/33244
+	* tree-ssa-sink.c (statement_sink_location): Don't sink into
+	empty loop latches.
+
 2010-09-08  Richard Guenther  <rguenther@suse.de>
 
 	PR tree-optimization/45578
diff --git a/gcc/testsuite/ChangeLog b/gcc/testsuite/ChangeLog
index 1f2cb92..7b40021 100644
--- a/gcc/testsuite/ChangeLog
+++ b/gcc/testsuite/ChangeLog
@@ -1,3 +1,8 @@
+2010-09-08  Michael Matz  <matz@suse.de>
+
+	PR tree-optimization/33244
+	* gfortran.dg/vect/fast-math-vect-8.f90: New test.
+
 2010-09-08  Richard Guenther  <rguenther@suse.de>
 
 	PR tree-optimization/45578
diff --git a/gcc/testsuite/gfortran.dg/vect/fast-math-vect-8.f90 b/gcc/testsuite/gfortran.dg/vect/fast-math-vect-8.f90
new file mode 100644
index 0000000..918943f
--- /dev/null
+++ b/gcc/testsuite/gfortran.dg/vect/fast-math-vect-8.f90
@@ -0,0 +1,93 @@
+! { dg-do compile } 
+! { dg-require-effective-target vect_float } 
+
+module solv_cap
+
+  implicit none
+
+  public  :: init_solve
+
+  integer, parameter, public :: dp = 4
+
+  real(kind=dp), private :: Pi, Mu0, c0, eps0
+  logical,       private :: UseFFT, UsePreco
+  real(kind=dp), private :: D1, D2
+  integer,       private, save :: Ng1=0, Ng2=0
+  integer,       private, pointer,     dimension(:,:)  :: Grid
+  real(kind=dp), private, allocatable, dimension(:,:)  :: G
+
+contains
+
+  subroutine init_solve(Grid_in, GrSize1, GrSize2, UseFFT_in, UsePreco_in)
+    integer, intent(in), target, dimension(:,:) :: Grid_in
+    real(kind=dp), intent(in)  :: GrSize1, GrSize2
+    logical,       intent(in)  :: UseFFT_in, UsePreco_in
+    integer                    :: i, j
+
+    Pi = acos(-1.0_dp)
+    Mu0 = 4e-7_dp * Pi
+    c0 = 299792458
+    eps0 = 1 / (Mu0 * c0**2)
+
+    UseFFT = UseFFT_in
+    UsePreco = UsePreco_in
+
+    if(Ng1 /= 0 .and. allocated(G) ) then
+      deallocate( G )
+    end if
+
+    Grid => Grid_in
+    Ng1 = size(Grid, 1)
+    Ng2 = size(Grid, 2)
+    D1 = GrSize1/Ng1
+    D2 = GrSize2/Ng2
+
+    allocate( G(0:Ng1,0:Ng2) )
+
+    write(unit=*, fmt=*) "Calculating G"
+    do i=0,Ng1
+      do j=0,Ng2
+        G(j,i) = Ginteg( -D1/2,-D2/2, D1/2,D2/2, i*D1,j*D2 )
+      end do
+    end do
+
+    if(UseFFT) then
+      write(unit=*, fmt=*) "Transforming G"
+      call FourirG(G,1)
+    end if
+
+    return
+
+
+  contains
+  function Ginteg(xq1,yq1, xq2,yq2, xp,yp)  result(G)
+    real(kind=dp), intent(in) :: xq1,yq1, xq2,yq2, xp,yp
+    real(kind=dp)             :: G
+    real(kind=dp)             :: x1,x2,y1,y2,t
+    x1 = xq1-xp
+    x2 = xq2-xp
+    y1 = yq1-yp
+    y2 = yq2-yp
+
+    if (x1+x2 < 0) then
+      t = -x1
+      x1 = -x2
+      x2 = t
+    end if
+    if (y1+y2 < 0) then
+      t = -y1
+      y1 = -y2
+      y2 = t
+    end if
+
+    G = (x2*y2)-(x1*y2)-(x2*y1)+(x1*y1)
+
+    return
+  end function Ginteg
+
+  end subroutine init_solve
+
+end module solv_cap
+
+! { dg-final { scan-tree-dump-times "vectorized 1 loops" 1 "vect" } } 
+! { dg-final { cleanup-tree-dump "vect" } } 
diff --git a/gcc/tree-ssa-sink.c b/gcc/tree-ssa-sink.c
index 436047b..bfa1a31 100644
--- a/gcc/tree-ssa-sink.c
+++ b/gcc/tree-ssa-sink.c
@@ -429,6 +429,12 @@ statement_sink_location (gimple stmt, basic_block frombb,
       || sinkbb->loop_father != frombb->loop_father)
     return false;
 
+  /* If the latch block is empty, don't make it non-empty by sinking
+     something into it.  */
+  if (sinkbb == frombb->loop_father->latch
+      && empty_block_p (sinkbb))
+    return false;
+
   /* Move the expression to a post dominator can't reduce the number of
      executions.  */
   if (dominated_by_p (CDI_POST_DOMINATORS, frombb, sinkbb))