diff --git a/gcc/config/i386/i386.cc b/gcc/config/i386/i386.cc
index 169b2dd24b65a03636e00c14948d8bf53b7ce02f..cb3c14bbb346ac05bc1bcbf41c15612b98be9b0c 100644
--- a/gcc/config/i386/i386.cc
+++ b/gcc/config/i386/i386.cc
@@ -22975,7 +22975,7 @@ ix86_autovectorize_vector_modes (vector_modes *modes, bool all)
   if (TARGET_SSE2)
     modes->safe_push (V4QImode);
 
-  return 0;
+  return ix86_vect_compare_costs ? VECT_COMPARE_COSTS : 0;
 }
 
 /* Implemenation of targetm.vectorize.get_mask_mode.  */
@@ -23230,6 +23230,23 @@ ix86_vector_costs::add_stmt_cost (int count, vect_cost_for_stmt kind,
 	    stmt_cost = ix86_cost->add;
 	  break;
 	default:
+	  if (truth_value_p (subcode))
+	    {
+	      if (SSE_FLOAT_MODE_SSEMATH_OR_HF_P (mode))
+		/* CMPccS? instructions are cheap, so use sse_op.  While they
+		   produce a mask which may need to be turned to 0/1 by and,
+		   expect that this will be optimized away in a common case.  */
+		stmt_cost = ix86_cost->sse_op;
+	      else if (X87_FLOAT_MODE_P (mode))
+		/* fcmp + setcc.  */
+		stmt_cost = ix86_cost->fadd + ix86_cost->add;
+	      else if (VECTOR_MODE_P (mode))
+		stmt_cost = ix86_vec_cost (mode, ix86_cost->sse_op);
+	      else
+		/* setcc.  */
+		stmt_cost = ix86_cost->add;
+	      break;
+	    }
 	  break;
 	}
     }
diff --git a/gcc/config/i386/i386.opt b/gcc/config/i386/i386.opt
index ed31cab0a46a7af8def157ec43b65bcdfc778edf..d8cd58f5291829ef79bc80273f970a0800fd1be0 100644
--- a/gcc/config/i386/i386.opt
+++ b/gcc/config/i386/i386.opt
@@ -1219,6 +1219,10 @@ Do not use GOT to access external symbols.
 Target Joined UInteger Var(x86_stlf_window_ninsns) Init(64) Param
 Instructions number above which STFL stall penalty can be compensated.
 
+-param=ix86-vect-compare-costs=
+Target Joined UInteger Var(ix86_vect_compare_costs) Init(1) IntegerRange(0, 1) Param Optimization
+Whether x86 vectorizer cost modeling compares costs of different vector sizes.
+
 mgather
 Target Alias(mtune-ctrl=, use_gather, ^use_gather)
 Enable vectorization for gather instruction.
diff --git a/gcc/doc/invoke.texi b/gcc/doc/invoke.texi
index baf066400daf290b54c4b3db3756666f8b22ef5d..7df9e16fdf322fe6093ce8dc3721228b069ebe50 100644
--- a/gcc/doc/invoke.texi
+++ b/gcc/doc/invoke.texi
@@ -15302,6 +15302,11 @@ The following choices of @var{name} are available on i386 and x86_64 targets:
 @item x86-stlf-window-ninsns
 Instructions number above which STFL stall penalty can be compensated.
 
+@item ix86-vect-compare-costs
+Whether x86 vectorizer cost modeling compares costs of different vector sizes.
+The value may be 0 or 1.  The default is 1; set it to 0 to restore the
+previous behavior.
+
 @end table
 
 @end table
diff --git a/gcc/testsuite/gcc.dg/vect/costmodel/x86_64/costmodel-pr120398.c b/gcc/testsuite/gcc.dg/vect/costmodel/x86_64/costmodel-pr120398.c
new file mode 100644
index 0000000000000000000000000000000000000000..39be0b61efd8cb3b373bb638661c72aea248a7bc
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/vect/costmodel/x86_64/costmodel-pr120398.c
@@ -0,0 +1,21 @@
+/* { dg-do compile } */
+
+static float
+muladd (float x, float y, float z)
+{
+  return x * y + z;
+}
+
+float
+g (float x[], long n)
+{
+  float r0 = 0, r1 = 0;
+  for (; n; x += 2, n--)
+    {
+      r0 = muladd (x[0], x[0], r0);
+      r1 = muladd (x[1], x[1], r1);
+    }
+  return r0 + r1;
+}
+
+/* { dg-final { scan-tree-dump "Choosing vector mode V8QI" "vect" } } */
diff --git a/gcc/testsuite/gcc.dg/vect/costmodel/x86_64/costmodel-pr123603.c b/gcc/testsuite/gcc.dg/vect/costmodel/x86_64/costmodel-pr123603.c
new file mode 100644
index 0000000000000000000000000000000000000000..b9844867ac1db62fe65e306edafdf12887a18abe
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/vect/costmodel/x86_64/costmodel-pr123603.c
@@ -0,0 +1,16 @@
+/* { dg-do compile } */
+/* This test relies on the EulerGCC-specific c86-4g-m7 cost model.  */
+/* { dg-additional-options "-mtune=c86-4g-m7" } */
+
+void foo (int *block)
+{
+  for (int i = 0; i < 3; ++i)
+    {
+      int a = block[i * 9];
+      int b = block[i * 9 + 1];
+      block[i * 9] = a + 10;
+      block[i * 9 + 1] = b + 10;
+    }
+}
+
+/* { dg-final { scan-tree-dump "optimized: loop vectorized using 8 byte vectors" "vect" } } */
diff --git a/gcc/testsuite/gcc.target/i386/vect-alignment-peeling-1.c b/gcc/testsuite/gcc.target/i386/vect-alignment-peeling-1.c
index 4aa536ba86c93526a40744e42be1f86d737af2cf..39ca8b13fdf545dad5cac76f4f13a04dbd7505ea 100644
--- a/gcc/testsuite/gcc.target/i386/vect-alignment-peeling-1.c
+++ b/gcc/testsuite/gcc.target/i386/vect-alignment-peeling-1.c
@@ -85,6 +85,6 @@ int main()
       __builtin_abort ();
 }
 
-/* { dg-final { scan-tree-dump-times "Alignment of access forced using peeling" 4 "vect" } } */ 
+/* { dg-final { scan-tree-dump-times "Alignment of access forced using peeling" 7 "vect" } } */
 /* Verify all vector accesses are emitted as aligned.  */
 /* { dg-final { scan-assembler-not "movup" } } */
diff --git a/gcc/testsuite/gcc.target/i386/vect-alignment-peeling-2.c b/gcc/testsuite/gcc.target/i386/vect-alignment-peeling-2.c
index 834bf0f770d66d1513dacb9cfb66387a75da3b76..9557e620a18e9128e641d78737393dc7e6692751 100644
--- a/gcc/testsuite/gcc.target/i386/vect-alignment-peeling-2.c
+++ b/gcc/testsuite/gcc.target/i386/vect-alignment-peeling-2.c
@@ -85,6 +85,6 @@ int main()
       __builtin_abort ();
 }
 
-/* { dg-final { scan-tree-dump-times "Alignment of access forced using peeling" 4 "vect" } } */ 
+/* { dg-final { scan-tree-dump-times "Alignment of access forced using peeling" 7 "vect" } } */
 /* Verify all vector accesses are emitted as aligned.  */
 /* { dg-final { scan-assembler-not "movup" } } */