Replace _mm_cvtps_epi32(x) with _mm_cvttps_epi32(_mm_add_ps(0.5f), x).

tests/PMFloatTest.cpp

 #include "SkPMFloat.h"
 #include "Test.h"
 
 DEF_TEST(SkPMFloat, r) {
     // Test SkPMColor <-> SkPMFloat
     SkPMColor c = SkPreMultiplyColor(0xFFCC9933);
     SkPMFloat pmf(c);
     REPORTER_ASSERT(r, SkScalarNearlyEqual(255.0f, pmf.a()));
     REPORTER_ASSERT(r, SkScalarNearlyEqual(204.0f, pmf.r()));
     REPORTER_ASSERT(r, SkScalarNearlyEqual(153.0f, pmf.g()));
     REPORTER_ASSERT(r, SkScalarNearlyEqual( 51.0f, pmf.b()));
     REPORTER_ASSERT(r, c == pmf.get());
 
-    // Test rounding.  (Don't bother testing .5... we don't care which way it goes.)
-    pmf = SkPMFloat(254.6f, 204.3f, 153.1f, 50.8f);
+    // Test rounding.
+    pmf = SkPMFloat(254.5f, 203.5f, 153.1f, 50.8f);
     REPORTER_ASSERT(r, c == pmf.get());
 
     // Test clamping.
     SkPMFloat clamped(SkPMFloat(510.0f, 153.0f, 1.0f, -0.2f).clamped());
     REPORTER_ASSERT(r, SkScalarNearlyEqual(255.0f, clamped.a()));
     REPORTER_ASSERT(r, SkScalarNearlyEqual(153.0f, clamped.r()));
     REPORTER_ASSERT(r, SkScalarNearlyEqual(  1.0f, clamped.g()));
     REPORTER_ASSERT(r, SkScalarNearlyEqual(  0.0f, clamped.b()));
 
     // Test SkPMFloat <-> Sk4f conversion.
@@ skipping 18 matching lines @@
     SkPMFloat::To4PMColors(back, floats);
     for (int i = 0; i < 4; i++) {
         REPORTER_ASSERT(r, back[i] == colors[i]);
     }
 
     SkPMFloat::ClampTo4PMColors(back, floats);
     for (int i = 0; i < 4; i++) {
         REPORTER_ASSERT(r, back[i] == colors[i]);
     }
 }