windows compile fix

RenderKit · Jan 7, 2015 · 2d8dbe6 · 2d8dbe6
1 parent c079c1c
commit 2d8dbe6
Show file tree

Hide file tree

Showing 6 changed files with 15 additions and 11 deletions.
diff --git a/common/math/color.h b/common/math/color.h
@@ -14,7 +14,7 @@
 // limitations under the License.                                           //
 // ======================================================================== //
 
-#pragma once
+#pragma once // FIXME: delete color files
 
 #include "../sys/constants.h"
 #include "../simd/sse.h"
@@ -201,7 +201,7 @@ namespace embree
 
   __forceinline const Color select( bool s, const Color& t, const Color& f ) {
     __m128 mask = s ? _mm_castsi128_ps(_mm_cmpeq_epi32(_mm_setzero_si128(), _mm_setzero_si128())) : _mm_setzero_ps();
-    return _mm_blendv_ps(f, t, mask);
+    return blendv_ps(f, t, mask);
   }
 
   ////////////////////////////////////////////////////////////////////////////////

diff --git a/common/math/vec3fa.h b/common/math/vec3fa.h
@@ -99,7 +99,7 @@ namespace embree
     return _mm_and_ps(a.m128, mask);
   }
   __forceinline const Vec3fa sign ( const Vec3fa& a ) {
-    return _mm_blendv_ps(Vec3fa(one), -Vec3fa(one), _mm_cmplt_ps (a,Vec3fa(zero))); 
+    return blendv_ps(Vec3fa(one), -Vec3fa(one), _mm_cmplt_ps (a,Vec3fa(zero))); 
   }
   __forceinline const Vec3fa rcp  ( const Vec3fa& a ) {
     const Vec3fa r = _mm_rcp_ps(a.m128);
@@ -111,7 +111,7 @@ namespace embree
     __m128 r = _mm_rsqrt_ps(a.m128);
     return _mm_add_ps(_mm_mul_ps(_mm_set1_ps(1.5f),r), _mm_mul_ps(_mm_mul_ps(_mm_mul_ps(a, _mm_set1_ps(-0.5f)), r), _mm_mul_ps(r, r)));
   }
-  __forceinline const Vec3fa zero_fix(const Vec3fa& a) { return _mm_blendv_ps(a, _mm_set1_ps(1E-10f), _mm_cmpeq_ps (a.m128, _mm_setzero_ps())); }
+  __forceinline const Vec3fa zero_fix(const Vec3fa& a) { return blendv_ps(a, _mm_set1_ps(1E-10f), _mm_cmpeq_ps (a.m128, _mm_setzero_ps())); }
   __forceinline const Vec3fa rcp_safe(const Vec3fa& a) { return rcp(zero_fix(a)); }
 
   __forceinline Vec3fa log ( const Vec3fa& a ) { 
@@ -254,11 +254,11 @@ namespace embree
 
   __forceinline const Vec3fa select( bool s, const Vec3fa& t, const Vec3fa& f ) {
     __m128 mask = s ? _mm_castsi128_ps(_mm_cmpeq_epi32(_mm_setzero_si128(), _mm_setzero_si128())) : _mm_setzero_ps();
-    return _mm_blendv_ps(f, t, mask);
+    return blendv_ps(f, t, mask);
   }
 
   __forceinline const Vec3fa select( const Vec3ba& s, const Vec3fa& t, const Vec3fa& f ) {
-    return _mm_blendv_ps(f, t, s);
+    return blendv_ps(f, t, s);
   }
 
   __forceinline int maxDim ( const Vec3fa& a ) 

diff --git a/common/simd/smmintrin_emu.h b/common/simd/smmintrin_emu.h
@@ -14,7 +14,7 @@
 // limitations under the License.                                           //
 // ======================================================================== //
 
-#pragma once
+#pragma once // FIXME: remove this file?
 
 #ifndef __GNUC__
 #pragma message (" --- Intel remark: SSE4 intrinsics are emulated with SSE3 ---")

diff --git a/common/simd/sse.h b/common/simd/sse.h
@@ -22,8 +22,12 @@
 
 namespace embree 
 {
-#if !defined(__SSE4_1__)
-  __forceinline __m128 _mm_blendv_ps( __m128 f, __m128 t, __m128 mask ) { 
+#if defined(__SSE4_1__)
+  __forceinline __m128 blendv_ps( __m128 f, __m128 t, __m128 mask ) { 
+    return _mm_blendv_ps(f,t,mask);
+  }
+#else
+  __forceinline __m128 blendv_ps( __m128 f, __m128 t, __m128 mask ) { 
     return _mm_or_ps(_mm_and_ps(mask, t), _mm_andnot_ps(mask, f)); 
   }
 #endif

diff --git a/common/simd/ssef.h b/common/simd/ssef.h
@@ -98,7 +98,7 @@ namespace embree
   __forceinline const ssef operator +( const ssef& a ) { return a; }
   __forceinline const ssef operator -( const ssef& a ) { return _mm_xor_ps(a.m128, _mm_castsi128_ps(_mm_set1_epi32(0x80000000))); }
   __forceinline const ssef abs       ( const ssef& a ) { return _mm_and_ps(a.m128, _mm_castsi128_ps(_mm_set1_epi32(0x7fffffff))); }
-  __forceinline const ssef sign      ( const ssef& a ) { return _mm_blendv_ps(ssef(one), -ssef(one), _mm_cmplt_ps (a,ssef(zero))); }
+  __forceinline const ssef sign      ( const ssef& a ) { return blendv_ps(ssef(one), -ssef(one), _mm_cmplt_ps (a,ssef(zero))); }
   __forceinline const ssef signmsk   ( const ssef& a ) { return _mm_and_ps(a.m128,_mm_castsi128_ps(_mm_set1_epi32(0x80000000))); }
 
   __forceinline const ssef rcp  ( const ssef& a ) {

diff --git a/common/simd/ssei.h b/common/simd/ssei.h
@@ -200,7 +200,7 @@ namespace embree
   }
 
 #if defined(__SSE4_1__) 
-#if defined(__clang__) || defined(_MSC_VER) && !defined(__INTEL_COMPILER)
+#if defined(__clang__) || defined(_MSC_VER) && !defined(__INTEL_COMPILER) // FIXME: can this get removed?
   __forceinline const ssei select(const int mask, const ssei& t, const ssei& f) {
 	  return select(sseb(mask), t, f);
   }