Cycles: attempt to fix internal compile error with some visual studio builds

6 files changed, 14 insertions, 35 deletions

diff --git a/intern/cycles/CMakeLists.txt b/intern/cycles/CMakeLists.txt
index 9f067bede33..a8293ac683c 100644
--- a/intern/cycles/CMakeLists.txt
+++ b/intern/cycles/CMakeLists.txt
@@ -13,9 +13,10 @@ if(WITH_RAYOPTIMIZATION AND SUPPORT_SSE_BUILD)
 endif()
 
 if(WIN32 AND MSVC)
-	set(CYCLES_SSE2_KERNEL_FLAGS "/arch:SSE2 -D_CRT_SECURE_NO_WARNINGS /fp:fast /EHsc")
-	set(CYCLES_SSE3_KERNEL_FLAGS "/arch:SSE2 -D_CRT_SECURE_NO_WARNINGS /fp:fast /EHsc")
-	set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} /fp:fast -D_CRT_SECURE_NO_WARNINGS /EHsc")
+	set(CYCLES_SSE2_KERNEL_FLAGS "/arch:SSE2 -D_CRT_SECURE_NO_WARNINGS /fp:fast /EHsc /Ox /GS-")
+	# there is no /arch:SSE3, but intrinsics are available anyway
+	set(CYCLES_SSE3_KERNEL_FLAGS "/arch:SSE2 -D_CRT_SECURE_NO_WARNINGS /fp:fast /EHsc /Ox /GS-")
+	set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} /fp:fast -D_CRT_SECURE_NO_WARNINGS /EHsc /Ox /GS-")
 elseif(CMAKE_COMPILER_IS_GNUCC)
 	set(CYCLES_SSE2_KERNEL_FLAGS "-ffast-math -msse -msse2 -mfpmath=sse")
 	set(CYCLES_SSE3_KERNEL_FLAGS "-ffast-math -msse -msse2 -msse3 -mssse3 -mfpmath=sse")
diff --git a/intern/cycles/SConscript b/intern/cycles/SConscript
index 43f838157b9..57f44ae9d52 100644
--- a/intern/cycles/SConscript
+++ b/intern/cycles/SConscript
@@ -79,11 +79,12 @@ if env['WITH_BF_RAYOPTIMIZATION']:
     sse3_cxxflags = Split(env['CXXFLAGS'])
 
     if env['OURPLATFORM'] == 'win32-vc':
-        sse2_cxxflags.append('/arch:SSE2 -D_CRT_SECURE_NO_WARNINGS /fp:fast /EHsc'.split())
-        sse3_cxxflags.append('/arch:SSE2 -D_CRT_SECURE_NO_WARNINGS /fp:fast /EHsc'.split())
+        sse2_cxxflags.append('/arch:SSE /arch:SSE2 -D_CRT_SECURE_NO_WARNINGS /fp:fast /EHsc /Ox /Gs-'.split())
+        # there is no /arch:SSE3, but intrinsics are available anyway
+        sse3_cxxflags.append('/arch:SSE /arch:SSE2 -D_CRT_SECURE_NO_WARNINGS /fp:fast /EHsc /Ox /Gs-'.split())
     elif env['OURPLATFORM'] == 'win64-vc':
-        sse2_cxxflags.append('-D_CRT_SECURE_NO_WARNINGS /fp:fast /EHsc'.split())
-        sse3_cxxflags.append('-D_CRT_SECURE_NO_WARNINGS /fp:fast /EHsc'.split())
+        sse2_cxxflags.append('-D_CRT_SECURE_NO_WARNINGS /fp:fast /EHsc /Ox /Gs-'.split())
+        sse3_cxxflags.append('-D_CRT_SECURE_NO_WARNINGS /fp:fast /EHsc /Ox /Gs-'.split())
     else:
         sse2_cxxflags.append('-ffast-math -msse -msse2 -mfpmath=sse'.split())
         sse3_cxxflags.append('-ffast-math -msse -msse2 -msse3 -mssse3 -mfpmath=sse'.split())
diff --git a/intern/cycles/bvh/bvh.cpp b/intern/cycles/bvh/bvh.cpp
index 69ccf2588c9..dd647fec2d3 100644
--- a/intern/cycles/bvh/bvh.cpp
+++ b/intern/cycles/bvh/bvh.cpp
@@ -579,6 +579,7 @@ void RegularBVH::pack_nodes(const array<int>& prims, const BVHNode *root)
 	int nextNodeIdx = 0;
 
 	vector<BVHStackEntry> stack;
+	stack.reserve(BVHParams::MAX_DEPTH*2);
 	stack.push_back(BVHStackEntry(root, nextNodeIdx++));
 
 	while(stack.size()) {
@@ -776,6 +777,7 @@ void QBVH::pack_nodes(const array<int>& prims, const BVHNode *root)
 	int nextNodeIdx = 0;
 
 	vector<BVHStackEntry> stack;
+	stack.reserve(BVHParams::MAX_DEPTH*2);
 	stack.push_back(BVHStackEntry(root, nextNodeIdx++));
 
 	while(stack.size()) {
diff --git a/intern/cycles/kernel/kernel_jitter.h b/intern/cycles/kernel/kernel_jitter.h
index 15d2151228f..17704b48cee 100644
--- a/intern/cycles/kernel/kernel_jitter.h
+++ b/intern/cycles/kernel/kernel_jitter.h
@@ -37,7 +37,7 @@ __device_inline int cmj_fast_mod_pow2(int a, int b)
 /* a must be > 0 and b must be > 1 */
 __device_inline int cmj_fast_div_pow2(int a, int b)
 {
-#ifdef __KERNEL_SSE2__
+#if defined(__KERNEL_SSE2__) && !defined(_MSC_VER)
 	return a >> __builtin_ctz(b);
 #else
 	return a/b;
@@ -46,7 +46,7 @@ __device_inline int cmj_fast_div_pow2(int a, int b)
 
 __device_inline uint cmj_w_mask(uint w)
 {
-#ifdef __KERNEL_SSE2__
+#if defined(__KERNEL_SSE2__) && !defined(_MSC_VER)
 	return ((1 << (32 - __builtin_clz(w))) - 1);
 #else
 	w |= w >> 1;
diff --git a/intern/cycles/util/util_math.h b/intern/cycles/util/util_math.h
index 68668d88d44..05ed8a2d3f6 100644
--- a/intern/cycles/util/util_math.h
+++ b/intern/cycles/util/util_math.h
@@ -165,7 +165,7 @@ __device_inline float clamp(float a, float mn, float mx)
 
 __device_inline int float_to_int(float f)
 {
-#ifdef __KERNEL_SSE2__
+#if defined(__KERNEL_SSE2__) && !defined(_MSC_VER)
 	return _mm_cvtt_ss2si(_mm_load_ss(&f));
 #else
 	return (int)f;
diff --git a/intern/cycles/util/util_types.h b/intern/cycles/util/util_types.h
index 32529e47c23..7ad0403e6c6 100644
--- a/intern/cycles/util/util_types.h
+++ b/intern/cycles/util/util_types.h
@@ -23,11 +23,6 @@
 
 #include <stdlib.h>
 
-	// Needed for _BistScanForward and _BitScanReverse on MSVC 2012 
-#	if defined(_MSC_VER) && _MSC_VER >= 1700
-#		include <intrin.h>
-#	endif
-
 #endif
 
 /* Qualifiers for kernel code shared by CPU and GPU */
@@ -511,26 +506,6 @@ template<size_t i0, size_t i1, size_t i2, size_t i3> __device_inline const __m12
 }
 #endif
 
-#if defined(__KERNEL_SSE2__) && defined(_MSC_VER)
-
-/* count zeros from start or end of integer bits */
-
-__device_inline uint32_t __builtin_ctz(uint32_t i)
-{
-	unsigned long r = 0;
-	_BitScanForward(&r, i);
-	return (uint32_t)r;
-}
-
-__device_inline uint32_t __builtin_clz(uint32_t i)
-{
-	unsigned long r = 0;
-	_BitScanReverse(&r, i);
-	return (uint32_t)r;
-}
-
-#endif
-
 CCL_NAMESPACE_END
 
 #endif /* __UTIL_TYPES_H__ */