Merge commit 'origin/master^' into blender2.8

6 files changed, 75 insertions, 8 deletions

diff --git a/intern/cycles/kernel/closure/bsdf.h b/intern/cycles/kernel/closure/bsdf.h
index e3beff6675a..d8ff69ca241 100644
--- a/intern/cycles/kernel/closure/bsdf.h
+++ b/intern/cycles/kernel/closure/bsdf.h
@@ -36,20 +36,42 @@ CCL_NAMESPACE_BEGIN
 
 /* Returns the square of the roughness of the closure if it has roughness,
  * 0 for singular closures and 1 otherwise. */
-ccl_device_inline float bsdf_get_roughness_squared(const ShaderClosure *sc)
+ccl_device_inline float bsdf_get_specular_roughness_squared(const ShaderClosure *sc)
 {
 	if(CLOSURE_IS_BSDF_SINGULAR(sc->type)) {
 		return 0.0f;
 	}
 
 	if(CLOSURE_IS_BSDF_MICROFACET(sc->type)) {
-		MicrofacetBsdf *bsdf = (MicrofacetBsdf*) sc;
+		MicrofacetBsdf *bsdf = (MicrofacetBsdf*)sc;
 		return bsdf->alpha_x*bsdf->alpha_y;
 	}
 
 	return 1.0f;
 }
 
+ccl_device_inline float bsdf_get_roughness_squared(const ShaderClosure *sc)
+{
+	/* This version includes diffuse, mainly for baking Principled BSDF
+	 * where specular and metallic zero otherwise does not bake the
+	 * specified roughness parameter. */
+	if(sc->type == CLOSURE_BSDF_OREN_NAYAR_ID) {
+		OrenNayarBsdf *bsdf = (OrenNayarBsdf*)sc;
+		return sqr(sqr(bsdf->roughness));
+	}
+
+	if(sc->type == CLOSURE_BSDF_PRINCIPLED_DIFFUSE_ID) {
+		PrincipledDiffuseBsdf *bsdf = (PrincipledDiffuseBsdf*)sc;
+		return sqr(sqr(bsdf->roughness));
+	}
+
+	if(CLOSURE_IS_BSDF_DIFFUSE(sc->type)) {
+		return 0.0f;
+	}
+
+	return bsdf_get_specular_roughness_squared(sc);
+}
+
 ccl_device_forceinline int bsdf_sample(KernelGlobals *kg,
                                        ShaderData *sd,
                                        const ShaderClosure *sc,
@@ -176,7 +198,7 @@ ccl_device_forceinline int bsdf_sample(KernelGlobals *kg,
 		float threshold_squared = kernel_data.background.transparent_roughness_squared_threshold;
 
 		if(threshold_squared >= 0.0f) {
-			if(bsdf_get_roughness_squared(sc) <= threshold_squared) {
+			if(bsdf_get_specular_roughness_squared(sc) <= threshold_squared) {
 				label |= LABEL_TRANSMIT_TRANSPARENT;
 			}
 		}
diff --git a/intern/cycles/kernel/kernel_passes.h b/intern/cycles/kernel/kernel_passes.h
index 1933d695f92..a42a8e9812f 100644
--- a/intern/cycles/kernel/kernel_passes.h
+++ b/intern/cycles/kernel/kernel_passes.h
@@ -140,7 +140,7 @@ ccl_device_inline void kernel_update_denoising_features(KernelGlobals *kg,
 		/* All closures contribute to the normal feature, but only diffuse-like ones to the albedo. */
 		normal += sc->N * sc->sample_weight;
 		sum_weight += sc->sample_weight;
-		if(bsdf_get_roughness_squared(sc) > sqr(0.075f)) {
+		if(bsdf_get_specular_roughness_squared(sc) > sqr(0.075f)) {
 			albedo += sc->weight;
 			sum_nonspecular_weight += sc->sample_weight;
 		}
diff --git a/intern/cycles/kernel/kernel_path_state.h b/intern/cycles/kernel/kernel_path_state.h
index 479bb22d780..8a358e51f94 100644
--- a/intern/cycles/kernel/kernel_path_state.h
+++ b/intern/cycles/kernel/kernel_path_state.h
@@ -164,6 +164,7 @@ ccl_device_inline void path_state_next(KernelGlobals *kg, ccl_addr_space PathSta
 #endif
 }
 
+#ifdef __VOLUME__
 ccl_device_inline bool path_state_volume_next(KernelGlobals *kg, ccl_addr_space PathState *state)
 {
 	/* For volume bounding meshes we pass through without counting transparent
@@ -180,6 +181,7 @@ ccl_device_inline bool path_state_volume_next(KernelGlobals *kg, ccl_addr_space
 
 	return true;
 }
+#endif
 
 ccl_device_inline uint path_state_ray_visibility(KernelGlobals *kg, ccl_addr_space PathState *state)
 {
diff --git a/intern/cycles/util/util_sseb.h b/intern/cycles/util/util_sseb.h
index 93c22aafdcd..977976c3fc0 100644
--- a/intern/cycles/util/util_sseb.h
+++ b/intern/cycles/util/util_sseb.h
@@ -119,7 +119,7 @@ __forceinline const sseb unpacklo( const sseb& a, const sseb& b ) { return _mm_u
 __forceinline const sseb unpackhi( const sseb& a, const sseb& b ) { return _mm_unpackhi_ps(a, b); }
 
 template<size_t i0, size_t i1, size_t i2, size_t i3> __forceinline const sseb shuffle( const sseb& a ) {
-	return _mm_shuffle_epi32(a, _MM_SHUFFLE(i3, i2, i1, i0));
+	return _mm_castsi128_ps(_mm_shuffle_epi32(a, _MM_SHUFFLE(i3, i2, i1, i0)));
 }
 
 template<> __forceinline const sseb shuffle<0, 1, 0, 1>( const sseb& a ) {
diff --git a/intern/cycles/util/util_thread.cpp b/intern/cycles/util/util_thread.cpp
index 3dcb09804b0..c66aa484264 100644
--- a/intern/cycles/util/util_thread.cpp
+++ b/intern/cycles/util/util_thread.cpp
@@ -26,7 +26,11 @@ thread::thread(function<void(void)> run_cb, int group)
     joined_(false),
 	group_(group)
 {
+#if (__cplusplus > 199711L) || (defined(_MSC_VER) && _MSC_VER >= 1800)
+	thread_ = std::thread(&thread::run, this);
+#else
 	pthread_create(&pthread_id_, NULL, run, (void*)this);
+#endif
 }
 
 thread::~thread()
@@ -60,7 +64,17 @@ void *thread::run(void *arg)
 bool thread::join()
 {
 	joined_ = true;
+#if (__cplusplus > 199711L) || (defined(_MSC_VER) && _MSC_VER >= 1800)
+	try {
+		thread_.join();
+		return true;
+	}
+	catch (const std::system_error&) {
+		return false;
+	}
+#else
 	return pthread_join(pthread_id_, NULL) == 0;
+#endif
 }
 
 CCL_NAMESPACE_END
diff --git a/intern/cycles/util/util_thread.h b/intern/cycles/util/util_thread.h
index 1e91fb8a706..4d8f464359c 100644
--- a/intern/cycles/util/util_thread.h
+++ b/intern/cycles/util/util_thread.h
@@ -24,10 +24,16 @@
 #  include <functional>
 #else
 #  include <boost/thread.hpp>
+#  include <pthread.h>
 #endif
-#include <pthread.h>
 #include <queue>
 
+#ifdef _WIN32
+#  include "util_windows.h"
+#else
+#  include <pthread.h>
+#endif
+
 #ifdef __APPLE__
 #  include <libkern/OSAtomic.h>
 #endif
@@ -60,7 +66,11 @@ public:
 
 protected:
 	function<void(void)> run_cb_;
+#if (__cplusplus > 199711L) || (defined(_MSC_VER) && _MSC_VER >= 1800)
+	std::thread thread_;
+#else
 	pthread_t pthread_id_;
+#endif
 	bool joined_;
 	int group_;
 };
@@ -81,7 +91,24 @@ public:
 	inline void unlock() {
 		OSSpinLockUnlock(&spin_);
 	}
-#else  /* __APPLE__ */
+#elif defined(_WIN32)
+	inline thread_spin_lock() {
+		const DWORD SPIN_COUNT = 50000;
+		InitializeCriticalSectionAndSpinCount(&cs_, SPIN_COUNT);
+	}
+
+	inline ~thread_spin_lock() {
+		DeleteCriticalSection(&cs_);
+	}
+
+	inline void lock() {
+		EnterCriticalSection(&cs_);
+	}
+
+	inline void unlock() {
+		LeaveCriticalSection(&cs_);
+	}
+#else
 	inline thread_spin_lock() {
 		pthread_spin_init(&spin_, 0);
 	}
@@ -97,10 +124,12 @@ public:
 	inline void unlock() {
 		pthread_spin_unlock(&spin_);
 	}
-#endif  /* __APPLE__ */
+#endif
 protected:
 #ifdef __APPLE__
 	OSSpinLock spin_;
+#elif defined(_WIN32)
+	CRITICAL_SECTION cs_;
 #else
 	pthread_spinlock_t spin_;
 #endif