Cycles: CUDA support for rendering scenes that don't fit on GPU.

In that case it can now fall back to CPU memory, at the cost of reduced
performance. For scenes that fit in GPU memory, this commit should not
cause any noticeable slowdowns.

We don't use all physical system RAM, since that can cause OS instability.
We leave at least half of system RAM or 4GB to other software, whichever
is smaller.

For image textures in host memory, performance was maybe 20-30% slower
in our tests (although this is highly hardware and scene dependent). Once
other type of data doesn't fit on the GPU, performance can be e.g. 10x
slower, and at that point it's probably better to just render on the CPU.

Differential Revision: https://developer.blender.org/D2056

2 files changed, 23 insertions, 0 deletions

diff --git a/intern/cycles/util/util_system.cpp b/intern/cycles/util/util_system.cpp
index a942d738b8a..9b1b9a60c30 100644
--- a/intern/cycles/util/util_system.cpp
+++ b/intern/cycles/util/util_system.cpp
@@ -292,5 +292,26 @@ bool system_cpu_support_avx2()
 
 #endif
 
+size_t system_physical_ram()
+{
+#ifdef _WIN32
+	MEMORYSTATUSEX ram;
+	ram.dwLength = sizeof (ram);
+	GlobalMemoryStatusEx(&ram);
+	return ram.ullTotalPhys * 1024;
+#elif defined(__APPLE__)
+	uint64_t ram = 0;
+	size_t len = sizeof(ram);
+	if (sysctlbyname("hw.memsize", &ram, &len, NULL, 0) == 0) {
+		return ram;
+	}
+	return 0;
+#else
+	size_t ps = sysconf(_SC_PAGESIZE);
+	size_t pn = sysconf(_SC_PHYS_PAGES);
+	return ps * pn;
+#endif
+}
+
 CCL_NAMESPACE_END
 
diff --git a/intern/cycles/util/util_system.h b/intern/cycles/util/util_system.h
index db7a45b2d59..e55dd6dd136 100644
--- a/intern/cycles/util/util_system.h
+++ b/intern/cycles/util/util_system.h
@@ -42,6 +42,8 @@ bool system_cpu_support_sse41();
 bool system_cpu_support_avx();
 bool system_cpu_support_avx2();
 
+size_t system_physical_ram();
+
 CCL_NAMESPACE_END
 
 #endif /* __UTIL_SYSTEM_H__ */