From 040fa75d7b9d796d2818d731934caf77e4288b7f Mon Sep 17 00:00:00 2001
From: Brecht Van Lommel <brechtvanlommel@gmail.com>
Date: Tue, 9 Aug 2016 01:00:57 +0200
Subject: [PATCH] Fix Cycles CUDA adaptive kernel not working correctly after
 recent closure changes.

---
 intern/cycles/kernel/kernel_compat_cuda.h   |  1 +
 intern/cycles/kernel/kernel_compat_opencl.h |  1 +
 intern/cycles/kernel/kernel_types.h         | 13 ++++---------
 3 files changed, 6 insertions(+), 9 deletions(-)

diff --git a/intern/cycles/kernel/kernel_compat_cuda.h b/intern/cycles/kernel/kernel_compat_cuda.h
index a039b414006..063220b542e 100644
--- a/intern/cycles/kernel/kernel_compat_cuda.h
+++ b/intern/cycles/kernel/kernel_compat_cuda.h
@@ -47,6 +47,7 @@
 #define ccl_may_alias
 #define ccl_addr_space
 #define ccl_restrict __restrict__
+#define ccl_align(n) __align__(n)
 
 /* No assert supported for CUDA */
 
diff --git a/intern/cycles/kernel/kernel_compat_opencl.h b/intern/cycles/kernel/kernel_compat_opencl.h
index 8505cb85576..2ae89dde7c4 100644
--- a/intern/cycles/kernel/kernel_compat_opencl.h
+++ b/intern/cycles/kernel/kernel_compat_opencl.h
@@ -40,6 +40,7 @@
 #define ccl_local __local
 #define ccl_private __private
 #define ccl_restrict restrict
+#define ccl_align(n) __attribute__((aligned(n)))
 
 #ifdef __SPLIT_KERNEL__
 #  define ccl_addr_space __global
diff --git a/intern/cycles/kernel/kernel_types.h b/intern/cycles/kernel/kernel_types.h
index 3923804ec96..f3b10c21b9d 100644
--- a/intern/cycles/kernel/kernel_types.h
+++ b/intern/cycles/kernel/kernel_types.h
@@ -656,23 +656,18 @@ typedef struct AttributeDescriptor {
  * ShaderClosure has a fixed size, and any extra space must be allocated
  * with closure_alloc_extra().
  *
- * float3 is 12 bytes on CUDA and 16 bytes on CPU/OpenCL, we set the data
- * size to ensure ShaderClosure is 80 bytes total everywhere. */
+ * We pad the struct to 80 bytes and ensure it is aligned to 16 bytes, which
+ * we assume to be the maximum required alignment for any struct. */
 
 #define SHADER_CLOSURE_BASE \
 	float3 weight; \
 	ClosureType type; \
 	float sample_weight \
 
-typedef ccl_addr_space struct ShaderClosure {
+typedef ccl_addr_space struct ccl_align(16) ShaderClosure {
 	SHADER_CLOSURE_BASE;
 
-	/* pad to 80 bytes, data types are aligned to own size */
-#ifdef __KERNEL_CUDA__
-	float data[15];
-#else
-	float data[14];
-#endif
+	float data[14]; /* pad to 80 bytes */
 } ShaderClosure;
 
 /* Shader Context