Cleanup: make vector types make/print functions consistent between CPU and GPU

Now all the same ones are available on CPU and GPU, which was previously not
possible due to lack of operator overloadng in OpenCL. Print functions are
no-ops on some GPUs.

Ref D15535

1 files changed, 28 insertions, 29 deletions

diff --git a/intern/cycles/util/types_int4_impl.h b/intern/cycles/util/types_int4_impl.h
index 2aab29df941..067794e67b4 100644
--- a/intern/cycles/util/types_int4_impl.h
+++ b/intern/cycles/util/types_int4_impl.h
@@ -56,52 +56,51 @@ __forceinline int &int4::operator[](int i)
 }
 #  endif
 
-ccl_device_inline int4 make_int4(int i)
+ccl_device_inline int4 make_int4(int x, int y, int z, int w)
 {
 #  ifdef __KERNEL_SSE__
-  int4 a(_mm_set1_epi32(i));
+  return int4(_mm_set_epi32(w, z, y, x));
 #  else
-  int4 a = {i, i, i, i};
+  return {x, y, z, w};
 #  endif
-  return a;
 }
 
-ccl_device_inline int4 make_int4(int x, int y, int z, int w)
+#endif /* __KERNEL_NATIVE_VECTOR_TYPES__ */
+
+ccl_device_inline int4 make_int4(int i)
 {
-#  ifdef __KERNEL_SSE__
-  int4 a(_mm_set_epi32(w, z, y, x));
-#  else
-  int4 a = {x, y, z, w};
-#  endif
-  return a;
+#ifdef __KERNEL_SSE__
+  return int4(_mm_set1_epi32(i));
+#else
+  return make_int4(i, i, i, i);
+#endif
 }
 
-ccl_device_inline int4 make_int4(const float3 &f)
+ccl_device_inline int4 make_int4(const float3 f)
 {
-#  ifdef __KERNEL_SSE__
-  int4 a(_mm_cvtps_epi32(f.m128));
-#  elif defined(__KERNEL_ONEAPI__)
-  int4 a = {(int)f.x, (int)f.y, (int)f.z, 0};
-#  else
-  int4 a = {(int)f.x, (int)f.y, (int)f.z, (int)f.w};
-#  endif
-  return a;
+#if defined(__KERNEL_GPU__)
+  return make_int4((int)f.x, (int)f.y, (int)f.z, 0);
+#elif defined(__KERNEL_SSE__)
+  return int4(_mm_cvtps_epi32(f.m128));
+#else
+  return make_int4((int)f.x, (int)f.y, (int)f.z, (int)f.w);
+#endif
 }
 
-ccl_device_inline int4 make_int4(const float4 &f)
+ccl_device_inline int4 make_int4(const float4 f)
 {
-#  ifdef __KERNEL_SSE__
-  int4 a(_mm_cvtps_epi32(f.m128));
-#  else
-  int4 a = {(int)f.x, (int)f.y, (int)f.z, (int)f.w};
-#  endif
-  return a;
+#ifdef __KERNEL_SSE__
+  return int4(_mm_cvtps_epi32(f.m128));
+#else
+  return make_int4((int)f.x, (int)f.y, (int)f.z, (int)f.w);
+#endif
 }
 
-ccl_device_inline void print_int4(const char *label, const int4 &a)
+ccl_device_inline void print_int4(ccl_private const char *label, const int4 a)
 {
+#ifdef __KERNEL_PRINTF__
   printf("%s: %d %d %d %d\n", label, a.x, a.y, a.z, a.w);
+#endif
 }
-#endif /* __KERNEL_NATIVE_VECTOR_TYPES__ */
 
 CCL_NAMESPACE_END