Cycles: Kernel address space changes for MSL

This is the first of a sequence of changes to support compiling Cycles kernels as MSL (Metal Shading Language) in preparation for a Metal GPU device implementation.

MSL requires that all pointer types be declared with explicit address space attributes (device, thread, etc...). There is already precedent for this with Cycles' address space macros (ccl_global, ccl_private, etc...), therefore the first step of MSL-enablement is to apply these consistently. Line-for-line this represents the largest change required to enable MSL. Applying this change first will simplify future patches as well as offering the emergent benefit of enhanced descriptiveness.

The vast majority of deltas in this patch fall into one of two cases:

- Ensuring ccl_private is specified for thread-local pointer types
- Ensuring ccl_global is specified for device-wide pointer types

Additionally, the ccl_addr_space qualifier can be removed. Prior to Cycles X, ccl_addr_space was used as a context-dependent address space qualifier, but now it is either redundant (e.g. in struct typedefs), or can be replaced by ccl_global in the case of pointer types. Associated function variants (e.g. lcg_step_float_addrspace) are also redundant.

In cases where address space qualifiers are chained with "const", this patch places the address space qualifier first. The rationale for this is that the choice of address space is likely to have the greater impact on runtime performance and overall architecture.

The final part of this patch is the addition of a metal/compat.h header. This is partially complete and will be extended in future patches, paving the way for the full Metal implementation.

Ref T92212

Reviewed By: brecht

Maniphest Tasks: T92212

Differential Revision: https://developer.blender.org/D12864

5 files changed, 126 insertions, 5 deletions

diff --git a/intern/cycles/kernel/device/cpu/compat.h b/intern/cycles/kernel/device/cpu/compat.h
index bfd936c7bbd..888c0d5d872 100644
--- a/intern/cycles/kernel/device/cpu/compat.h
+++ b/intern/cycles/kernel/device/cpu/compat.h
@@ -32,8 +32,6 @@
 #include "util/util_texture.h"
 #include "util/util_types.h"
 
-#define ccl_addr_space
-
 /* On x86_64, versions of glibc < 2.16 have an issue where expf is
  * much slower than the double version.  This was fixed in glibc 2.16.
  */
diff --git a/intern/cycles/kernel/device/cuda/compat.h b/intern/cycles/kernel/device/cuda/compat.h
index 3c85a8e7bd2..685c7a5b753 100644
--- a/intern/cycles/kernel/device/cuda/compat.h
+++ b/intern/cycles/kernel/device/cuda/compat.h
@@ -59,7 +59,6 @@ typedef unsigned long long uint64_t;
 #define ccl_gpu_shared __shared__
 #define ccl_private
 #define ccl_may_alias
-#define ccl_addr_space
 #define ccl_restrict __restrict__
 #define ccl_loop_no_unroll
 #define ccl_align(n) __align__(n)
diff --git a/intern/cycles/kernel/device/hip/compat.h b/intern/cycles/kernel/device/hip/compat.h
index 95338fe7d6e..089976d84e4 100644
--- a/intern/cycles/kernel/device/hip/compat.h
+++ b/intern/cycles/kernel/device/hip/compat.h
@@ -52,7 +52,6 @@ typedef unsigned long long uint64_t;
 #define ccl_gpu_shared __shared__
 #define ccl_private
 #define ccl_may_alias
-#define ccl_addr_space
 #define ccl_restrict __restrict__
 #define ccl_loop_no_unroll
 #define ccl_align(n) __align__(n)
diff --git a/intern/cycles/kernel/device/metal/compat.h b/intern/cycles/kernel/device/metal/compat.h
new file mode 100644
index 00000000000..77cea30914c
--- /dev/null
+++ b/intern/cycles/kernel/device/metal/compat.h
@@ -0,0 +1,126 @@
+/*
+ * Copyright 2011-2013 Blender Foundation
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#pragma once
+
+#define __KERNEL_GPU__
+#define __KERNEL_METAL__
+#define CCL_NAMESPACE_BEGIN
+#define CCL_NAMESPACE_END
+
+#ifndef ATTR_FALLTHROUGH
+#  define ATTR_FALLTHROUGH
+#endif
+
+#include <metal_atomic>
+#include <metal_pack>
+#include <metal_stdlib>
+#include <simd/simd.h>
+
+using namespace metal;
+
+#pragma clang diagnostic ignored "-Wunused-variable"
+#pragma clang diagnostic ignored "-Wsign-compare"
+
+/* Qualifiers */
+
+#define ccl_device
+#define ccl_device_inline ccl_device
+#define ccl_device_forceinline ccl_device
+#define ccl_device_noinline ccl_device __attribute__((noinline))
+#define ccl_device_noinline_cpu ccl_device
+#define ccl_global device
+#define ccl_static_constant static constant constexpr
+#define ccl_device_constant constant
+#define ccl_constant const device
+#define ccl_gpu_shared threadgroup
+#define ccl_private thread
+#define ccl_may_alias
+#define ccl_restrict __restrict
+#define ccl_loop_no_unroll
+#define ccl_align(n) alignas(n)
+#define ccl_optional_struct_init
+
+/* No assert supported for Metal */
+
+#define kernel_assert(cond)
+
+/* make_type definitions with Metal style element initializers */
+#ifdef make_float2
+#  undef make_float2
+#endif
+#ifdef make_float3
+#  undef make_float3
+#endif
+#ifdef make_float4
+#  undef make_float4
+#endif
+#ifdef make_int2
+#  undef make_int2
+#endif
+#ifdef make_int3
+#  undef make_int3
+#endif
+#ifdef make_int4
+#  undef make_int4
+#endif
+#ifdef make_uchar4
+#  undef make_uchar4
+#endif
+
+#define make_float2(x, y) float2(x, y)
+#define make_float3(x, y, z) float3(x, y, z)
+#define make_float4(x, y, z, w) float4(x, y, z, w)
+#define make_int2(x, y) int2(x, y)
+#define make_int3(x, y, z) int3(x, y, z)
+#define make_int4(x, y, z, w) int4(x, y, z, w)
+#define make_uchar4(x, y, z, w) uchar4(x, y, z, w)
+
+/* Math functions */
+
+#define __uint_as_float(x) as_type<float>(x)
+#define __float_as_uint(x) as_type<uint>(x)
+#define __int_as_float(x) as_type<float>(x)
+#define __float_as_int(x) as_type<int>(x)
+#define __float2half(x) half(x)
+#define powf(x, y) pow(float(x), float(y))
+#define fabsf(x) fabs(float(x))
+#define copysignf(x, y) copysign(float(x), float(y))
+#define asinf(x) asin(float(x))
+#define acosf(x) acos(float(x))
+#define atanf(x) atan(float(x))
+#define floorf(x) floor(float(x))
+#define ceilf(x) ceil(float(x))
+#define hypotf(x, y) hypot(float(x), float(y))
+#define atan2f(x, y) atan2(float(x), float(y))
+#define fmaxf(x, y) fmax(float(x), float(y))
+#define fminf(x, y) fmin(float(x), float(y))
+#define fmodf(x, y) fmod(float(x), float(y))
+#define sinhf(x) sinh(float(x))
+#define coshf(x) cosh(float(x))
+#define tanhf(x) tanh(float(x))
+
+/* Use native functions with possibly lower precision for performance,
+ * no issues found so far. */
+#define trigmode fast
+#define sinf(x) trigmode::sin(float(x))
+#define cosf(x) trigmode::cos(float(x))
+#define tanf(x) trigmode::tan(float(x))
+#define expf(x) trigmode::exp(float(x))
+#define sqrtf(x) trigmode::sqrt(float(x))
+#define logf(x) trigmode::log(float(x))
+
+#define NULL 0
diff --git a/intern/cycles/kernel/device/optix/compat.h b/intern/cycles/kernel/device/optix/compat.h
index fb9e094b535..c9ec9be05df 100644
--- a/intern/cycles/kernel/device/optix/compat.h
+++ b/intern/cycles/kernel/device/optix/compat.h
@@ -58,7 +58,6 @@ typedef unsigned long long uint64_t;
 #define ccl_gpu_shared __shared__
 #define ccl_private
 #define ccl_may_alias
-#define ccl_addr_space
 #define ccl_restrict __restrict__
 #define ccl_loop_no_unroll
 #define ccl_align(n) __align__(n)