Cycles: add HIP device support for AMD GPUs

NOTE: this feature is not ready for user testing, and not yet enabled in daily
builds. It is being merged now for easier collaboration on development.

HIP is a heterogenous compute interface allowing C++ code to be executed on
GPUs similar to CUDA. It is intended to bring back AMD GPU rendering support
on Windows and Linux.

https://github.com/ROCm-Developer-Tools/HIP.

As of the time of writing, it should compile and run on Linux with existing
HIP compilers and driver runtimes. Publicly available compilers and drivers
for Windows will come later.

See task T91571 for more details on the current status and work remaining
to be done.

Credits:

Sayak Biswas (AMD)
Arya Rafii (AMD)
Brian Savery (AMD)

Differential Revision: https://developer.blender.org/D12578

1 files changed, 121 insertions, 0 deletions

diff --git a/intern/cycles/kernel/device/hip/compat.h b/intern/cycles/kernel/device/hip/compat.h
new file mode 100644
index 00000000000..3644925d5be
--- /dev/null
+++ b/intern/cycles/kernel/device/hip/compat.h
@@ -0,0 +1,121 @@
+/*
+ * Copyright 2011-2021 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_HIP__
+#define CCL_NAMESPACE_BEGIN
+#define CCL_NAMESPACE_END
+
+#ifndef ATTR_FALLTHROUGH
+#  define ATTR_FALLTHROUGH
+#endif
+
+#ifdef __HIPCC_RTC__
+typedef unsigned int uint32_t;
+typedef unsigned long long uint64_t;
+#else
+#  include <stdint.h>
+#endif
+
+#ifdef CYCLES_HIPBIN_CC
+#  define FLT_MIN 1.175494350822287507969e-38f
+#  define FLT_MAX 340282346638528859811704183484516925440.0f
+#  define FLT_EPSILON 1.192092896e-07F
+#endif
+
+/* Qualifiers */
+
+#define ccl_device __device__ __inline__
+#define ccl_device_inline __device__ __inline__
+#define ccl_device_forceinline __device__ __forceinline__
+#define ccl_device_noinline __device__ __noinline__
+#define ccl_device_noinline_cpu ccl_device
+#define ccl_global
+#define ccl_static_constant __constant__
+#define ccl_device_constant __constant__ __device__
+#define ccl_constant const
+#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)
+#define ccl_optional_struct_init
+
+#define kernel_assert(cond)
+
+/* Types */
+#ifdef __HIP__
+#  include "hip/hip_fp16.h"
+#  include "hip/hip_runtime.h"
+#endif
+
+#ifdef _MSC_VER
+#  include <immintrin.h>
+#endif
+
+#define ccl_gpu_thread_idx_x (threadIdx.x)
+#define ccl_gpu_block_dim_x (blockDim.x)
+#define ccl_gpu_block_idx_x (blockIdx.x)
+#define ccl_gpu_grid_dim_x (gridDim.x)
+#define ccl_gpu_warp_size (warpSize)
+
+#define ccl_gpu_global_id_x() (ccl_gpu_block_idx_x * ccl_gpu_block_dim_x + ccl_gpu_thread_idx_x)
+#define ccl_gpu_global_size_x() (ccl_gpu_grid_dim_x * ccl_gpu_block_dim_x)
+
+/* GPU warp synchronizaton */
+
+#define ccl_gpu_syncthreads() __syncthreads()
+#define ccl_gpu_ballot(predicate) __ballot(predicate)
+#define ccl_gpu_shfl_down_sync(mask, var, detla) __shfl_down(var, detla)
+#define ccl_gpu_popc(x) __popc(x)
+
+/* GPU texture objects */
+typedef hipTextureObject_t ccl_gpu_tex_object;
+
+template<typename T>
+ccl_device_forceinline T ccl_gpu_tex_object_read_2D(const ccl_gpu_tex_object texobj,
+                                                    const float x,
+                                                    const float y)
+{
+  return tex2D<T>(texobj, x, y);
+}
+
+template<typename T>
+ccl_device_forceinline T ccl_gpu_tex_object_read_3D(const ccl_gpu_tex_object texobj,
+                                                    const float x,
+                                                    const float y,
+                                                    const float z)
+{
+  return tex3D<T>(texobj, x, y, z);
+}
+
+/* Use fast math functions */
+
+#define cosf(x) __cosf(((float)(x)))
+#define sinf(x) __sinf(((float)(x)))
+#define powf(x, y) __powf(((float)(x)), ((float)(y)))
+#define tanf(x) __tanf(((float)(x)))
+#define logf(x) __logf(((float)(x)))
+#define expf(x) __expf(((float)(x)))
+
+/* Types */
+
+#include "util/util_half.h"
+#include "util/util_types.h"