diff options
author | Patrick Mours <pmours@nvidia.com> | 2020-06-08 18:16:10 +0300 |
---|---|---|
committer | Patrick Mours <pmours@nvidia.com> | 2020-06-08 18:55:49 +0300 |
commit | 9f7d84b656fbb56966620ecc249ce5bc7089a1d1 (patch) | |
tree | d0a022feae43f6db2166cf5214b56cce99b96a60 /intern/cycles/device/cuda/device_cuda.h | |
parent | 0a907657d4d525d320e0c8518f583b7210736214 (diff) |
Cycles: Add support for P2P memory distribution (e.g. via NVLink)
This change modifies the multi-device implementation to support memory distribution
across devices, to reduce the overall memory footprint of large scenes and allow scenes to
fit entirely into combined GPU memory that previously had to fall back to host memory.
Reviewed By: brecht
Differential Revision: https://developer.blender.org/D7426
Diffstat (limited to 'intern/cycles/device/cuda/device_cuda.h')
-rw-r--r-- | intern/cycles/device/cuda/device_cuda.h | 3 |
1 files changed, 3 insertions, 0 deletions
diff --git a/intern/cycles/device/cuda/device_cuda.h b/intern/cycles/device/cuda/device_cuda.h index 3f23f0fe4c5..9f31ed12cf4 100644 --- a/intern/cycles/device/cuda/device_cuda.h +++ b/intern/cycles/device/cuda/device_cuda.h @@ -51,6 +51,7 @@ class CUDADevice : public Device { size_t map_host_used; size_t map_host_limit; int can_map_host; + int pitch_alignment; int cuDevId; int cuDevArchitecture; bool first_error; @@ -111,6 +112,8 @@ class CUDADevice : public Device { bool support_device(const DeviceRequestedFeatures & /*requested_features*/); + bool check_peer_access(Device *peer_device); + bool use_adaptive_compilation(); bool use_split_kernel(); |