[lang] Refactor allocation logic for SNodeTreeBufferManager #7795
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
✅ Deploy Preview for docsite-preview canceled.
|
jim19930609
force-pushed
the
memory_pool_refactor_8
branch
from
b192299 to
63e377f
Compare
Closed
|
/rebase |
taichi-gardener
force-pushed
the
memory_pool_refactor_8
branch
from
8c882f4 to
b0930eb
Compare
quadpixels
pushed a commit
to quadpixels/taichi
that referenced
this pull request
May 13, 2023
…ev#7795) Issue: taichi-dev#7599 ### Brief Summary <!-- copilot:summary --> ### <samp>🤖 Generated by Copilot at b192299</samp> This pull request refactors and simplifies the memory management classes and interfaces in the RHI module, by introducing `HostMemoryPool` and `DeviceMemoryPool` classes that use the unified allocator and the caching allocator respectively. It also adapts the CC, CPU, CUDA, and AMDGPU backends to use the new memory pool classes, and updates the relevant header files, source files, and CMakeLists files. ### Walkthrough <!-- copilot:walkthrough --> ### <samp>🤖 Generated by Copilot at b192299</samp> * Rename `MemoryPool` class to `HostMemoryPool` and move its implementation from `memory_pool.cpp` to `host_memory_pool.cpp` ([link](https://github.com/taichi-dev/taichi/pull/7795/files?diff=unified&w=0#diff-3b567f822ae335185b4ee459dbe761b0f385b8502caea7e51e9d4a12c064c7e5L12), [link](https://github.com/taichi-dev/taichi/pull/7795/files?diff=unified&w=0#diff-4a2f42ebf872b21247efc079652e216f1b006b8f5564744813a5436f9db10befL7-R7), [link](https://github.com/taichi-dev/taichi/pull/7795/files?diff=unified&w=0#diff-c7f11648081cc60e309965f3b13b2644536f2da81d1eb9d45eeeafc87346cb07R1-R135), [link](https://github.com/taichi-dev/taichi/pull/7795/files?diff=unified&w=0#diff-77849776a4e61703fdf4a076fd01bb3d70b0edf16d96879fc7a6bb051b1c00e1)) * Update the include paths of `memory_pool.h` to `host_memory_pool.h` in `cc_program.cpp`, `program.cpp`, `unified_allocator.cpp`, and `cpu_device.cpp` ([link](https://github.com/taichi-dev/taichi/pull/7795/files?diff=unified&w=0#diff-0a875d6f7e430ca9b888f9203cb94c9862c2c08cd3153b6ceba055fabec3a47eL2-R2), [link](https://github.com/taichi-dev/taichi/pull/7795/files?diff=unified&w=0#diff-d2572e3acfa01c9e64a50f4f311338bedac2c1aef1d6ece19a980cda3c3e71abL18-R18), [link](https://github.com/taichi-dev/taichi/pull/7795/files?diff=unified&w=0#diff-6cdadd09403135efbb53bd51e0c282e0ac6d50fd8784f31a65044771aa7627c0L3-R4), [link](https://github.com/taichi-dev/taichi/pull/7795/files?diff=unified&w=0#diff-e49f3cb85e1e9d5e83043e56f4d5d69c371bf8f40b4b653e31fc24eaea7e30c7L3-R3)) * Remove the `arch` argument and field from `HostMemoryPool` and `UnifiedAllocator` classes, and update their constructors and calls accordingly ([link](https://github.com/taichi-dev/taichi/pull/7795/files?diff=unified&w=0#diff-0092a6dc164cbb577e2900a6ac31599e7583362a892ddb36347f0ba9a58fc275L40-R37), [link](https://github.com/taichi-dev/taichi/pull/7795/files?diff=unified&w=0#diff-6cdadd09403135efbb53bd51e0c282e0ac6d50fd8784f31a65044771aa7627c0L34-R32), [link](https://github.com/taichi-dev/taichi/pull/7795/files?diff=unified&w=0#diff-9a3937a6437de08c0a5c79e202d18bcd154cbce83265d171f07ba50996a8947cL28-R37), [link](https://github.com/taichi-dev/taichi/pull/7795/files?diff=unified&w=0#diff-6cdadd09403135efbb53bd51e0c282e0ac6d50fd8784f31a65044771aa7627c0L82-R75), [link](https://github.com/taichi-dev/taichi/pull/7795/files?diff=unified&w=0#diff-d2572e3acfa01c9e64a50f4f311338bedac2c1aef1d6ece19a980cda3c3e71abL338), [link](https://github.com/taichi-dev/taichi/pull/7795/files?diff=unified&w=0#diff-d2572e3acfa01c9e64a50f4f311338bedac2c1aef1d6ece19a980cda3c3e71abL357-R356), [link](https://github.com/taichi-dev/taichi/pull/7795/files?diff=unified&w=0#diff-e49f3cb85e1e9d5e83043e56f4d5d69c371bf8f40b4b653e31fc24eaea7e30c7L15), [link](https://github.com/taichi-dev/taichi/pull/7795/files?diff=unified&w=0#diff-e49f3cb85e1e9d5e83043e56f4d5d69c371bf8f40b4b653e31fc24eaea7e30c7L22-R22), [link](https://github.com/taichi-dev/taichi/pull/7795/files?diff=unified&w=0#diff-e49f3cb85e1e9d5e83043e56f4d5d69c371bf8f40b4b653e31fc24eaea7e30c7L56-R55)) * Add `DeviceMemoryPool` class in `device_memory_pool.h` and `device_memory_pool.cpp`, which uses `CachingAllocator` to allocate and release memory on the device ([link](https://github.com/taichi-dev/taichi/pull/7795/files?diff=unified&w=0#diff-aa21493a95ff3932837037225cd6f333d54c7ddcd3e9a8774fd65a0a096c21a1R9), [link](https://github.com/taichi-dev/taichi/pull/7795/files?diff=unified&w=0#diff-cd7ce2b821966dececd4c4d49390a1c3e750be38c980d6448787b4dbf50f329dR1-R151), [link](https://github.com/taichi-dev/taichi/pull/7795/files?diff=unified&w=0#diff-dbd6bdc1bd636021e4226b81e8ebdaccca472d22f9e54b90013b84260710d048R1-R43)) * Add constructors for `CudaDevice` and `AmdgpuDevice` classes, which initialize the device memory pool instance with different `merge_upon_release` flags ([link](https://github.com/taichi-dev/taichi/pull/7795/files?diff=unified&w=0#diff-7919f5d7e33aafc72f27ed93febc58a0ac77c220ae718bf78ef134dad3790654R8-R12), [link](https://github.com/taichi-dev/taichi/pull/7795/files?diff=unified&w=0#diff-01cbaa214a08d98caa33977f217853d54c85d51c03e2dab610909f530b63ee7fR84), [link](https://github.com/taichi-dev/taichi/pull/7795/files?diff=unified&w=0#diff-e0843d34a17298595cd26f5b47b1933a0d18c5c795b0c91b73ec8541a2cdd3f9R9-R13), [link](https://github.com/taichi-dev/taichi/pull/7795/files?diff=unified&w=0#diff-a7b85199b3c0a668f23bf0b723c126009e774437499d555de255883c1d1f4648R70)) * Update `CudaDevice` and `AmdgpuDevice` classes to use `DeviceMemoryPool` instead of directly calling the CUDA or AMDGPU driver to allocate and release memory, and remove the unused caching allocator instances and comments ([link](https://github.com/taichi-dev/taichi/pull/7795/files?diff=unified&w=0#diff-7919f5d7e33aafc72f27ed93febc58a0ac77c220ae718bf78ef134dad3790654L17-R31), [link](https://github.com/taichi-dev/taichi/pull/7795/files?diff=unified&w=0#diff-7919f5d7e33aafc72f27ed93febc58a0ac77c220ae718bf78ef134dad3790654L31-L34), [link](https://github.com/taichi-dev/taichi/pull/7795/files?diff=unified&w=0#diff-7919f5d7e33aafc72f27ed93febc58a0ac77c220ae718bf78ef134dad3790654L49-R52), [link](https://github.com/taichi-dev/taichi/pull/7795/files?diff=unified&w=0#diff-7919f5d7e33aafc72f27ed93febc58a0ac77c220ae718bf78ef134dad3790654L76-R79), [link](https://github.com/taichi-dev/taichi/pull/7795/files?diff=unified&w=0#diff-01cbaa214a08d98caa33977f217853d54c85d51c03e2dab610909f530b63ee7fL7), [link](https://github.com/taichi-dev/taichi/pull/7795/files?diff=unified&w=0#diff-01cbaa214a08d98caa33977f217853d54c85d51c03e2dab610909f530b63ee7fL139), [link](https://github.com/taichi-dev/taichi/pull/7795/files?diff=unified&w=0#diff-e0843d34a17298595cd26f5b47b1933a0d18c5c795b0c91b73ec8541a2cdd3f9L17-L24), [link](https://github.com/taichi-dev/taichi/pull/7795/files?diff=unified&w=0#diff-e0843d34a17298595cd26f5b47b1933a0d18c5c795b0c91b73ec8541a2cdd3f9R28-R38), [link](https://github.com/taichi-dev/taichi/pull/7795/files?diff=unified&w=0#diff-e0843d34a17298595cd26f5b47b1933a0d18c5c795b0c91b73ec8541a2cdd3f9L49-R59), [link](https://github.com/taichi-dev/taichi/pull/7795/files?diff=unified&w=0#diff-e0843d34a17298595cd26f5b47b1933a0d18c5c795b0c91b73ec8541a2cdd3f9L77-R86), [link](https://github.com/taichi-dev/taichi/pull/7795/files?diff=unified&w=0#diff-a7b85199b3c0a668f23bf0b723c126009e774437499d555de255883c1d1f4648L9), [link](https://github.com/taichi-dev/taichi/pull/7795/files?diff=unified&w=0#diff-a7b85199b3c0a668f23bf0b723c126009e774437499d555de255883c1d1f4648L113)) * Update `CachingAllocator` class to remove the `device` field and argument, and pass the `device` argument to the `merge_and_insert` and `allocate_llvm_runtime_memory_jit` functions instead ([link](https://github.com/taichi-dev/taichi/pull/7795/files?diff=unified&w=0#diff-2db035897bca403c2ce8a2b5ce2ff1de0ecc77b6fa97f9dae0bb7fd6416de2adL6-R7), [link](https://github.com/taichi-dev/taichi/pull/7795/files?diff=unified&w=0#diff-2db035897bca403c2ce8a2b5ce2ff1de0ecc77b6fa97f9dae0bb7fd6416de2adR34), [link](https://github.com/taichi-dev/taichi/pull/7795/files?diff=unified&w=0#diff-2db035897bca403c2ce8a2b5ce2ff1de0ecc77b6fa97f9dae0bb7fd6416de2adL54-R55), [link](https://github.com/taichi-dev/taichi/pull/7795/files?diff=unified&w=0#diff-06b92f3a1b8326081cb6b99afd424e2c0be4d4f3df97ca913a8978c9c9b5d2d2L15-R18), [link](https://github.com/taichi-dev/taichi/pull/7795/files?diff=unified&w=0#diff-06b92f3a1b8326081cb6b99afd424e2c0be4d4f3df97ca913a8978c9c9b5d2d2L25)) * Update the assertion in `unified_allocator.cpp` to use `HostMemoryPool::page_size` instead of `MemoryPool::page_size` ([link](https://github.com/taichi-dev/taichi/pull/7795/files?diff=unified&w=0#diff-6cdadd09403135efbb53bd51e0c282e0ac6d50fd8784f31a65044771aa7627c0L90-R82)) * Update the comment for `HostMemoryPool` class in `host_memory_pool.h` ([link](https://github.com/taichi-dev/taichi/pull/7795/files?diff=unified&w=0#diff-0092a6dc164cbb577e2900a6ac31599e7583362a892ddb36347f0ba9a58fc275L12-R30)) * Remove the unused local variable `arch` in `finalize` function in `program.cpp` ([link](https://github.com/taichi-dev/taichi/pull/7795/files?diff=unified&w=0#diff-d2572e3acfa01c9e64a50f4f311338bedac2c1aef1d6ece19a980cda3c3e71abL338))
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Issue: #7599
Brief Summary
🤖 Generated by Copilot at b192299
This pull request refactors and simplifies the memory management classes and interfaces in the RHI module, by introducing
HostMemoryPoolandDeviceMemoryPoolclasses that use the unified allocator and the caching allocator respectively. It also adapts the CC, CPU, CUDA, and AMDGPU backends to use the new memory pool classes, and updates the relevant header files, source files, and CMakeLists files.Walkthrough
🤖 Generated by Copilot at b192299
MemoryPoolclass toHostMemoryPooland move its implementation frommemory_pool.cpptohost_memory_pool.cpp(link, link, link, link)memory_pool.htohost_memory_pool.hincc_program.cpp,program.cpp,unified_allocator.cpp, andcpu_device.cpp(link, link, link, link)archargument and field fromHostMemoryPoolandUnifiedAllocatorclasses, and update their constructors and calls accordingly (link, link, link, link, link, link, link, link, link)DeviceMemoryPoolclass indevice_memory_pool.handdevice_memory_pool.cpp, which usesCachingAllocatorto allocate and release memory on the device (link, link, link)CudaDeviceandAmdgpuDeviceclasses, which initialize the device memory pool instance with differentmerge_upon_releaseflags (link, link, link, link)CudaDeviceandAmdgpuDeviceclasses to useDeviceMemoryPoolinstead of directly calling the CUDA or AMDGPU driver to allocate and release memory, and remove the unused caching allocator instances and comments (link, link, link, link, link, link, link, link, link, link, link, link)CachingAllocatorclass to remove thedevicefield and argument, and pass thedeviceargument to themerge_and_insertandallocate_llvm_runtime_memory_jitfunctions instead (link, link, link, link, link)unified_allocator.cppto useHostMemoryPool::page_sizeinstead ofMemoryPool::page_size(link)HostMemoryPoolclass inhost_memory_pool.h(link)archinfinalizefunction inprogram.cpp(link)