| commit | e44ba9108a9672f6ecbee92820f6a78237ea6a8d | [log] [tgz] |
|---|---|---|
| author | Scott Todd <scotttodd@google.com> | Tue Aug 15 08:15:37 2023 -0700 |
| committer | GitHub <noreply@github.com> | Tue Aug 15 08:15:37 2023 -0700 |
| tree | 2c521cce9159b009fb17b7b2b11cafde12676a70 | |
| parent | 48f7394c6e24171da688095ee3427e0f36f271c8 [diff] |
A few fixes for compiling experimental/cuda2. (#14667)
* Remove trailing semicolon in cuda2/cuda_dynamic_symbol_table.h to fix
```
[build]
D:\dev\projects\iree\experimental/cuda2/cuda_dynamic_symbol_table.h(87):
error C2059: syntax error: ';'
[build]
D:\dev\projects\iree\experimental/cuda2/cuda_dynamic_symbols.h(33):
error C2059: syntax error: '}'
[build]
D:\dev\projects\iree\experimental/cuda2/cuda_dynamic_symbols.h(40):
error C2143: syntax error: missing ')' before '*'
```
* Make cuda2 flags use unique names to fix
```
[build]
iree_experimental_cuda2_registration_registration.lib(driver_module.c.obj)
: error LNK2005: iree_flag_register_cuda_async_allocations_ already
defined in
iree_hal_drivers_cuda_registration_registration.lib(driver_module.c.obj)
[build]
iree_experimental_cuda2_registration_registration.lib(driver_module.c.obj)
: error LNK2005: iree_flag_register_cuda_tracing_ already defined in
iree_hal_drivers_cuda_registration_registration.lib(driver_module.c.obj)
[build] tools\iree-run-module.exe : fatal error LNK1169: one or more
multiply defined symbols found
```
---
I see some runtime errors on my Windows machine (2080 TI, CUDA 12.2,
Driver version 536.25, using `--iree-hal-cuda-llvm-target-arch=sm_60`):
```
λ ..\iree-build\tools\iree-run-module.exe --module=..\iree-tmp\2023_08_14\model_EfficientNetB7PT_alias.vmfb --device=cuda2 --function=forward --input=1x3x600x600xf32
EXEC @forward
D:\dev\projects\iree\experimental\cuda2\graph_command_buffer.c:421: INVALID_ARGUMENT; CUDA error 'CUDA_ERROR_INVALID_VALUE' (1): invalid argument; cuGraphAddMemsetNode; while invoking native function hal.command_buffer.fill_buffer; while calling import;
[ 1] native hal.command_buffer.fill_buffer:0 -
[ 0] bytecode module.forward:9022 [
<eval_with_key>.29:2509:12,
```IREE (Intermediate Representation Execution Environment, pronounced as “eerie”) is an MLIR-based end-to-end compiler and runtime that lowers Machine Learning (ML) models to a unified IR that scales up to meet the needs of the datacenter and down to satisfy the constraints and special considerations of mobile and edge deployments.
See our website for project details, user guides, and instructions on building from source.
IREE is still in its early phase. We have settled down on the overarching infrastructure and are actively improving various software components as well as project logistics. It is still quite far from ready for everyday use and is made available without any support at the moment. With that said, we welcome any kind of feedback on any communication channels!
See our website for more information.
IREE is licensed under the terms of the Apache 2.0 License with LLVM Exceptions. See LICENSE for more information.