mirage-project/mirage
Mirage: Automatically Generating Fast GPU Kernels without Programming in Triton/CUDA
C++Apache-2.0
Issues
- 0
- 1
WSL2环境下由于找不到CUDA_CUDNN_LIBRARY无法通过pip安装mirage
#140 opened by WuTianyi321 - 1
Does mirage support dynamic shape ?
#134 opened by Leaf996 - 0
[Transpiler] List of TODOs
#60 opened by jiazhihao - 0
- 0
[Transpiler] relax stensors' innermost dimension's alignment to reduce shared memory usage
#131 opened by jiazhihao - 1
H100 crash
#129 opened by goddice - 0
[Transpiler] the Transpiler should return an error flag when the planned shared memory usage exceeds hardware capacity
#116 opened by jiazhihao - 0
- 3
- 0
- 3
- 1
[Bug][Segmentation fault] Some of the search demos result in segmentation fault
#110 opened by jiazhihao - 4
- 0
ImportError: libz3.so.4.13: cannot open shared object file: No such file or directory
#17 opened by gknagoing - 0
- 1
module 'mirage' has no attribute 'new_graph'
#35 opened by mert-cemri - 0
Update INSTALL.md
#77 opened by jiazhihao - 1
- 0
- 2
- 4
- 2
Accuracy of the SiLU op is incorrect
#99 opened by zk1998 - 1
- 0
Does mirage support the model training process and the backward process of gradient calculation?
#103 opened by daneren - 4
Potential Mirage projects
#18 opened by jiazhihao - 0
Support kernel level reduction in transpiler
#96 opened by wmdi - 0
- 0
Improve the display of search statistics
#84 opened by wmdi - 0
Too large thread block
#86 opened by jiazhihao - 1
Add algebraic patterns for SILU and RMSNorm
#69 opened by jiazhihao - 1
- 1
[Search] The search is only parallelized across two threads for the Gated MLP example
#75 opened by jiazhihao - 0
- 0
[Transpiler] CUDA Compilation Error
#64 opened by jiazhihao - 0
[Search] Multithreaded DFS
#62 opened by wmdi - 0
[Transpiler] CUDA compilation error
#58 opened by jiazhihao - 1
[Transpiler] invalid redeclaration of type name
#56 opened by jiazhihao - 0
- 1
- 1
- 1
[Bug] Undefined symbol `dmemlayout_to_cmemlayout` when building from source
#24 opened by interestingLSY - 0
Using Mirage for different models
#27 opened by ramyaprabhu-alt - 1
[Compile Warning] operator type mismatch
#12 opened by jiazhihao - 0
Fix the search procedure
#7 opened by jiazhihao - 0
Support for CUDA 12.2
#11 opened by sam-h-bean - 3
Manual build instructions fail
#8 opened by catid - 1
- 0
- 0
List of TODOs for the cutlass backend
#6 opened by jiazhihao