Issues
- 8
build error
#13195 opened - 5
lowering stablehlo to ptx
#13175 opened - 5
- 4
Wrong output from JAX test after onednn change
#13054 opened - 0
- 2
- 4
MHLO Extraction from XLA Compiler
#12847 opened - 3
- 4
Porting XLA to different backends.
#12524 opened - 9
- 3
BF16 matmul slower than F32 matmul on T4 GPU
#12429 opened - 2
OpenCL Support.
#12413 opened - 4
- 0
Workable example on normal DNN model
#12240 opened - 1
gpu_fused_mha_test fails at HEAD on H100
#12175 opened - 1
Support builds with cuDNN 9
#12136 opened - 1
- 1
XLA documentation for Windows
#12028 opened - 7
Build from source fails
#12025 opened - 6
- 1
- 4
- 2
gpu_hlo_cost_analysis NumOfDevices always return 0
#11872 opened - 2
call ptxas become defunct, cause xla hung
#11824 opened - 1
[XLA:CPU] oneDNN Softmax gives inaccurate results
#11772 opened - 2
- 3
Integer overflow in ZlibOutputBuffer causes crash
#11668 opened - 9
- 2
No static version for linking NCCL (libnccl)
#11604 opened - 10
How to build for TPU ?
#11599 opened - 1
Compiling with `build --config nonccl` fails
#11596 opened - 2
`PjRtCApiBuffer::AcquireExternalReference` does not support `WaitUntilBufferReadyOnStream`
#11535 opened - 2
- 1
- 10
Incorrect boundary behaviour of `igamma`, `igammac`
#11151 opened - 0
- 0
cuDNN fMHA failures return unhelpful error messages
#11129 opened - 0
- 6
GpuLayoutAssignment pass unexpectedly replaced the host memory space color of the copy-start/copy-done instructions
#10826 opened - 12
Build fails with ROCm on Gentoo Linux
#10793 opened - 0
XLA Backend support for third party AI Accelerator
#10690 opened - 8
Potentially Misbehaving GPU
#10660 opened - 1
Building on Redhat 8.9
#10643 opened - 8
- 0
Building on FreeBSD 14
#10630 opened - 7
- 4
PJRT C API does not support GetDefaultLayout, however it is called (always) by DLPackManagedTensorToBuffer
#10613 opened - 6
build failed with undefined reference problem
#10592 opened - 8
[xla:cpu] use Intel AMX for int8 and bf16?
#10556 opened - 12