Search results

[Regression] CollectivePipeliner GPU pass present in JAX 0.4.38 is missing from JAX 0.10.1 — pipelining flags accepted but produce no pass stages; class symbols absent from binaries

Status: Open.

#44228 In openxla/xla;

· rsdcastro opened on Jun 12, 2026

[RFC] The Stable TopK Adapter: Integrating Unstable TopK Kernels into XLA

enhancementNew feature or requestNew feature or request

questionFurther information is requestedFurther information is requested

Status: Open.

#43913 In openxla/xla;

· apivovarov opened on Jun 9, 2026

tf.raw_ops.Where with tf.experimental.numpy.compress output causes XLA runtime error: "Buffers have different size"

err: RuntimeRuntime ErrorRuntime Error

Status: Open.

#43451 In openxla/xla;

· jasminetrail opened on May 29, 2026

tf.raw_ops.LRN with float16 input causes XLA mixed-precision error: "Seen floating point types of different precisions"

err:performancePerformance issuesPerformance issues

Status: Open.

#43450 In openxla/xla;

· jasminetrail opened on May 29, 2026

tf.image.sobel_edges crashes with cuDNN status 1002 (CUDNN_BACKEND_TENSOR_DESCRIPTOR finalize failure) on GPU

GPUXLA on GPUXLA on GPU

NVIDIA-GPUXLA on Nvidia GPUXLA on Nvidia GPU

Status: Open.

#43449 In openxla/xla;

· jasminetrail opened on May 29, 2026

[XLA] InternalError: mixed floating-point precision in TanhGrad during XLA compilation — f32 multiply sees mismatched types despite no explicit mixed precision

err:performancePerformance issuesPerformance issues

Status: Open.

#43052 In openxla/xla;

· jasminetrail opened on May 21, 2026

XLA algebraic simplifier folds inverse function pairs, ignoring IEEE 754 intermediate overflow/underflow

CPURelated to XLA on CPURelated to XLA on CPU

err:performancePerformance issuesPerformance issues

GPUXLA on GPUXLA on GPU

Status: Open.

#42732 In openxla/xla;

· wuyii8941 opened on May 16, 2026

SPMD partitioner produces incorrect all-reduce ordering for scatter-add backward

err:performancePerformance issuesPerformance issues

NVIDIA-GPUXLA on Nvidia GPUXLA on Nvidia GPU

Status: Open.

#42481 In openxla/xla;

· young-geng opened on May 12, 2026

Is it possible to run optimized HLOs on cloud TPUs?

Google-TPUXLA on Google TPUsXLA on Google TPUs

questionFurther information is requestedFurther information is requested

stat:awaiting response from contributorAwaiting response from contributor/authorAwaiting response from contributor/author

Status: Open.

#41992 In openxla/xla;

· IKACE opened on May 4, 2026

Use explicit batching dims in torch.gather lowering to stablehlo.gather instead of the implicit batch dims workaround

enhancementNew feature or requestNew feature or request

Status: Open.

#41902 In openxla/xla;

· hshahTT opened on May 1, 2026

# XLA FMA contraction silently changes a*x + b*x semantics, producing results many orders of magnitude wrong

bugSomething isn't workingSomething isn't working

err:performancePerformance issuesPerformance issues

stat:awaiting response from contributorAwaiting response from contributor/authorAwaiting response from contributor/author

Status: Open.

#41579 In openxla/xla;

· wuyii8941 opened on Apr 25, 2026

[GPU] Allow avoiding scale swizzling in cuDNN BlockScaledDot (BlockScalingRewriter) for pre-swizzled weights

GPUXLA on GPUXLA on GPU

Status: Open.

#41201 In openxla/xla;

· loupicaaa opened on Apr 20, 2026