[TRTLLM-10184][chore] Remove legacy XQA precompiled path by pengbowang-nv · Pull Request #14941 · NVIDIA/TensorRT-LLM

pengbowang-nv · 2026-06-04T06:20:14Z

XQA has moved to JIT path since #10335 .

This PR removed the precompile path and related cubins from repo.

Description

Test Coverage

PR Checklist

Please review the following before submitting your PR:

PR description clearly explains what and why. If using CodeRabbit's summary, please make sure it makes sense.
PR Follows TRT-LLM CODING GUIDELINES to the best of your knowledge.
Test cases are provided for new code paths (see test instructions)
If PR introduces API changes, an appropriate PR label is added - either api-compatible or api-breaking. For api-breaking, include BREAKING in the PR title.
Any new dependencies have been scanned for license and vulnerabilities
CODEOWNERS updated if ownership changes
Documentation updated as needed
Update tava architecture diagram if there is a significant design change in PR.
The reviewers assigned automatically/manually are appropriate for the PR.
Please check this after reviewing the above items as appropriate for this PR.

GitHub Bot Help

To see a list of available CI bot commands, please comment /bot help.

Summary by CodeRabbit

Refactor
- Simplified internal XQA JIT compilation logic by removing conditional environment-based control and associated build infrastructure.

pengbowang-nv · 2026-06-04T06:29:28Z

/bot run --disable-fail-fast

tensorrt-cicd · 2026-06-04T06:35:58Z

PR_Github #51992 [ run ] triggered by Bot. Commit: 4640251 Link to invocation

pengbowang-nv · 2026-06-04T14:55:36Z

/bot run --disable-fail-fast

tensorrt-cicd · 2026-06-04T15:01:43Z

PR_Github #52092 [ run ] triggered by Bot. Commit: 4640251 Link to invocation

tensorrt-cicd · 2026-06-04T15:05:47Z

PR_Github #51992 [ run ] completed with state ABORTED. Commit: 4640251

Link to invocation

…gle-GPU test_single_request_chat_multiple_images[pd_disagg-qwen3_30b_a3b_fp8] in test_mm_encoder_standalone.py fails at setup on the pre-merge DGX_B200 single-GPU stage with a NIXL CacheTransceiver init assertion (status == NIXL_SUCCESS, transferAgent.cpp:614) when the pd_disagg LLM constructs its disaggregated KV-cache transfer agent. This is a fleet-wide failure on the single-GPU pre-merge stage, observed across many unrelated PRs (e.g. NVIDIA#13978, NVIDIA#13925, NVIDIA#14841, NVIDIA#14599, NVIDIA#14524, NVIDIA#14941, NVIDIA#14398); it passes only intermittently (~1/3) depending on node. Waiving until the single-GPU NIXL EPD-disagg path is fixed or the variant is gated to multi-GPU. NVBug: https://nvbugs/6269683 Signed-off-by: venkywonka <23023424+venkywonka@users.noreply.github.com>

tensorrt-cicd · 2026-06-05T00:30:46Z

PR_Github #52092 [ run ] completed with state SUCCESS. Commit: 4640251
/LLM/main/L0_MergeRequest_PR pipeline #41422 completed with status: 'FAILURE'

CI Report

⚠️ Action Required:

Please check the failed tests and fix your PR
If you cannot view the failures, ask the CI triggerer to share details
Once fixed, request an NVIDIA team member to trigger CI again

CI Agent Failure Analysis

Link to invocation

pengbowang-nv · 2026-06-05T03:01:30Z

/bot run --disable-fail-fast

tensorrt-cicd · 2026-06-05T03:07:56Z

PR_Github #52223 [ run ] triggered by Bot. Commit: 4640251 Link to invocation

tensorrt-cicd · 2026-06-05T07:42:43Z

PR_Github #52223 [ run ] completed with state SUCCESS. Commit: 4640251
/LLM/main/L0_MergeRequest_PR pipeline #41541 completed with status: 'FAILURE'

CI Report

⚠️ Action Required:

Please check the failed tests and fix your PR
If you cannot view the failures, ask the CI triggerer to share details
Once fixed, request an NVIDIA team member to trigger CI again

CI Agent Failure Analysis

Link to invocation

pengbowang-nv · 2026-06-05T07:43:29Z

/bot run --disable-fail-fast

tensorrt-cicd · 2026-06-05T07:49:41Z

PR_Github #52301 [ run ] triggered by Bot. Commit: 4640251 Link to invocation

tensorrt-cicd · 2026-06-05T12:22:48Z

PR_Github #52301 [ run ] completed with state SUCCESS. Commit: 4640251
/LLM/main/L0_MergeRequest_PR pipeline #41609 completed with status: 'FAILURE'

CI Report

⚠️ Action Required:

Please check the failed tests and fix your PR
If you cannot view the failures, ask the CI triggerer to share details
Once fixed, request an NVIDIA team member to trigger CI again

CI Agent Failure Analysis

Link to invocation

pengbowang-nv · 2026-06-05T14:34:48Z

/bot run --disable-fail-fast

tensorrt-cicd · 2026-06-05T14:40:24Z

PR_Github #52351 [ run ] triggered by Bot. Commit: 4640251 Link to invocation

pengbowang-nv · 2026-06-05T16:21:34Z

/bot run --disable-fail-fast

tensorrt-cicd · 2026-06-05T16:27:15Z

PR_Github #52379 [ run ] triggered by Bot. Commit: 4640251 Link to invocation

tensorrt-cicd · 2026-06-05T16:27:28Z

PR_Github #52351 [ run ] completed with state ABORTED. Commit: 4640251

Link to invocation

tensorrt-cicd · 2026-06-05T21:02:44Z

PR_Github #52379 [ run ] completed with state SUCCESS. Commit: 4640251
/LLM/main/L0_MergeRequest_PR pipeline #41674 completed with status: 'FAILURE'

CI Report

⚠️ Action Required:

Please check the failed tests and fix your PR
If you cannot view the failures, ask the CI triggerer to share details
Once fixed, request an NVIDIA team member to trigger CI again

CI Agent Failure Analysis

Link to invocation

pengbowang-nv · 2026-06-08T12:24:58Z

/bot run --disable-fail-fast

tensorrt-cicd · 2026-06-08T12:30:35Z

PR_Github #52770 [ run ] triggered by Bot. Commit: 51a6af5 Link to invocation

tensorrt-cicd · 2026-06-09T00:27:19Z

PR_Github #52770 [ run ] completed with state SUCCESS. Commit: 51a6af5
/LLM/main/L0_MergeRequest_PR pipeline #42030 completed with status: 'FAILURE'

CI Report

⚠️ Action Required:

Please check the failed tests and fix your PR
If you cannot view the failures, ask the CI triggerer to share details
Once fixed, request an NVIDIA team member to trigger CI again

CI Agent Failure Analysis

Link to invocation

pengbowang-nv · 2026-06-09T03:41:12Z

/bot run --disable-fail-fast

tensorrt-cicd · 2026-06-09T03:46:43Z

PR_Github #52935 [ run ] triggered by Bot. Commit: 51a6af5 Link to invocation

tensorrt-cicd · 2026-06-09T06:10:23Z

PR_Github #52935 [ run ] completed with state SUCCESS. Commit: 51a6af5
/LLM/main/L0_MergeRequest_PR pipeline #42180 completed with status: 'FAILURE'

CI Report

⚠️ Action Required:

Please check the failed tests and fix your PR
If you cannot view the failures, ask the CI triggerer to share details
Once fixed, request an NVIDIA team member to trigger CI again

CI Agent Failure Analysis

Link to invocation

jhaotingc

LGTM and I do agree with @yuxianq 's point of removing the Impl wrapper.
But the scale of the PR would be big and it could be a follow-up PR.
Thanks for the work!

Signed-off-by: Pengbo Wang <221450789+pengbowang-nv@users.noreply.github.com>

pengbowang-nv · 2026-06-10T13:24:21Z

/bot run --disable-fail-fast

tensorrt-cicd · 2026-06-10T13:29:59Z

PR_Github #53313 [ run ] triggered by Bot. Commit: 358191d Link to invocation

pengbowang-nv · 2026-06-11T03:12:01Z

/bot kill

tensorrt-cicd · 2026-06-11T03:17:46Z

PR_Github #53457 [ kill ] triggered by Bot. Commit: 358191d Link to invocation

tensorrt-cicd · 2026-06-11T03:23:37Z

PR_Github #53313 [ run ] completed with state ABORTED. Commit: 358191d

Link to invocation

tensorrt-cicd · 2026-06-11T03:23:42Z

PR_Github #53457 [ kill ] completed with state SUCCESS. Commit: 358191d
Successfully killed previous jobs for commit 358191d

Link to invocation

pengbowang-nv · 2026-06-11T07:57:20Z

/bot run --disable-fail-fast

tensorrt-cicd · 2026-06-11T08:02:54Z

PR_Github #53523 [ run ] triggered by Bot. Commit: 358191d Link to invocation

tensorrt-cicd · 2026-06-11T14:15:45Z

PR_Github #53523 [ run ] completed with state SUCCESS. Commit: 358191d
/LLM/main/L0_MergeRequest_PR pipeline #42680 completed with status: 'FAILURE'

CI Report

⚠️ Action Required:

Please check the failed tests and fix your PR
If you cannot view the failures, ask the CI triggerer to share details
Once fixed, request an NVIDIA team member to trigger CI again

CI Agent Failure Analysis

Link to invocation

pengbowang-nv · 2026-06-11T14:41:45Z

/bot run --disable-fail-fast

tensorrt-cicd · 2026-06-11T14:47:33Z

PR_Github #53590 [ run ] triggered by Bot. Commit: 358191d Link to invocation

tensorrt-cicd · 2026-06-11T16:20:23Z

PR_Github #53590 [ run ] completed with state SUCCESS. Commit: 358191d
/LLM/main/L0_MergeRequest_PR pipeline #42736 completed with status: 'SUCCESS'

CI Report

Link to invocation

github-actions Bot assigned pengbowang-nv Jun 4, 2026

pengbowang-nv changed the title ~~[TRTLLM-10184][chore] remove XQA precompiled legacy path~~ [TRTLLM-10184][chore] Remove legacy XQA precompiled path Jun 4, 2026

pengbowang-nv force-pushed the dev-refactor-remove-xqa-legacy branch from 4669b12 to 4640251 Compare June 4, 2026 06:25

venkywonka mentioned this pull request Jun 4, 2026

[None][test] Waive qwen3_30b_a3b_fp8 pd_disagg mm-encoder test on single-GPU (NIXL setup failure) #14968

Open

pengbowang-nv force-pushed the dev-refactor-remove-xqa-legacy branch from 4640251 to 51a6af5 Compare June 8, 2026 03:27

pengbowang-nv requested a review from yuxianq June 9, 2026 07:48

yuxianq reviewed Jun 9, 2026

View reviewed changes

Comment thread cpp/tensorrt_llm/kernels/decoderMaskedMultiheadAttention/decoderXQAImpl.h Outdated

jhaotingc approved these changes Jun 9, 2026

View reviewed changes

pengbowang-nv added 9 commits June 9, 2026 20:29

remove XQA precompiled legacy path

e9f08f3

Signed-off-by: Pengbo Wang <221450789+pengbowang-nv@users.noreply.github.com>

fix hash key overflow

3b2d7be

Signed-off-by: Pengbo Wang <221450789+pengbowang-nv@users.noreply.github.com>

combine abstract impl define and jit impl

a5e3745

Signed-off-by: Pengbo Wang <221450789+pengbowang-nv@users.noreply.github.com>

split resource and remove unused code

ac26d95

Signed-off-by: Pengbo Wang <221450789+pengbowang-nv@users.noreply.github.com>

remove abstract impl and use just real jit impl

8902a2f

Signed-off-by: Pengbo Wang <221450789+pengbowang-nv@users.noreply.github.com>

fix copyright year

2dacaad

Signed-off-by: Pengbo Wang <221450789+pengbowang-nv@users.noreply.github.com>

remove impl abstraction and directly use runner

dc4ec3f

Signed-off-by: Pengbo Wang <221450789+pengbowang-nv@users.noreply.github.com>

change coding style for torched files

f24c10d

Signed-off-by: Pengbo Wang <221450789+pengbowang-nv@users.noreply.github.com>

fix template

d262572

Signed-off-by: Pengbo Wang <221450789+pengbowang-nv@users.noreply.github.com>

pengbowang-nv force-pushed the dev-refactor-remove-xqa-legacy branch from 51a6af5 to d262572 Compare June 10, 2026 10:15

fix copyright year for trtGptModel.h

358191d

Signed-off-by: Pengbo Wang <221450789+pengbowang-nv@users.noreply.github.com>

yuxianq reviewed Jun 12, 2026

View reviewed changes

Comment thread cpp/tensorrt_llm/kernels/decoderMaskedMultiheadAttention/decoderXQARunnerUtils.h

yuxianq approved these changes Jun 12, 2026

View reviewed changes

pengbowang-nv merged commit 44550bc into NVIDIA:main Jun 12, 2026
7 checks passed

pengbowang-nv deleted the dev-refactor-remove-xqa-legacy branch June 12, 2026 07:58

Uh oh!

Conversation

pengbowang-nv commented Jun 4, 2026 • edited by coderabbitai Bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Description

Test Coverage

PR Checklist

GitHub Bot Help

Summary by CodeRabbit

Uh oh!

pengbowang-nv commented Jun 4, 2026

Uh oh!

tensorrt-cicd commented Jun 4, 2026

Uh oh!

pengbowang-nv commented Jun 4, 2026

Uh oh!

tensorrt-cicd commented Jun 4, 2026

Uh oh!

tensorrt-cicd commented Jun 4, 2026

Uh oh!

tensorrt-cicd commented Jun 5, 2026

Uh oh!

pengbowang-nv commented Jun 5, 2026

Uh oh!

tensorrt-cicd commented Jun 5, 2026

Uh oh!

tensorrt-cicd commented Jun 5, 2026

Uh oh!

pengbowang-nv commented Jun 5, 2026

Uh oh!

tensorrt-cicd commented Jun 5, 2026

Uh oh!

tensorrt-cicd commented Jun 5, 2026

Uh oh!

pengbowang-nv commented Jun 5, 2026

Uh oh!

tensorrt-cicd commented Jun 5, 2026

Uh oh!

pengbowang-nv commented Jun 5, 2026

Uh oh!

tensorrt-cicd commented Jun 5, 2026

Uh oh!

tensorrt-cicd commented Jun 5, 2026

Uh oh!

tensorrt-cicd commented Jun 5, 2026

Uh oh!

pengbowang-nv commented Jun 8, 2026

Uh oh!

tensorrt-cicd commented Jun 8, 2026

Uh oh!

tensorrt-cicd commented Jun 9, 2026

Uh oh!

pengbowang-nv commented Jun 9, 2026

Uh oh!

tensorrt-cicd commented Jun 9, 2026

Uh oh!

tensorrt-cicd commented Jun 9, 2026

Uh oh!

Uh oh!

jhaotingc left a comment

Choose a reason for hiding this comment

Uh oh!

pengbowang-nv commented Jun 10, 2026

Uh oh!

tensorrt-cicd commented Jun 10, 2026

Uh oh!

pengbowang-nv commented Jun 11, 2026

Uh oh!

tensorrt-cicd commented Jun 11, 2026

Uh oh!

tensorrt-cicd commented Jun 11, 2026

Uh oh!

tensorrt-cicd commented Jun 11, 2026

Uh oh!

pengbowang-nv commented Jun 11, 2026

Uh oh!

tensorrt-cicd commented Jun 11, 2026

Uh oh!

tensorrt-cicd commented Jun 11, 2026

Uh oh!

pengbowang-nv commented Jun 4, 2026 •

edited by coderabbitai Bot

Loading