Skip to content

uadk: add the empty size for the hash table row size#751

Merged
Liulongfang merged 5 commits intoLinaro:developfrom
tomismyfriend:develop
Feb 9, 2026
Merged

uadk: add the empty size for the hash table row size#751
Liulongfang merged 5 commits intoLinaro:developfrom
tomismyfriend:develop

Conversation

@tomismyfriend
Copy link

Adjusted the rehash descriptors counta_vld, agg_col_bit_map, Agg_Oid, Agg_Out_Type, Col_Data_Type, and Col_Data_Info. These descriptors are consistent with those generated by the hash aggregation task. In addition, an extra 4 bytes are added when calculating the row size to ensure that each hash table contains 4 bytes of empty information.

Adjusted the rehash descriptors counta_vld, agg_col_bit_map,
Agg_Oid, Agg_Out_Type, Col_Data_Type, and Col_Data_Info.
These descriptors are consistent with those generated by the
hash aggregation task. In addition, an extra 4 bytes are added
when calculating the row size to ensure that each hash table
contains 4 bytes of empty information.

Signed-off-by: Wenkai Lin <linwenkai6@hisilicon.com>
Signed-off-by: Zhushuai Yin <yinzhushuai@huawei.com>
@lin755
Copy link
Collaborator

lin755 commented Feb 5, 2026

lgtm

1 similar comment
@zhushuai12
Copy link

lgtm

Zongyu Wu and others added 4 commits February 5, 2026 17:17
Add error warning when CRC errors occur;When using the same
ctx, the context data of the previous service flow that has
ended needs to be cleared;An error message is added to report
related information to zip module;The minimum output length
of the lz77_zstd_price algorithm should be 4096+16+800+insize.

Signed-off-by: Zongyu Wu <wuzongyu1@huawei.com>
Signed-off-by: Chenghai Huang <huangchenghai2@huawei.com>
When the sgl pool is busy, the hisi_qm_get_hw_sgl function
returns an error, causing the operation to fail.
Now, this function returns the code -WD_EBUSY to inform
the user to wait until the sgl pool is available again.

Signed-off-by: Wenkai Lin <linwenkai6@hisilicon.com>
1.In the original approach, using sched_getcpu() followed by
numa_node_of_cpu() requires two system calls, resulting in low
efficiency.By adopting the new getcpu() method, only one system
call is needed, and in some cases, the information can even be
directly obtained from process data without any system call.
2.Use getcpu() to directly obtain the node id,instead of first
obtaining the cpu id and then the node id, to reduce the number
of system calls.

Signed-off-by: Longfang Liu <liulongfang@huawei.com>
Signed-off-by: Weili Qian <qianweili@huawei.com>
Set the fd for soft ctx to avoid requesting reserved memory.

Signed-off-by: Weili Qian <qianweili@huawei.com>
Copy link
Collaborator

@Liulongfang Liulongfang left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

looks good to me

@Liulongfang Liulongfang merged commit 1ac16bf into Linaro:develop Feb 9, 2026
2 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

4 participants