[RISCV] Add thread-pointer relaxations for local-exec TLS model#1534
Open
Jonathon Penix (jonathonpenix) wants to merge 3 commits into
Open
[RISCV] Add thread-pointer relaxations for local-exec TLS model#1534Jonathon Penix (jonathonpenix) wants to merge 3 commits into
Jonathon Penix (jonathonpenix) wants to merge 3 commits into
Conversation
Implemented as defined in the psABI at [1] Broadly, local-exec TLS accesses may be relaxed from the full sequence (ex: lui/add/lw) to a single tp-relative instruction (ex: a tp-relative lw) if the offset is within +-2KB. We check for offset < 2048 as I don't think a negative tp offset is meaningful in "normal" operation. While here, I'm also removing the usage of applyTprelAdd--as far as I'm aware there are no other psABI-defined relaxations around this relocation, so this should be obsolete. A few misc. items to note: * This can misrelax the sequence if part is marked relaxable and part is not. I'm not sure there's a great way to prevent this that works in general and this is a) in line with some other RISC-V relaxations in eld and b) is consistent with what lld and bfd do. So I didn't worry about trying to handle this. * You can also see weird things happen if you write (probably erroneous) assembly with different addends. As far as I can tell, eld is relaxing such sequences consistently with lld/bfd, so I left it be. * I *think* it should be possible to change the tp-relative offset *after* the relaxation happens, causing a mislink. And I don't think we have any error checking that would catch these cases currently. This should be consistent with ex: the TLSDESC short-form relaxations, so I'm hoping to address this as a follow-up patch. [1] https://github.com/riscv-non-isa/riscv-elf-psabi-doc/blob/e03d44ae2f0e1144f9498c2896b5ae25b0449398/riscv-elf.adoc?plain=1#L2531 Signed-off-by: Jonathon Penix <jpenix@qti.qualcomm.com>
Signed-off-by: Jonathon Penix <jpenix@qti.qualcomm.com>
Signed-off-by: Jonathon Penix <jpenix@qti.qualcomm.com>
Jonathon Penix (jonathonpenix)
marked this pull request as ready for review
July 20, 2026 23:43
Jonathon Penix (jonathonpenix)
requested a review
from Shankar Easwaran (quic-seaswara)
as a code owner
July 20, 2026 23:43
Contributor
Author
|
Just linking this comment/request for changes from Shankar: #1428 (comment) Will come back to/update this accordingly |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Implemented as defined in the psABI at [1]
Broadly, local-exec TLS accesses may be relaxed from the full sequence (ex: lui/add/lw) to a single tp-relative instruction (ex: a tp-relative lw) if the offset is within +-2KB. We check for offset < 2048 as I don't think a negative tp offset is meaningful in "normal" operation.
While here, I'm also removing the usage of applyTprelAdd--as far as I'm aware there are no other psABI-defined relaxations around this relocation, so this should be obsolete.
A few misc. items to note:
[1] https://github.com/riscv-non-isa/riscv-elf-psabi-doc/blob/e03d44ae2f0e1144f9498c2896b5ae25b0449398/riscv-elf.adoc?plain=1#L2531