Skip to content

feat(vpto): add sequential soft post-update analysis #1018

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
yexiaosu wants to merge 3 commits into
base: main
Choose a base branch
from
Open

feat(vpto): add sequential soft post-update analysis #1018

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
716f624
feat(vpto): add sequential soft post-update analysis
yexiaosu Jul 28, 2026
5dbb060
Preserve i32 offsets in soft post-update
yexiaosu Jul 28, 2026
3e10509
Fix sequential run restart handling
yexiaosu Jul 28, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
251 changes: 173 additions & 78 deletions docs/designs/vpto-soft-postupdate-design-zh.md
View file
Open in desktop

Large diffs are not rendered by default.

9 changes: 4 additions & 5 deletions include/PTO/Transforms/Passes.td
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1180,11 +1180,10 @@ def VPTOSoftPostUpdate
: Pass<"vpto-soft-postupdate", "ModuleOp"> {
let summary = "Convert fixed-stride VPTO memory ops to post-update form";
let description = [{
Identifies VPTO load/store operations (vlds, vsts, vsstb) inside scf.for
loops whose effective address advances by a loop-invariant stride per
iteration, and rewrites them to use the post-update (updated_base) result.
The base pointer is carried through scf.for iter_args, eliminating explicit
address arithmetic in the loop body.
Identifies VPTO load/store operations (vlds, vsts, vsstb) whose effective
address advances by a fixed symbolic stride. Loop-carried accesses use
scf.for iter_args; remaining arithmetic runs within each block, including
loop bodies, are rewritten as updated_base pointer chains.
}];
let constructor = "mlir::pto::createVPTOSoftPostUpdatePass()";
let dependentDialects = ["mlir::func::FuncDialect",
Expand Down
Loading
Loading