Propagate layout with reshape after lowering

[pfultz2]

Motivation

Even though reshape propagates the permutation after lowering its get lowered to contiguous->reshape_lazy, which will always use standard shape. Instead this adds a pass to insert a layout so it will propagate the permutation in reshape_lazy.

Technical Details

Changelog Category

Add a CHANGELOG.md entry for any option other than Not Applicable

• • Added: New functionality.
• • Changed: Changes to existing functionality.
• • Removed: Functionality or support that has been removed. (Compared to a previous release)
• • Optimized: Component performance that has been optimized or improved.
• • Resolved Issues: Known issues from a previous version that have been resolved.
• • Not Applicable: This PR is not to be included in the changelog.

[Copilot]

Pull request overview

This PR adds a GPU backend pass to preserve/propagate non-standard (permuted) layouts through a lowered reshape sequence, addressing cases where eliminate_contiguous keeps a standardizing gpu::contiguous before reshape_lazy and thereby discards the original permutation.

Changes:

• Introduces gpu::propagate_reshape_layout pass that replaces the surviving pre-reshape_lazy standardizing gpu::contiguous with a layout (via gpu::precompile_op) + allocate to repack into the alias-friendly order.
• Hooks the new pass into the GPU backend pipeline immediately after eliminate_contiguous.
• Adds a GPU unit test covering both the permutation-propagation case and a no-op case.

Reviewed changes

Copilot reviewed 5 out of 5 changed files in this pull request and generated no comments.

Show a summary per file

File	Description
test/gpu/propagate_reshape_layout.cpp	Adds coverage to ensure the pass inserts layout when permutation would otherwise be lost, and is a no-op when there’s no permutation to propagate.
src/targets/gpu/target.cpp	Inserts propagate_reshape_layout into the GPU compilation pipeline after eliminate_contiguous.
src/targets/gpu/propagate_reshape_layout.cpp	Implements the matcher-based rewrite from gpu::contiguous -> reshape_lazy into layout -> reshape_lazy for non-standard inputs.
src/targets/gpu/include/migraphx/gpu/propagate_reshape_layout.hpp	Declares the new GPU pass and documents its intent and placement.
src/targets/gpu/CMakeLists.txt	Adds the new pass source file to the GPU library build.

[CharlieL7]

LGTM. Though I'm wondering if there is a way to keep track of memory layout vs. shape. It is difficult to follow what's happening to both.