CANN: optimize rope cache

[hipudding]

Make sure to read the contributing guidelines before submitting a PR

[Copilot]

Pull Request Overview

This PR optimizes the rope cache implementation in the CANN backend by changing the initialization strategy from pre-allocating a fixed-size cache to dynamically allocating memory only when needed.

• Changes the initial max_prompt_length from a hardcoded value of 65536 to 0
• Removes the upfront allocation of sin/cos cache memory
• Adds proper memory deallocation before reallocating when cache needs to grow

Reviewed Changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 2 comments.

File	Description
ggml/src/ggml-cann/common.h	Initialize max_prompt_length to 0 instead of 65536
ggml/src/ggml-cann/aclnn_ops.cpp	Remove upfront cache allocation and add proper cleanup before reallocation

---

_{Tip: Customize your code reviews with copilot-instructions.md. Create the file or learn how to get started.}

[Copilot]

The conditional check for ctx.cos_ptr != nullptr is missing braces around the ACL_CHECK(aclrtFree(ctx.cos_ptr)) call on line 2226. This creates a logic error where the free operation is always executed regardless of the null check.

Suggested change

	ACL_CHECK(aclrtFree(ctx.cos_ptr));
	if (ctx.cos_ptr != nullptr) {
	ACL_CHECK(aclrtFree(ctx.cos_ptr));
	}

[hipudding]

It's fine because there's only one line in if.

[hipudding]

This optimize has already done in #15501