Vectorize 5 scalar-unrolled primitives in LibMatrixMult using Java Vector API

[Copilot]

Five performance-critical loop primitives in LibMatrixMult were still using manual 8-wide scalar unrolling while the surrounding code already adopted jdk.incubator.vector. This upgrades all five to use DoubleVector.SPECIES_PREFERRED with the same scalar-tail + vector-loop pattern already established in the class.

Changed methods (LibMatrixMult.java)

Method	Operation
vectAdd(double[], double[], int, int, int)	c[i] += a[i]
vectAdd4(double[]×4, double[], int, int, int)	c[i] += a1[i]+a2[i]+a3[i]+a4[i]
vectAddInPlace(double, double[], int, int)	c[i] += scalar (broadcast)
vectMultiplyInPlace(double, double[], int, int)	c[i] *= scalar (broadcast)
vectSubtract(double[], double[], int, int, int)	c[i] -= a[i]

Before (vectAdd as example):

for (int j = bn; j < len; j+=8, ai+=8, ci+=8) {
    c[ci+0] += a[ai+0]; c[ci+1] += a[ai+1];
    // ... 6 more manual lines
}

After:

for (int j = bn; j < len; j+=vLen, ai+=vLen, ci+=vLen) {
    DoubleVector aVec = DoubleVector.fromArray(SPECIES, a, ai);
    DoubleVector cVec = DoubleVector.fromArray(SPECIES, c, ci);
    cVec.add(aVec).intoArray(c, ci);
}

New test class (VectorPrimitivesPerformanceTest)

• Correctness tests for each method covering non-power-of-two lengths and non-zero offsets (edge cases for the scalar tail)
• Throughput tests on 1 M doubles with warm-up, printing per-call latency to CI output for regression tracking

Original prompt

/create-agent implement the vector api for the top five of these. & verify the performance

[Chronological Review: The conversation began with the user requesting an understanding of the basic structure of the project. The user then expressed interest in the Java Vector API, specifically regarding optimizations. The user requested to proceed with this understanding, followed by a command to implement the vector API for the top five optimizations and verify performance. The conversation was ongoing when the token budget was exceeded.]

[Intent Mapping:

1. "understand the basic structure of this project" - User seeks foundational knowledge of the project.
2. "I would like to understand the java vector api for this project. optimizations, possible." - User wants to learn about the Java Vector API and its optimizations.
3. "let's do that please" - User requests to proceed with the exploration of the Java Vector API.
4. "/create-agent implement the vector api for the top five of these. & verify the performance" - User commands the implementation of the vector API for the top five optimizations and performance verification.]

[Technical Inventory:

• Java Vector API: A set of classes and interfaces that provide a mechanism for vectorized operations in Java, aimed at improving performance.
• MemoryEstimates.java: A file in the project that likely deals with memory estimation, possibly relevant to the optimizations being discussed.]

[Code Archaeology:

• Current file: /workspaces/systemds/src/main/java/org/apache/sysds/utils/MemoryEstimates.java - This file is being referenced but no specific code changes or discussions about it have been documented yet.]

[Progress Assessment:

• Completed Tasks: User has expressed understanding of the project structure and interest in the Java Vector API.
• Partially Complete Work: The user has requested the implementation of the vector API and performance verification, which is pending.]

[Context Validation: All critical information for continuation is captured, including user requests and the current focus on the Java Vector API.]

[Recent Commands Analysis:

• Last Agent Commands: "/create-agent implement the vector api for the top five of these. & verify the performance" - This command indicates the user's intention to implement specific optimizations using the Java Vector API.
• Tool Results Summary: No specific tool results were returned as the command was a directive rather than a tool execution.
• Pre-Summary State: The agent was actively working on implementing the vector API for optimizations when the token budget was exceeded.
• Operation Context: The command was executed to align with the user's goal of enhancing performance through the Java Vector API.]

1. Conversation Overview: - Primary Objectives: - "understand the basic structure of this project" - "I would like to understand the java vector api for this project. optimizations, possible." - "let's do that please" - "/create-agent implement the vector api for the top five of these. & verify the performance" - Session Context: The conversation has focused on understanding the project structure and the Java Vector API, culminating in a command to implement optimizations. - User Intent Evolution: The user transitioned from seeking basic knowledge to requesting specific implementations and performance verification.

2. Technical Foundation:

• Java Vector API: A mechanism for vectorized operations in Java aimed at improving performance.

3. Codebase Status:

• File Name: MemoryEstimates.java
  • Purpose: Likely deals with memory estimation relevant to optimizations.
  • Current State: No specific modifications discussed yet.
  • Key Code Segments: Not documented yet.
  • Dependencies: Related to the overall performance optimizations being explored.

4. Problem Resolution:

• Issues Encountered: None documented yet.
• Solutions Implemented: None implemented yet.
• Debugging Context: Ongoing exploration of the Java Vector API.
• Lessons Learned: User is focused on performance optimizations.

5. Progress Tracking:

• Completed Tasks: Understanding of project structure and interest in Java Vector API.
• Partially Complete Work: Implementation of the vector API and performance verification is pending.

6. Active Work State:

• Current Focus: Implementing the vector API for optimizations.
• Recent Context: User has requested to implement the vector API for the top five optimizations and verify performance.
• Working Code: No specific code snippets discussed recently.
• Immediate Context: User's command to implement the vector API.

7. Recent Operations:

• Last Agent Commands: "/create-agent implement the vector api for the top five of these. & verify the performance"
• Tool Results Summary: No specific results returned as it was a command.
• Pre-Summary State: The agent was focused on implementing the vector API for optimizations.
• Operation Context: The command aligns with the ...

Created from VS Code.