Mailing-List: contact reviews-help@spark.apache.org; run by ezmlm
Precedence: bulk
From: hvanhovell <git@git.apache.org>
To: reviews@spark.apache.org
Reply-To: reviews@spark.apache.org
References: <git-pr-19222-spark@git.apache.org>
In-Reply-To: <git-pr-19222-spark@git.apache.org>
Subject: [GitHub] spark issue #19222: [SPARK-10399][CORE][SQL] Introduce multiple MemoryBlocks...
Content-Type: text/plain
Message-Id: <20170922140609.135AEF5615@git1-us-west.apache.org>
Date: Fri, 22 Sep 2017 14:06:09 +0000 (UTC)
archived-at: Fri, 22 Sep 2017 14:06:11 -0000

Github user hvanhovell commented on the issue:

    https://github.com/apache/spark/pull/19222
  
    Circling back to the inheritance discussion. My worry is that this will introduce a lot polymorphic call sites in very performance critical code. Even if you tag on final to each method, the call site will still be monomorphic. There are two options from my end:
    
    1. Convince me wrong, by posting benchmark before and after benchmark results.
    2. Use a single class that combines the various implementations. This is very easy to do and will have relatively little overhead.
    
    Also cc @rednaxelafx
    
    In all I really think this work is worth the effort. It should make working with Spark internals a lot easier/safer.


---

---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org