Yes, there was an error. Scott Jones helped me fix it. I am working with 
Randy to make sure the updated version is the one that's re-blogged. Sorry 
for the error!

On Wednesday, June 8, 2016 at 1:14:44 PM UTC-7, Andre Bieler wrote:
>
> "Switching Float to Int in `test2()` makes the vector of arrays faster 
> than `test1()` and `test*3*()`"  that is in my previous post. sorry for 
> the typo...
>
> On Wednesday, June 8, 2016 at 4:08:52 PM UTC-4, Andre Bieler wrote:
>>
>> I am somewhat confused.
>>
>> When reading the blog post on:
>>
>>
>> http://www.juliabloggers.com/using-julias-type-system-for-hidden-performance-gains/
>>
>>
>> the first benchmark says that: "*Notice that the vector of matrices is 
>> 300x slower on an i7 computer.*"
>> Note that the Matrices in this vector (`test2()`) are holding elements of 
>> type float, while for the other (`test1()`, `test3()`) it is ints.
>>
>> Switching Float to Int in `test2()` makes the vector of arrays faster 
>> than `test1()` and `test2()`. Which is consistent with the findings of the 
>> same author in his original blog post.
>>
>>  
>> http://www.stochasticlifestyle.com/using-julias-type-system-hidden-performance-chunkedarrays-growablearrays-ellipsisnotation/
>>  
>> <http://www.stochasticlifestyle.com/using-julias-type-system-hidden-performance-chunkedarrays-growablearrays-ellipsisnotation/>
>>
>> sooo...? :)
>>
>> this is on julia 0.4.5 for my own benchmarks, which agree with the 
>> results from the second link.
>>
>

Reply via email to