RFR (L) JDK-8195100: Use a low latency hashtable for SymbolTable

[Gerard Ziemski]

> On Aug 2, 2018, at 10:40 AM, Ioi Lam <ioi.lam at oracle.com> wrote:
> 
> 
> 
> On 8/2/18 8:26 AM, Gerard Ziemski wrote:
>> 
>>> symbolTable.cpp:
>>> 
>>>  342   // Otherwise, add the symbol to table. Copy to a C string first.
>>>  343   char stack_buf[ON_STACK_BUFFER_LENGTH];
>>>  344   ResourceMark rm(THREAD);
>>>  345   if (len <= ON_STACK_BUFFER_LENGTH) {
>>>  346     buffer = stack_buf;
>>>  347   } else {
>>>  348     buffer = NEW_RESOURCE_ARRAY_IN_THREAD(THREAD, char, len);
>>>  349   }
>>>  350   for (int i=0; i<len; i++) {
>>>  351     buffer[i] = name[i];
>>>  352   }
>>>  353   // Make sure there is no safepoint in the code above since name can't move.
>>>  354   // We can't include the code in NoSafepointVerifier because of the
>>>  355   // ResourceMark.
>>> 
>>> The above is dead code and should be removed.
>> That’s not dead code, but in fact you did find an issue here - we should be using the “buffer”, not “name” when we add to the table. Thanks!
>> 
>> 
> Why do you need to make a copy here? buffer and name seems to have the exact same contents.

I didn’t write that code, but I assume the name comes in some temp memory that can go away at any time, so we need a solid copy.


cheers