[Mesa-dev] [PATCH] glsl/cse: Maintain a list of free ae_entry objects

[Ian Romanick]

On 03/27/2015 11:33 AM, Kenneth Graunke wrote:
> On Thursday, March 26, 2015 08:52:33 PM Ian Romanick wrote:
>> From: Ian Romanick <ian.d.romanick at intel.com>
>>
>> The CSE algorithm will continuously allocate new ae_entry objects.  As
>> each new basic block is exited, all of the previously allocated objects
>> are dumped.  Instead, put them in a free list and re-use them in the
>> next basic block.  Reduce, reuse, recycle!
>>
>> Signed-off-by: Ian Romanick <ian.d.romanick at intel.com>
>> Cc: Jordan Justen <jordan.l.justen at intel.com>
>>
>> ---
>>  src/glsl/opt_cse.cpp | 63 
> +++++++++++++++++++++++++++++++++++++++++++++-------
>>  1 file changed, 55 insertions(+), 8 deletions(-)
>>
>> diff --git a/src/glsl/opt_cse.cpp b/src/glsl/opt_cse.cpp
>> index 623268e..425eebc 100644
>> --- a/src/glsl/opt_cse.cpp
>> +++ b/src/glsl/opt_cse.cpp
>> @@ -63,6 +63,17 @@ public:
>>        var = NULL;
>>     }
>>  
>> +   void init(ir_instruction *base_ir, ir_rvalue **val)
>> +   {
>> +      this->val = val;
>> +      this->base_ir = base_ir;
>> +      this->var = NULL;
>> +
>> +      assert(val);
>> +      assert(*val);
>> +      assert(base_ir);
>> +   }
>> +
>>     /**
>>      * The pointer to the expression that we might be able to reuse
>>      *
>> @@ -116,6 +127,18 @@ private:
>>     ir_rvalue *try_cse(ir_rvalue *rvalue);
>>     void add_to_ae(ir_rvalue **rvalue);
>>  
>> +   /**
>> +    * Move all nodes from the ae list to the free list
>> +    */
>> +   void empty_ae_list();
>> +
>> +   /**
>> +    * Get and initialize a new ae_entry
>> +    *
>> +    * This will either come from the free list or be freshly allocated.
>> +    */
>> +   ae_entry *get_ae_entry(ir_rvalue **rvalue);
>> +
>>     /** List of ae_entry: The available expressions to reuse */
>>     exec_list *ae;
>>  
>> @@ -126,6 +149,11 @@ private:
>>      * right.
>>      */
>>     exec_list *validate_instructions;
>> +
>> +   /**
>> +    * List of available-for-use ae_entry objects.
>> +    */
>> +   exec_list free_ae_entries;
>>  };
>>  
>>  /**
>> @@ -322,6 +350,25 @@ cse_visitor::try_cse(ir_rvalue *rvalue)
>>     return NULL;
>>  }
>>  
>> +void
>> +cse_visitor::empty_ae_list()
>> +{
>> +   free_ae_entries.append_list(ae);
> 
> If you're just trying to save memory...it sure looks like you can just
> ralloc_free(mem_ctx); mem_ctx = ralloc_context(NULL) here.  Which would
> be much simpler.
> 
> But I suppose this does reduce the number of malloc calls considerably,
> so it's probably more efficient.

It was a bit of both.  The original code was clearly "doing it wrong,"
and using an object pool seemed like the obvious solution.  Doing an
extra ralloc_free and ralloc_context here will increase the time spent
in ralloc, and that doesn't seem desirable.

Now, if I really thought CSE was a significant performance bottleneck,
I'd change get_ae_entry to use pop_tail to get the most recently added
(and likely cache-hot) ae_entry.  I'd probably also keep the ae_entry
list across calls and free them all at the end of the optimization loop.
 It doesn't seem worth the effort.

There are a bunch of places throughout Mesa that could benefit from an
object pool.  What may be worth the effort is making a general mechanism
that could be used in all the places.  That couldn't be more than a half
day of effort.

>> +}
>> +
>> +ae_entry *
>> +cse_visitor::get_ae_entry(ir_rvalue **rvalue)
>> +{
>> +   ae_entry *entry = (ae_entry *) free_ae_entries.pop_head();
>> +   if (entry) {
>> +      entry->init(base_ir, rvalue);
>> +   } else {
>> +      entry = new(mem_ctx) ae_entry(base_ir, rvalue);
>> +   }
>> +
>> +   return entry;
>> +}
>> +


-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 181 bytes
Desc: OpenPGP digital signature
URL: <http://lists.freedesktop.org/archives/mesa-dev/attachments/20150327/554ef7e2/attachment.sig>