[Mesa-dev] [PATCH] llvmpipe: increase fs shader variant instruction cache limit by factor 4

[Roland Scheidegger]

Am 11.10.2013 16:21, schrieb Brian Paul:
> On 10/11/2013 07:11 AM, sroland at vmware.com wrote:
>> From: Roland Scheidegger <sroland at vmware.com>
>>
>> The previous limit of of 128*1024 was reported to cause frequent
>> recompiles
>> in some apps due to shader variant thrashing on IRC in some apps leading
>> to noticeable lags.
>> Note that the LP_MAX_SHADER_VARIANTS limit (1024) was more or less
>> impossible
>> to reach, since even simple fragment shaders without texturing
>> (glxgears) used
>> more than twice than 128 instructions, hence the instruction limit
>> would have
>> always been reached first (excluding things like trivial shaders not
>> writing
>> color). Even with the new limit it is VERY likely the instruction
>> limit is hit
>> first.
>> Should help with such lags due to recompiles (though other shader
>> types have
>> their own limits, LP_MAX_SETUP_VARIANTS and DRAW_MAX_SHADER_VARIANTS, in
>> particular the latter seems a bit small (128)).
>> ---
>>   src/gallium/drivers/llvmpipe/lp_limits.h |    2 +-
>>   1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/src/gallium/drivers/llvmpipe/lp_limits.h
>> b/src/gallium/drivers/llvmpipe/lp_limits.h
>> index af31b35..8cfab5a 100644
>> --- a/src/gallium/drivers/llvmpipe/lp_limits.h
>> +++ b/src/gallium/drivers/llvmpipe/lp_limits.h
>> @@ -79,7 +79,7 @@
>>    * Max number of instructions (for all fragment shaders combined per
>> context)
>>    * that will be kept around.
>>    */
>> -#define LP_MAX_SHADER_INSTRUCTIONS (128*1024)
>> +#define LP_MAX_SHADER_INSTRUCTIONS (512*LP_MAX_SHADER_VARIANTS)
>>
>>   /**
>>    * Max number of setup variants that will be kept around.
>>
> 
> Reviewed-by: Brian Paul <brianp at vmware.com>
> 
> Maybe the comment on LP_MAX_SHADER_INSTRUCTIONS should indicate that
> these are LLVM IR instructions, not TGSI instructions, not machine
> instructions.  I had to dig for a while to find that.

Oh yes that's a good idea. I think ideally we'd just count compiled size
of the shaders but IIRC that was actually difficult to do.

Roland