[Mesa-dev] [PATCH] Revert "radeon/llvm: enable unsafe math for graphics shaders"

[Grigori Goronzy]

Am 2015-02-18 09:13, schrieb Michel Dänzer:
> On 18.02.2015 16:52, Grigori Goronzy wrote:
>> Hi,
>> 
>> AFAIR not enabling this makes LLVM generate really slow code in some
>> common cases. Maybe this is just a bug in LLVM/R600 triggered by 
>> unsafe
>> FP math optimization or some optimization is too eager. Other drivers 
>> do
>> fine with these types of optimization.
> 
> It can be enabled again after fixing the problem exposed by The Talos
> Principle.
> 

Right. I just want to avoid the situation that this workaround gets 
committed and then forgotten about. I guess it's up to me then. :)

> 
>> On 17.02.2015 09:15, Michel Dänzer wrote:
>>> From: Michel Dänzer <michel.daenzer at amd.com>
>>> 
>>> This reverts commit 0e9cdedd2e3943bdb7f3543a3508b883b167e427.
>>> 
>>> It caused the grass to disappear in The Talos Principle.
>>> 
>>> Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=89069
>>> Signed-off-by: Michel Dänzer <michel at daenzer.net>
>>> ---
>>>  src/gallium/drivers/radeon/radeon_llvm_emit.c | 4 ----
>>>  1 file changed, 4 deletions(-)
>>> 
>>> diff --git a/src/gallium/drivers/radeon/radeon_llvm_emit.c 
>>> b/src/gallium/drivers/radeon/radeon_llvm_emit.c
>>> index 0f9dbab..624077c 100644
>>> --- a/src/gallium/drivers/radeon/radeon_llvm_emit.c
>>> +++ b/src/gallium/drivers/radeon/radeon_llvm_emit.c
>>> @@ -80,10 +80,6 @@ void radeon_llvm_shader_type(LLVMValueRef F, 
>>> unsigned type)
>>>  	sprintf(Str, "%1d", llvm_type);
>>> 
>>>  	LLVMAddTargetDependentFunctionAttr(F, "ShaderType", Str);
>>> -
>>> -	if (type != TGSI_PROCESSOR_COMPUTE) {
>>> -		LLVMAddTargetDependentFunctionAttr(F, "unsafe-fp-math", "true");
>>> -	}
>>>  }
>>> 
>>>  static void init_r600_target()
>>>