[llvm] [AMDGPU] Add support for point sample accel out of order returns (PR #127991)

[Jay Foad via llvm-commits]

================
@@ -1811,20 +1817,23 @@ def FeatureISAVersion11_5_0 : FeatureSet<
   !listconcat(FeatureISAVersion11_Common.Features,
     [FeatureSALUFloatInsts,
      FeatureDPPSrc1SGPR,
-     FeatureRequiredExportPriority])>;
+     FeatureRequiredExportPriority,
+     FeaturePointSampleAccel])>;
 
 def FeatureISAVersion11_5_1 : FeatureSet<
   !listconcat(FeatureISAVersion11_Common.Features,
     [FeatureSALUFloatInsts,
      FeatureDPPSrc1SGPR,
      Feature1_5xVGPRs,
-     FeatureRequiredExportPriority])>;
+     FeatureRequiredExportPriority,
+     FeaturePointSampleAccel])>;
 
 def FeatureISAVersion11_5_2 : FeatureSet<
   !listconcat(FeatureISAVersion11_Common.Features,
     [FeatureSALUFloatInsts,
      FeatureDPPSrc1SGPR,
-     FeatureRequiredExportPriority])>;
+     FeatureRequiredExportPriority,
+     FeaturePointSampleAccel])>;
 
 def FeatureISAVersion11_5_3 : FeatureSet<
----------------
jayfoad wrote:

It's really not the same for all RDNA3.5 chips. (Sorry for the delayed response, I see you've raised this in #139880 now.)

https://github.com/llvm/llvm-project/pull/127991