commit c72ece6c2cb92edfd040d92b7fa6ca62bdaa8293
author Konstantin Zhuravlyov <kzhuravl_dev@outlook.com> Wed May 16 20:47:48 2018 +0000
committer Konstantin Zhuravlyov <kzhuravl_dev@outlook.com> Wed May 16 20:47:48 2018 +0000
tree c0c78d4655f32f89fc0c18fbfb2fef0a0a41f791
parent 1f5eb86b51fdc236bd37161b8074da6a93fdea96

AMDGPU : Recalculate SGPRs when trap handler is supported

Differential Revision: https://reviews.llvm.org/D29911

llvm-svn: 332523

llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.cpp

diff --git a/llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.cpp b/llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.cpp
index 74dac65..3896bcc 100644
--- a/llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.cpp
+++ b/llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.cpp
@@ -358,9 +358,11 @@
 
   if (WavesPerEU >= getMaxWavesPerEU(Features))
     return 0;
-  unsigned MinNumSGPRs =
-      alignDown(getTotalNumSGPRs(Features) / (WavesPerEU + 1),
-                getSGPRAllocGranule(Features)) + 1;
+
+  unsigned MinNumSGPRs = getTotalNumSGPRs(Features) / (WavesPerEU + 1);
+  if (Features.test(FeatureTrapHandler))
+    MinNumSGPRs -= std::min(MinNumSGPRs, (unsigned)TRAP_NUM_SGPRS);
+  MinNumSGPRs = alignDown(MinNumSGPRs, getSGPRAllocGranule(Features)) + 1;
   return std::min(MinNumSGPRs, getAddressableNumSGPRs(Features));
 }
 
@@ -369,11 +371,13 @@
   assert(WavesPerEU != 0);
 
   IsaVersion Version = getIsaVersion(Features);
-  unsigned MaxNumSGPRs = alignDown(getTotalNumSGPRs(Features) / WavesPerEU,
-                                   getSGPRAllocGranule(Features));
   unsigned AddressableNumSGPRs = getAddressableNumSGPRs(Features);
   if (Version.Major >= 8 && !Addressable)
     AddressableNumSGPRs = 112;
+  unsigned MaxNumSGPRs = getTotalNumSGPRs(Features) / WavesPerEU;
+  if (Features.test(FeatureTrapHandler))
+    MaxNumSGPRs -= std::min(MaxNumSGPRs, (unsigned)TRAP_NUM_SGPRS);
+  MaxNumSGPRs = alignDown(MaxNumSGPRs, getSGPRAllocGranule(Features));
   return std::min(MaxNumSGPRs, AddressableNumSGPRs);
 }
 

llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.h

diff --git a/llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.h b/llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.h
index 5459ddf..54932d2 100644
--- a/llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.h
+++ b/llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.h
@@ -42,7 +42,8 @@
 enum {
   // The closed Vulkan driver sets 96, which limits the wave count to 8 but
   // doesn't spill SGPRs as much as when 80 is set.
-  FIXED_NUM_SGPRS_FOR_INIT_BUG = 96
+  FIXED_NUM_SGPRS_FOR_INIT_BUG = 96,
+  TRAP_NUM_SGPRS = 16
 };
 
 /// Instruction set architecture version.