[llvm] 1cc5290 - [AMDGPU] Factor out getNumUsedPhysRegs(). NFC. (#112624)

via llvm-commits llvm-commits at lists.llvm.org
Thu Oct 17 00:47:23 PDT 2024


Author: Stanislav Mekhanoshin
Date: 2024-10-17T00:47:19-07:00
New Revision: 1cc5290a30a0d6dffeb2e0f475558fcf3ded8e1f

URL: https://github.com/llvm/llvm-project/commit/1cc5290a30a0d6dffeb2e0f475558fcf3ded8e1f
DIFF: https://github.com/llvm/llvm-project/commit/1cc5290a30a0d6dffeb2e0f475558fcf3ded8e1f.diff

LOG: [AMDGPU] Factor out getNumUsedPhysRegs(). NFC. (#112624)

I will need it from one more place.

Added: 
    

Modified: 
    llvm/lib/Target/AMDGPU/AMDGPUResourceUsageAnalysis.cpp
    llvm/lib/Target/AMDGPU/SIRegisterInfo.cpp
    llvm/lib/Target/AMDGPU/SIRegisterInfo.h

Removed: 
    


################################################################################
diff  --git a/llvm/lib/Target/AMDGPU/AMDGPUResourceUsageAnalysis.cpp b/llvm/lib/Target/AMDGPU/AMDGPUResourceUsageAnalysis.cpp
index 1ee3c40d69a3b3..9087442caf8d0b 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPUResourceUsageAnalysis.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPUResourceUsageAnalysis.cpp
@@ -146,44 +146,10 @@ AMDGPUResourceUsageAnalysis::analyzeResourceUsage(
   // count easily.
   // A tail call isn't considered a call for MachineFrameInfo's purposes.
   if (!FrameInfo.hasCalls() && !FrameInfo.hasTailCall()) {
-    MCPhysReg HighestVGPRReg = AMDGPU::NoRegister;
-    for (MCPhysReg Reg : reverse(AMDGPU::VGPR_32RegClass.getRegisters())) {
-      if (MRI.isPhysRegUsed(Reg)) {
-        HighestVGPRReg = Reg;
-        break;
-      }
-    }
-
-    if (ST.hasMAIInsts()) {
-      MCPhysReg HighestAGPRReg = AMDGPU::NoRegister;
-      for (MCPhysReg Reg : reverse(AMDGPU::AGPR_32RegClass.getRegisters())) {
-        if (MRI.isPhysRegUsed(Reg)) {
-          HighestAGPRReg = Reg;
-          break;
-        }
-      }
-      Info.NumAGPR = HighestAGPRReg == AMDGPU::NoRegister
-                         ? 0
-                         : TRI.getHWRegIndex(HighestAGPRReg) + 1;
-    }
-
-    MCPhysReg HighestSGPRReg = AMDGPU::NoRegister;
-    for (MCPhysReg Reg : reverse(AMDGPU::SGPR_32RegClass.getRegisters())) {
-      if (MRI.isPhysRegUsed(Reg)) {
-        HighestSGPRReg = Reg;
-        break;
-      }
-    }
-
-    // We found the maximum register index. They start at 0, so add one to get
-    // the number of registers.
-    Info.NumVGPR = HighestVGPRReg == AMDGPU::NoRegister
-                       ? 0
-                       : TRI.getHWRegIndex(HighestVGPRReg) + 1;
-    Info.NumExplicitSGPR = HighestSGPRReg == AMDGPU::NoRegister
-                               ? 0
-                               : TRI.getHWRegIndex(HighestSGPRReg) + 1;
-
+    Info.NumVGPR = TRI.getNumUsedPhysRegs(MRI, AMDGPU::VGPR_32RegClass);
+    Info.NumExplicitSGPR = TRI.getNumUsedPhysRegs(MRI, AMDGPU::SGPR_32RegClass);
+    if (ST.hasMAIInsts())
+      Info.NumAGPR = TRI.getNumUsedPhysRegs(MRI, AMDGPU::AGPR_32RegClass);
     return Info;
   }
 

diff  --git a/llvm/lib/Target/AMDGPU/SIRegisterInfo.cpp b/llvm/lib/Target/AMDGPU/SIRegisterInfo.cpp
index de9cbe403ab618..8de16974a3e79b 100644
--- a/llvm/lib/Target/AMDGPU/SIRegisterInfo.cpp
+++ b/llvm/lib/Target/AMDGPU/SIRegisterInfo.cpp
@@ -3851,3 +3851,12 @@ SIRegisterInfo::getSubRegAlignmentNumBits(const TargetRegisterClass *RC,
   }
   return 0;
 }
+
+unsigned
+SIRegisterInfo::getNumUsedPhysRegs(const MachineRegisterInfo &MRI,
+                                   const TargetRegisterClass &RC) const {
+  for (MCPhysReg Reg : reverse(RC.getRegisters()))
+    if (MRI.isPhysRegUsed(Reg))
+      return getHWRegIndex(Reg) + 1;
+  return 0;
+}

diff  --git a/llvm/lib/Target/AMDGPU/SIRegisterInfo.h b/llvm/lib/Target/AMDGPU/SIRegisterInfo.h
index 99fa632c0300be..e12a41371c7fd7 100644
--- a/llvm/lib/Target/AMDGPU/SIRegisterInfo.h
+++ b/llvm/lib/Target/AMDGPU/SIRegisterInfo.h
@@ -457,6 +457,11 @@ class SIRegisterInfo final : public AMDGPUGenRegisterInfo {
   // No check if the subreg is supported by the current RC is made.
   unsigned getSubRegAlignmentNumBits(const TargetRegisterClass *RC,
                                      unsigned SubReg) const;
+
+  // \returns a number of registers of a given \p RC used in a function.
+  // Does not go inside function calls.
+  unsigned getNumUsedPhysRegs(const MachineRegisterInfo &MRI,
+                              const TargetRegisterClass &RC) const;
 };
 
 namespace AMDGPU {


        


More information about the llvm-commits mailing list