Skip site navigation (1)Skip section navigation (2) 
Date:      Mon, 30 Sep 2019 19:24:21 +0000 (UTC)
From:      Dimitry Andric <dim@FreeBSD.org>
To:        src-committers@freebsd.org, svn-src-all@freebsd.org, svn-src-head@freebsd.org
Subject:   svn commit: r352915 - head/contrib/llvm/lib/Target/X86
Message-ID:  <201909301924.x8UJOLFI041312@repo.freebsd.org>
next in thread | raw e-mail | index | archive | help 
Author: dim
Date: Mon Sep 30 19:24:21 2019
New Revision: 352915
URL: https://svnweb.freebsd.org/changeset/base/352915

Log:
  Pull in r357528 from upstream llvm trunk (by Craig Topper):
  
    [X86] Check MI.isConvertibleTo3Addr() before calling
    convertToThreeAddress in X86FixupLEAs.
  
    X86FixupLEAs just assumes convertToThreeAddress will return nullptr
    for any instruction that isn't convertible.
  
    But the code in convertToThreeAddress for X86 assumes that any
    instruction coming in has at least 2 operands and that the second one
    is a register. But those properties aren't guaranteed of all
    instructions. We should check the instruction property first.
  
  Pull in r365720 from upstream llvm trunk (by Craig Topper):
  
    [X86] Don't convert 8 or 16 bit ADDs to LEAs on Atom in FixupLEAPass.
  
    We use the functions that convert to three address to do the
    conversion, but changing an 8 or 16 bit will cause it to create a
    virtual register. This can't be done after register allocation where
    this pass runs.
  
    I've switched the pass completely to a white list of instructions
    that can be converted to LEA instead of a blacklist that was
    incorrect. This will avoid surprises if we enhance the three address
    conversion function to include additional instructions in the future.
  
    Fixes PR42565.
  
  This should fix assertions/segfaults when compiling certain ports with
  CPUTYPE=atom.
  
  PR:		240928
  MFC after:	3 days

Modified:
  head/contrib/llvm/lib/Target/X86/X86FixupLEAs.cpp

Modified: head/contrib/llvm/lib/Target/X86/X86FixupLEAs.cpp
==============================================================================
--- head/contrib/llvm/lib/Target/X86/X86FixupLEAs.cpp	Mon Sep 30 19:05:14 2019	(r352914)
+++ head/contrib/llvm/lib/Target/X86/X86FixupLEAs.cpp	Mon Sep 30 19:24:21 2019	(r352915)
@@ -154,6 +154,15 @@ FixupLEAPass::postRAConvertToLEA(MachineFunction::iter
     MFI->insert(MBBI, NewMI); // Insert the new inst
     return NewMI;
   }
+  }
+
+  if (!MI.isConvertibleTo3Addr())
+    return nullptr;
+
+  switch (MI.getOpcode()) {
+  default:
+    // Only convert instructions that we've verified are safe.
+    return nullptr;
   case X86::ADD64ri32:
   case X86::ADD64ri8:
   case X86::ADD64ri32_DB:
@@ -162,24 +171,24 @@ FixupLEAPass::postRAConvertToLEA(MachineFunction::iter
   case X86::ADD32ri8:
   case X86::ADD32ri_DB:
   case X86::ADD32ri8_DB:
-  case X86::ADD16ri:
-  case X86::ADD16ri8:
-  case X86::ADD16ri_DB:
-  case X86::ADD16ri8_DB:
     if (!MI.getOperand(2).isImm()) {
       // convertToThreeAddress will call getImm()
       // which requires isImm() to be true
       return nullptr;
     }
     break;
-  case X86::ADD16rr:
-  case X86::ADD16rr_DB:
-    if (MI.getOperand(1).getReg() != MI.getOperand(2).getReg()) {
-      // if src1 != src2, then convertToThreeAddress will
-      // need to create a Virtual register, which we cannot do
-      // after register allocation.
-      return nullptr;
-    }
+  case X86::SHL64ri:
+  case X86::SHL32ri:
+  case X86::INC64r:
+  case X86::INC32r:
+  case X86::DEC64r:
+  case X86::DEC32r:
+  case X86::ADD64rr:
+  case X86::ADD64rr_DB:
+  case X86::ADD32rr:
+  case X86::ADD32rr_DB:
+    // These instructions are all fine to convert.
+    break;
   }
   return TII->convertToThreeAddress(MFI, MI, nullptr);
 }

Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?201909301924.x8UJOLFI041312>