Rebased PNaCl localmods in LLVM to 223109

lib/Target/X86/X86InstrCompiler.td

 //===- X86InstrCompiler.td - Compiler Pseudos and Patterns -*- tablegen -*-===//
 //
 //                     The LLVM Compiler Infrastructure
 //
 // This file is distributed under the University of Illinois Open Source
 // License. See LICENSE.TXT for details.
 //
 //===----------------------------------------------------------------------===//
 //
 // This file describes the various pseudo instructions used by the compiler,
@@ skipping 75 matching lines @@
 
 // The VAARG_64 pseudo-instruction takes the address of the va_list,
 // and places the address of the next argument into a register.
 let Defs = [EFLAGS] in
 def VAARG_64 : I<0, Pseudo,
                  (outs GR64:$dst),
                  (ins i8mem:$ap, i32imm:$size, i8imm:$mode, i32imm:$align),
                  "#VAARG_64 $dst, $ap, $size, $mode, $align",
                  [(set GR64:$dst,
                     (X86vaarg64 addr:$ap, imm:$size, imm:$mode, imm:$align)),
-                  (implicit EFLAGS)]>;
-
+                  (implicit EFLAGS)]>,
+                 Requires<[IsNotNaCl]>;
 // Dynamic stack allocation yields a _chkstk or _alloca call for all Windows
 // targets.  These calls are needed to probe the stack when allocating more than
 // 4k bytes in one go. Touching the stack at 4K increments is necessary to
 // ensure that the guard pages used by the OS virtual memory manager are
 // allocated in correct sequence.
 // The main point of having separate instruction are extra unmodelled effects
 // (compared to ordinary calls) like stack pointer change.
 
 let Defs = [EAX, ESP, EFLAGS], Uses = [ESP] in
   def WIN_ALLOCA : I<0, Pseudo, (outs), (ins),
@@ skipping 330 matching lines @@
 def TLS_addr64 : I<0, Pseudo, (outs), (ins i64mem:$sym),
                    "# TLS_addr64",
                   [(X86tlsaddr tls64addr:$sym)]>,
                   Requires<[In64BitMode]>;
 def TLS_base_addr64 : I<0, Pseudo, (outs), (ins i64mem:$sym),
                    "# TLS_base_addr64",
                   [(X86tlsbaseaddr tls64baseaddr:$sym)]>,
                   Requires<[In64BitMode]>;
 }
 
+// @LOCALMOD-BEGIN
+// NaCl TLS support
+let usesCustomInserter = 1 in {
+  def THREAD_POINTER_FROM_GS :
+    I<0, Pseudo, (outs GR32:$dst), (ins),
+      "# get thread pointer from %gs:0",
+      [(set GR32:$dst, (X86thread_pointer_from_gs))]>;
+}
+// @LOCALMOD-END
+
 // Darwin TLS Support
 // For i386, the address of the thunk is passed on the stack, on return the
 // address of the variable is in %eax.  %ecx is trashed during the function
 // call.  All other registers are preserved.
 let Defs = [EAX, ECX, EFLAGS],
     Uses = [ESP],
     usesCustomInserter = 1 in
 def TLSCall_32 : I<0, Pseudo, (outs), (ins i32mem:$sym),
                 "# TLSCall_32",
                 [(X86TLSCall addr:$sym)]>,
@@ skipping 559 matching lines @@
 def : Pat<(i64 (X86Wrapper tglobaltlsaddr :$dst)),
           (MOV64ri32 tglobaltlsaddr :$dst)>;
 // This corresponds to add $foo@tpoff, %rax
 def : Pat<(add GR64:$src1, (X86Wrapper tglobaltlsaddr :$dst)),
           (ADD64ri32 GR64:$src1, tglobaltlsaddr :$dst)>;
 
 
 // Direct PC relative function call for small code model. 32-bit displacement
 // sign extended to 64-bit.
 def : Pat<(X86call (i64 tglobaladdr:$dst)),
-          (CALL64pcrel32 tglobaladdr:$dst)>;
+          (CALL64pcrel32 tglobaladdr:$dst)>, Requires<[IsNotNaCl]>;
 def : Pat<(X86call (i64 texternalsym:$dst)),
-          (CALL64pcrel32 texternalsym:$dst)>;
+          (CALL64pcrel32 texternalsym:$dst)>, Requires<[IsNotNaCl]>;
 
 // Tailcall stuff. The TCRETURN instructions execute after the epilog, so they
 // can never use callee-saved registers. That is the purpose of the GR64_TC
 // register classes.
 //
 // The only volatile register that is never used by the calling convention is
 // %r11. This happens when calling a vararg function with 6 arguments.
 //
 // Match an X86tcret that uses less than 7 volatile registers.
 def X86tcret_6regs : PatFrag<(ops node:$ptr, node:$off),
                              (X86tcret node:$ptr, node:$off), [{
   // X86tcret args: (*chain, ptr, imm, regs..., glue)
   unsigned NumRegs = 0;
   for (unsigned i = 3, e = N->getNumOperands(); i != e; ++i)
     if (isa<RegisterSDNode>(N->getOperand(i)) && ++NumRegs > 6)
       return false;
   return true;
 }]>;
 
 def : Pat<(X86tcret ptr_rc_tailcall:$dst, imm:$off),
           (TCRETURNri ptr_rc_tailcall:$dst, imm:$off)>,
           Requires<[Not64BitMode]>;
 
 // FIXME: This is disabled for 32-bit PIC mode because the global base
 // register which is part of the address mode may be assigned a
 // callee-saved register.
 def : Pat<(X86tcret (load addr:$dst), imm:$off),
           (TCRETURNmi addr:$dst, imm:$off)>,
-          Requires<[Not64BitMode, IsNotPIC]>;
+          Requires<[Not64BitMode, IsNotPIC, IsNotNaCl]>; // @LOCALMOD
 
 def : Pat<(X86tcret (i32 tglobaladdr:$dst), imm:$off),
           (TCRETURNdi texternalsym:$dst, imm:$off)>,
           Requires<[Not64BitMode]>;
 
 def : Pat<(X86tcret (i32 texternalsym:$dst), imm:$off),
           (TCRETURNdi texternalsym:$dst, imm:$off)>,
           Requires<[Not64BitMode]>;
 
 def : Pat<(X86tcret ptr_rc_tailcall:$dst, imm:$off),
           (TCRETURNri64 ptr_rc_tailcall:$dst, imm:$off)>,
           Requires<[In64BitMode]>;
 
 // Don't fold loads into X86tcret requiring more than 6 regs.
 // There wouldn't be enough scratch registers for base+index.
 def : Pat<(X86tcret_6regs (load addr:$dst), imm:$off),
           (TCRETURNmi64 addr:$dst, imm:$off)>,
-          Requires<[In64BitMode]>;
+          Requires<[In64BitMode, IsNotNaCl]>;
 
 def : Pat<(X86tcret (i64 tglobaladdr:$dst), imm:$off),
           (TCRETURNdi64 tglobaladdr:$dst, imm:$off)>,
-          Requires<[In64BitMode]>;
+          Requires<[In64BitMode, IsNotNaCl]>;
 
 def : Pat<(X86tcret (i64 texternalsym:$dst), imm:$off),
           (TCRETURNdi64 texternalsym:$dst, imm:$off)>,
-          Requires<[In64BitMode]>;
+          Requires<[In64BitMode, IsNotNaCl]>;
 
 // Normal calls, with various flavors of addresses.
 def : Pat<(X86call (i32 tglobaladdr:$dst)),
           (CALLpcrel32 tglobaladdr:$dst)>;
 def : Pat<(X86call (i32 texternalsym:$dst)),
           (CALLpcrel32 texternalsym:$dst)>;
 def : Pat<(X86call (i32 imm:$dst)),
           (CALLpcrel32 imm:$dst)>, Requires<[CallImmAddr]>;
 
 // Comparisons.
@@ skipping 427 matching lines @@
             (MOVZX32_NOREXrr8
               (EXTRACT_SUBREG (i16 (COPY_TO_REGCLASS GR16:$src, GR16_ABCD)),
                               sub_8bit_hi)),
             sub_32bit)>;
 
 // h-register extract and store.
 def : Pat<(store (i8 (trunc_su (srl_su GR64:$src, (i8 8)))), addr:$dst),
           (MOV8mr_NOREX
             addr:$dst,
             (EXTRACT_SUBREG (i64 (COPY_TO_REGCLASS GR64:$src, GR64_ABCD)),
-                            sub_8bit_hi))>;
+                            sub_8bit_hi))>, Requires<[IsNotNaCl]>; // @LOCALMOD
 def : Pat<(store (i8 (trunc_su (srl_su GR32:$src, (i8 8)))), addr:$dst),
           (MOV8mr_NOREX
             addr:$dst,
             (EXTRACT_SUBREG (i32 (COPY_TO_REGCLASS GR32:$src, GR32_ABCD)),
                             sub_8bit_hi))>,
-      Requires<[In64BitMode]>;
+      Requires<[In64BitMode, IsNotNaCl]>; // @LOCALMOD
 def : Pat<(store (i8 (trunc_su (srl_su GR16:$src, (i8 8)))), addr:$dst),
           (MOV8mr_NOREX
             addr:$dst,
             (EXTRACT_SUBREG (i16 (COPY_TO_REGCLASS GR16:$src, GR16_ABCD)),
                             sub_8bit_hi))>,
-      Requires<[In64BitMode]>;
+      Requires<[In64BitMode, IsNotNaCl]>; // @LOCALMOD
 
 
 // (shl x, 1) ==> (add x, x)
 // Note that if x is undef (immediate or otherwise), we could theoretically
 // end up with the two uses of x getting different values, producing a result
 // where the least significant bit is not 0. However, the probability of this
 // happening is considered low enough that this is officially not a
 // "real problem".
 def : Pat<(shl GR8 :$src1, (i8 1)), (ADD8rr  GR8 :$src1, GR8 :$src1)>;
 def : Pat<(shl GR16:$src1, (i8 1)), (ADD16rr GR16:$src1, GR16:$src1)>;
@@ skipping 298 matching lines @@
 def : Pat<(cttz_zero_undef GR64:$src), (BSF64rr GR64:$src)>;
 def : Pat<(cttz_zero_undef (loadi16 addr:$src)), (BSF16rm addr:$src)>;
 def : Pat<(cttz_zero_undef (loadi32 addr:$src)), (BSF32rm addr:$src)>;
 def : Pat<(cttz_zero_undef (loadi64 addr:$src)), (BSF64rm addr:$src)>;
 
 // When HasMOVBE is enabled it is possible to get a non-legalized
 // register-register 16 bit bswap. This maps it to a ROL instruction.
 let Predicates = [HasMOVBE] in {
  def : Pat<(bswap GR16:$src), (ROL16ri GR16:$src, (i8 8))>;
 }