Use set operations instead of plain lists to enumerate register classes.

This simplifies many of the target description files since it is common
for register classes to be related or contain sequences of numbered
registers.

I have verified that this doesn't change the files generated by TableGen
for ARM and X86. It alters the allocation order of MBlaze GPR and Mips
FGR32 registers, but I believe the change is benign.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133105 91177308-0d34-0410-b5e6-96231b3b80d8
diff --git a/docs/WritingAnLLVMBackend.html b/docs/WritingAnLLVMBackend.html
index 39ac34b..5e3d070 100644
--- a/docs/WritingAnLLVMBackend.html
+++ b/docs/WritingAnLLVMBackend.html
@@ -706,8 +706,7 @@
 <div class="doc_code">
 <pre>
 class RegisterClass&lt;string namespace,
-list&lt;ValueType&gt; regTypes, int alignment,
-                    list&lt;Register&gt; regList&gt; {
+list&lt;ValueType&gt; regTypes, int alignment, dag regList&gt; {
   string Namespace = namespace;
   list&lt;ValueType&gt; RegTypes = regTypes;
   int Size = 0;  // spill size, in bits; zero lets tblgen pick the size
@@ -717,7 +716,7 @@
   // default value 1 means a single instruction
   // A negative value means copying is extremely expensive or impossible
   int CopyCost = 1;  
-  list&lt;Register&gt; MemberList = regList;
+  dag MemberList = regList;
   
   // for register classes that are subregisters of this class
   list&lt;RegisterClass&gt; SubRegClassList = [];  
@@ -749,9 +748,11 @@
     memory.</li>
 
 <li>The final argument, <tt>regList</tt>, specifies which registers are in this
-    class.  If an <tt>allocation_order_*</tt> method is not specified,
-    then <tt>regList</tt> also defines the order of allocation used by the
-    register allocator.</li>
+    class. If an alternative allocation order method is not specified, then
+    <tt>regList</tt> also defines the order of allocation used by the register
+    allocator. Besides simply listing registers with <tt>(add R0, R1, ...)</tt>,
+    more advanced set operators are available. See
+    <tt>include/llvm/Target/Target.td</tt> for more information.</li>
 </ul>
 
 <p>
@@ -761,44 +762,31 @@
 '<tt>SP</tt>'. <tt>FPRegs</tt> defines a group of 32 single-precision
 floating-point registers (<tt>F0</tt> to <tt>F31</tt>); <tt>DFPRegs</tt> defines
 a group of 16 double-precision registers
-(<tt>D0-D15</tt>). For <tt>IntRegs</tt>, the <tt>MethodProtos</tt>
-and <tt>MethodBodies</tt> methods are used by TableGen to insert the specified
-code into generated output.
+(<tt>D0-D15</tt>).
 </p>
 
 <div class="doc_code">
 <pre>
-def FPRegs : RegisterClass&lt;"SP", [f32], 32,
-  [F0, F1, F2, F3, F4, F5, F6, F7, F8, F9, F10, F11, F12, F13, F14, F15,
-   F16, F17, F18, F19, F20, F21, F22, F23, F24, F25, F26, F27, F28, F29, F30, F31]&gt;;
+// F0, F1, F2, ..., F31
+def FPRegs : RegisterClass&lt;"SP", [f32], 32, (sequence "F%u", 0, 31)&gt;;
 
 def DFPRegs : RegisterClass&lt;"SP", [f64], 64,
-  [D0, D1, D2, D3, D4, D5, D6, D7, D8, D9, D10, D11, D12, D13, D14, D15]&gt;;
+                            (add D0, D1, D2, D3, D4, D5, D6, D7, D8,
+                                 D9, D10, D11, D12, D13, D14, D15)&gt;;
 &nbsp;
 def IntRegs : RegisterClass&lt;"SP", [i32], 32,
-    [L0, L1, L2, L3, L4, L5, L6, L7,
-     I0, I1, I2, I3, I4, I5,
-     O0, O1, O2, O3, O4, O5, O7,
-     G1,
-     // Non-allocatable regs:
-     G2, G3, G4, 
-     O6,        // stack ptr
-    I6,        // frame ptr
-     I7,        // return address
-     G0,        // constant zero
-     G5, G6, G7 // reserved for kernel
-    ]&gt; {
-  let MethodProtos = [{
-    iterator allocation_order_end(const MachineFunction &amp;MF) const;
-  }];
-  let MethodBodies = [{
-    IntRegsClass::iterator
-    IntRegsClass::allocation_order_end(const MachineFunction &amp;MF) const {
-      return end() - 10  // Don't allocate special registers
-         -1;
-    }
-  }];
-}
+    (add L0, L1, L2, L3, L4, L5, L6, L7,
+         I0, I1, I2, I3, I4, I5,
+         O0, O1, O2, O3, O4, O5, O7,
+         G1,
+         // Non-allocatable regs:
+         G2, G3, G4,
+         O6,        // stack ptr
+         I6,        // frame ptr
+         I7,        // return address
+         G0,        // constant zero
+         G5, G6, G7 // reserved for kernel
+    )&gt;;
 </pre>
 </div>
 
@@ -820,10 +808,7 @@
 register implementation. The code below shows only the generated integer
 registers and associated register classes. The order of registers
 in <tt>IntRegs</tt> reflects the order in the definition of <tt>IntRegs</tt> in
-the target description file. Take special note of the use
-of <tt>MethodBodies</tt> in <tt>SparcRegisterInfo.td</tt> to create code in
-<tt>SparcGenRegisterInfo.inc</tt>. <tt>MethodProtos</tt> generates similar code
-in <tt>SparcGenRegisterInfo.h.inc</tt>.
+the target description file.
 </p>
 
 <div class="doc_code">
@@ -866,13 +851,7 @@
   static const TargetRegisterClass* const IntRegsSuperclasses [] = {
     NULL
   };
-...
-  IntRegsClass::iterator
-  IntRegsClass::allocation_order_end(const MachineFunction &amp;MF) const {
-     return end()-10  // Don't allocate special registers
-         -1;
-  }
-  
+
   IntRegsClass::IntRegsClass() : TargetRegisterClass(IntRegsRegClassID, 
     IntRegsVTs, IntRegsSubclasses, IntRegsSuperclasses, IntRegsSubRegClasses, 
     IntRegsSuperRegClasses, 4, 4, 1, IntRegs, IntRegs + 32) {}
@@ -880,6 +859,13 @@
 </pre>
 </div>
 
+<p>
+The register allocators will avoid using reserved registers, and callee saved
+registers are not used until all the volatile registers have been used.  That
+is usually good enough, but in some cases it may be necessary to provide custom
+allocation orders.
+</p>
+
 </div>
 
 <!-- ======================================================================= -->