Split the Add, Sub, and Mul instruction opcodes into separate

integer and floating-point opcodes, introducing
FAdd, FSub, and FMul.

For now, the AsmParser, BitcodeReader, and IRBuilder all preserve
backwards compatability, and the Core LLVM APIs preserve backwards
compatibility for IR producers. Most front-ends won't need to change
immediately.

This implements the first step of the plan outlined here:
http://nondot.org/sabre/LLVMNotes/IntegerOverflow.txt


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@72897 91177308-0d34-0410-b5e6-96231b3b80d8

test/Assembler/2002-04-07-HexFloatConstants.ll

@@ -11,6 +11,6 @@
 ; RUN: diff %t.1 %t.2
 
 define double @test() {
-        %tmp = mul double 7.200000e+101, 0x427F4000             ; <double> [#uses=1]
+        %tmp = fmul double 7.200000e+101, 0x427F4000             ; <double> [#uses=1]
         ret double %tmp
 }

test/Assembler/2002-04-07-InfConstant.ll

@@ -3,7 +3,7 @@
 ; RUN: llvm-as < %s | llvm-dis | llvm-as | llvm-dis | grep 0x7FF0000000000000
 
 define float @test() {
-        %tmp = mul float 0x7FF0000000000000, 1.000000e+01               ; <float> [#uses=1]
+        %tmp = fmul float 0x7FF0000000000000, 1.000000e+01               ; <float> [#uses=1]
         ret float %tmp
 }