compiler.cfg.two-operand: sometimes we can eliminate a copy in the x = y <op> y case

basis/compiler/cfg/two-operand/two-operand-tests.factor

@@ -35,3 +35,18 @@ compiler.cfg.registers cpu.architecture namespaces tools.test ;
         T{ ##sub-float f 1 2 3 }
     } (convert-two-operand)
 ] unit-test
+
+[
+    V{
+        T{ ##copy f 1 2 double-float-rep }
+        T{ ##mul-float f 1 1 1 }
+    }
+] [
+    H{
+        { 1 double-float-rep }
+        { 2 double-float-rep }
+    } clone representations set
+    {
+        T{ ##mul-float f 1 2 2 }
+    } (convert-two-operand)
+] unit-test

basis/compiler/cfg/two-operand/two-operand.factor

@@ -48,8 +48,10 @@ GENERIC: convert-two-operand* ( insn -- )
 
 M: two-operand-insn convert-two-operand*
     [ [ dst>> ] [ src1>> ] bi emit-copy ]
-    [ dup dst>> >>src1 , ]
+    [
-    bi ;
+        dup [ src1>> ] [ src2>> ] bi = [ dup dst>> >>src2 ] when
+        dup dst>> >>src1 ,
+    ] bi ;
 
 M: ##not convert-two-operand*
     [ [ dst>> ] [ src>> ] bi emit-copy ]