diff --git a/Ghidra/Features/Decompiler/src/decompile/cpp/subflow.cc b/Ghidra/Features/Decompiler/src/decompile/cpp/subflow.cc
index 42d185d7c8..c88d3bdf49 100644
--- a/Ghidra/Features/Decompiler/src/decompile/cpp/subflow.cc
+++ b/Ghidra/Features/Decompiler/src/decompile/cpp/subflow.cc
@@ -731,10 +731,15 @@ bool SubvariableFlow::traceBackward(ReplaceVarnode *rvn)
       addNewConstant(rop,0,(uintb)0);
       return true;
     }
-    if ((newmask<<sa) != rvn->mask)
-      break;			// subvariable is truncated by shift
-    rop = createOp(CPUI_COPY,1,rvn);
-    if (!createLink(rop,newmask,0,op->getIn(0))) return false;
+    if ((newmask<<sa) == rvn->mask) {
+      rop = createOp(CPUI_COPY,1,rvn);
+      if (!createLink(rop,newmask,0,op->getIn(0))) return false;
+      return true;
+    }
+    if ((rvn->mask & 1)==0) return false; // Can't assume zeroes are shifted into least sig bits
+    rop = createOp(CPUI_INT_LEFT,2,rvn);
+    if (!createLink(rop,rvn->mask,0,op->getIn(0))) return false;
+    addConstant(rop,calc_mask(op->getIn(1)->getSize()),1,op->getIn(1)); // Preserve the shift amount
     return true;
   case CPUI_INT_RIGHT:
     if (!op->getIn(1)->isConstant()) break; // Dynamic shift
diff --git a/Ghidra/Framework/SoftwareModeling/data/languages/compiler_spec.rxg b/Ghidra/Framework/SoftwareModeling/data/languages/compiler_spec.rxg
index 37d7842e45..8ba5b316b6 100644
--- a/Ghidra/Framework/SoftwareModeling/data/languages/compiler_spec.rxg
+++ b/Ghidra/Framework/SoftwareModeling/data/languages/compiler_spec.rxg
@@ -406,6 +406,11 @@
         <optional>
           <attribute name="stackspill"/>
         </optional>
+        <optional>
+          <attribute name="backfill">
+            <ref name="boolean_type"/>
+          </attribute>
+        </optional>
       </element>
       <element name="hidden_return">
         <optional>
diff --git a/Ghidra/Processors/ARM/data/languages/ARM.cspec b/Ghidra/Processors/ARM/data/languages/ARM.cspec
index a7fcff8138..c116a64d37 100644
--- a/Ghidra/Processors/ARM/data/languages/ARM.cspec
+++ b/Ghidra/Processors/ARM/data/languages/ARM.cspec
@@ -58,6 +58,30 @@
         <pentry minsize="1" maxsize="4" metatype="float">
           <register name="s7"/>
         </pentry>
+        <pentry minsize="1" maxsize="4" metatype="float">
+          <register name="s8"/>
+        </pentry>
+        <pentry minsize="1" maxsize="4" metatype="float">
+          <register name="s9"/>
+        </pentry>
+        <pentry minsize="1" maxsize="4" metatype="float">
+          <register name="s10"/>
+        </pentry>
+        <pentry minsize="1" maxsize="4" metatype="float">
+          <register name="s11"/>
+        </pentry>
+        <pentry minsize="1" maxsize="4" metatype="float">
+          <register name="s12"/>
+        </pentry>
+        <pentry minsize="1" maxsize="4" metatype="float">
+          <register name="s13"/>
+        </pentry>
+        <pentry minsize="1" maxsize="4" metatype="float">
+          <register name="s14"/>
+        </pentry>
+        <pentry minsize="1" maxsize="4" metatype="float">
+          <register name="s15"/>
+        </pentry>
         <pentry minsize="1" maxsize="4" extension="inttype">
           <register name="r0"/>
         </pentry>
@@ -73,17 +97,92 @@
         <pentry minsize="1" maxsize="500" align="4">
           <addr offset="0" space="stack"/>
         </pentry>
+        <!-- Variadic calls do not use floating-point registers -->
+        <rule>
+          <datatype name="float"/>
+          <varargs/>
+          <join align="true"/>
+        </rule>
+        <!-- Homogeneous float aggregates become regular structs in variadic calls -->
+        <rule>
+          <datatype name="homogeneous-float-aggregate"/>
+          <varargs/>
+          <join align="true"/>
+        </rule>
+        <rule>
+          <datatype name="homogeneous-float-aggregate"/>
+          <join storage="float" align="true" stackspill="false"/>
+        </rule>
+        <rule>
+          <datatype name="homogeneous-float-aggregate"/>
+          <goto_stack/>                	<!-- Don't consume general purpose registers -->
+          <consume_extra storage="float"/> <!-- Once the stack has been used, don't go back to registers -->
+        </rule>
+        <rule>
+          <datatype name="float"/>
+          <join storage="float" align="true" backfill="true" stackspill="false"/>
+        </rule>
+        <rule>
+          <datatype name="float"/>
+          <goto_stack/>					<!-- Don't consume general purpose registers -->
+          <consume_extra storage="float"/> <!-- Once the stack has been used, don't go back to registers -->
+        </rule>
+        <rule>
+          <datatype name="any"/>
+          <join align="true"/>          <!-- Chunk from general purpose registers -->
+        </rule>
       </input>
       <output>
-        <pentry minsize="1" maxsize="8" metatype="float">
-          <register name="d0"/>
+        <pentry minsize="1" maxsize="4" metatype="float">
+          <register name="s0"/>
+        </pentry>
+        <pentry minsize="1" maxsize="4" metatype="float">
+          <register name="s1"/>
+        </pentry>
+        <pentry minsize="1" maxsize="4" metatype="float">
+          <register name="s2"/>
+        </pentry>
+        <pentry minsize="1" maxsize="4" metatype="float">
+          <register name="s3"/>
+        </pentry>
+        <pentry minsize="1" maxsize="4" metatype="float">
+          <register name="s4"/>
+        </pentry>
+        <pentry minsize="1" maxsize="4" metatype="float">
+          <register name="s5"/>
+        </pentry>
+        <pentry minsize="1" maxsize="4" metatype="float">
+          <register name="s6"/>
+        </pentry>
+        <pentry minsize="1" maxsize="4" metatype="float">
+          <register name="s7"/>
         </pentry>
         <pentry minsize="1" maxsize="4" extension="inttype">
           <register name="r0"/>
         </pentry>
-        <pentry minsize="5" maxsize="8">
-          <addr space="join" piece1="r1" piece2="r0"/>
+        <pentry minsize="1" maxsize="4" extension="inttype">
+          <register name="r1"/>
         </pentry>
+        <rule>
+          <datatype name="homogeneous-float-aggregate"/>
+          <join storage="float"/>
+        </rule>
+        <rule>
+          <datatype name="float"/>
+          <join storage="float"/>
+        </rule>
+        <rule>
+          <datatype name="struct" minsize="5"/>
+          <hidden_return/>
+        </rule>
+        <rule>
+          <datatype name="union" minsize="5"/>
+          <hidden_return/>
+        </rule>
+        <rule>
+          <datatype name="any"/>
+          <join/>
+        </rule>
       </output>
       <unaffected>
         <register name="r4"/>
@@ -103,9 +202,15 @@
         <register name="d14"/>
         <register name="d15"/>
         <register name="sp"/>
+        <register name="lr"/>
+        <register name="pc"/>
       </unaffected>
       <killedbycall>
+          <register name="r0"/>
           <register name="r1"/>
+          <register name="r2"/>
+          <register name="r3"/>
+          <register name="r12"/>
           <register name="d0"/>
           <register name="d1"/>
           <register name="d2"/>
@@ -117,6 +222,69 @@
       </killedbycall>
     </prototype>
   </default_proto>
+  <prototype name="__stdcall_softfp" extrapop="0" stackshift="0">
+  <!-- For binaries without hardware floating-point support (-mfloat-abi=soft),
+       or binaries with soft-float compatible interfaces (-mfloat-abi=softfp) -->
+      <input>
+        <pentry minsize="1" maxsize="4">
+          <register name="r0"/>
+        </pentry>
+        <pentry minsize="1" maxsize="4">
+          <register name="r1"/>
+        </pentry>
+        <pentry minsize="1" maxsize="4">
+          <register name="r2"/>
+        </pentry>
+        <pentry minsize="1" maxsize="4">
+          <register name="r3"/>
+        </pentry>
+        <pentry minsize="1" maxsize="500" align="4">
+          <addr offset="0" space="stack"/>
+        </pentry>
+        <rule>
+          <datatype name="any"/>
+          <join align="true"/>          <!-- Chunk from general purpose registers -->
+        </rule>
+      </input>
+      <output>
+        <pentry minsize="1" maxsize="4" extension="inttype">
+          <register name="r0"/>
+        </pentry>
+       <pentry minsize="1" maxsize="4" extension="inttype">
+          <register name="r1"/>
+        </pentry>
+        <rule>
+          <datatype name="struct" minsize="5"/>
+          <hidden_return/>
+        </rule>
+        <rule>
+          <datatype name="union" minsize="5"/>
+          <hidden_return/>
+        </rule>
+        <rule>
+          <datatype name="any"/>
+          <join/>
+        </rule>
+      </output>
+      <unaffected>
+        <register name="r4"/>
+        <register name="r5"/>
+        <register name="r6"/>
+        <register name="r7"/>
+        <register name="r8"/>
+        <register name="r9"/>
+        <register name="r10"/>
+        <register name="r11"/>
+        <register name="sp"/>
+      </unaffected>
+      <killedbycall>
+          <register name="r0"/>
+          <register name="r1"/>
+          <register name="r2"/>
+          <register name="r3"/>
+          <register name="r12"/>
+      </killedbycall>
+    </prototype>
   <prototype name="processEntry" extrapop="0" stackshift="0">
     <input pointermax="4">
       <pentry minsize="1" maxsize="4">
diff --git a/Ghidra/Processors/ARM/data/languages/ARM_win.cspec b/Ghidra/Processors/ARM/data/languages/ARM_win.cspec
index 79e4532da5..c9e0c331df 100644
--- a/Ghidra/Processors/ARM/data/languages/ARM_win.cspec
+++ b/Ghidra/Processors/ARM/data/languages/ARM_win.cspec
@@ -37,7 +37,105 @@
   <default_proto>
     <prototype name="__stdcall" extrapop="0" stackshift="0">
       <input>
-        <!-- we cannot accurately model the allocation scheme when parameters are larger than 4 bytes -->
+        <pentry minsize="1" maxsize="4" metatype="float">
+          <register name="s0"/>
+        </pentry>
+        <pentry minsize="1" maxsize="4" metatype="float">
+          <register name="s1"/>
+        </pentry>
+        <pentry minsize="1" maxsize="4" metatype="float">
+          <register name="s2"/>
+        </pentry>
+        <pentry minsize="1" maxsize="4" metatype="float">
+          <register name="s3"/>
+        </pentry>
+        <pentry minsize="1" maxsize="4" metatype="float">
+          <register name="s4"/>
+        </pentry>
+        <pentry minsize="1" maxsize="4" metatype="float">
+          <register name="s5"/>
+        </pentry>
+        <pentry minsize="1" maxsize="4" metatype="float">
+          <register name="s6"/>
+        </pentry>
+        <pentry minsize="1" maxsize="4" metatype="float">
+          <register name="s7"/>
+        </pentry>
+        <pentry minsize="1" maxsize="4" metatype="float">
+          <register name="s8"/>
+        </pentry>
+        <pentry minsize="1" maxsize="4" metatype="float">
+          <register name="s9"/>
+        </pentry>
+        <pentry minsize="1" maxsize="4" metatype="float">
+          <register name="s10"/>
+        </pentry>
+        <pentry minsize="1" maxsize="4" metatype="float">
+          <register name="s11"/>
+        </pentry>
+        <pentry minsize="1" maxsize="4" metatype="float">
+          <register name="s12"/>
+        </pentry>
+        <pentry minsize="1" maxsize="4" metatype="float">
+          <register name="s13"/>
+        </pentry>
+        <pentry minsize="1" maxsize="4" metatype="float">
+          <register name="s14"/>
+        </pentry>
+        <pentry minsize="1" maxsize="4" metatype="float">
+          <register name="s15"/>
+        </pentry>
+        <pentry minsize="1" maxsize="4" extension="inttype">
+          <register name="r0"/>
+        </pentry>
+        <pentry minsize="1" maxsize="4" extension="inttype">
+          <register name="r1"/>
+        </pentry>
+        <pentry minsize="1" maxsize="4" extension="inttype">
+          <register name="r2"/>
+        </pentry>
+        <pentry minsize="1" maxsize="4" extension="inttype">
+          <register name="r3"/>
+        </pentry>
+        <pentry minsize="1" maxsize="500" align="4">
+          <addr offset="0" space="stack"/>
+        </pentry>
+		<!-- Variadic functions do not use floating-point registers -->
+        <rule>
+          <datatype name="float"/>
+          <varargs/>
+          <join align="true"/>
+        </rule>
+        <!-- Homogeneous float aggregates become regular structs in variadic calls -->
+        <rule>
+          <datatype name="homogeneous-float-aggregate"/>
+          <varargs/>
+          <join align="true"/>
+        </rule>
+        <rule>
+          <datatype name="homogeneous-float-aggregate"/>
+          <join storage="float" align="true" stackspill="false"/>
+        </rule>
+        <rule>
+          <datatype name="homogeneous-float-aggregate"/>
+          <goto_stack/>                	<!-- Don't consume general purpose registers -->
+		  <consume_extra storage="float"/> <!-- Once the stack has been used, don't go back to registers -->
+        </rule>
+        <rule>
+          <datatype name="float"/>
+          <join storage="float" align="true" backfill="true" stackspill="false"/>
+        </rule>
+        <rule>
+          <datatype name="float"/>
+          <goto_stack/>					<!-- Don't consume general purpose registers -->
+          <consume_extra storage="float"/> <!-- Once the stack has been used, don't go back to registers -->
+        </rule>
+        <rule>
+          <datatype name="any"/>
+          <join align="true"/>          <!-- Chunk from general purpose registers -->
+        </rule>
+      </input>
+      <output>
         <pentry minsize="1" maxsize="4" metatype="float">
           <register name="s0"/>
         </pentry>
@@ -68,26 +166,26 @@
         <pentry minsize="1" maxsize="4" extension="inttype">
           <register name="r1"/>
         </pentry>
-        <pentry minsize="1" maxsize="4" extension="inttype">
-          <register name="r2"/>
-        </pentry>
-        <pentry minsize="1" maxsize="4" extension="inttype">
-          <register name="r3"/>
-        </pentry>
-        <pentry minsize="1" maxsize="500" align="4">
-          <addr offset="0" space="stack"/>
-        </pentry>
-      </input>
-      <output>
-        <pentry minsize="1" maxsize="16" metatype="float">
-          <register name="q0"/>
-        </pentry>
-        <pentry minsize="1" maxsize="4" extension="inttype">
-          <register name="r0"/>
-        </pentry>
-        <pentry minsize="5" maxsize="8">
-          <addr space="join" piece1="r1" piece2="r0"/>
-        </pentry>
+        <rule>
+          <datatype name="homogeneous-float-aggregate"/>
+          <join storage="float"/>
+        </rule>
+        <rule>
+          <datatype name="float"/>
+          <join storage="float"/>
+        </rule>
+        <rule>
+          <datatype name="struct" minsize="5"/>
+          <hidden_return/>
+        </rule>
+        <rule>
+          <datatype name="union" minsize="5"/>
+          <hidden_return/>
+        </rule>
+        <rule>
+          <datatype name="any"/>
+          <join/>
+        </rule>
       </output>
       <unaffected>
         <register name="r4"/>