Fix two small bugs and update the Readme for release.

Author: Pascal Beyer

README.md

@@ -28,6 +28,7 @@ local functions, etc, but all of these should be considered in the "design"-stag
 
 See the [release page](https://github.com/PascalBeyer/Headerless-C-Compiler/releases) for pre-build binaries. To compile from source invoke the `build_msvc.bat` with `cl.exe`,
 `link.exe` and `ucrt.lib` in your path (for example, from an [x64 Native Tools Command Prompt](https://learn.microsoft.com/en-us/cpp/build/building-on-the-command-line?view=msvc-170&viewFallbackFrom=vs-2019)).
+For a release build of the compiler you can use `build_msvc.bat release`.
 
 To use standard library or Windows library functions it is necessary to install a version of the [Windows SDK](https://developer.microsoft.com/en-us/windows/downloads/windows-sdk/).
 
@@ -191,9 +192,7 @@ by re-iterating the syntax tree of all declarations and functions that made it i
 
 ## Current State and Next Steps
 
-At this point, most C-code that compiles using MSVC also compiles using `hlc`. Notable exceptions include code, which uses
-[`__declspec(allocate("<section>"))`](https://learn.microsoft.com/en-us/cpp/cpp/allocate?view=msvc-170)
-or [structured exception handling](https://learn.microsoft.com/en-us/windows/win32/debug/structured-exception-handling).
+At this point, most C-code that compiles using MSVC also compiles using `hlc`.
 
 Code-wise, most of the C-components (`preprocess.c`, `parse.c`, `emit_x64.c`) are hardened through test suites and fuzz-testing.
 Some larger refactors are on the "eventually"-list, like separating out the concept of `tokens` and `preprocessor` tokens.

build_msvc.bat

@@ -8,7 +8,11 @@ IF NOT EXIST build mkdir build
 set warning_options= -Wall -wd4201 -wd4820  -wd4711 -wd4200 -wd4204 -wd4214 -wd4324 -wd5045 -wd4061 -wd4221 -wd4366
 :: ********************************************************************************************************
 
-set compiler_flags= -Od -nologo -fp:fast -GR- -EHa- -Zo -Oi -D_Debug  %warning_options% -FC -Z7 -GS- -Gs9999999 /diagnostics:column /TC
+set optimization= -Od
+
+if "%1" == "release" set optimization= -O2
+
+set compiler_flags= %optimization% -nologo -fp:fast -GR- -EHa- -Zo -Oi -D_Debug  %warning_options% -FC -Z7 -GS- -Gs9999999 /diagnostics:column /TC
 
 set linker_flags= -HEAP:0,0 -STACK:0x100000,0x100000 -incremental:no -opt:ref Advapi32.lib kernel32.lib ucrt.lib /NODEFAULTLIB /ENTRY:_start /SUBSYSTEM:console /NOLOGO /DEBUG 
 

examples/tinycc.bat

@@ -5,7 +5,7 @@ cd tinycc\win32 || exit /b 1
 git clean -xdf
 
 set TCC_C=..\tcc.c
-build-tcc.bat -c hlc.exe -t 64 || exit /b 1
+call build-tcc.bat -c hlc.exe -t 64 || exit /b 1
 
 cd ..\..
 

src/emit_x64.c

@@ -1440,7 +1440,7 @@ func struct emit_location *emit_load_into_specific_gpr(struct context *context,
 func struct emit_location *emit_load_float_into_specific_register(struct context *context, struct emit_location *loc, enum register_encoding reg){
 
 
-    if(loc->state == EMIT_LOCATION_loaded && loc->loaded_register == reg) return loc;
+    if(loc->state == EMIT_LOCATION_loaded && loc->loaded_register == reg && loc->register_kind == REGISTER_KIND_xmm) return loc;
 
     struct prefixes sse_prefix;
     if(loc->size == 32){
@@ -1452,6 +1452,11 @@ func struct emit_location *emit_load_float_into_specific_register(struct context
     allocate_specific_register(context, REGISTER_KIND_xmm, reg);
     struct emit_location *ret = emit_location_loaded(context, REGISTER_KIND_xmm, reg, loc->size);
 
+    // @cleanup: This is stupid, we should rework the back-end to keep track of the register_kind again.
+    if(loc->state == EMIT_LOCATION_loaded && loc->register_kind != REGISTER_KIND_xmm){
+        spill_register(context, REGISTER_KIND_gpr, loc->loaded_register);
+    }
+    
     switch(loc->state){
         case EMIT_LOCATION_loaded:{
             assert(loc->register_kind == REGISTER_KIND_xmm);

src/main.c

@@ -4866,7 +4866,9 @@ globals.typedef_##postfix = (struct ast_type){
 
     // Ensure that all thread wrote in `thread_infos[thread_index].context`.
     for(u32 thread_index = 1; thread_index < thread_count; thread_index++){
-        while(atomic_load(struct context *, thread_infos[thread_index].context) == null){};
+        while(atomic_load(struct context *, thread_infos[thread_index].context) == null){
+            _mm_pause();
+        };
     }
 
     assert(globals.work_queue_tokenize_files.work_entries_in_flight == 0);

tests/compile/returning_assignment_of_double.c

@@ -0,0 +1,10 @@
+
+double function_returning_double(double number){
+    double a;
+    return a = number;
+}
+
+
+int main(){
+    return function_returning_double(0);
+}