Updated Makefile, and introduced a simple decode-to-length benchmark/test tool, requires my psflib library

Author: kode54

Makefile

@@ -1,26 +1,32 @@
 
-CFLAGS = -c -m32
+CFLAGS = -g -c
 
 OBJS_RECOMPILER_32 = r4300/x86/assemble.o r4300/x86/gbc.o r4300/x86/gcop0.o r4300/x86/gcop1.o r4300/x86/gcop1_d.o r4300/x86/gcop1_l.o r4300/x86/gcop1_s.o r4300/x86/gcop1_w.o r4300/x86/gr4300.o r4300/x86/gregimm.o r4300/x86/gspecial.o r4300/x86/gtlb.o r4300/x86/regcache.o r4300/x86/rjump.o
 
 OBJS_RECOMPILER_64 = r4300/x86_64/assemble.o r4300/x86_64/gbc.o r4300/x86_64/gcop0.o r4300/x86_64/gcop1.o r4300/x86_64/gcop1_d.o r4300/x86_64/gcop1_l.o r4300/x86_64/gcop1_s.o r4300/x86_64/gcop1_w.o r4300/x86_64/gr4300.o r4300/x86_64/gregimm.o r4300/x86_64/gspecial.o r4300/x86_64/gtlb.o r4300/x86_64/regcache.o r4300/x86_64/rjump.o
 
-OBJS = ai/ai_controller.o api/callbacks.o main/main.o main/rom.o main/savestates.o main/util.o memory/memory.o pi/cart_rom.o pi/pi_controller.o r4300/cached_interp.o r4300/cp0.o r4300/cp1.o r4300/exception.o r4300/interupt.o r4300/mi_controller.o r4300/pure_interp.o r4300/r4300.o r4300/r4300_core.o r4300/recomp.o r4300/reset.o r4300/tlb.o rdp/rdp_core.o ri/rdram.o ri/rdram_detection_hack.o ri/ri_controller.o rsp/rsp_core.o rsp_hle/alist.o rsp_hle/alist_audio.o rsp_hle/alist_naudio.o rsp_hle/alist_nead.o rsp_hle/audio.o rsp_hle/cicx105.o rsp_hle/hle.o rsp_hle/jpeg.o rsp_hle/memory.o rsp_hle/mp3.o rsp_hle/musyx.o rsp_hle/plugin.o rsp_lle/rsp.o si/cic.o si/game_controller.o si/n64_cic_nus_6105.o si/pif.o si/si_controller.o usf/usf.o usf/resampler.o vi/vi_controller.o $(OBJS_RECOMPILER_32)
+OBJS = ai/ai_controller.o api/callbacks.o main/main.o main/rom.o main/savestates.o main/util.o memory/memory.o pi/cart_rom.o pi/pi_controller.o r4300/cached_interp.o r4300/cp0.o r4300/cp1.o r4300/exception.o r4300/interupt.o r4300/mi_controller.o r4300/pure_interp.o r4300/r4300.o r4300/r4300_core.o r4300/recomp.o r4300/reset.o r4300/tlb.o rdp/rdp_core.o ri/rdram.o ri/rdram_detection_hack.o ri/ri_controller.o rsp/rsp_core.o rsp_hle/alist.o rsp_hle/alist_audio.o rsp_hle/alist_naudio.o rsp_hle/alist_nead.o rsp_hle/audio.o rsp_hle/cicx105.o rsp_hle/hle.o rsp_hle/jpeg.o rsp_hle/memory.o rsp_hle/mp3.o rsp_hle/musyx.o rsp_hle/plugin.o rsp_lle/rsp.o si/cic.o si/game_controller.o si/n64_cic_nus_6105.o si/pif.o si/si_controller.o usf/usf.o usf/resampler.o vi/vi_controller.o $(OBJS_RECOMPILER_64)
 
 OPTS = -O3 -DDYNAREC -I.
 ROPTS = -O3 -DARCH_MIN_SSE2 -I.
 
-all: liblazyusf.a
+all: liblazyusf.a bench
 
 liblazyusf.a : $(OBJS)
 	$(AR) rcs $@ $^
 
+bench : test/bench.o liblazyusf.a
+	$(CC) -g -o $@ $^ ../psflib/libpsflib.a -lz
+
 .c.o:
 	$(CC) $(CFLAGS) $(OPTS) -o $@ $*.c
 
 rsp_lle/rsp.o: rsp_lle/rsp.c
 	$(CC) $(CFLAGS) $(ROPTS) -o $@ $^
 
+test/bench.o: test/bench.c
+	$(CC) $(CFLAGS) $(OPTS) -I../psflib -o $@ $^
+
 clean:
-	rm -f $(OBJS) liblazyusf.a > /dev/null
+	rm -f $(OBJS) liblazyusf.a test/bench.o bench > /dev/null
 

README.md

@@ -3,8 +3,15 @@ This is a half assed attempt to replace Project64 with Mupen64plus, for a hoped
 DONE:
 * Get it working at all
 * Verify old x86_64 recompiler works
-* Convert x86 recompiler
+* Converted x86 recompiler
 
 TODO:
-* Get x86 recompiler working with MSVC 32 bit
 * Get new recompiler converted and working
+
+Usage notes:
+
+If your platform is x86 or x86_64, you may define ARCH_MIN_SSE2 for the rsp.c inside the rsp_lle subdirectory. You may also define DYNAREC and include the contents of either the r4300/x86 or r4300/x86_64 directories, and exclude the r4300/empty_dynarec.c file.
+
+If you are not either of the above architectures, you include r4300/empty_dynarec.c and do not define DYNAREC, in which case the fastest you get is a cached interpreter, which "compiles" blocks of opcode function pointers and their pre-decoded parameters.
+
+In either case, you do not include the contents of the new_dynarec folder, since that code is not ready for use with this library yet.

r4300/x86/rjump.c

@@ -90,7 +90,7 @@ void dyna_start(usf_state_t * state, void *code)
      pop ebp
    }
 #elif defined(__GNUC__) && defined(__i386__)
-  #if defined(__PIC__)
+  #if defined(__PIC__) && !defined(__APPLE__)
     /* for -fPIC (shared libraries) */
     #ifndef __GNUC_PREREQ
     #  if defined __GNUC__ && defined __GNUC_MINOR__

test/bench.c

@@ -0,0 +1,147 @@
+#include <stdlib.h>
+#include <string.h>
+
+#include "../usf/usf.h"
+
+#include <psflib.h>
+
+unsigned char * state = 0;
+
+unsigned int enable_compare = 0;
+unsigned int enable_fifo_full = 0;
+
+unsigned long length_ms = 0;
+
+static void * stdio_fopen( const char * path )
+{
+    return fopen( path, "rb" );
+}
+
+static size_t stdio_fread( void *p, size_t size, size_t count, void *f )
+{
+    return fread( p, size, count, (FILE*) f );
+}
+
+static int stdio_fseek( void * f, int64_t offset, int whence )
+{
+    return fseek( (FILE*) f, offset, whence );
+}
+
+static int stdio_fclose( void * f )
+{
+    return fclose( (FILE*) f );
+}
+
+static long stdio_ftell( void * f )
+{
+    return ftell( (FILE*) f );
+}
+
+static psf_file_callbacks stdio_callbacks =
+{
+    "\\/:",
+    stdio_fopen,
+    stdio_fread,
+    stdio_fseek,
+    stdio_fclose,
+    stdio_ftell
+};
+
+static int usf_loader(void * context, const uint8_t * exe, size_t exe_size,
+                      const uint8_t * reserved, size_t reserved_size)
+{
+    if ( exe && exe_size > 0 ) return -1;
+
+    return usf_upload_section( state, reserved, reserved_size );
+}
+
+#define BORK_TIME 0xC0CAC01A
+
+static unsigned long parse_time_crap(const char *input)
+{
+    unsigned long value = 0;
+    unsigned long multiplier = 1000;
+    const char * ptr = input;
+    unsigned long colon_count = 0;
+    
+    while (*ptr && ((*ptr >= '0' && *ptr <= '9') || *ptr == ':'))
+    {
+        colon_count += *ptr == ':';
+        ++ptr;
+    }
+    if (colon_count > 2) return BORK_TIME;
+    if (*ptr && *ptr != '.' && *ptr != ',') return BORK_TIME;
+    if (*ptr) ++ptr;
+    while (*ptr && *ptr >= '0' && *ptr <= '9') ++ptr;
+    if (*ptr) return BORK_TIME;
+    
+    ptr = strrchr(input, ':');
+    if (!ptr)
+        ptr = input;
+    for (;;)
+    {
+        char * end;
+        if (ptr != input) ++ptr;
+        if (multiplier == 1000)
+        {
+            double temp = strtod(ptr, &end);
+            if (temp >= 60.0) return BORK_TIME;
+            value = (long)(temp * 1000.0f);
+        }
+        else
+        {
+            unsigned long temp = strtoul(ptr, &end, 10);
+            if (temp >= 60 && multiplier < 3600000) return BORK_TIME;
+            value += temp * multiplier;
+        }
+        if (ptr == input) break;
+        ptr -= 2;
+        while (ptr > input && *ptr != ':') --ptr;
+        multiplier *= 60;
+    }
+    
+    return value;
+}
+
+static int usf_info(void * context, const char * name, const char * value)
+{
+    if (!strcasecmp(name, "length") || !strcasecmp(name, "fade"))
+        length_ms += parse_time_crap(value);
+    else if (!strcasecmp(name, "_enablecompare") && *value)
+        enable_compare = 1;
+    else if (!strcasecmp(name, "_enablefifofull") && *value)
+        enable_fifo_full = 1;
+    
+    return 0;
+}
+
+int main(int argc, char ** argv)
+{
+	if ( argc == 2 || argc == 3 )
+	{
+        int32_t sample_rate;
+        
+        state = (unsigned char *) malloc(usf_get_state_size());
+        
+        usf_clear(state);
+ 
+		if ( psf_load( argv[1], &stdio_callbacks, 0x21, usf_loader, 0, usf_info, 0, 1 ) <= 0 )
+            return 1;
+        
+        usf_set_compare(state, enable_compare);
+        usf_set_fifo_full(state, enable_fifo_full);
+        
+        if (argc == 3)
+            usf_set_hle_audio(state, 1);
+        
+        usf_render(state, 0, 0, &sample_rate);
+        
+        usf_render(state, 0, length_ms * sample_rate / 1000, &sample_rate);
+        
+        usf_shutdown(state);
+        
+        free(state);
+	}
+    
+    return 0;
+}