Forums

Dear Bernd,

I am still unable to run the optimizations and turn on the inline functions. None of the inline functions contain any stdli function just pure asignments, reads, simple arithmetic and bitwise operations.

In fact, the file containing the main function and those functions for export in AS3 did have memset and memcpy. I tried your suggestion and put the code above the functions calling memset and memcpy. It did not work soe I put the code in a header which is included topmost in each C file. The only system header I use is malloc.h and it is included topmost. In other C file I use pow, sin and log10 from math.h but I removed it and made the same thing:

//shared.h

#ifndef _SHARED_H_

#define _SHARED_H_

#include <malloc.h>

static void * custom_memmove( void * destination, const void * source, unsigned int num ) {

static float custom_pow(float x, int y) {

It still behave the same way as if nothing was changed (works incorrectly - displays jerk which does not move, whereby the image is supposed to move)

As I am porting an emulator (Sega Mega Drive) I use manu arrays of function pointers for implementing the opcodes of the CPU's. Could this be an issue?

I did a workaround for the floating point problem but processing is very slow so I hear only bzzt bzzt but this is for now out of scope. The emulator compiled with gcc runs at 300 fps on a 1.3 GHz machine, whereby my non optimized AVM2 code compiled by alchemy produces 14 fps. The pure rendering is super fast and the problem lies in the computational power of AVM. The frame buffer and the enulation are generated in the C code and only the pixels are copied to AS3, where they are plotted in a BitmapData. On 2.0 GHz Dual core I achieved only 21 fps. Goal is 60 fps to have smooth audio and video. But this is offtopic. After all everything works (slow) without optimization, and I would somehow turn it on. Suggestions?

Here is the file with the main function:

#include "shared.h"
#include "AS3.h"

#define FRAMEBUFFER_LENGTH    (320*240*4)
static uint8* framebuffer;
static uint32  audioSamples;

AS3_Val sega_rom(void* self, AS3_Val args)
{
    int size, offset, i;
    uint8 hardware;
    uint8 country;
    uint8 header[0x200];
    uint8 *ptr;
    AS3_Val length;
    AS3_Val ba;
   
    AS3_ArrayValue(args, "AS3ValType", &ba);

    country = 0;
    offset = 0;

    length = AS3_GetS(ba, "length");
    size = AS3_IntValue(length);

    ptr = (uint8*)malloc(size);

    AS3_SetS(ba, "position", AS3_Int(0));
    AS3_ByteArray_readBytes(ptr, ba, size);
   
    //FILE* f = fopen("boris_dump.bin", "wb");
    //fwrite(ptr, size, 1, f);
    //fclose(f);

    if((size / 512) & 1)
    {
        size -= 512;
        offset += 512;

        memcpy(header, ptr, 512);

        for(i = 0; i < (size / 0x4000); i += 1)
        {
            deinterleave_block(ptr + offset + (i * 0x4000));
        }
    }

    memset(cart_rom, 0, 0x400000);
    if(size > 0x400000) size = 0x400000;
    memcpy(cart_rom, ptr + offset, size);

    /* Free allocated file data */
    free(ptr);

    hardware = 0;

    for (i = 0x1f0; i < 0x1ff; i++)
        switch (cart_rom[i]) {
     case 'U':
         hardware |= 4;
         break;
     case 'J':
         hardware |= 1;
         break;
     case 'E':
         hardware |= 8;
         break;
    }

    if (cart_rom[0x1f0] >= '1' && cart_rom[0x1f0] <= '9') {
        hardware = cart_rom[0x1f0] - '0';
    } else if (cart_rom[0x1f0] >= 'A' && cart_rom[0x1f0] <= 'F') {
        hardware = cart_rom[0x1f0] - 'A' + 10;
    }

    if (country) hardware=country; //simple autodetect override

    //From PicoDrive
    if (hardware&8)        
    {
        hw=0xc0; vdp_pal=1;
    } // Europe
    else if (hardware&4)    
    {
        hw=0x80; vdp_pal=0;
    } // USA
    else if (hardware&2)    
    {
        hw=0x40; vdp_pal=1;
    } // Japan PAL
    else if (hardware&1)      
    {
        hw=0x00; vdp_pal=0;
    } // Japan NTSC
    else
        hw=0x80; // USA

    if (vdp_pal) {
        vdp_rate = 50;
        lines_per_frame = 312;
    } else {
        vdp_rate = 60;
        lines_per_frame = 262;
    };

    /*SRAM*/   
    if(cart_rom[0x1b1] == 'A' && cart_rom[0x1b0] == 'R')
    {
        save_start = cart_rom[0x1b4] << 24 | cart_rom[0x1b5] << 16 |
            cart_rom[0x1b6] << 8  | cart_rom[0x1b7] << 0;
        save_len = cart_rom[0x1b8] << 24 | cart_rom[0x1b9] << 16 |
            cart_rom[0x1ba] << 8  | cart_rom[0x1bb] << 0;
        // Make sure start is even, end is odd, for alignment
        // A ROM that I came across had the start and end bytes of
        // the save ram the same and wouldn't work.  Fix this as seen
        // fit, I know it could probably use some work. [PKH]
        if(save_start != save_len)
        {
            if(save_start & 1) --save_start;
            if(!(save_len & 1)) ++save_len;
            save_len -= (save_start - 1);
            saveram = (unsigned char*)malloc(save_len);
            // If save RAM does not overlap main ROM, set it active by default since
            // a few games can't manage to properly switch it on/off.
            if(save_start >= (unsigned)size)
                save_active = 1;
        }
        else
        {
            save_start = save_len = 0;
            saveram = NULL;
        }
    }
    else
    {
        save_start = save_len = 0;
        saveram = NULL;
    }
    return AS3_Int(0);
}

AS3_Val sega_init(void* self, AS3_Val args)
{
    system_init();
    audioSamples = (44100 / vdp_rate)*2;
    framebuffer = (uint8*)malloc(FRAMEBUFFER_LENGTH);
    return AS3_Int(vdp_rate);
}

AS3_Val sega_reset(void* self, AS3_Val args)
{
    system_reset();
    return AS3_Int(0);
}

AS3_Val sega_frame(void* self, AS3_Val args)
{
    uint32 width;
    uint32 height;
    uint32 x, y;
    uint32 di, si, r;
    uint16 p;
    AS3_Val fb_ba;
   
    AS3_ArrayValue(args, "AS3ValType", &fb_ba);
   
    system_frame(0);
   
    AS3_SetS(fb_ba, "position", AS3_Int(0));
   
    width = (reg[12] & 1) ? 320 : 256;
    height = (reg[1] & 8) ? 240 : 224;

    for(y=0;y<240;y++)
    {
        for(x=0;x<320;x++)
        {
            di = 1280*y + x<<2;
           
            si = (y << 10) + ((x + bitmap.viewport.x) << 1);

            p = *((uint16*)(bitmap.data + si));
            framebuffer[di + 3] = (uint8)((p & 0x1f) << 3);
            framebuffer[di + 2] = (uint8)(((p >> 5) & 0x1f) << 3);
            framebuffer[di + 1] = (uint8)(((p >> 10) & 0x1f) << 3);
        }
    }

    AS3_ByteArray_writeBytes(fb_ba, framebuffer, FRAMEBUFFER_LENGTH);
    AS3_SetS(fb_ba, "position", AS3_Int(0));
   
    r = (width << 16) | height;
    return AS3_Int(r);
}

AS3_Val sega_audio(void* self, AS3_Val args)
{
    AS3_Val ab_ba;
    AS3_ArrayValue(args, "AS3ValType", &ab_ba);
   
    AS3_SetS(ab_ba, "position", AS3_Int(0));
    AS3_ByteArray_writeBytes(ab_ba, snd.buffer, audioSamples*sizeof(int16));
    AS3_SetS(ab_ba, "position", AS3_Int(0));
   
    return AS3_Int(0);
}

int main()
{
    AS3_Val romMethod = AS3_Function(NULL, sega_rom);
    AS3_Val initMethod = AS3_Function(NULL, sega_init);
    AS3_Val resetMethod = AS3_Function(NULL, sega_reset);
    AS3_Val frameMethod = AS3_Function(NULL, sega_frame);
    AS3_Val audioMethod = AS3_Function(NULL, sega_audio);

    // construct an object that holds references to the functions
    AS3_Val result = AS3_Object("sega_rom: AS3ValType, sega_init: AS3ValType, sega_reset: AS3ValType, sega_frame: AS3ValType, sega_audio: AS3ValType",
        romMethod, initMethod, resetMethod, frameMethod, audioMethod);

    // Release
    AS3_Release(romMethod);
    AS3_Release(initMethod);
    AS3_Release(resetMethod);
    AS3_Release(frameMethod);
    AS3_Release(audioMethod);

    // notify that we initialized -- THIS DOES NOT RETURN!
    AS3_LibInit(result);

    // should never get here!
    return 0;
}

Hello Bernd,

I have tried this but results are the same. Though, I was able to enable the optimization in antoher way:

I have all my code in a single file flash.c, then I could use: "gcc flash.c -O2 -Wall -swc -o emu.swc" and my code ran correctly, but still as slow as before. The only difference is that the swf file got twice smaller. Inline functions did not worked as before. Suprisingly, they are not executed due to an if statement I use, so they only take part in the final code without being executed at all in runtime. Something with code generation having inline functions is wrong in alchemy.

Finally, I suppose avm2.0 is still > 20x slower than native code. My code uses sin, pow and log10 only on init and in runtime it uses only reads, assignments, simple arithmetic and bitwise operations. Not stdlib functions are used at runtime even a single memset or memcpy in the core emulation code. So, as I said, avm has still a poor performance for real time apps being about > 20x slower than native application in my test. Even a debug compilation in Visual Studio without any optimization or inline functions drawing in software mode is still 9 times faster than avm2.

I really had a hope for bringing Sega Mega Drive to browser multiplayer gaming...

Here is my example without controls though... just a speed test

http://www.mailaufzeit.de/flash.html

Please all users tell the FPS count and the processor your are using.

Vielen Dank, aber optimizing the generated code is too much effort. I request a speed of at least 60 fps which is 3x speedup. I do not think that manual edits will do the job...and the C code is ~50 000 lines of code...too much.

Btw, findest Du die Idee mit dem Emulator gut?

Wow, you are telling very interesting stories...

Unfortunatelly, I did not understand a word. What to reimplement and where and how? Please provide me a step by step description! How to compile with -avm2-use-memuser - again step by step?

What I found in the generated file is:

public class MemUser
{
  public final function _mr32(addr:int):int { gstate.ds.position = addr; return gstate.ds.readInt(); }
  public final function _mru16(addr:int):int { gstate.ds.position = addr; return gstate.ds.readUnsignedShort(); }
  public final function _mrs16(addr:int):int { gstate.ds.position = addr; return gstate.ds.readShort(); }
  public final function _mru8(addr:int):int { gstate.ds.position = addr; return gstate.ds.readUnsignedByte(); }
  public final function _mrs8(addr:int):int { gstate.ds.position = addr; return gstate.ds.readByte(); }
  public final function _mrf(addr:int):Number { gstate.ds.position = addr; return gstate.ds.readFloat(); }
  public final function _mrd(addr:int):Number { gstate.ds.position = addr; return gstate.ds.readDouble(); }
  public final function _mw32(addr:int, val:int):void { gstate.ds.position = addr; gstate.ds.writeInt(val); }
  public final function _mw16(addr:int, val:int):void { gstate.ds.position = addr; gstate.ds.writeShort(val); }
  public final function _mw8(addr:int, val:int):void { gstate.ds.position = addr; gstate.ds.writeByte(val); }
  public final function _mwf(addr:int, val:Number):void { gstate.ds.position = addr; gstate.ds.writeFloat(val); }
  public final function _mwd(addr:int, val:Number):void { gstate.ds.position = addr; gstate.ds.writeDouble(val); }
}

and those methods are used in each push, pop! I am waiting for your answer ungeduldigly.

Danke im Voraus!

Some update - I'm always using -avm2-use-memuser, but for fun I tried without it and slow down was slightly above 3x! Would this mean that the ASM reimplementation may reach at least 3x speed up?

We are waiting even more ungeduldigly for your ASM code and instructions on how to use it.

Gruss

Boris

I have changed the MemUser class with the ASM version and I got 4 fps slower SWF.

My AS file is called 1620.achacks.as so

I modified the gcc compilation to:

#sys(@llc, "-o=".($last = "1620.achacks.as"), @ll, @oo); //Do not create the as file to avoid overwrite
 
$last = "1620.achacks.as";

And renamed all $$.achacks strings to 1620.achacks

gstate.ds.write is referenced extra in memset, memmove and memcpy. But these functions are used only on init where speed is irrelevant.

It is terrible...

Thank you Bernd

Dear Bernd,

please provide the whole bash script from compiling the modified as file to the creation of the SWC. I am a windows guy with no knoledge about bash and its black magic. I was able to build with alchemy only strictly following the instructions the manual and you provide. Would you take a look at the AS files if I send them to you per email? It'd be very kind.

Regards

Boris

Dear friends,

there are several news regarding the issues discussed in this topic:

1. Floating points - there is no error while working with floats/doubles as I've initiallly thought. I wanted to write a float array to a AS byte array in C:

static AS3_Val sega_audio(void* self, AS3_Val args)
{
    uint32 i;
    AS3_Val ab_ba;
    AS3_ArrayValue(args, "AS3ValType", &ab_ba);
   
    AS3_SetS(ab_ba, "position", AS3_Int(0));
    for(i=0;i<audioSamples;i++)
        audiobuffer[i] = (float)snd.output[i] / 32768.0f;
   
    AS3_ByteArray_writeBytes(ab_ba, audiobuffer, audioSamples*sizeof(float));
    AS3_SetS(ab_ba, "position", AS3_Int(0));
   
    return AS3_Int(0);
}

Result was nosy and incorrect sound. Inspecting the floats -> incorrect numbers. Thanks to Bernd, who pointed me to the code of Doom ported for alchemy, I have discovered that Flash is actually holding the data in byte arrays in little endian (reversed). So all the hours of trying fixing the audio code were invane. The solution was a simple float reverse function:

STATIC_INLINE float FloatSwap (float f)
{
    union
    {
        float f;
        uint8 b[4];
    } dat1, dat2;

    dat1.f = f;
    dat2.b[0] = dat1.b[3];
    dat2.b[1] = dat1.b[2];
    dat2.b[2] = dat1.b[1];
    dat2.b[3] = dat1.b[0];
   
    return dat2.f;
}

static AS3_Val sega_audio(void* self, AS3_Val args)
{
    uint32 i;
    AS3_Val ab_ba;
    AS3_ArrayValue(args, "AS3ValType", &ab_ba);
   
    AS3_SetS(ab_ba, "position", AS3_Int(0));
    for(i=0;i<audioSamples;i++)
        audiobuffer[i] = FloatSwap((float)snd.output[i] / 32768.0f);
   
    AS3_ByteArray_writeBytes(ab_ba, audiobuffer, audioSamples*4);
    AS3_SetS(ab_ba, "position", AS3_Int(0));
   
    return AS3_Int(0);
}

So, my floating point problems were RESOLVED!

2. Inline funtions - Well, I was able to resolve the issue of not working code with inlines for one particular project (see above). Actually, solutions are two, while the second is not very useful for large projects with many inline functions because you will spend your life time in correcting the code as an idiot, which is on the other side error prown. Here are the solutions:

Solution 1: Make your code as clear as possible regarding the includes! Do not make a single header including all others and then use it everywhere. This may be not a problem for MSVC or normal GCC but for the alchemy version this might (but not must!) be a problem.

Solution 2: Convert your inline functions to macros. This will achieve same speed as inline functions. But you may need not to forget that inlines are functions so each rule for copying the arguments is still there! Another problem is for the inline functions returning a value and have more than one statement. You must then rewrite it, and eventually make all needed changes in your code where you call the function (now a macro). This may be painful - please believe me!

This issue was RESOLVED, too! But please see point 3 below!

3. GCC Optimization problems - here things are still not really clear for me. Well I am porting a sega mega drive emulator to flash, which is still very slow. Just for the experiment I have ported one emulator for the predecessor of mega drive - the sega master system. Initially, it was slow, too. I was compilling separately each C file with -O3 and linking them together in a SWC which is later used in the actual AS project. It was again slow. I turned on the inline functions - again slow. Then I have included all C files in the main.c file and triend again compiling with:

gcc main.c -O3 -Wall -swc -o emu.swc

Great speed up was obvious, but still not enough to have smooth graphics and sound. Then I discovered that my approach for game (main) loop was wrong. Basicly it was:

private var thread:Timer = new Timer(1);

thread.start();
thread.addEventListener(TimerEvent.TIMER, run);

private function run(e:TimerEvent):void

{

     if(is time so that this will run at 60 fps)

     {

           //One frame is done here

     }

     else

          return;

}

This is WRONG!!!

Basicly, a game must run at 50 or 60 fps depending on if it is PAL or NTSC TV standard. After looking at the code of Doom I have seen that they do something else to draw the frames:

addEventListener(Event.ENTER_FRAME, onFrame);

private function onFrame(event:Event):void

{

     //One frame is done here

}

So, I did too. But fps were never > 24 fps. Then I discovered that a sprite object has a stage member where you can get/set frameRate. I plotted it and it was 24. So before executing the game I set it to 50 or 60 depending on the TV standard. And it worked! Full speed and sound and everything (but still everything included in one single file) and built with:

gcc main.c -O3 -Wall -swc -o emu.swc

Compiling separately leads to slow code.

Anyway, I was very happy and spammed Bernd with cheerful messages. I've then tried the same with the megadrive emulator but in reverse order -> stop using the thread timer, fixing the sound and then include everything in the main.c and compile it.

No visual speed changes were detected after changing to ENTER_FRAME, but sound worked (slow but correct). After including everything in the main file, code compiled without errors and warning but stopped working (no image, no sound), though still running. Then, I reversed back to separate compilation but with enabled inline functions. NO SPEED UP! On native platforms enabling them gives at least 2x faster run. With -O3 turned on - 6x speedup. Here nothing. I build the code with a file containg the following:

echo Compilling main.c
gcc -c main.c -O3 -o main.o
echo Compilling fm.c
gcc -c fm.c -O3 -o fm.o
echo Compilling genesis.c
gcc -c genesis.c -O3 -o genesis.o
echo Compilling input.c
gcc -c input.c -O3 -o input.o
echo Compilling io.c
gcc -c io.c -O3 -o io.o
echo Compilling loadrom.c
gcc -c loadrom.c -O3 -o loadrom.o
echo Compilling m68kcpu.c
gcc -c m68kcpu.c -O3 -o m68kcpu.o
echo Compilling m68kopac.c
gcc -c m68kopac.c -O3 -o m68kopac.o
echo Compilling m68kopdm.c
gcc -c m68kopdm.c -O3 -o m68kopdm.o
echo Compilling m68kopnz.c
gcc -c m68kopnz.c -O3 -o m68kopnz.o
echo Compilling m68kops.c
gcc -c m68kops.c -O3 -o m68kops.o
echo Compilling mem68k.c
gcc -c mem68k.c -O3 -o mem68k.o
echo Compilling membnk.c
gcc -c membnk.c -O3 -o membnk.o
echo Compilling memvdp.c
gcc -c memvdp.c -O3 -o memvdp.o
echo Compilling memz80.c
gcc -c memz80.c -O3 -o memz80.o
echo Compilling render.c
gcc -c render.c -O3 -o render.o
echo Compilling sn76496.c
gcc -c sn76496.c -O3 -o sn76496.o
echo Compilling sound.c
gcc -c sound.c -O3 -o sound.o
echo Compilling system.c
gcc -c system.c -O3 -o system.o
echo Compilling vdp.c
gcc -c vdp.c -O3 -o vdp.o
echo Compilling z80.c
gcc -c z80.c -O3 -o z80.o

echo Building SWC file...
gcc *.o -Wall -swc -O3 -o emu.swc
echo Building SWF file...
mxmlc.exe -library-path+=emu.swc --target-player=10.0.0 -default-size 320 240 Emu.as
echo Removing temporary data...
for file in $( ls *.o )
    do
        rm $file
    done
echo Ready!

Is this correct? Ideas how to make it work faster?

Bernd, did you have time to check the AS code?