Next Previous Contents

3. Format Conversion

There are a few steps involved in getting your buffers / images converted. Initialisation, request for conversion, palette setting and the actual conversion.

The routines below will normally be called in the following order to do what they are supposed to do:

You will have to call Hermes_ConverterPalette every time before copying because your palette might have changed. And if there is a conversion from 8 bit to a direct colour format then HERMES has to update its lookup tables.

WARNING: Do not attempt to save time by not calling Hermes_ConverterPalette or Hermes_ConverterRequest in your program's main loop. You have nothing to be afraid of, HERMES caches everything for you. If nothing changes, HERMES returns immediately. Omitting the calls will only cause problems.

3.1 HermesHandle Hermes_ConverterInstance(unsigned long flags)

This will generate a new converter instance and return it to you. If an error occurs, 0 will be returned instead. Don't forget to check. At the moment, you can pass the following flags:

Additional notes: Hermes keeps a dynamic array of converter instances that will grow if it runs out of space. You can change the default size and growth of the array in HermConf.h before compiling.

Keep the handle this function returns somewhere safe, you will need it for any other conversion function (and you can use it for locking if you want to be thread safe!)

3.2 void Hermes_ConverterReturn(HermesHandle handle)

Return the conversion routine specified by handle to HERMES to free up its spaces. This will get rid of all caches, lookup tables, etc. associated with this converter. Try not to create and destroy converters too often, it is costly. After all, the main effort in HERMES was to shift everything from loop time to initialisation time.

HERMES will ignore attempts to return non-existing handles.

3.3 int Hermes_ConverterRequest(...)

Complete function declaration: Hermes_ConverterRequest(HermesHandle handle, HermesFormat *source, HermesFormat *dest).

This will instruct Hermes to find a converter from the given source format to the given destination format and store it in the structure associated with handle. (Get a converter handle using Hermes_ConverterInstance).

As mentioned before, don't be shy, call it every time before you call Hermes_ConverterCopy unless you are really sure that the size of your window, colour format etc. doesn't change. You don't have to be afraid as any calls without changes will return success immediately.

The way HERMES will find your converter is the following:

This function will return a non-zero value if a converter could be found, otherwise zero will be returned.

3.4 int Hermes_ConverterPalette(...)

Complete function declaration: Hermes_ConverterPalette(HermesHandle handle, HermesHandle sourcepal, HermesHandle destpal)

handle is the converter handle you got by using Hermes_ConverterInstance(). sourcepal and destpal are two palette handles which you have to get using Hermes_PaletteInstance().

This function will instruct HERMES to use the specified palettes for format conversion. At the moment, only sourcepal is of interest. It will be used whenever you want to convert from an indexed 8 bit format to any destination format. You may pass the sourcepal handle as destpal for the time being, it has no effect.

If converting from 8 bit to a non-indexed format, HERMES has to create a lookup table. In order to prevent any cycle-wasting, HERMES will cache those lookup tables for you. Whenever you change the palette, however - using Hermes_PaletteSet - the lookup tables have to be recalculated. In any case, generating lookup tables is a minor effort.

If any of the handles you pass are invalid or some other error occurs, HERMES will return zero on this function. Otherwise any non-zero value is returned.

3.5 int Hermes_ConverterCopy(...)

Complete function declaration: Hermes_ConverterCopy(HermesHandle handle, void *s_pixels,int s_x,int s_y,int s_width,int s_height,int s_pitch, void *d_pixels,int d_x,int d_y,int d_width,int d_height,int d_pitch)

Now, this is the real things and it has so many parameters that I might as well make a list out of them :)

And, before I forget it, coordinates and widths up there are in pixels, not bytes! (That should prevent you from making mistakes now :)

The pitch of a buffer is basically the number of bytes in a scanline. This is normally equal to the width of the scanline in pixels times the number of bytes in a pixel. However, one some platforms this is not true.. and somebody might want to align their scanlines and they would find this quite comfortable.

In other words, the pitch is the amount in bytes to add to the beginning of a scanline in order to get to the beginning of the next scanline. Read again: amount in BYTES this time, NOT pixels. Beware!

There are a few things that can cause this function to return 0 (which indicates an error):

  1. You specified an invalid handle
  2. The source width or source height does not match the destination width or destination height AND no stretch converter can be found

From this, the policy of this function should be apparent but shall be made a bit more explicit:

It is probably also worthwhile noting that even though this all has a Converter prefix, if the source and destination formats are equal, then straight copying will be used.


Next Previous Contents