Move EmuGrArea to Nano-X server side to optimize the render speed

[toncho11]

The objective is speed.
The GrArea is not available on ELKS, so instead the EmuGrArea is used in nxjpeg. GrArea draws generic rectangles of raw pixels, while GrDrawIndexedSpanBuffer draws pre-quantized palette-index rows more efficiently by collapsing them into horizontal color runs.

This is a very experimental move of EmuGrArea to server side of Nano-X. It should reside in src/server/srvdraw.c.

#include "srvdraw.h"
#include "device.h"
#include "devfont.h"
#include "devdraw.h"
#include "serv.h"

/* Draws a block of indexed pixels by grouping consecutive pixels
 * of the same color into horizontal runs, then drawing each run
 * in one operation (faster than per-pixel drawing).
 */
void srv_DrawIndexedSpanBuffer(GR_CLIENT *cl, GR_DRAW_INDEXED_SPAN_BUFFER_REQ *req)
{
    GR_WINDOW *wp = GsFindWindow(req->wid);
    if (!wp) return;

    /* read raw data buffer sent by client */
    const unsigned char *buf = (const unsigned char *)GsReadData(req->pitch * req->h);

    if (!buf) return;

    int x = req->x;
    int y = req->y;
    int w = req->w;
    int h = req->h;
    int pitch = req->pitch;

    /* Loop through scanlines */
    for (int iy = 0; iy < h; ++iy) {

        const unsigned char *row = buf + iy * pitch;

        int run_start = 0;
        unsigned char run_idx = row[0];

        /* Scan through each pixel to detect runs */
        for (int ix = 1; ix < w; ++ix) {

            if (row[ix] != run_idx) {

                GdSetForegroundColorIndex(run_idx);

                GdLine(x + run_start, y + iy,
                       x + ix - 1,    y + iy,
                       wp->output);

                run_start = ix;
                run_idx = row[ix];
            }
        }

        /* Draw final run */
        GdSetForegroundColorIndex(run_idx);
        GdLine(x + run_start, y + iy,
               x + w - 1,      y + iy,
               wp->output);
    }

    /* Update window region */
    GsExposeArea(wp, x, y, w, h);
}

[toncho11]

With color setting optimization:

/* Draws a block of indexed pixels by grouping consecutive pixels
 * of the same color into horizontal runs, then drawing each run
 * in one operation (faster than per-pixel drawing).
 */
void srv_DrawIndexedSpanBuffer(GR_CLIENT *cl, GR_DRAW_INDEXED_SPAN_BUFFER_REQ *req)
{
    GR_WINDOW *wp = GsFindWindow(req->wid);
    if (!wp) return;

    /* read raw data buffer sent by client */
    const unsigned char *buf =
        (const unsigned char *)GsReadData(req->pitch * req->h);

    if (!buf) return;

    int x = req->x;
    int y = req->y;
    int w = req->w;
    int h = req->h;
    int pitch = req->pitch;

    unsigned char last_color = 0xff;

    /* Loop through scanlines */
    for (int iy = 0; iy < h; ++iy) {

        const unsigned char *row = buf + iy * pitch;

        int run_start = 0;
        unsigned char run_idx = row[0];

        /* Scan through each pixel to detect runs */
        for (int ix = 1; ix < w; ++ix) {

            if (row[ix] != run_idx) {

                if (run_idx != last_color) {
                    GdSetForegroundColorIndex(run_idx);
                    last_color = run_idx;
                }

                GdLine(x + run_start, y + iy,
                       x + ix - 1,    y + iy,
                       wp->output);

                run_start = ix;
                run_idx = row[ix];
            }
        }

        /* Draw final run */
        if (run_idx != last_color) {
            GdSetForegroundColorIndex(run_idx);
            last_color = run_idx;
        }

        GdLine(x + run_start, y + iy,
               x + w - 1,      y + iy,
               wp->output);
    }

    /* Update window region */
    GsExposeArea(wp, x, y, w, h);
}