private static unsafe void V128ForLoop([NoAlias] u8 *src, [NoAlias] u32 *dst, int count)
{
// Input:
//
// 0123 4567 8901 2345
// RGBR GBRG BRGB RGBR
// Output:
// RGBA RGBA RGBA RGBA
var shuffle = setr_epi8(0, 1, 2, -1, 3, 4, 5, -1, 6, 7, 8, -1, 9, 10, 11, -1);
var alpha = set1_epi32(0xFF << 24);
int i = 0;
var alignedCount = count & ~3;
for (; i < alignedCount; i += 4)
{
var v0 = loadu_ps(src + i * 3);
var v1 = shuffle_epi8(v0, shuffle);
var v2 = or_ps(v1, alpha);
storeu_ps((dst + i), v2);
}
for (; i < count; i++)
{
dst[i] = (u32)src[i * 3 + 0] << 0 |
(u32)src[i * 3 + 1] << 8 |
(u32)src[i * 3 + 2] << 16 |
(u32)0xFF << 24;
}
}
private static unsafe void BurstForLoop([NoAlias] u8 *src, [NoAlias] u32 *dst, int count)
{
for (int i = 0; i < count; i++)
{
dst[i] = (u32)src[i * 3 + 0] << 0 |
(u32)src[i * 3 + 1] << 8 |
(u32)src[i * 3 + 2] << 16 |
(u32)0xFF << 24;
}
}
private static unsafe void V256ForLoop([NoAlias] u8 *src, [NoAlias] u32 *dst, int count)
{
// Input
// u128 0 1
// u64 0 1 2 3
// u32 0 1 2 3 4 5 6 7
// u16 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5
// u8 0123 4567 8901 2345 6789 0123 4567 8901
// RGBR GBRG BRGB RGBR GBRG BRGB ---- ----
// Output
// RGBA RGBA RGBA RGBA RGBA RGBA RBGA RGBA
// Path
// v0 = 0123 4567 89AB CDEF GHIJ KLMN ---- ----
// v1 = 0123 4567 89AB ---- CDEF GHIJ KLMN ----
// v2 = 012- 345- 678- 9AB- CDE- FGH- IJK- LMN-
// v3 = 012α 345α 678α 9ABα CDEα FGHα IJKα LMNα
var alignedCount = count & ~7;
var permute = mm256_setr_epi32(0, 1, 2, 0xFF, 3, 4, 5, 0xFF);
var shuffleV128 = setr_epi8(0, 1, 2, -1, 3, 4, 5, -1, 6, 7, 8, -1, 9, 10, 11, -1);
var shuffleV256 = mm256_setr_m128(shuffleV128, shuffleV128);
var alpha = mm256_set1_epi32(0xFF << 24);
int i = 0;
for (; i < alignedCount; i += 8)
{
var v0 = mm256_loadu_ps(src + i * 3);
var v1 = mm256_permutevar8x32_epi32(v0, permute);
var v2 = mm256_shuffle_epi8(v1, shuffleV256);
var v3 = mm256_or_ps(v2, alpha);
mm256_storeu_ps(dst + i, v3);
}
for (; i < count; i++)
{
dst[i] = (u32)src[i * 3 + 0] << 0 |
(u32)src[i * 3 + 1] << 8 |
(u32)src[i * 3 + 2] << 16 |
(u32)0xFF << 24;
}
}
//-------------------------------------------------
// char_expand - expand the raw data for a
// character into a bitmap
//-------------------------------------------------
void char_expand(char32_t chnum, glyph gl)
{
LOG("render_font::char_expand: expanding character {0}\n", chnum);
rgb_t fgcol = (gl.color != 0 ? gl.color : new rgb_t(0xff, 0xff, 0xff, 0xff));
rgb_t bgcol = new rgb_t(0x00, 0xff, 0xff, 0xff);
bool is_cmd = ((chnum >= COMMAND_UNICODE) && (chnum < COMMAND_UNICODE + MAX_GLYPH_FONT));
if (is_cmd)
{
throw new emu_unimplemented();
#if false
// punt if nothing there
if (gl.bmwidth == 0 || gl.bmheight == 0 || gl.rawdata == null)
{
return;
}
// allocate a new bitmap of the size we need
gl.bitmap.allocate(gl.bmwidth, m_height_cmd);
gl.bitmap.fill(0);
// extract the data
const char *ptr = gl.rawdata;
byte accum = 0;
byte accumbit = 7;
for (int y = 0; y < gl.bmheight; y++)
{
int desty = y + m_height_cmd + m_yoffs_cmd - gl.yoffs - gl.bmheight;
u32 *dest = (desty >= 0 && desty < m_height_cmd) ? &gl.bitmap.pix(desty, 0) : nullptr;
{
for (int x = 0; x < gl.bmwidth; x++)
{
if (accumbit == 7)
{
accum = *ptr++;
}
if (dest != null)
{
*dest++ = (accum & (1 << accumbit)) ? color : rgb_t(0x00, 0xff, 0xff, 0xff);
}
accumbit = (accumbit - 1) & 7;
}
}
}
#endif
}
else if (m_format == format.OSD)
{
throw new emu_unimplemented();
#if false
// if we're an OSD font, query the info
#endif
}
else if (gl.bmwidth == 0 || gl.bmheight == 0 || gl.rawdata == null)
{
// abort if nothing there
LOG("render_font::char_expand: empty bitmap bounds or no raw data\n");
return;
}
else
{
// other formats need to parse their data
LOG("render_font::char_expand: building bitmap from raw data\n");
// allocate a new bitmap of the size we need
gl.bitmap.allocate(gl.bmwidth, m_height);
gl.bitmap.fill(0);
// extract the data
Pointer <u8> ptr = new Pointer <u8>(gl.rawdata); //const char *ptr = gl.rawdata;
u8 accum = 0;
u8 accumbit = 7;
for (int y = 0; y < gl.bmheight; y++)
{
int desty = y + m_height + m_yoffs - gl.yoffs - gl.bmheight;
PointerU32 dest = ((0 <= desty) && (m_height > desty)) ? gl.bitmap.pix(desty) : null; //u32 *dest(((0 <= desty) && (m_height > desty)) ? &gl.bitmap.pix(desty) : nullptr);
if (m_format == format.TEXT)
{
if (dest != null)
{
for (int x = 0; gl.bmwidth > x;)
{
// scan for the next hex digit
int bits = -1;
while ('\r' != ptr[0] && '\n' != ptr[0] && 0 > bits) // while (('\r' != *ptr) && ('\n' != *ptr) && (0 > bits))
{
if (ptr[0] >= '0' && ptr[0] <= '9')
{
bits = ptr[0] - '0'; //bits = *ptr++ - '0';
ptr++;
}
else if (ptr[0] >= 'A' && ptr[0] <= 'F')
{
bits = ptr[0] - 'A' + 10;
ptr++;
}
else if (ptr[0] >= 'a' && ptr[0] <= 'f')
{
bits = ptr[0] - 'a' + 10;
ptr++;
}
else
{
ptr++;
}
}
// expand the four bits
dest[0] = (bits & 8) != 0 ? fgcol : bgcol; dest++; //*dest++ = (bits & 8) ? fgcol : bgcol;
if (gl.bmwidth > ++x)
{
dest[0] = (bits & 4) != 0 ? fgcol : bgcol; dest++;
} //*dest++ = (bits & 4) ? fgcol : bgcol;
if (gl.bmwidth > ++x)
{
dest[0] = (bits & 2) != 0 ? fgcol : bgcol; dest++;
} //*dest++ = (bits & 2) ? fgcol : bgcol;
if (gl.bmwidth > ++x)
{
dest[0] = (bits & 1) != 0 ? fgcol : bgcol; dest++;
} //*dest++ = (bits & 1) ? fgcol : bgcol;
x++;
}
// advance to the next line
ptr = next_line(ptr);
}
}
else if (m_format == format.CACHED)
{
for (int x = 0; x < gl.bmwidth; x++)
{
if (accumbit == 7)
{
accum = ptr[0];
ptr++;
}
if (dest != null)
{
dest[0] = (accum & (1 << accumbit)) != 0 ? fgcol : bgcol; dest++;
} //*dest++ = (accum & (1 << accumbit)) ? fgcol : bgcol;
accumbit = (u8)((accumbit - 1) & 7);
}
}
}
}
// wrap a texture around the bitmap
gl.texture = m_manager.texture_alloc(render_texture.hq_scale);
gl.texture.set_bitmap(gl.bitmap, gl.bitmap.cliprect(), texture_format.TEXFORMAT_ARGB32);
}
