void syscall_gate() {
__asm__("SYSTEM_CALL:");
__asm__("pushl $0");
store_reg();
svc();
restore_reg();
__asm__("add $4, %esp");
iret();
}
static void __stdcall
transpose_bgra2(const uint8_t* srcp, uint8_t* dstp, const int rowsize,
const int height, const int src_pitch, const int dst_pitch) noexcept
{
const int h = height / 4 * 4;
const int r = rowsize / 32 * 32;
const uint8_t* s = srcp;
for (int y = 0; y < h; y += 4) {
uint8_t* d = dstp + 4 * y;
for (int x = 0; x < r; x += 32) {
__m128i a03 = load_reg(s + x + 0 + 0 * src_pitch);
__m128i a47 = load_reg(s + x + 16 + 0 * src_pitch);
__m128i b03 = load_reg(s + x + 0 + 1 * src_pitch);
__m128i b47 = load_reg(s + x + 16 + 1 * src_pitch);
__m128i c03 = load_reg(s + x + 0 + 2 * src_pitch);
__m128i c47 = load_reg(s + x + 16 + 2 * src_pitch);
__m128i d03 = load_reg(s + x + 0 + 3 * src_pitch);
__m128i d47 = load_reg(s + x + 16 + 3 * src_pitch);
__m128i ab01 = unpacklo32(a03, b03);
__m128i cd01 = unpacklo32(c03, d03);
__m128i ab23 = unpackhi32(a03, b03);
__m128i cd23 = unpackhi32(c03, d03);
__m128i ab45 = unpacklo32(a47, b47);
__m128i cd45 = unpacklo32(c47, d47);
__m128i ab67 = unpackhi32(a47, b47);
__m128i cd67 = unpackhi32(c47, d47);
store_reg(d + 0 * dst_pitch, unpacklo64(ab01, cd01));
store_reg(d + 1 * dst_pitch, unpackhi64(ab01, cd01));
store_reg(d + 2 * dst_pitch, unpacklo64(ab23, cd23));
store_reg(d + 3 * dst_pitch, unpackhi64(ab23, cd23));
store_reg(d + 4 * dst_pitch, unpacklo64(ab45, cd45));
store_reg(d + 5 * dst_pitch, unpackhi64(ab45, cd45));
store_reg(d + 6 * dst_pitch, unpacklo64(ab67, cd67));
store_reg(d + 7 * dst_pitch, unpackhi64(ab67, cd67));
d += 8 * dst_pitch;
}
s += 4 * src_pitch;
}
if (r == rowsize && h == height)
return;
int width = rowsize / 4;
const uint32_t* s4 = (uint32_t*)srcp;
uint32_t* d4 = (uint32_t*)dstp;
int sp = src_pitch / 4;
int dp = dst_pitch / 4;
for (int y = h; y < height; ++y) {
for (int x = 0; x < width; ++x) {
d4[y + x * dp] = s4[x + y * sp];
}
}
if (r < rowsize) {
for (int y = 0; y < height; ++y) {
for (int x = r / 4; x < width; ++x) {
d4[y + x * dp] = s4[x + y * sp];
}
}
}
}
static void __stdcall
transpose_bgra(const uint8_t* srcp, uint8_t* dstp, const int rowsize,
const int height, const int src_pitch, const int dst_pitch) noexcept
{
const int h = height / 4 * 4;
const int r = rowsize / 16 * 16;
const uint8_t* s = srcp;
for (int y = 0; y < h; y += 4) {
uint8_t* d = dstp + 4 * y;
for (int x = 0; x < r; x += 16) {
__m128i s0 = load_reg(s + x + 0 * src_pitch);
__m128i s1 = load_reg(s + x + 1 * src_pitch);
__m128i s2 = load_reg(s + x + 2 * src_pitch);
__m128i s3 = load_reg(s + x + 3 * src_pitch);
__m128i ab01 = unpacklo32(s0, s1);
__m128i ab23 = unpackhi32(s0, s1);
__m128i cd01 = unpacklo32(s2, s3);
__m128i cd23 = unpackhi32(s2, s3);
__m128i abcd0 = unpacklo64(ab01, cd01);
__m128i abcd1 = unpackhi64(ab01, cd01);
__m128i abcd2 = unpacklo64(ab23, cd23);
__m128i abcd3 = unpackhi64(ab23, cd23);
store_reg(d + 0 * dst_pitch, abcd0);
store_reg(d + 1 * dst_pitch, abcd1);
store_reg(d + 2 * dst_pitch, abcd2);
store_reg(d + 3 * dst_pitch, abcd3);
d += 4 * dst_pitch;
}
s += 4 * src_pitch;
}
if (r == rowsize && h == height)
return;
int width = rowsize / 4;
const uint32_t* s4 = (uint32_t*)srcp;
uint32_t* d4 = (uint32_t*)dstp;
int sp = src_pitch / 4;
int dp = dst_pitch / 4;
for (int y = h; y < height; ++y) {
for (int x = 0; x < width; ++x) {
d4[y + x * dp] = s4[x + y * sp];
}
}
if (r < rowsize) {
for (int y = 0; y < height; ++y) {
for (int x = r / 4; x < width; ++x) {
d4[y + x * dp] = s4[x + y * sp];
}
}
}
}
