twiddle: fully parameterize texture twiddling function

I think the original version is more readable, but the newer version
is better overall because it doesn't require reading from dst, and is
able to directly write to a 32-bit dst.
This commit is contained in:
Zack Buhman 2023-12-20 23:54:32 +08:00
parent 7e03eee3ed
commit a8cceee46e
2 changed files with 71 additions and 19 deletions

View File

@ -176,16 +176,20 @@ inline void inflate_character(const uint8_t * src, const uint8_t c)
auto texture = reinterpret_cast<volatile uint32_t *>(mem->texture);
uint32_t offset = 8 * 8 * character_index;
/*
union {
uint8_t u8[8 * 8];
uint32_t u32[8 * 8 / 4];
} temp2;
//twiddle::texture(&texture[offset], temp, 8, 8);
twiddle::texture_4bpp(temp2.u8, temp, 8, 8);
twiddle::texure_4bpp(temp2.u8, temp, 8, 8);
for (uint32_t i = 0; i < 8 * 8 / 4; i++) {
texture[(offset / 4) + i] = temp2.u32[i];
}
*/
twiddle::texture2<4>(&texture[offset / 4], temp, 8, 8);
}
void inflate_font(const uint8_t * src)

View File

@ -1,5 +1,5 @@
#include <cstdint>
#include <array>
#include <tuple>
namespace twiddle {
@ -33,7 +33,7 @@ constexpr inline uint32_t from_xy(uint32_t x, uint32_t y)
// bits : 19-0
uint32_t twiddle_ix = 0;
for (int i = 0; i <= (19 / 2); i++) {
for (int i = 0; i <= (20 / 2); i++) {
twiddle_ix |= ((y >> i) & 1) << (i * 2 + 0);
twiddle_ix |= ((x >> i) & 1) << (i * 2 + 1);
}
@ -58,6 +58,7 @@ static_assert(from_xy(0b000, 0b101) == 17);
static_assert(from_xy(0b000, 0b110) == 20);
static_assert(from_xy(0b000, 0b111) == 21);
/*
constexpr inline std::array<uint32_t, 2>
from_ix(uint32_t curve_ix)
{
@ -72,23 +73,42 @@ from_ix(uint32_t curve_ix)
return x_y;
}
*/
static_assert(from_ix(0) == std::array<uint32_t, 2>{{0b000, 0b000}});
static_assert(from_ix(2) == std::array<uint32_t, 2>{{0b001, 0b000}});
static_assert(from_ix(8) == std::array<uint32_t, 2>{{0b010, 0b000}});
static_assert(from_ix(10) == std::array<uint32_t, 2>{{0b011, 0b000}});
static_assert(from_ix(32) == std::array<uint32_t, 2>{{0b100, 0b000}});
static_assert(from_ix(34) == std::array<uint32_t, 2>{{0b101, 0b000}});
static_assert(from_ix(40) == std::array<uint32_t, 2>{{0b110, 0b000}});
static_assert(from_ix(42) == std::array<uint32_t, 2>{{0b111, 0b000}});
constexpr inline std::tuple<uint32_t, uint32_t>
from_ix(uint32_t curve_ix)
{
uint32_t y = (curve_ix >> 0) & 0x55555555;
y = (y | (y >> 1)) & 0x33333333;
y = (y | (y >> 2)) & 0x0f0f0f0f;
y = (y | (y >> 4)) & 0x00ff00ff;
y = (y | (y >> 8)) & 0x0000ffff;
static_assert(from_ix(1) == std::array<uint32_t, 2>{{0b000, 0b001}});
static_assert(from_ix(4) == std::array<uint32_t, 2>{{0b000, 0b010}});
static_assert(from_ix(5) == std::array<uint32_t, 2>{{0b000, 0b011}});
static_assert(from_ix(16) == std::array<uint32_t, 2>{{0b000, 0b100}});
static_assert(from_ix(17) == std::array<uint32_t, 2>{{0b000, 0b101}});
static_assert(from_ix(20) == std::array<uint32_t, 2>{{0b000, 0b110}});
static_assert(from_ix(21) == std::array<uint32_t, 2>{{0b000, 0b111}});
uint32_t x = (curve_ix >> 1) & 0x55555555;
x = (x | (x >> 1)) & 0x33333333;
x = (x | (x >> 2)) & 0x0f0f0f0f;
x = (x | (x >> 4)) & 0x00ff00ff;
x = (x | (x >> 8)) & 0x0000ffff;
return {x, y};
}
static_assert(from_ix(0) == std::tuple<uint32_t, uint32_t>{0b000, 0b000});
static_assert(from_ix(2) == std::tuple<uint32_t, uint32_t>{0b001, 0b000});
static_assert(from_ix(8) == std::tuple<uint32_t, uint32_t>{0b010, 0b000});
static_assert(from_ix(10) == std::tuple<uint32_t, uint32_t>{0b011, 0b000});
static_assert(from_ix(32) == std::tuple<uint32_t, uint32_t>{0b100, 0b000});
static_assert(from_ix(34) == std::tuple<uint32_t, uint32_t>{0b101, 0b000});
static_assert(from_ix(40) == std::tuple<uint32_t, uint32_t>{0b110, 0b000});
static_assert(from_ix(42) == std::tuple<uint32_t, uint32_t>{0b111, 0b000});
static_assert(from_ix(1) == std::tuple<uint32_t, uint32_t>{0b000, 0b001});
static_assert(from_ix(4) == std::tuple<uint32_t, uint32_t>{0b000, 0b010});
static_assert(from_ix(5) == std::tuple<uint32_t, uint32_t>{0b000, 0b011});
static_assert(from_ix(16) == std::tuple<uint32_t, uint32_t>{0b000, 0b100});
static_assert(from_ix(17) == std::tuple<uint32_t, uint32_t>{0b000, 0b101});
static_assert(from_ix(20) == std::tuple<uint32_t, uint32_t>{0b000, 0b110});
static_assert(from_ix(21) == std::tuple<uint32_t, uint32_t>{0b000, 0b111});
template <typename T>
void texture(volatile T * dst, const T * src, const uint32_t width, const uint32_t height)
@ -116,4 +136,32 @@ void texture_4bpp(volatile T * dst, const T * src, const uint32_t width, const u
}
}
template <int B, typename T, typename U>
void texture2(volatile T * dst, const U * src, const uint32_t width, const uint32_t height)
{
constexpr uint32_t t_bits = (sizeof (T)) * 8;
constexpr uint32_t bits_per_pixel = B;
static_assert(t_bits >= bits_per_pixel);
static_assert((t_bits / bits_per_pixel) * bits_per_pixel == t_bits);
constexpr uint32_t pixels_per_t = t_bits / bits_per_pixel;
static_assert(pixels_per_t == 1 || pixels_per_t == 2 || pixels_per_t == 4 || pixels_per_t == 8);
T dst_val = 0;
for (uint32_t curve_ix = 0; curve_ix < (width * height); curve_ix++) {
auto [x, y] = from_ix(curve_ix);
const U src_val = src[y * width + x];
if constexpr (pixels_per_t == 1) {
dst[curve_ix] = src_val;
} else {
const uint32_t curve_ix_mod = curve_ix & (pixels_per_t - 1);
dst_val |= src_val << (bits_per_pixel * curve_ix_mod);
if (curve_ix_mod == (pixels_per_t - 1)) {
dst[curve_ix / pixels_per_t] = dst_val;
dst_val = 0;
}
}
}
}
}