izzy/xemu
0
0
Fork 0
mirror of https://github.com/izzy2lost/xemu.git synced 2026-07-06 00:20:22 -07:00
Code Issues Packages Projects Releases Wiki Activity

Add NEON optimizations for S3TC

Browse Source
This commit is contained in:
izzy2lost
2026-03-20 21:44:03 -04:00
parent 0cff80455c
commit dd5ea8f934
1 changed files with 95 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files
hw/xbox/nv2a/pgraph/s3tc.c
+95
View File
@@ -23,6 +23,11 @@
*/
#include "qemu/osdep.h"
#ifdef __aarch64__
#include <arm_neon.h>
#endif
#include "s3tc.h"
static void decode_bc1_colors(uint16_t c0, uint16_t c1, uint8_t r[4],
@@ -62,6 +67,88 @@ static void decode_bc1_colors(uint16_t c0, uint16_t c1, uint8_t r[4],
}
}
#ifdef __aarch64__
static inline uint8x8_t s3tc_make_lookup_table(uint8_t v0, uint8_t v1,
uint8_t v2, uint8_t v3)
{
uint32_t packed = v0 | ((uint32_t)v1 << 8) | ((uint32_t)v2 << 16) |
((uint32_t)v3 << 24);
return vreinterpret_u8_u32(vdup_n_u32(packed));
}
static inline uint8x8_t s3tc_make_index_vector(uint8_t packed_indices)
{
uint64_t indices = (packed_indices & 0x03) |
(((uint64_t)((packed_indices >> 2) & 0x03)) << 8) |
(((uint64_t)((packed_indices >> 4) & 0x03)) << 16) |
(((uint64_t)((packed_indices >> 6) & 0x03)) << 24);
return vcreate_u8(indices);
}
static inline uint8x8_t s3tc_load_alpha_row(const uint8_t *alpha_row)
{
uint32_t packed = alpha_row[0] | ((uint32_t)alpha_row[1] << 8) |
((uint32_t)alpha_row[2] << 16) |
((uint32_t)alpha_row[3] << 24);
return vreinterpret_u8_u32(vdup_n_u32(packed));
}
static inline void s3tc_store_rgba_row(uint8_t *dst, uint8x8_t r,
uint8x8_t g, uint8x8_t b, uint8x8_t a)
{
uint8x8_t rg = vzip1_u8(r, g);
uint8x8_t ba = vzip1_u8(b, a);
uint16x4_t rgba_lo =
vzip1_u16(vreinterpret_u16_u8(rg), vreinterpret_u16_u8(ba));
uint16x4_t rgba_hi =
vzip2_u16(vreinterpret_u16_u8(rg), vreinterpret_u16_u8(ba));
vst1q_u8(dst, vcombine_u8(vreinterpret_u8_u16(rgba_lo),
vreinterpret_u8_u16(rgba_hi)));
}
static bool write_block_to_texture_neon(uint8_t *converted_data, uint32_t indices,
int i, int j, int width, int height,
int z_pos_factor, const uint8_t r[4],
const uint8_t g[4], const uint8_t b[4],
const uint8_t a[16],
bool separate_alpha)
{
int x0 = i * 4;
int y0 = j * 4;
uint8x8_t r_table;
uint8x8_t g_table;
uint8x8_t b_table;
uint8x8_t a_table;
if (x0 + 4 > width || y0 + 4 > height) {
return false;
}
r_table = s3tc_make_lookup_table(r[0], r[1], r[2], r[3]);
g_table = s3tc_make_lookup_table(g[0], g[1], g[2], g[3]);
b_table = s3tc_make_lookup_table(b[0], b[1], b[2], b[3]);
a_table = s3tc_make_lookup_table(a[0], a[1], a[2], a[3]);
for (int row = 0; row < 4; row++) {
uint8_t packed_row_indices = (indices >> (row * 8)) & 0xFF;
uint8x8_t row_indices = s3tc_make_index_vector(packed_row_indices);
uint8x8_t row_r = vtbl1_u8(r_table, row_indices);
uint8x8_t row_g = vtbl1_u8(g_table, row_indices);
uint8x8_t row_b = vtbl1_u8(b_table, row_indices);
uint8x8_t row_a = separate_alpha
? s3tc_load_alpha_row(a + row * 4)
: vtbl1_u8(a_table, row_indices);
uint8_t *dst = converted_data +
(z_pos_factor + (y0 + row) * width + x0) * 4;
s3tc_store_rgba_row(dst, row_r, row_g, row_b, row_a);
}
return true;
}
#endif
static void write_block_to_texture(uint8_t *converted_data, uint32_t indices,
int i, int j, int width, int height,
int z_pos_factor, uint8_t r[4],
@@ -74,6 +161,14 @@ static void write_block_to_texture(uint8_t *converted_data, uint32_t indices,
int x1 = x0 + 4,
y1 = y0 + 4;
#ifdef __aarch64__
if (write_block_to_texture_neon(converted_data, indices, i, j, width,
height, z_pos_factor, r, g, b, a,
separate_alpha)) {
return;
}
#endif
for (int y = y0; y < y1 && y < height; y++) {
int y_index = 4 * (y - y0);
int z_plus_y_pos_factor = z_pos_factor + y * width;