MrMonopoly/raylib
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

use optionnal LUT for uint8_t -> float conversion

Browse Source
This commit is contained in:
Bigfoot71 2026-03-18 21:23:50 +01:00
parent 7f2c9a0827
commit bdc2a6e26e
1 changed files with 25 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

25
src/external/rlsw.h vendored
View File

@ -153,6 +153,17 @@
#define SW_MAX_TEXTURES 128 #define SW_MAX_TEXTURES 128
#endif #endif
// Enables the use of a lookup table for uint8_t to float conversion
// Requires an additional 1KB of global memory
// Disabled when SIMD intrinsics are enabled
#ifndef SW_USE_COLOR_LUT
#if RLSW_USE_SIMD_INTRINSICS
#define SW_USE_COLOR_LUT false
#else
#define SW_USE_COLOR_LUT true
#endif
#endif
// Under normal circumstances, clipping a polygon can add at most one vertex per clipping plane // Under normal circumstances, clipping a polygon can add at most one vertex per clipping plane
// Considering the largest polygon involved is a quadrilateral (4 vertices), // Considering the largest polygon involved is a quadrilateral (4 vertices),
// and that clipping occurs against both the frustum (6 planes) and the scissors (4 planes), // and that clipping occurs against both the frustum (6 planes) and the scissors (4 planes),
@ -1060,6 +1071,10 @@ typedef struct {
//---------------------------------------------------------------------------------- //----------------------------------------------------------------------------------
static sw_context_t RLSW = { 0 }; static sw_context_t RLSW = { 0 };
#if SW_USE_COLOR_LUT
static float SW_LUT_U8_TO_F32[256] = { 0 };
#endif
//---------------------------------------------------------------------------------- //----------------------------------------------------------------------------------
// Internal Constants Definition // Internal Constants Definition
//---------------------------------------------------------------------------------- //----------------------------------------------------------------------------------
@ -1677,6 +1692,12 @@ static inline void sw_pixel_color8_to_color(float *SW_RESTRICT dst, const uint8_
vfloat32m1_t vnorm = __riscv_vfmul_vf_f32m1(vsrc_f32, SW_INV_255, vl); // Multiply by 1/255.0 to normalize vfloat32m1_t vnorm = __riscv_vfmul_vf_f32m1(vsrc_f32, SW_INV_255, vl); // Multiply by 1/255.0 to normalize
__riscv_vse32_v_f32m1(dst, vnorm, vl); // Store result __riscv_vse32_v_f32m1(dst, vnorm, vl); // Store result
#elif SW_USE_COLOR_LUT
dst[0] = SW_LUT_U8_TO_F32[src[0]];
dst[1] = SW_LUT_U8_TO_F32[src[1]];
dst[2] = SW_LUT_U8_TO_F32[src[2]];
dst[3] = SW_LUT_U8_TO_F32[src[3]];
#else #else
dst[0] = (float)src[0]*SW_INV_255; dst[0] = (float)src[0]*SW_INV_255;
dst[1] = (float)src[1]*SW_INV_255; dst[1] = (float)src[1]*SW_INV_255;
@ -3936,6 +3957,10 @@ bool swInit(int w, int h)
RLSW.polyMode = SW_FILL; RLSW.polyMode = SW_FILL;
RLSW.cullFace = SW_BACK; RLSW.cullFace = SW_BACK;
#if SW_USE_COLOR_LUT
for (int i = 0; i < 256; i++) SW_LUT_U8_TO_F32[i] = (float)i*SW_INV_255;
#endif
SW_LOG("INFO: RLSW: Software renderer initialized successfully\n"); SW_LOG("INFO: RLSW: Software renderer initialized successfully\n");
#if defined(SW_HAS_FMA_AVX) && defined(SW_HAS_FMA_AVX2) #if defined(SW_HAS_FMA_AVX) && defined(SW_HAS_FMA_AVX2)
SW_LOG("INFO: RLSW: Using SIMD instructions: FMA AVX\n"); SW_LOG("INFO: RLSW: Using SIMD instructions: FMA AVX\n");