bilbo/r500
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

matrix_cubesphere_specular: rewrite specular fragment shader

Browse Source
This commit is contained in:
Zack Buhman 2025-10-31 20:15:11 -05:00
parent 30c018a748
commit 2b90395b2d
4 changed files with 86 additions and 81 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
drm/matrix_cubesphere_specular.cpp
View File

@ -365,8 +365,8 @@ int _3d_cube_inner(int ix, mat4x4 trans, mat4x4 world_trans, vec4 light_pos, vec
vec2 t = model->texture[obj->triangle[i][j].texture];
vec3 n = model->normal[obj->triangle[i][j].normal];
fprintf(stderr, "% 2.03f, % 2.03f, % 2.03f, % 2.03f, % 2.03f, % 2.03f,\n",
p.x, p.y, p.z, n.x, n.y, n.z);
//fprintf(stderr, "% 2.03f, % 2.03f, % 2.03f, % 2.03f, % 2.03f, % 2.03f,\n",
//p.x, p.y, p.z, n.x, n.y, n.z);
ib[ix++].f32 = p.x;
ib[ix++].f32 = p.y;

133
drm/matrix_cubesphere_specular.fs.asm
View File

@ -9,94 +9,97 @@
TEX TEX_SEM_WAIT TEX_SEM_ACQUIRE
temp[4].rgba = LD tex[0].rgba temp[4].rgaa ;
--
-- normal = normalize(normal)
-- normal = (1.0 / sqrt(dot(normal, normal))) * normal
src0.rgb = temp[1] :
--
src0.rgb = temp[1] : -- normal
DP3 src0.rgb src0.rgb ,
temp[1].a = DP ;
src0.a = temp[1] :
temp[1].a = RSQ |src0.a| ;
src0.a = temp[1], src0.rgb = temp[1] :
src0.a = temp[1] ,
src0.rgb = temp[1] : -- normal
temp[1].rgb = MAD src0.rgb src0.aaa src0.000 ;
-- light_dir = normalize((f_light_pos - f_world_pos))
src1.rgb = temp[2] , -- f_light_pos
src0.rgb = temp[0] , -- f_world_pos
srcp.rgb = neg : -- (f_light_pos - f_world_pos)
DP3 srcp.rgb srcp.rgb ,
--
-- light_dir = light_pos - world_pos
--
src0.rgb = temp[2] , -- light pos
src1.rgb = temp[0] : -- world pos
temp[2].rgb = MAD src0.111 src0.rgb -src1.rgb ;
--
-- light_dir = normalize(light_dir)
--
src0.rgb = temp[2] : -- light_dir
DP3 src0.rgb src0.rgb ,
temp[2].a = DP ;
src0.a = temp[2] :
temp[2].a = RSQ |src0.a| ;
src0.a = temp[2], src0.rgb = temp[2] :
src0.a = temp[2] ,
src0.rgb = temp[2] : -- light_dir
temp[2].rgb = MAD src0.rgb src0.aaa src0.000 ;
-- diff = dot(normal, light_dir)
src0.rgb = temp[2] ,
src1.rgb = temp[1] :
DP3 src0.rgb src1.rgb ,
temp[5].a = DP ;
-- diff = max(diff, 0)
src0.a = temp[5] :
temp[5].a = MAX src0.a src0.0 ;
-- intensity = diff + 0.125
src0.a = temp[5] ,
src1.a = float(32) : -- 0.125
temp[5].a = MAD src0.a src0.1 src1.a ;
--
-- view_dir = view_pos - world_pos
--
src0.rgb = temp[3] , -- view pos
src1.rgb = temp[0] : -- world pos
temp[3].rgb = MAD src0.111 src0.rgb -src1.rgb ;
--
-- specular
-- view_dir = normalize(view_dir)
--
-- temp[3] -- view pos (world space)
-- view_dir = normalize(f_view_pos - f_world_pos)
src1.rgb = temp[3] , -- f_light_pos
src0.rgb = temp[0] , -- f_world_pos
srcp.rgb = neg : -- (f_light_pos - f_world_pos)
DP3 srcp.rgb srcp.rgb ,
src0.rgb = temp[3] : -- view dir
DP3 src0.rgb src0.rgb ,
temp[3].a = DP ;
src0.a = temp[3] :
temp[3].a = RSQ |src0.a| ;
src0.a = temp[3], src0.rgb = temp[3] :
src0.a = temp[3] ,
src0.rgb = temp[3] : -- view dir
temp[3].rgb = MAD src0.rgb src0.aaa src0.000 ;
-- reflect(I, N)
-- I - 2.0 * dot(N, I) * N
-- reflect_dir = reflect(-light_dir, norm)
-- reflect_dir = reflect(-temp[2], temp[1])
-- I - 2.0 * dot(N, I) * N
-- - (2.0 * dot(N, I)) * temp[1] + -temp[2]
src0.rgb = temp[1] , -- N=normal
src1.rgb = temp[2] : -- I=light_dir dot(N, -I)
temp[6].r = DP3 src0.rgb -src1.rgb ;
src0.rgb = temp[6] ,
src1.rgb = float(64) : -- 2.0
temp[6].r = MAD src0.r00 src1.r00 src0.000 ;
src0.rgb = temp[6] ,
src1.rgb = temp[1] , -- N
src2.rgb = temp[2] : -- I
temp[6].rgb = MAD -src0.rrr src1.rgb -src2.rgb ;
--
-- reflect_dir = reflect(light_dir, normal)
--
-- dotLN = dot(-light_dir, normal)
src0.rgb = temp[2] , -- light dir
src1.rgb = temp[1] : -- normal
DP3 -src0.rgb src1.rgb ,
temp[5].a = DP ;
-- dotLN = 2.0 * dotLN
src0.a = temp[5] , -- dotLN
src1.a = float(64) : -- 2.0
temp[5].a = MAD src0.a src1.a src0.0 ;
-- dotLN = -dotLN * normal + -light_dir
src0.a = temp[5] , -- dotLN
src1.rgb = temp[1] , -- normal
src2.rgb = temp[2] : -- light dir
temp[5].rgb = MAD -src0.aaa src1.rgb -src2.rgb ;
--
-- spec = max(dot(view_dir, reflect_dir), 0.0)
src0.rgb = temp[6] , -- reflect_dir
src1.rgb = temp[3] : -- view_dir
temp[6].r = DP3 src0.rgb src1.rgb ;
src0.rgb = temp[6] :
temp[6].a = MAX src0.r src0.0 ;
--
src0.rgb = temp[3] ,
src1.rgb = temp[5] :
DP3 src0.rgb src1.rgb ,
temp[5].a = DP ;
src0.a = temp[5] :
temp[5].a = MAX src0.a src0.0 ;
--
-- spec = pow(spec, 32)
src0.a = temp[6] :
temp[6].a = LN2 src0.a ;
src0.a = temp[6] ,
src1.a = float(72) : -- 32
temp[6].a = MAD src0.a src1.a src1.0 ;
src0.a = temp[6] :
temp[6].a = EX2 src0.a ;
--
src0.a = temp[5] : -- spec
temp[5].a = LN2 src0.a ;
src0.a = temp[5] , -- spec
src1.a = float(96) : -- 32
temp[5].a = MAD src0.a src1.a src1.0 ;
src0.a = temp[5] : -- spec
temp[5].a = EX2 src0.a ;
OUT TEX_SEM_WAIT
src0.a = temp[4],
src0.rgb = temp[4] ,
src1.a = temp[6] ,
src1.rgb = temp[6] :
out[0].a = MAX src0.a src0.a ,
out[0].rgb = MAD src0.111 src1.aaa src1.000 ;
src1.a = temp[5] ,
src1.rgb = temp[5] :
out[0].a = MAX src1.1 src0.1 ,
out[0].rgb = MAD src1.111 src1.aaa src1.000 ;

12
drm/matrix_cubesphere_specular.vs.asm
View File

@ -37,12 +37,14 @@ temp[3].z = VE_DOT const[6].xyz0 input[2].xyz0 ;
-- position (clip space)
out[0].xyzw = VE_ADD temp[1].xyzw const[0].0000 ;
-- position (world space)
out[1].xyzw = VE_ADD temp[2].xyzw const[0].0000 ;
out[1].xyzw = VE_ADD temp[2].xyzw const[0].0000 ; -- 0 world pos
-- normal
out[2].xyzw = VE_ADD temp[3].xyz0 const[0].0000 ;
out[2].xyzw = VE_ADD temp[3].xyz0 const[0].0000 ; -- 1 normal
-- light pos (world space)
out[3].xyzw = VE_ADD const[8].xyzw const[8].0000 ;
out[3].xyzw = VE_ADD const[8].xyzw const[8].0000 ; -- 2 light pos / light dir
-- view pos (world space)
out[4].xyzw = VE_ADD const[9].xyzw const[9].0000 ;
out[4].xyzw = VE_ADD const[9].xyzw const[9].0000 ; -- 3 view pos / view dir
-- texture
out[5].xyzw = VE_ADD input[1].xy00 const[0].0000 ;
out[5].xyzw = VE_ADD input[1].xy00 const[0].0000 ; -- 4 texture
-- 5 reflect dir

2
regs/assembler/vs/validator.py
View File

@ -382,7 +382,7 @@ def validate_instruction(ins):
if len(opcodes) == 2:
return validate_dual_math_instruction(ins.operations, opcodes)
else:
assert len(opcodes) == 1
assert len(opcodes) == 1, (opcodes, ins)
return validate_instruction_inner(ins.operations[0], opcodes[0])
if __name__ == "__main__":