Sprinter/mame
2
0
Fork 0
mirror of https://github.com/holub/mame synced 2025-07-03 17:08:39 +03:00
Code Issues Actions 1 Packages Projects Releases Wiki Activity

voodoo: Few more SSE optimizations. (nw)

Browse Source
This commit is contained in:
Ted Green 2017-10-15 12:11:10 -06:00
parent 89157c45ec
commit 1c424550a1
4 changed files with 68 additions and 44 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

89
src/devices/video/vooddefs.h
View File

@ -1044,20 +1044,20 @@ static inline void ATTR_FORCE_INLINE alphaBlend(uint32_t FBZMODE, uint32_t ALPHA
{
default: /* reserved */
case 0: /* AZERO */
srcScale.set(srcAlphaScale, 0, 0, 0);
srcScale.zero();
//(RR) = (GG) = (BB) = 0;
break;
case 1: /* ASRC_ALPHA */
ta = sa + 1;
srcScale.set(srcAlphaScale, ta, ta, ta);
srcScale.set_all(ta);
//(RR) = (sr * (sa + 1)) >> 8;
//(GG) = (sg * (sa + 1)) >> 8;
//(BB) = (sb * (sa + 1)) >> 8;
break;
case 2: /* A_COLOR */
srcScale.set(srcAlphaScale-1, dr, dg, db);
srcScale.set(dr, dr, dg, db);
srcScale.add_imm(1);
//(RR) = (sr * (dr + 1)) >> 8;
//(GG) = (sg * (dg + 1)) >> 8;
@ -1066,26 +1066,26 @@ static inline void ATTR_FORCE_INLINE alphaBlend(uint32_t FBZMODE, uint32_t ALPHA
case 3: /* ADST_ALPHA */
ta = da + 1;
srcScale.set(srcAlphaScale, ta, ta, ta);
srcScale.set_all(ta);
//(RR) = (sr * (da + 1)) >> 8;
//(GG) = (sg * (da + 1)) >> 8;
//(BB) = (sb * (da + 1)) >> 8;
break;
case 4: /* AONE */
srcScale.set(srcAlphaScale, 256, 256, 256);
srcScale.set_all(256);
break;
case 5: /* AOMSRC_ALPHA */
ta = (0x100 - sa);
srcScale.set(srcAlphaScale, ta, ta, ta);
srcScale.set_all(ta);
//(RR) = (sr * (0x100 - sa)) >> 8;
//(GG) = (sg * (0x100 - sa)) >> 8;
//(BB) = (sb * (0x100 - sa)) >> 8;
break;
case 6: /* AOM_COLOR */
srcScale.set(srcAlphaScale, (0x100 - dr), (0x100 - dg), (0x100 - db));
srcScale.set((0x100 - dr), (0x100 - dr), (0x100 - dg), (0x100 - db));
//(RR) = (sr * (0x100 - dr)) >> 8;
//(GG) = (sg * (0x100 - dg)) >> 8;
//(BB) = (sb * (0x100 - db)) >> 8;
@ -1093,7 +1093,7 @@ static inline void ATTR_FORCE_INLINE alphaBlend(uint32_t FBZMODE, uint32_t ALPHA
case 7: /* AOMDST_ALPHA */
ta = (0x100 - da);
srcScale.set(srcAlphaScale, ta, ta, ta);
srcScale.set_all(ta);
//(RR) = (sr * (0x100 - da)) >> 8;
//(GG) = (sg * (0x100 - da)) >> 8;
//(BB) = (sb * (0x100 - da)) >> 8;
@ -1102,12 +1102,14 @@ static inline void ATTR_FORCE_INLINE alphaBlend(uint32_t FBZMODE, uint32_t ALPHA
case 15: /* ASATURATE */
ta = (sa < (0x100 - da)) ? sa : (0x100 - da);
ta++;
srcScale.set(srcAlphaScale, ta, ta, ta);
srcScale.set_all(ta);
//(RR) = (sr * (ta + 1)) >> 8;
//(GG) = (sg * (ta + 1)) >> 8;
//(BB) = (sb * (ta + 1)) >> 8;
break;
}
// Set srcScale alpha
srcScale.set_a16(srcAlphaScale);
/* blend the dest alpha */
if (ALPHAMODE_DSTALPHABLEND(ALPHAMODE) == 4)
@ -1121,12 +1123,12 @@ static inline void ATTR_FORCE_INLINE alphaBlend(uint32_t FBZMODE, uint32_t ALPHA
{
default: /* reserved */
case 0: /* AZERO */
destScale.set(destAlphaScale, 0, 0, 0);
destScale.zero();
break;
case 1: /* ASRC_ALPHA */
ta = sa + 1;
destScale.set(destAlphaScale, ta, ta, ta);
destScale.set_all(ta);
//(RR) += (dr * (sa + 1)) >> 8;
//(GG) += (dg * (sa + 1)) >> 8;
//(BB) += (db * (sa + 1)) >> 8;
@ -1135,7 +1137,6 @@ static inline void ATTR_FORCE_INLINE alphaBlend(uint32_t FBZMODE, uint32_t ALPHA
case 2: /* A_COLOR */
destScale.set(srcColor);
destScale.add_imm(1);
destScale.set_a(destAlphaScale);
//(RR) += (dr * (sr + 1)) >> 8;
//(GG) += (dg * (sg + 1)) >> 8;
//(BB) += (db * (sb + 1)) >> 8;
@ -1143,14 +1144,14 @@ static inline void ATTR_FORCE_INLINE alphaBlend(uint32_t FBZMODE, uint32_t ALPHA
case 3: /* ADST_ALPHA */
ta = da + 1;
destScale.set(destAlphaScale, ta, ta, ta);
destScale.set_all(ta);
//(RR) += (dr * (da + 1)) >> 8;
//(GG) += (dg * (da + 1)) >> 8;
//(BB) += (db * (da + 1)) >> 8;
break;
case 4: /* AONE */
destScale.set(destAlphaScale, 256, 256, 256);
destScale.set_all(256);
//(RR) += dr;
//(GG) += dg;
//(BB) += db;
@ -1158,16 +1159,15 @@ static inline void ATTR_FORCE_INLINE alphaBlend(uint32_t FBZMODE, uint32_t ALPHA
case 5: /* AOMSRC_ALPHA */
ta = (0x100 - sa);
destScale.set(destAlphaScale, ta, ta, ta);
destScale.set_all(ta);
//(RR) += (dr * (0x100 - sa)) >> 8;
//(GG) += (dg * (0x100 - sa)) >> 8;
//(BB) += (db * (0x100 - sa)) >> 8;
break;
case 6: /* AOM_COLOR */
destScale.set(0x100, 0x100, 0x100, 0x100);
destScale.set_all(0x100);
destScale.sub(srcColor);
destScale.set_a(destAlphaScale);
//destScale.set(destAlphaScale, (0x100 - color.rgb.r), (0x100 - color.rgb.g), (0x100 - color.rgb.b));
//(RR) += (dr * (0x100 - sr)) >> 8;
//(GG) += (dg * (0x100 - sg)) >> 8;
@ -1176,7 +1176,7 @@ static inline void ATTR_FORCE_INLINE alphaBlend(uint32_t FBZMODE, uint32_t ALPHA
case 7: /* AOMDST_ALPHA */
ta = (0x100 - da);
destScale.set(destAlphaScale, ta, ta, ta);
destScale.set_all(ta);
//(RR) += (dr * (0x100 - da)) >> 8;
//(GG) += (dg * (0x100 - da)) >> 8;
//(BB) += (db * (0x100 - da)) >> 8;
@ -1185,14 +1185,15 @@ static inline void ATTR_FORCE_INLINE alphaBlend(uint32_t FBZMODE, uint32_t ALPHA
case 15: /* A_COLORBEFOREFOG */
destScale.set(preFog);
destScale.add_imm(1);
destScale.set_a(destAlphaScale);
//destScale.set((rgb_t) (((destAlphaScale-1)<<24) | (preFog.u & 0x00ffffff)));
//destScale.add_imm(1);
//(RR) += (dr * (prefogr + 1)) >> 8;
//(GG) += (dg * (prefogg + 1)) >> 8;
//(BB) += (db * (prefogb + 1)) >> 8;
break;
}
// Set destScale alpha
destScale.set_a16(destAlphaScale);
// Main blend
rgbaint_t destColor(da, dr, dg, db);
@ -1354,7 +1355,7 @@ static inline void ATTR_FORCE_INLINE applyFogging(voodoo_device *vd, uint32_t fb
/* if fog_add is zero, we start with the fog color */
if (FOGMODE_FOG_ADD(fogModeReg))
fogColorLocal.set(0, 0, 0, 0);
fogColorLocal.zero();
//fr = fg = fb = 0;
/* if fog_mult is zero, we subtract the incoming color */
@ -1449,7 +1450,7 @@ static inline void ATTR_FORCE_INLINE applyFogging(voodoo_device *vd, uint32_t fb
//CLAMP((RR), 0x00, 0xff);
//CLAMP((GG), 0x00, 0xff);
//CLAMP((BB), 0x00, 0xff);
fogColorLocal.merge_alpha(color);
fogColorLocal.merge_alpha16(color);
color.set(fogColorLocal);
}
}
@ -2412,15 +2413,15 @@ inline bool ATTR_FORCE_INLINE voodoo_device::combineColor(voodoo_device *vd, sta
switch (FBZCP_CC_ASELECT(FBZCOLORPATH))
{
case 0: /* iterated alpha */
c_other.merge_alpha(srcColor);
c_other.merge_alpha16(srcColor);
break;
case 1: /* texture alpha */
c_other.merge_alpha(TEXELARGB);
c_other.merge_alpha16(TEXELARGB);
break;
case 2: /* color1 alpha */
c_other.set_a(vd->reg[color1].rgb.a);
c_other.set_a16(vd->reg[color1].rgb.a);
break;
default: /* reserved - voodoo3 LFB Alpha*/
@ -2456,18 +2457,18 @@ inline bool ATTR_FORCE_INLINE voodoo_device::combineColor(voodoo_device *vd, sta
{
default:
case 0: /* iterated alpha */
c_local.merge_alpha(srcColor);
c_local.merge_alpha16(srcColor);
break;
case 1: /* color0 alpha */
c_local.set_a(vd->reg[color0].rgb.a);
c_local.set_a16(vd->reg[color0].rgb.a);
break;
case 2: /* clamped iterated Z[27:20] */
{
int temp;
CLAMPED_Z(ITERZ, FBZCOLORPATH, temp);
c_local.set_a((uint8_t) temp);
c_local.set_a16((uint8_t) temp);
break;
}
@ -2475,7 +2476,7 @@ inline bool ATTR_FORCE_INLINE voodoo_device::combineColor(voodoo_device *vd, sta
{
int temp;
CLAMPED_W(ITERW, FBZCOLORPATH, temp); /* Voodoo 2 only */
c_local.set_a((uint8_t) temp);
c_local.set_a16((uint8_t) temp);
break;
}
}
@ -2491,7 +2492,7 @@ inline bool ATTR_FORCE_INLINE voodoo_device::combineColor(voodoo_device *vd, sta
if (FBZCP_CCA_ZERO_OTHER(FBZCOLORPATH))
blend_color.zero_alpha();
else
blend_color.merge_alpha(c_other);
blend_color.merge_alpha16(c_other);
/* subtract a/c_local */
if (FBZCP_CC_SUB_CLOCAL(FBZCOLORPATH) || (FBZCP_CCA_SUB_CLOCAL(FBZCOLORPATH)))
@ -2506,7 +2507,7 @@ inline bool ATTR_FORCE_INLINE voodoo_device::combineColor(voodoo_device *vd, sta
if (!FBZCP_CCA_SUB_CLOCAL(FBZCOLORPATH))
sub_val.zero_alpha();
else
sub_val.merge_alpha(c_local);
sub_val.merge_alpha16(c_local);
blend_color.sub(sub_val);
}
@ -2550,15 +2551,15 @@ inline bool ATTR_FORCE_INLINE voodoo_device::combineColor(voodoo_device *vd, sta
case 1: /* a_local */
case 3: /* a_local */
blend_factor.merge_alpha(c_local);
blend_factor.merge_alpha16(c_local);
break;
case 2: /* a_other */
blend_factor.merge_alpha(c_other);
blend_factor.merge_alpha16(c_other);
break;
case 4: /* texture alpha */
blend_factor.merge_alpha(TEXELARGB);
blend_factor.merge_alpha16(TEXELARGB);
break;
}
@ -2598,7 +2599,7 @@ inline bool ATTR_FORCE_INLINE voodoo_device::combineColor(voodoo_device *vd, sta
add_val.zero_alpha();
else
//color.rgb.a += c_local.rgb.a;
add_val.merge_alpha(c_local);
add_val.merge_alpha16(c_local);
/* clamp */
//CLAMP(color.rgb.a, 0x00, 0xff);
@ -3038,7 +3039,7 @@ inline rgbaint_t ATTR_FORCE_INLINE voodoo_device::tmu_state::combineTexture(cons
if (TEXMODE_TCA_ZERO_OTHER(TEXMODE))
blend_color.zero_alpha();
else
blend_color.merge_alpha(c_other);
blend_color.merge_alpha16(c_other);
if (TEXMODE_TC_SUB_CLOCAL(TEXMODE) || TEXMODE_TCA_SUB_CLOCAL(TEXMODE))
{
@ -3053,7 +3054,7 @@ inline rgbaint_t ATTR_FORCE_INLINE voodoo_device::tmu_state::combineTexture(cons
if (!TEXMODE_TCA_SUB_CLOCAL(TEXMODE))
sub_val.zero_alpha();
else
sub_val.merge_alpha(c_local);
sub_val.merge_alpha16(c_local);
blend_color.sub(sub_val);
}
@ -3105,15 +3106,15 @@ inline rgbaint_t ATTR_FORCE_INLINE voodoo_device::tmu_state::combineTexture(cons
break;
case 1: /* c_local */
blend_factor.merge_alpha(c_local);
blend_factor.merge_alpha16(c_local);
break;
case 2: /* a_other */
blend_factor.merge_alpha(c_other);
blend_factor.merge_alpha16(c_other);
break;
case 3: /* a_local */
blend_factor.merge_alpha(c_local);
blend_factor.merge_alpha16(c_local);
break;
case 4: /* LOD (detail factor) */
@ -3125,12 +3126,12 @@ inline rgbaint_t ATTR_FORCE_INLINE voodoo_device::tmu_state::combineTexture(cons
tmp = (((detailbias - lod) << detailscale) >> 8);
if (tmp > detailmax)
tmp = detailmax;
blend_factor.set_a(tmp);
blend_factor.set_a16(tmp);
}
break;
case 5: /* LOD fraction */
blend_factor.set_a(lod & 0xff);
blend_factor.set_a16(lod & 0xff);
break;
}
@ -3169,7 +3170,7 @@ inline rgbaint_t ATTR_FORCE_INLINE voodoo_device::tmu_state::combineTexture(cons
if (!TEXMODE_TCA_ADD_ACLOCAL(TEXMODE))
add_val.zero_alpha();
else
add_val.merge_alpha(c_local);
add_val.merge_alpha16(c_local);
/* clamp */
//result.rgb.r = (tr < 0) ? 0 : (tr > 0xff) ? 0xff : tr;

6
src/emu/video/rgbgen.h
View File

@ -55,6 +55,7 @@ public:
return rgb_t(a, r, g, b);
}
void set_a16(const s32 value) { m_a = value; }
void set_a(const s32 value) { m_a = value; }
void set_r(const s32 value) { m_r = value; }
void set_g(const s32 value) { m_g = value; }
@ -349,6 +350,11 @@ public:
m_b = (m_b < b) ? 0xffffffff : 0;
}
void merge_alpha16(const rgbaint_t& alpha)
{
m_a = alpha.m_a;
}
void merge_alpha(const rgbaint_t& alpha)
{
m_a = alpha.m_a;

6
src/emu/video/rgbsse.h
View File

@ -58,6 +58,7 @@ public:
return _mm_cvtsi128_si32(_mm_packus_epi16(_mm_packs_epi32(m_value, _mm_setzero_si128()), _mm_setzero_si128()));
}
void set_a16(const s32 value) { m_value = _mm_insert_epi16(m_value, value, 6); }
#ifdef __SSE4_1__
void set_a(const s32 value) { m_value = _mm_insert_epi32(m_value, value, 3); }
void set_r(const s32 value) { m_value = _mm_insert_epi32(m_value, value, 2); }
@ -415,6 +416,11 @@ public:
return *this;
}
inline void merge_alpha16(const rgbaint_t& alpha)
{
m_value = _mm_insert_epi16(m_value, _mm_extract_epi16(alpha.m_value, 6), 6);
}
inline void merge_alpha(const rgbaint_t& alpha)
{
#ifdef __SSE4_1__

11
src/emu/video/rgbvmx.h
View File

@ -100,6 +100,12 @@ public:
return result;
}
void set_a16(const s32 value)
{
const VECS32 temp = { value, value, value, value };
m_value = vec_perm(m_value, temp, alpha_perm);
}
void set_a(const s32 value)
{
const VECS32 temp = { value, value, value, value };
@ -606,6 +612,11 @@ public:
return *this;
}
inline void merge_alpha16(const rgbaint_t& alpha)
{
m_value = vec_perm(m_value, alpha.m_value, alpha_perm);
}
inline void merge_alpha(const rgbaint_t& alpha)
{
m_value = vec_perm(m_value, alpha.m_value, alpha_perm);