gsdx-ogl: use SW blending when no barrier is required

Speed penality is small (only GPU) but it is more accurate
This commit is contained in:
Gregory Hainaut 2015-07-18 19:55:32 +02:00
parent 8c8fe633a5
commit 5c1b8986c6
1 changed files with 37 additions and 34 deletions

View File

@ -417,7 +417,7 @@ void GSRendererOGL::DrawPrims(GSTexture* rt, GSTexture* ds, GSTextureCache::Sour
// Format of the output
GIFRegALPHA ALPHA = context->ALPHA;
const GIFRegALPHA& ALPHA = context->ALPHA;
float afix = (float)context->ALPHA.FIX / 0x80;
// Blend
@ -452,16 +452,50 @@ void GSRendererOGL::DrawPrims(GSTexture* rt, GSTexture* ds, GSTextureCache::Sour
}
}
// Compute the blending equation to detect special case
int blend_sel = ((om_bsel.a * 3 + om_bsel.b) * 3 + om_bsel.c) * 3 + om_bsel.d;
int blend_flag = GSDeviceOGL::m_blendMapD3D9[blend_sel].bogus;
// SW Blend is (nearly) free. Let's use it.
int free_blend = m_sw_blending && ((blend_flag & NO_BAR) || (m_prim_overlap == PRIM_OVERLAP_NO));
// Color clip
bool colclip_wrap = env.COLCLAMP.CLAMP == 0 && !tex && PRIM->PRIM != GS_POINTLIST;
bool acc_colclip_wrap = env.COLCLAMP.CLAMP == 0 && (m_sw_blending >= ACC_BLEND_CCLIP || free_blend);
if ((ALPHA.A == ALPHA.B) || !om_bsel.abe) { // Optimize-away colclip
// No addition neither substraction so no risk of overflow the [0:255] range.
colclip_wrap = false;
acc_colclip_wrap = false;
}
if (acc_colclip_wrap) {
colclip_wrap = false;
ps_sel.colclip = 3;
GL_INS("COLCLIP SW ENABLED (blending is %d/%d/%d/%d)", ALPHA.A, ALPHA.B, ALPHA.C, ALPHA.D);
} else if (colclip_wrap) {
ps_sel.colclip = 1;
GL_INS("COLCLIP ENABLED (blending is %d/%d/%d/%d)", ALPHA.A, ALPHA.B, ALPHA.C, ALPHA.D);
}
bool impossible_blend = m_sw_blending && (blend_flag & A_MAX);
bool all_blend_sw = (m_sw_blending >= ACC_BLEND_ULTRA)
|| (m_sw_blending >= ACC_BLEND_FULL && !( (ALPHA.A == ALPHA.B) || (ALPHA.C == 2 && afix <= 1.002f) ));
bool sw_blending = free_blend // Free case
|| impossible_blend || all_blend_sw // Impossible blend or all
|| acc_colclip_wrap // accurate colclip
|| ps_sel.fbmask; // accurate fbmask
if (ps_sel.dfmt == 1) {
if (ALPHA.C == 1) {
if (ALPHA.C == 1 && !sw_blending) {
// 24 bits no alpha channel so use 1.0f fix factor as equivalent
ALPHA.C = 2;
om_bsel.c = 2;
afix = 1.0f;
}
// Disable writing of the alpha channel
om_csel.wa = 0;
}
// DATE
if (DATE) {
if (GLLoader::found_GL_ARB_texture_barrier && (m_prim_overlap == PRIM_OVERLAP_NO)) {
DATE_GL45 = true;
@ -480,8 +514,6 @@ void GSRendererOGL::DrawPrims(GSTexture* rt, GSTexture* ds, GSTextureCache::Sour
}
}
// DATE
if (DATE_GL45) {
gl_TextureBarrier();
dev->PSSetShaderResource(3, rt);
@ -613,22 +645,6 @@ void GSRendererOGL::DrawPrims(GSTexture* rt, GSTexture* ds, GSTextureCache::Sour
om_dssel.date = 1;
}
bool colclip_wrap = env.COLCLAMP.CLAMP == 0 && !tex && PRIM->PRIM != GS_POINTLIST;
bool acc_colclip_wrap = env.COLCLAMP.CLAMP == 0 && (m_sw_blending >= ACC_BLEND_CCLIP || (m_prim_overlap == PRIM_OVERLAP_NO));
if ((ALPHA.A == ALPHA.B) || !om_bsel.abe) { // Optimize-away colclip
// No addition neither substraction so no risk of overflow the [0:255] range.
colclip_wrap = false;
acc_colclip_wrap = false;
}
if (acc_colclip_wrap) {
colclip_wrap = false;
ps_sel.colclip = 3;
GL_INS("COLCLIP SW ENABLED (blending is %d/%d/%d/%d)", ALPHA.A, ALPHA.B, ALPHA.C, ALPHA.D);
} else if (colclip_wrap) {
ps_sel.colclip = 1;
GL_INS("COLCLIP ENABLED (blending is %d/%d/%d/%d)", ALPHA.A, ALPHA.B, ALPHA.C, ALPHA.D);
}
ps_sel.fba = context->FBA.FBA;
if (PRIM->FGE)
@ -764,19 +780,6 @@ void GSRendererOGL::DrawPrims(GSTexture* rt, GSTexture* ds, GSTextureCache::Sour
}
// SW Blending
// Compute the blending equation to detect special case
int blend_sel = ((om_bsel.a * 3 + om_bsel.b) * 3 + om_bsel.c) * 3 + om_bsel.d;
int blend_flag = GSDeviceOGL::m_blendMapD3D9[blend_sel].bogus;
bool impossible_blend = m_sw_blending && (blend_flag & A_MAX);
bool all_blend_sw = (m_sw_blending >= ACC_BLEND_ULTRA)
|| (m_sw_blending >= ACC_BLEND_FULL && !( (ALPHA.A == ALPHA.B) || (ALPHA.C == 2 && afix <= 1.002f) ));
bool sw_blending = (m_prim_overlap == PRIM_OVERLAP_NO) // Free case
|| impossible_blend || all_blend_sw // Impossible blend or all
|| acc_colclip_wrap // accurate colclip
|| ps_sel.fbmask; // accurate fbmask
// GL42 interact very badly with sw blending. GL42 uses the primitiveID to find the primitive
// that write the bad alpha value. Sw blending will force the draw to run primitive by primitive
// (therefore primitiveID will be constant to 1)