ProcessVertexData = ProcessVertexDataSSE;
}
else
+#elif defined(__ARM_NEON__)
+ if( !g_curRomInfo.bPrimaryDepthHack && options.enableHackForGames != HACK_FOR_NASCAR && options.enableHackForGames != HACK_FOR_ZELDA_MM && !options.bWinFrameMode)
+ {
+ ProcessVertexData = ProcessVertexDataNEON;
+ }
+ else
#endif
{
ProcessVertexData = ProcessVertexDataNoSSE;
bool bHalfTxtScale=false;
extern uint32 lastSetTile;
+#ifndef __ARM_NEON__
+static void multiply_subtract2(float *d, const float *m1, const float *m2, const float *s)
+{
+ int i;
+ for (i = 0; i < 2; i++)
+ d[i] = m1[i] * m2[i] - s[i];
+}
+#else
+extern "C" void multiply_subtract2(float *d, const float *m1, const float *m2, const float *s);
+#endif
+
void InitVertex(uint32 dwV, uint32 vtxIndex, bool bTexture, bool openGL)
{
VTX_DUMP(TRACE2("Init vertex (%d) to vtx buf[%d]:", dwV, vtxIndex));
}
else
{
- float tex0u = g_fVtxTxtCoords[dwV].x *gRSP.tex0scaleX - gRSP.tex0OffsetX ;
- float tex0v = g_fVtxTxtCoords[dwV].y *gRSP.tex0scaleY - gRSP.tex0OffsetY ;
+ TexCord tex0;
+ multiply_subtract2(&tex0.u, &g_fVtxTxtCoords[dwV].x, &gRSP.tex0scaleX, &gRSP.tex0OffsetX);
if( CRender::g_pRender->IsTexel1Enable() )
{
- float tex1u = g_fVtxTxtCoords[dwV].x *gRSP.tex1scaleX - gRSP.tex1OffsetX ;
- float tex1v = g_fVtxTxtCoords[dwV].y *gRSP.tex1scaleY - gRSP.tex1OffsetY ;
+ TexCord tex1;
+ multiply_subtract2(&tex1.u, &g_fVtxTxtCoords[dwV].x, &gRSP.tex1scaleX, &gRSP.tex1OffsetX);
- CRender::g_pRender->SetVertexTextureUVCoord(v, tex0u, tex0v, tex1u, tex1v);
- VTX_DUMP(TRACE2(" (tex0): %f, %f", tex0u,tex0v));
- VTX_DUMP(TRACE2(" (tex1): %f, %f", tex1u,tex1v));
+ CRender::g_pRender->SetVertexTextureUVCoord(v, tex0, tex1);
+ VTX_DUMP(TRACE2(" (tex0): %f, %f", tex0.u,tex0.v));
+ VTX_DUMP(TRACE2(" (tex1): %f, %f", tex1.u,tex1.v));
}
else
{
- CRender::g_pRender->SetVertexTextureUVCoord(v, tex0u, tex0v);
- VTX_DUMP(TRACE2(" (tex0): %f, %f", tex0u,tex0v));
+ CRender::g_pRender->SetVertexTextureUVCoord(v, tex0);
+ VTX_DUMP(TRACE2(" (tex0): %f, %f", tex0.u,tex0.v));
}
}
DEBUGGER_PAUSE_AND_DUMP(NEXT_VERTEX_CMD,{TRACE0("Paused at Vertex Cmd");});
}
+/* NEON code */
+
+#include "RenderBase_neon.h"
+
+extern "C" void pv_neon(XVECTOR4 *g_vtxTransformed, XVECTOR4 *g_vecProjected,
+ uint32 *g_dwVtxDifColor, VECTOR2 *g_fVtxTxtCoords,
+ float *g_fFogCoord, uint32 *g_clipFlag2,
+ uint32 dwNum, int neon_state,
+ const FiddledVtx *vtx,
+ const Light *gRSPlights, const float *fRSPAmbientLightRGBA,
+ const XMATRIX *gRSPworldProject, const XMATRIX *gRSPmodelViewTop,
+ uint32 gRSPnumLights, float gRSPfFogMin,
+ uint32 primitiveColor, uint32 primitiveColor_);
+
+void ProcessVertexDataNEON(uint32 dwAddr, uint32 dwV0, uint32 dwNum)
+{
+ if (gRSP.bTextureGen && gRSP.bLightingEnable) {
+ ProcessVertexDataNoSSE(dwAddr, dwV0,dwNum);
+ return;
+ }
+
+ // assumtions:
+ // - g_clipFlag is not used at all
+ // - g_fFogCoord is not used at all
+ // - g_vtxNonTransformed is not used after ProcessVertexData*() returns
+ // - g_normal - same
+
+ int neon_state = 0;
+ if ( gRSP.bLightingEnable )
+ neon_state |= PV_NEON_ENABLE_LIGHT;
+ if ( (gRDP.geometryMode & G_SHADE) || gRSP.ucode >= 5 )
+ neon_state |= PV_NEON_ENABLE_SHADE;
+ if ( gRSP.bFogEnabled )
+ neon_state |= PV_NEON_ENABLE_FOG;
+ if ( gRDP.geometryMode & G_FOG )
+ neon_state |= PV_NEON_FOG_ALPHA;
+
+ uint32 i;
+
+ UpdateCombinedMatrix();
+
+ // This function is called upon SPvertex
+ // - do vertex matrix transform
+ // - do vertex lighting
+ // - do texture cooridinate transform if needed
+ // - calculate normal vector
+
+ // Output: - g_vecProjected[i] -> transformed vertex x,y,z
+ // - g_vecProjected[i].w -> saved vertex 1/w
+ // - g_vtxTransformed[i]
+ // - g_dwVtxDifColor[i] -> vertex color
+ // - g_fVtxTxtCoords[i] -> vertex texture cooridinates
+ // - g_fFogCoord[i] -> unused
+ // - g_clipFlag2[i]
+
+ const FiddledVtx * pVtxBase = (const FiddledVtx*)(g_pRDRAMu8 + dwAddr);
+ g_pVtxBase = (FiddledVtx *)pVtxBase;
+
+ gRSPmodelViewTop._14 = gRSPmodelViewTop._24 =
+ gRSPmodelViewTop._34 = 0;
+
+ // SP_Timing(RSP_GBI0_Vtx);
+ status.SPCycleCount += Timing_RSP_GBI0_Vtx * dwNum;
+
+#if 1
+ i = dwV0;
+ pv_neon(&g_vtxTransformed[i], &g_vecProjected[i],
+ &g_dwVtxDifColor[i], &g_fVtxTxtCoords[i],
+ &g_fFogCoord[i], &g_clipFlag2[i],
+ dwNum, neon_state, &pVtxBase[i - dwV0],
+ gRSPlights, gRSP.fAmbientColors,
+ &gRSPworldProject, &gRSPmodelViewTop,
+ gRSPnumLights, gRSPfFogMin,
+ gRDP.primitiveColor, gRDP.primitiveColor);
+#else
+ for (i = dwV0; i < dwV0 + dwNum; i++)
+ {
+ const FiddledVtx & vert = pVtxBase[i - dwV0];
+ XVECTOR3 vtx_raw; // was g_vtxNonTransformed
+
+ vtx_raw.x = (float)vert.x;
+ vtx_raw.y = (float)vert.y;
+ vtx_raw.z = (float)vert.z;
+
+ Vec3Transform(&g_vtxTransformed[i], &vtx_raw, &gRSPworldProject); // Convert to w=1
+
+ g_vecProjected[i].w = 1.0f / g_vtxTransformed[i].w;
+ g_vecProjected[i].x = g_vtxTransformed[i].x * g_vecProjected[i].w;
+ g_vecProjected[i].y = g_vtxTransformed[i].y * g_vecProjected[i].w;
+ g_vecProjected[i].z = g_vtxTransformed[i].z * g_vecProjected[i].w;
+
+ // RSP_Vtx_Clipping(i);
+ g_clipFlag2[i] = 0;
+ if( g_vecProjected[i].w > 0 )
+ {
+ if( g_vecProjected[i].x > 1 ) g_clipFlag2[i] |= X_CLIP_MAX;
+ if( g_vecProjected[i].x < -1 ) g_clipFlag2[i] |= X_CLIP_MIN;
+ if( g_vecProjected[i].y > 1 ) g_clipFlag2[i] |= Y_CLIP_MAX;
+ if( g_vecProjected[i].y < -1 ) g_clipFlag2[i] |= Y_CLIP_MIN;
+ }
+
+ if( neon_state & PV_NEON_ENABLE_LIGHT )
+ {
+ XVECTOR3 normal; // was g_normal
+ float r, g, b;
+
+ normal.x = (float)vert.norma.nx;
+ normal.y = (float)vert.norma.ny;
+ normal.z = (float)vert.norma.nz;
+
+ Vec3TransformNormal(normal, gRSPmodelViewTop);
+
+ r = gRSP.fAmbientLightR;
+ g = gRSP.fAmbientLightG;
+ b = gRSP.fAmbientLightB;
+
+ for (unsigned int l=0; l < gRSPnumLights; l++)
+ {
+ float fCosT = normal.x * gRSPlights[l].x + normal.y * gRSPlights[l].y + normal.z * gRSPlights[l].z;
+
+ if (fCosT > 0 )
+ {
+ r += gRSPlights[l].fr * fCosT;
+ g += gRSPlights[l].fg * fCosT;
+ b += gRSPlights[l].fb * fCosT;
+ }
+ }
+ if (r > 255) r = 255;
+ if (g > 255) g = 255;
+ if (b > 255) b = 255;
+ g_dwVtxDifColor[i] = ((vert.rgba.a<<24)|(((uint32)r)<<16)|(((uint32)g)<<8)|((uint32)b));
+ }
+ else if( neon_state & PV_NEON_ENABLE_SHADE )
+ {
+ IColor &color = *(IColor*)&g_dwVtxDifColor[i];
+ color.b = vert.rgba.r;
+ color.g = vert.rgba.g;
+ color.r = vert.rgba.b;
+ color.a = vert.rgba.a;
+ }
+ else
+ g_dwVtxDifColor[i] = gRDP.primitiveColor; // FLAT shade
+
+ // ReplaceAlphaWithFogFactor(i);
+ if( neon_state & PV_NEON_FOG_ALPHA )
+ {
+ // Use fog factor to replace vertex alpha
+ if( g_vecProjected[i].z > 1 )
+ *(((uint8*)&(g_dwVtxDifColor[i]))+3) = 0xFF;
+ // missing 'else' in original code??
+ else if( g_vecProjected[i].z < 0 )
+ *(((uint8*)&(g_dwVtxDifColor[i]))+3) = 0;
+ else
+ *(((uint8*)&(g_dwVtxDifColor[i]))+3) = (uint8)(g_vecProjected[i].z*255);
+ }
+
+ g_fVtxTxtCoords[i].x = (float)vert.tu;
+ g_fVtxTxtCoords[i].y = (float)vert.tv;
+ }
+#endif
+}
+
bool PrepareTriangle(uint32 dwV0, uint32 dwV1, uint32 dwV2)
{
if( status.isVertexShaderEnabled || status.bUseHW_T_L )
notaz.gp2x.de / mupen64plus-pandora.git / blobdiff