flycast/core/rend/soft/softrend.cpp

611 lines
14 KiB
C++

#include "hw\pvr\Renderer_if.h"
#include "hw\pvr\pvr_mem.h"
#include "oslib\oslib.h"
/*
SSE/MMX based softrend
Initial code by skmp and gigaherz
This is a rather weird very basic pvr softrend.
Renders in some kind of tile format (that I forget now),
and does depth and color, but no alpha, texture, or pixel
processing. All of the pipeline is based on quads.
*/
#include <mmintrin.h>
#include <xmmintrin.h>
#include <emmintrin.h>
BITMAPINFOHEADER bi = { sizeof(BITMAPINFOHEADER), 0, 0, 1, 32, BI_RGB };
struct softrend : Renderer
{
virtual bool Process(TA_context* ctx) {
//disable RTTs for now ..
if (ctx->rend.isRTT)
return false;
ctx->rend_inuse.Lock();
ctx->MarkRend();
if (!ta_parse_vdrc(ctx))
return false;
return true;
}
DECL_ALIGN(32) u32 render_buffer[640 * 480 * 2 * 4]; //Color + depth
DECL_ALIGN(32) u32 pixels[640 * 480 * 4];
static __m128i _mm_load_scaled(int v, int s)
{
return _mm_setr_epi32(v, v + s, v + s + s, v + s + s + s);
}
static __m128i _mm_broadcast(int v)
{
__m128i rv = _mm_cvtsi32_si128(v);
return _mm_shuffle_epi32(rv, 0);
}
static __m128 _mm_load_ps_r(float a, float b, float c, float d)
{
static __declspec(align(128)) float v[4];
v[0] = a;
v[1] = b;
v[2] = c;
v[3] = d;
return _mm_load_ps(v);
}
__forceinline int iround(float x)
{
return _mm_cvtt_ss2si(_mm_load_ss(&x));
}
int mmin(int a, int b, int c, int d)
{
int rv = min(a, b);
rv = min(c, rv);
return max(d, rv);
}
int mmax(int a, int b, int c, int d)
{
int rv = max(a, b);
rv = max(c, rv);
return min(d, rv);
}
//i think this gives false positives ...
//yup, if ANY of the 3 tests fail the ANY tests fails.
__forceinline void EvalHalfSpace(bool& all, bool& any, int cp, int sv, int lv)
{
//bool a00 = C1 + DX12 * y0 - DY12 * x0 > 0;
//bool a10 = C1 + DX12 * y0 - DY12 * x0 > qDY12;
//bool a01 = C1 + DX12 * y0 - DY12 * x0 > -qDX12;
//bool a11 = C1 + DX12 * y0 - DY12 * x0 > (qDY12-qDX12);
//C1 + DX12 * y0 - DY12 * x0 > 0
// + DX12 * y0 - DY12 * x0 > 0 - C1
//int pd=DX * y0 - DY * x0;
bool a = cp > sv; //needed for ANY
bool b = cp > lv; //needed for ALL
any &= a;
all &= b;
}
//return true if any is positive
__forceinline bool EvalHalfSpaceFAny(int cp12, int cp23, int cp31)
{
int svt = cp12; //needed for ANY
svt |= cp23;
svt |= cp31;
return svt>0;
}
__forceinline bool EvalHalfSpaceFAll(int cp12, int cp23, int cp31, int lv12, int lv23, int lv31)
{
int lvt = cp12 - lv12;
lvt |= cp23 - lv23;
lvt |= cp31 - lv31; //needed for all
return lvt>0;
}
__forceinline void PlaneMinMax(int& MIN, int& MAX, int DX, int DY, int q)
{
int q_fp = (q - 1) << 4;
int v1 = 0;
int v2 = q_fp*DY;
int v3 = -q_fp*DX;
int v4 = q_fp*(DY - DX);
MIN = min(v1, min(v2, min(v3, v4)));
MAX = max(v1, max(v2, max(v3, v4)));
}
struct PlaneStepper
{
__m128 ddx, ddy;
__m128 c;
void Setup(const Vertex &v1, const Vertex &v2, const Vertex &v3, int minx, int miny, int q
, float v1_a, float v2_a, float v3_a
, float v1_b, float v2_b, float v3_b
, float v1_c, float v2_c, float v3_c
, float v1_d, float v2_d, float v3_d)
{
// float v1_z=v1.z,v2_z=v2.z,v3_z=v3.z;
float Aa = ((v3_a - v1_a) * (v2.y - v1.y) - (v2_a - v1_a) * (v3.y - v1.y));
float Ba = ((v3.x - v1.x) * (v2_a - v1_a) - (v2.x - v1.x) * (v3_a - v1_a));
float Ab = ((v3_b - v1_b) * (v2.y - v1.y) - (v2_b - v1_b) * (v3.y - v1.y));
float Bb = ((v3.x - v1.x) * (v2_b - v1_b) - (v2.x - v1.x) * (v3_b - v1_b));
float Ac = ((v3_c - v1_c) * (v2.y - v1.y) - (v2_c - v1_c) * (v3.y - v1.y));
float Bc = ((v3.x - v1.x) * (v2_c - v1_c) - (v2.x - v1.x) * (v3_c - v1_c));
float Ad = ((v3_d - v1_d) * (v2.y - v1.y) - (v2_d - v1_d) * (v3.y - v1.y));
float Bd = ((v3.x - v1.x) * (v2_d - v1_d) - (v2.x - v1.x) * (v3_d - v1_d));
float C = ((v2.x - v1.x) * (v3.y - v1.y) - (v3.x - v1.x) * (v2.y - v1.y));
float ddx_s_a = -Aa / C;
float ddy_s_a = -Ba / C;
float ddx_s_b = -Ab / C;
float ddy_s_b = -Bb / C;
float ddx_s_c = -Ac / C;
float ddy_s_c = -Bc / C;
float ddx_s_d = -Ad / C;
float ddy_s_d = -Bd / C;
ddx = _mm_load_ps_r(ddx_s_a, ddx_s_b, ddx_s_c, ddx_s_d);
ddy = _mm_load_ps_r(ddy_s_a, ddy_s_b, ddy_s_c, ddy_s_d);
float c_s_a = (v1_a - ddx_s_a *v1.x - ddy_s_a*v1.y);
float c_s_b = (v1_b - ddx_s_b *v1.x - ddy_s_b*v1.y);
float c_s_c = (v1_c - ddx_s_c *v1.x - ddy_s_c*v1.y);
float c_s_d = (v1_d - ddx_s_d *v1.x - ddy_s_d*v1.y);
c = _mm_load_ps_r(c_s_a, c_s_b, c_s_c, c_s_d);
//z = z1 + dzdx * (minx - v1.x) + dzdy * (minx - v1.y);
//z = (z1 - dzdx * v1.x - v1.y*dzdy) + dzdx*inx + dzdy *iny;
}
__forceinline __m128 Ip(__m128 x, __m128 y) const
{
__m128 p1 = _mm_mul_ps(x, ddx);
__m128 p2 = _mm_mul_ps(y, ddy);
__m128 s1 = _mm_add_ps(p1, p2);
return _mm_add_ps(s1, c);
}
__forceinline __m128 InStep(__m128 bas) const
{
return _mm_add_ps(bas, ddx);
}
};
struct IPs
{
PlaneStepper ZUV;
PlaneStepper Col;
void Setup(const Vertex &v1, const Vertex &v2, const Vertex &v3, int minx, int miny, int q)
{
ZUV.Setup(v1, v2, v3, minx, miny, q,
v1.z, v2.z, v3.z,
v1.u, v2.u, v3.u,
v1.v, v2.v, v3.v,
0, -1, 1);
Col.Setup(v1, v2, v3, minx, miny, q,
v1.col[2], v2.col[2], v3.col[2],
v1.col[1], v2.col[1], v3.col[1],
v1.col[0], v2.col[0], v3.col[0],
v1.col[3], v2.col[3], v3.col[3]
);
}
};
IPs __declspec(align(64)) ip;
template<bool useoldmsk>
__forceinline void PixelFlush(__m128 x, __m128 y, u8* cb, __m128 oldmask)
{
x = _mm_shuffle_ps(x, x, 0);
__m128 invW = ip.ZUV.Ip(x, y);
__m128 u = ip.ZUV.InStep(invW);
__m128 v = ip.ZUV.InStep(u);
__m128 ws = ip.ZUV.InStep(v);
_MM_TRANSPOSE4_PS(invW, u, v, ws);
//invW : {z1,z2,z3,z4}
//u : {u1,u2,u3,u4}
//v : {v1,v2,v3,v4}
//wx : {?,?,?,?}
__m128* zb = (__m128*)&cb[640 * 480 * 4];
__m128 ZMask = _mm_cmpgt_ps(invW, *zb);
if (useoldmsk)
ZMask = _mm_and_ps(oldmask, ZMask);
u32 msk = _mm_movemask_ps(ZMask);//0xF
if (msk == 0)
return;
__m128i rv;
{
__m128 a = ip.Col.Ip(x, y);
__m128 b = ip.Col.InStep(a);
__m128 c = ip.Col.InStep(b);
__m128 d = ip.Col.InStep(c);
__m128i ui = _mm_cvttps_epi32(u);
__m128i vi = _mm_cvttps_epi32(v);
//(int)v<<x+(int)u
__m128i textadr = _mm_add_epi32(_mm_slli_epi32(vi, 8), ui);//texture addresses ! 4x of em !
//we need :
__m128i ab = _mm_packs_epi32(_mm_cvttps_epi32(a), _mm_cvttps_epi32(b));
__m128i cd = _mm_packs_epi32(_mm_cvttps_epi32(c), _mm_cvttps_epi32(d));
rv = _mm_packus_epi16(ab, cd);
//rv = _mm_xor_si128(rv,textadr);
}
//__m128i rv=ip.col;//_mm_xor_si128(_mm_cvtps_epi32(_mm_mul_ps(x,Z.c)),_mm_cvtps_epi32(y));
if (msk != 0xF)
{
rv = _mm_and_si128(rv, *(__m128i*)&ZMask);
rv = _mm_or_si128(_mm_andnot_si128(*(__m128i*)&ZMask, *(__m128i*)cb), rv);
invW = _mm_and_ps(invW, ZMask);
invW = _mm_or_ps(_mm_andnot_ps(ZMask, *zb), invW);
}
*zb = invW;
*(__m128i*)cb = rv;
}
//u32 nok,fok;
void Rendtriangle(const Vertex &v1, const Vertex &v2, const Vertex &v3, u32* colorBuffer)
{
const int stride = 640 * 4;
//Plane equation
// 28.4 fixed-point coordinates
const int Y1 = iround(16.0f * v1.y);
const int Y2 = iround(16.0f * v2.y);
const int Y3 = iround(16.0f * v3.y);
const int X1 = iround(16.0f * v1.x);
const int X2 = iround(16.0f * v2.x);
const int X3 = iround(16.0f * v3.x);
int sgn = 1;
// Deltas
{
//area: (X1-X3)*(Y2-Y3)-(Y1-Y3)*(X2-X3)
if (((X1 - X3)*(Y2 - Y3) - (Y1 - Y3)*(X2 - X3))>0)
sgn = -1;
}
const int DX12 = sgn*(X1 - X2);
const int DX23 = sgn*(X2 - X3);
const int DX31 = sgn*(X3 - X1);
const int DY12 = sgn*(Y1 - Y2);
const int DY23 = sgn*(Y2 - Y3);
const int DY31 = sgn*(Y3 - Y1);
// Fixed-point deltas
const int FDX12 = DX12 << 4;
const int FDX23 = DX23 << 4;
const int FDX31 = DX31 << 4;
const int FDY12 = DY12 << 4;
const int FDY23 = DY23 << 4;
const int FDY31 = DY31 << 4;
// Block size, standard 4x4 (must be power of two)
const int q = 4;
// Bounding rectangle
int minx = (mmin(X1, X2, X3, 0) + 0xF) >> 4;
int miny = (mmin(Y1, Y2, Y3, 0) + 0xF) >> 4;
// Start in corner of block
minx &= ~(q - 1);
miny &= ~(q - 1);
int spanx = ((mmax(X1, X2, X3, 640 << 4) + 0xF) >> 4) - minx;
int spany = ((mmax(Y1, Y2, Y3, 480 << 4) + 0xF) >> 4) - miny;
// Half-edge constants
int C1 = DY12 * X1 - DX12 * Y1;
int C2 = DY23 * X2 - DX23 * Y2;
int C3 = DY31 * X3 - DX31 * Y3;
// Correct for fill convention
if (DY12 < 0 || (DY12 == 0 && DX12 > 0)) C1++;
if (DY23 < 0 || (DY23 == 0 && DX23 > 0)) C2++;
if (DY31 < 0 || (DY31 == 0 && DX31 > 0)) C3++;
int MAX_12, MAX_23, MAX_31, MIN_12, MIN_23, MIN_31;
PlaneMinMax(MIN_12, MAX_12, DX12, DY12, q);
PlaneMinMax(MIN_23, MAX_23, DX23, DY23, q);
PlaneMinMax(MIN_31, MAX_31, DX31, DY31, q);
const int FDqX12 = FDX12 * q;
const int FDqX23 = FDX23 * q;
const int FDqX31 = FDX31 * q;
const int FDqY12 = FDY12 * q;
const int FDqY23 = FDY23 * q;
const int FDqY31 = FDY31 * q;
const int FDX12mq = FDX12 + FDY12*q;
const int FDX23mq = FDX23 + FDY23*q;
const int FDX31mq = FDX31 + FDY31*q;
int hs12 = C1 + FDX12 * miny - FDY12 * minx + FDqY12 - MIN_12;
int hs23 = C2 + FDX23 * miny - FDY23 * minx + FDqY23 - MIN_23;
int hs31 = C3 + FDX31 * miny - FDY31 * minx + FDqY31 - MIN_31;
MAX_12 -= MIN_12;
MAX_23 -= MIN_23;
MAX_31 -= MIN_31;
int C1_pm = MIN_12;
int C2_pm = MIN_23;
int C3_pm = MIN_31;
u8* cb_y = (u8*)colorBuffer;
cb_y += miny*stride + minx*(q * 4);
ip.Setup(v1, v2, v3, minx, miny, q);
__m128 y_ps = _mm_cvtepi32_ps(_mm_broadcast(miny));
__m128 minx_ps = _mm_cvtepi32_ps(_mm_load_scaled(minx - q, 1));
static __declspec(align(16)) float ones_ps[4] = { 1, 1, 1, 1 };
static __declspec(align(16)) float q_ps[4] = { q, q, q, q };
// Loop through blocks
for (int y = spany; y > 0; y -= q)
{
int Xhs12 = hs12;
int Xhs23 = hs23;
int Xhs31 = hs31;
u8* cb_x = cb_y;
__m128 x_ps = minx_ps;
for (int x = spanx; x > 0; x -= q)
{
Xhs12 -= FDqY12;
Xhs23 -= FDqY23;
Xhs31 -= FDqY31;
x_ps = _mm_add_ps(x_ps, *(__m128*)q_ps);
// Corners of block
bool any = EvalHalfSpaceFAny(Xhs12, Xhs23, Xhs31);
// Skip block when outside an edge
if (!any)
{
cb_x += q*q * 4;
continue;
}
bool all = EvalHalfSpaceFAll(Xhs12, Xhs23, Xhs31, MAX_12, MAX_23, MAX_31);
// Accept whole block when totally covered
if (all)
{
__m128 yl_ps = y_ps;
for (int iy = q; iy > 0; iy--)
{
PixelFlush<false>(x_ps, yl_ps, cb_x, x_ps);
yl_ps = _mm_add_ps(yl_ps, *(__m128*)ones_ps);
cb_x += sizeof(__m128);
}
}
else // Partially covered block
{
int CY1 = C1_pm + Xhs12;
int CY2 = C2_pm + Xhs23;
int CY3 = C3_pm + Xhs31;
__m128i pfdx12 = _mm_broadcast(FDX12);
__m128i pfdx23 = _mm_broadcast(FDX23);
__m128i pfdx31 = _mm_broadcast(FDX31);
__m128i pcy1 = _mm_load_scaled(CY1, -FDY12);
__m128i pcy2 = _mm_load_scaled(CY2, -FDY23);
__m128i pcy3 = _mm_load_scaled(CY3, -FDY31);
__m128i pzero = _mm_setzero_si128();
//bool ok=false;
__m128 yl_ps = y_ps;
for (int iy = q; iy > 0; iy--)
{
__m128i a = _mm_cmpgt_epi32(_mm_or_si128(_mm_or_si128(pcy1, pcy2), pcy3), pzero);
int msk = _mm_movemask_ps(*(__m128*)&a);
if (msk != 0)
{
PixelFlush<true>(x_ps, yl_ps, cb_x, *(__m128*)&a);
}
yl_ps = _mm_add_ps(yl_ps, *(__m128*)ones_ps);
cb_x += sizeof(__m128);
//CY1 += FDX12mq;
//CY2 += FDX23mq;
//CY3 += FDX31mq;
pcy1 = _mm_add_epi32(pcy1, pfdx12);
pcy2 = _mm_add_epi32(pcy2, pfdx23);
pcy3 = _mm_add_epi32(pcy3, pfdx31);
}
/*
if (!ok)
{
nok++;
}
else
{
fok++;
}*/
}
}
next_y:
hs12 += FDqX12;
hs23 += FDqX23;
hs31 += FDqX31;
cb_y += stride*q;
y_ps = _mm_add_ps(y_ps, *(__m128*)q_ps);
}
}
void RenderParamList(List<PolyParam>* param_list) {
Vertex* verts = pvrrc.verts.head();
u16* idx = pvrrc.idx.head();
PolyParam* params = param_list->head();
int param_count = param_list->used();
for (int i = 0; i < param_count; i++)
{
int vertex_count = params[i].count - 2;
u16* poly_idx = &idx[params[i].first];
for (int v = 0; v < vertex_count; v++) {
Rendtriangle(verts[poly_idx[v]], verts[poly_idx[v + 1]], verts[poly_idx[v + 2]], render_buffer);
}
}
}
virtual bool Render() {
bool is_rtt = pvrrc.isRTT;
memset(render_buffer, 0, sizeof(render_buffer));
if (pvrrc.verts.used()<3)
return false;
RenderParamList(&pvrrc.global_param_op);
RenderParamList(&pvrrc.global_param_pt);
RenderParamList(&pvrrc.global_param_tr);
/*
for (int y = 0; y < 480; y++) {
for (int x = 0; x < 640; x++) {
color_buffer[x + y * 640] = rand();
}
} */
return !is_rtt;
}
HWND hWnd;
HBITMAP hBMP = 0, holdBMP;
HDC hmem;
virtual bool Init() {
hWnd = (HWND)libPvr_GetRenderTarget();
bi.biWidth = 640;
bi.biHeight = 480;
RECT rect;
GetClientRect(hWnd, &rect);
HDC hdc = GetDC(hWnd);
FillRect(hdc, &rect, (HBRUSH)(COLOR_BACKGROUND));
bi.biSizeImage = bi.biWidth * bi.biHeight * 4;
hBMP = CreateCompatibleBitmap(hdc, bi.biWidth, bi.biHeight);
hmem = CreateCompatibleDC(hdc);
holdBMP = (HBITMAP)SelectObject(hmem, hBMP);
ReleaseDC(hWnd, hdc);
return true;
}
virtual void Resize(int w, int h) {
}
virtual void Term() {
if (hBMP) {
DeleteObject(SelectObject(hmem, holdBMP));
DeleteDC(hmem);
}
}
virtual void Present() {
__m128* psrc = (__m128*)render_buffer;
__m128* pdst = (__m128*)pixels;
const int stride = 640 / 4;
for (int y = 0; y<480; y += 4)
{
for (int x = 0; x<640; x += 4)
{
pdst[(480 - (y + 0))*stride + x / 4] = *psrc++;
pdst[(480 - (y + 1))*stride + x / 4] = *psrc++;
pdst[(480 - (y + 2))*stride + x / 4] = *psrc++;
pdst[(480 - (y + 3))*stride + x / 4] = *psrc++;
}
}
SetDIBits(hmem, hBMP, 0, 480, pixels, (BITMAPINFO*)&bi, DIB_RGB_COLORS);
RECT clientRect;
GetClientRect(hWnd, &clientRect);
HDC hdc = GetDC(hWnd);
int w = clientRect.right - clientRect.left;
int h = clientRect.bottom - clientRect.top;
int x = (w - 640) / 2;
int y = (h - 480) / 2;
BitBlt(hdc, x, y, 640 , 480 , hmem, 0, 0, SRCCOPY);
ReleaseDC(hWnd, hdc);
}
};
Renderer* rend_softrend() {
return new(_mm_malloc(sizeof(softrend), 32)) softrend();
}