2 * Copyright (C) 2011 Gilead Kutnick "Exophase" <exophase@gmail.com>
3 * Copyright (C) 2022 GraÅžvydas Ignotas "notaz" <notasas@gmail.com>
5 * This program is free software; you can redistribute it and/or
6 * modify it under the terms of the GNU General Public License as
7 * published by the Free Software Foundation; either version 2 of
8 * the License, or (at your option) any later version.
10 * This program is distributed in the hope that it will be useful,
11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
13 * General Public License for more details.
18 #include "psx_gpu_simd.h"
19 //#define ASM_PROTOTYPES
20 //#include "psx_gpu_simd.h"
22 #error "please define SIMD_BUILD if you want this gpu_neon C simd implementation"
25 typedef u8 gvu8 __attribute__((vector_size(16)));
26 typedef u16 gvu16 __attribute__((vector_size(16)));
27 typedef u32 gvu32 __attribute__((vector_size(16)));
28 typedef u64 gvu64 __attribute__((vector_size(16)));
29 typedef s8 gvs8 __attribute__((vector_size(16)));
30 typedef s16 gvs16 __attribute__((vector_size(16)));
31 typedef s32 gvs32 __attribute__((vector_size(16)));
32 typedef s64 gvs64 __attribute__((vector_size(16)));
34 typedef u8 gvhu8 __attribute__((vector_size(8)));
35 typedef u16 gvhu16 __attribute__((vector_size(8)));
36 typedef u32 gvhu32 __attribute__((vector_size(8)));
37 typedef u64 gvhu64 __attribute__((vector_size(8)));
38 typedef s8 gvhs8 __attribute__((vector_size(8)));
39 typedef s16 gvhs16 __attribute__((vector_size(8)));
40 typedef s32 gvhs32 __attribute__((vector_size(8)));
41 typedef s64 gvhs64 __attribute__((vector_size(8)));
69 // this may be tempting, but it causes gcc to do lots of stack spills
73 #if defined(__ARM_NEON) || defined(__ARM_NEON__)
76 #define gvaddhn_u32(d, a, b) d.u16 = vaddhn_u32(a.u32, b.u32)
77 #define gvaddw_s32(d, a, b) d.s64 = vaddw_s32(a.s64, b.s32)
78 #define gvabsq_s32(d, s) d.s32 = vabsq_s32(s.s32)
79 #define gvbic_n_u16(d, n) d.u16 = vbic_u16(d.u16, vmov_n_u16(n))
80 #define gvbifq(d, a, b) d.u8 = vbslq_u8(b.u8, d.u8, a.u8)
81 #define gvbit(d, a, b) d.u8 = vbsl_u8(b.u8, a.u8, d.u8)
82 #define gvceqq_u16(d, a, b) d.u16 = vceqq_u16(a.u16, b.u16)
83 #define gvcgt_s16(d, a, b) d.u16 = vcgt_s16(a.s16, b.s16)
84 #define gvclt_s16(d, a, b) d.u16 = vclt_s16(a.s16, b.s16)
85 #define gvcreate_s32(d, a, b) d.s32 = vcreate_s32((u32)(a) | ((u64)(b) << 32))
86 #define gvcreate_u32(d, a, b) d.u32 = vcreate_u32((u32)(a) | ((u64)(b) << 32))
87 #define gvcreate_s64(d, s) d.s64 = (gvhs64)vcreate_s64(s)
88 #define gvcreate_u64(d, s) d.u64 = (gvhu64)vcreate_u64(s)
89 #define gvcombine_u16(d, l, h) d.u16 = vcombine_u16(l.u16, h.u16)
90 #define gvcombine_u32(d, l, h) d.u32 = vcombine_u32(l.u32, h.u32)
91 #define gvcombine_s64(d, l, h) d.s64 = vcombine_s64((int64x1_t)l.s64, (int64x1_t)h.s64)
92 #define gvdup_l_u8(d, s, l) d.u8 = vdup_lane_u8(s.u8, l)
93 #define gvdup_l_u16(d, s, l) d.u16 = vdup_lane_u16(s.u16, l)
94 #define gvdup_l_u32(d, s, l) d.u32 = vdup_lane_u32(s.u32, l)
95 #define gvdupq_l_s64(d, s, l) d.s64 = vdupq_lane_s64((int64x1_t)s.s64, l)
96 #define gvdupq_l_u32(d, s, l) d.u32 = vdupq_lane_u32(s.u32, l)
97 #define gvdup_n_s64(d, n) d.s64 = vdup_n_s64(n)
98 #define gvdup_n_u8(d, n) d.u8 = vdup_n_u8(n)
99 #define gvdup_n_u16(d, n) d.u16 = vdup_n_u16(n)
100 #define gvdup_n_u32(d, n) d.u32 = vdup_n_u32(n)
101 #define gvdupq_n_u16(d, n) d.u16 = vdupq_n_u16(n)
102 #define gvdupq_n_u32(d, n) d.u32 = vdupq_n_u32(n)
103 #define gvdupq_n_s64(d, n) d.s64 = vdupq_n_s64(n)
104 #define gvhaddq_u16(d, a, b) d.u16 = vhaddq_u16(a.u16, b.u16)
105 #define gvmax_s16(d, a, b) d.s16 = vmax_s16(a.s16, b.s16)
106 #define gvmin_s16(d, a, b) d.s16 = vmin_s16(a.s16, b.s16)
107 #define gvminq_u8(d, a, b) d.u8 = vminq_u8(a.u8, b.u8)
108 #define gvminq_u16(d, a, b) d.u16 = vminq_u16(a.u16, b.u16)
109 #define gvmla_s32(d, a, b) d.s32 = vmla_s32(d.s32, a.s32, b.s32)
110 #define gvmla_u32(d, a, b) d.u32 = vmla_u32(d.u32, a.u32, b.u32)
111 #define gvmlaq_s32(d, a, b) d.s32 = vmlaq_s32(d.s32, a.s32, b.s32)
112 #define gvmlaq_u32(d, a, b) d.u32 = vmlaq_u32(d.u32, a.u32, b.u32)
113 #define gvmlal_s32(d, a, b) d.s64 = vmlal_s32(d.s64, a.s32, b.s32)
114 #define gvmlal_u8(d, a, b) d.u16 = vmlal_u8(d.u16, a.u8, b.u8)
115 #define gvmlsq_s32(d, a, b) d.s32 = vmlsq_s32(d.s32, a.s32, b.s32)
116 #define gvmlsq_l_s32(d, a, b, l) d.s32 = vmlsq_lane_s32(d.s32, a.s32, b.s32, l)
117 #define gvmov_l_s32(d, s, l) d.s32 = vset_lane_s32(s, d.s32, l)
118 #define gvmov_l_u32(d, s, l) d.u32 = vset_lane_u32(s, d.u32, l)
119 #define gvmovl_u8(d, s) d.u16 = vmovl_u8(s.u8)
120 #define gvmovl_s32(d, s) d.s64 = vmovl_s32(s.s32)
121 #define gvmovn_u16(d, s) d.u8 = vmovn_u16(s.u16)
122 #define gvmovn_u32(d, s) d.u16 = vmovn_u32(s.u32)
123 #define gvmovn_u64(d, s) d.u32 = vmovn_u64(s.u64)
124 #define gvmul_s32(d, a, b) d.s32 = vmul_s32(a.s32, b.s32)
125 #define gvmull_s16(d, a, b) d.s32 = vmull_s16(a.s16, b.s16)
126 #define gvmull_s32(d, a, b) d.s64 = vmull_s32(a.s32, b.s32)
127 #define gvmull_u8(d, a, b) d.u16 = vmull_u8(a.u8, b.u8)
128 #define gvmull_l_u32(d, a, b, l) d.u64 = vmull_lane_u32(a.u32, b.u32, l)
129 #define gvmlsl_s16(d, a, b) d.s32 = vmlsl_s16(d.s32, a.s16, b.s16)
130 #define gvneg_s32(d, s) d.s32 = vneg_s32(s.s32)
131 #define gvqadd_u8(d, a, b) d.u8 = vqadd_u8(a.u8, b.u8)
132 #define gvqsub_u8(d, a, b) d.u8 = vqsub_u8(a.u8, b.u8)
133 #define gvshl_u16(d, a, b) d.u16 = vshl_u16(a.u16, b.s16)
134 #define gvshlq_s64(d, a, b) d.s64 = vshlq_s64(a.s64, b.s64)
135 #define gvshlq_u32(d, a, b) d.u32 = vshlq_u32(a.u32, b.s32)
136 #define gvshlq_u64(d, a, b) d.u64 = vshlq_u64(a.u64, b.s64)
137 #define gvshrq_n_s16(d, s, n) d.s16 = vshrq_n_s16(s.s16, n)
138 #define gvshrq_n_u16(d, s, n) d.u16 = vshrq_n_u16(s.u16, n)
139 #define gvshl_n_u32(d, s, n) d.u32 = vshl_n_u32(s.u32, n)
140 #define gvshlq_n_u16(d, s, n) d.u16 = vshlq_n_u16(s.u16, n)
141 #define gvshlq_n_u32(d, s, n) d.u32 = vshlq_n_u32(s.u32, n)
142 #define gvshll_n_s8(d, s, n) d.s16 = vshll_n_s8(s.s8, n)
143 #define gvshll_n_u8(d, s, n) d.u16 = vshll_n_u8(s.u8, n)
144 #define gvshll_n_u16(d, s, n) d.u32 = vshll_n_u16(s.u16, n)
145 #define gvshr_n_u8(d, s, n) d.u8 = vshr_n_u8(s.u8, n)
146 #define gvshr_n_u16(d, s, n) d.u16 = vshr_n_u16(s.u16, n)
147 #define gvshr_n_u32(d, s, n) d.u32 = vshr_n_u32(s.u32, n)
148 #define gvshr_n_u64(d, s, n) d.u64 = (gvhu64)vshr_n_u64((uint64x1_t)s.u64, n)
149 #define gvshrn_n_s64(d, s, n) d.s32 = vshrn_n_s64(s.s64, n)
150 #define gvshrn_n_u16(d, s, n) d.u8 = vshrn_n_u16(s.u16, n)
151 #define gvshrn_n_u32(d, s, n) d.u16 = vshrn_n_u32(s.u32, n)
152 #define gvsli_n_u8(d, s, n) d.u8 = vsli_n_u8(d.u8, s.u8, n)
153 #define gvsri_n_u8(d, s, n) d.u8 = vsri_n_u8(d.u8, s.u8, n)
154 #define gvtstq_u16(d, a, b) d.u16 = vtstq_u16(a.u16, b.u16)
155 #define gvqshrun_n_s16(d, s, n) d.u8 = vqshrun_n_s16(s.s16, n)
156 #define gvqsubq_u8(d, a, b) d.u8 = vqsubq_u8(a.u8, b.u8)
157 #define gvqsubq_u16(d, a, b) d.u16 = vqsubq_u16(a.u16, b.u16)
159 #define gvget_lo(d, s) d.u16 = vget_low_u16(s.u16)
160 #define gvget_hi(d, s) d.u16 = vget_high_u16(s.u16)
161 #define gvlo(s) ({gvhreg t_; gvget_lo(t_, s); t_;})
162 #define gvhi(s) ({gvhreg t_; gvget_hi(t_, s); t_;})
164 #define gvset_lo(d, s) d.u16 = vcombine_u16(s.u16, gvhi(d).u16)
165 #define gvset_hi(d, s) d.u16 = vcombine_u16(gvlo(d).u16, s.u16)
167 #define gvtbl2_u8(d, a, b) { \
169 v_.val[0] = vget_low_u8(a.u8); v_.val[1] = vget_high_u8(a.u8); \
170 d.u8 = vtbl2_u8(v_, b.u8); \
173 #define gvzip_u8(d, a, b) { \
174 uint8x8x2_t v_ = vzip_u8(a.u8, b.u8); \
175 d.u8 = vcombine_u8(v_.val[0], v_.val[1]); \
177 #define gvzipq_u16(d0, d1, s0, s1) { \
178 uint16x8x2_t v_ = vzipq_u16(s0.u16, s1.u16); \
179 d0.u16 = v_.val[0]; d1.u16 = v_.val[1]; \
182 #define gvld1_u8(d, s) d.u8 = vld1_u8(s)
183 #define gvld1_u32(d, s) d.u32 = vld1_u32((const u32 *)(s))
184 #define gvld1q_u8(d, s) d.u8 = vld1q_u8(s)
185 #define gvld1q_u16(d, s) d.u16 = vld1q_u16(s)
186 #define gvld1q_u32(d, s) d.u32 = vld1q_u32((const u32 *)(s))
187 #define gvld2_dup(v0, v1, p) { \
188 uint8x8x2_t v_ = vld2_dup_u8(p); \
189 v0.u8 = v_.val[0]; v1.u8 = v_.val[1]; \
191 #define gvld2q_u8(v0, v1, p) { \
192 uint8x16x2_t v_ = vld2q_u8(p); \
193 v0.u8 = v_.val[0]; v1.u8 = v_.val[1]; \
196 #define gvst1_u8(v, p) \
198 #define gvst1q_u16(v, p) \
200 #define gvst1q_inc_u32(v, p, i) { \
201 vst1q_u32((u32 *)(p), v.u32); \
202 p += (i) / sizeof(*p); \
204 #define gvst2_u8(v0, v1, p) { \
206 v_.val[0] = v0.u8; v_.val[1] = v1.u8; \
209 #define gvst2_u16(v0, v1, p) { \
211 v_.val[0] = v0.u16; v_.val[1] = v1.u16; \
214 #define gvst2q_u8(v0, v1, p) { \
216 v_.val[0] = v0.u8; v_.val[1] = v1.u8; \
219 #define gvst4_4_inc_u32(v0, v1, v2, v3, p, i) { \
221 v_.val[0] = v0.u32; v_.val[1] = v1.u32; v_.val[2] = v2.u32; v_.val[3] = v3.u32; \
222 vst4_u32(p, v_); p += (i) / sizeof(*p); \
224 #define gvst4_pi_u16(v0, v1, v2, v3, p) { \
226 v_.val[0] = v0.u16; v_.val[1] = v1.u16; v_.val[2] = v2.u16; v_.val[3] = v3.u16; \
227 vst4_u16((u16 *)(p), v_); p += sizeof(v_) / sizeof(*p); \
229 #define gvst1q_pi_u32(v, p) \
230 gvst1q_inc_u32(v, p, sizeof(v))
231 // could use vst1q_u32_x2 but that's not always available
232 #define gvst1q_2_pi_u32(v0, v1, p) { \
233 gvst1q_inc_u32(v0, p, sizeof(v0)); \
234 gvst1q_inc_u32(v1, p, sizeof(v1)); \
238 - gcc > 9: (arm32) int64x1_t type produces ops on gp regs
239 (also u64 __attribute__((vector_size(8))) :( )
240 - gcc <11: (arm32) handles '<vec> == 0' poorly
244 #elif defined(__SSE2__)
245 #include <x86intrin.h>
248 #error "arch not supported or SIMD support was not enabled by your compiler"
251 // the below have intrinsics but they evaluate to basic operations on both gcc and clang
252 #define gvadd_s64(d, a, b) d.s64 = a.s64 + b.s64
253 #define gvadd_u8(d, a, b) d.u8 = a.u8 + b.u8
254 #define gvadd_u16(d, a, b) d.u16 = a.u16 + b.u16
255 #define gvadd_u32(d, a, b) d.u32 = a.u32 + b.u32
256 #define gvaddq_s64 gvadd_s64
257 #define gvaddq_u16 gvadd_u16
258 #define gvaddq_u32 gvadd_u32
259 #define gvand(d, a, b) d.u32 = a.u32 & b.u32
260 #define gvbic(d, a, b) d.u32 = a.u32 & ~b.u32
262 #define gveor(d, a, b) d.u32 = a.u32 ^ b.u32
264 #define gvceqz_u16(d, s) d.u16 = s.u16 == 0
265 #define gvceqzq_u16 gvceqz_u16
266 #define gvcltz_s16(d, s) d.s16 = s.s16 < 0
267 #define gvcltzq_s16 gvcltz_s16
268 #define gvsub_u16(d, a, b) d.u16 = a.u16 - b.u16
269 #define gvsub_u32(d, a, b) d.u32 = a.u32 - b.u32
270 #define gvsubq_u16 gvsub_u16
271 #define gvsubq_u32 gvsub_u32
272 #define gvorr(d, a, b) d.u32 = a.u32 | b.u32
277 #define gssub16(d, a, b) asm("ssub16 %0,%1,%2" : "=r"(d) : "r"(a), "r"(b))
278 #define gsmusdx(d, a, b) asm("smusdx %0,%1,%2" : "=r"(d) : "r"(a), "r"(b))
281 // gcc/config/arm/arm.c
283 #define gvadd_s64(d, a, b) asm("vadd.i64 %P0,%P1,%P2" : "=w"(d.s64) : "w"(a.s64), "w"(b.s64))
288 #define gssub16(d, a, b) d = (u16)((a) - (b)) | ((((a) >> 16) - ((b) >> 16)) << 16)
289 #define gsmusdx(d, a, b) d = ((s32)(s16)(a) * ((s32)(b) >> 16)) \
290 - (((s32)(a) >> 16) * (s16)(b))
294 // for compatibility with the original psx_gpu.c code
295 #define vec_2x64s gvreg
296 #define vec_2x64u gvreg
297 #define vec_4x32s gvreg
298 #define vec_4x32u gvreg
299 #define vec_8x16s gvreg
300 #define vec_8x16u gvreg
301 #define vec_16x8s gvreg
302 #define vec_16x8u gvreg
303 #define vec_1x64s gvhreg
304 #define vec_1x64u gvhreg
305 #define vec_2x32s gvhreg
306 #define vec_2x32u gvhreg
307 #define vec_4x16s gvhreg
308 #define vec_4x16u gvhreg
309 #define vec_8x8s gvhreg
310 #define vec_8x8u gvhreg
317 void cmpp(const char *name, const void *a_, const void *b_, size_t len)
319 const uint32_t *a = a_, *b = b_, masks[] = { 0, 0xff, 0xffff, 0xffffff };
322 for (i = 0; i < (len + 3)/4; i++) {
324 mask = left >= 4 ? ~0u : masks[left];
325 if ((a[i] ^ b[i]) & mask) {
326 printf("%s: %08x %08x [%03zx/%zu] #%d\n",
327 name, a[i] & mask, b[i] & mask, i*4, i, ccount);
333 #define ccmpf(n) cmpp(#n, &psx_gpu->n, &n##_c, sizeof(n##_c))
334 #define ccmpa(n,c) cmpp(#n, &psx_gpu->n, &n##_c, sizeof(n##_c[0]) * c)
336 void dump_r_(const char *name, void *dump, int is_q)
338 unsigned long long *u = dump;
339 //if (ccount > 1) return;
340 printf("%10s %016llx ", name, u[0]);
342 printf("%016llx", u[1]);
345 void __attribute__((noinline,noclone)) dump_r_d(const char *name, void *dump)
346 { dump_r_(name, dump, 0); }
347 void __attribute__((noinline,noclone)) dump_r_q(const char *name, void *dump)
348 { dump_r_(name, dump, 1); }
349 #define dumprd(n) { u8 dump_[8]; gvst1_u8(n, dump_); dump_r_d(#n, dump_); }
350 #define dumprq(n) { u16 dump_[8]; gvst1q_u16(n, dump_); dump_r_q(#n, dump_); }
353 void compute_all_gradients(psx_gpu_struct * __restrict__ psx_gpu,
354 const vertex_struct * __restrict__ a, const vertex_struct * __restrict__ b,
355 const vertex_struct * __restrict__ c)
357 union { double d; struct { u32 l; u32 h; } i; } divident, divider;
358 union { double d; gvhreg v; } d30;
361 compute_all_gradients_(psx_gpu, a, b, c);
364 // First compute the triangle area reciprocal and shift. The division will
365 // happen concurrently with much of the work which follows.
367 // load exponent of 62 into upper half of double
368 u32 shift = __builtin_clz(psx_gpu->triangle_area);
369 u32 triangle_area_normalized = psx_gpu->triangle_area << shift;
371 // load area normalized into lower half of double
372 divident.i.l = triangle_area_normalized >> 10;
373 divident.i.h = (62 + 1023) << 20;
375 divider.i.l = triangle_area_normalized << 20;
376 divider.i.h = ((1022 + 31) << 20) + (triangle_area_normalized >> 11);
378 d30.d = divident.d / divider.d; // d30 = ((1 << 62) + ta_n) / ta_n
380 // ((x1 - x0) * (y2 - y1)) - ((x2 - x1) * (y1 - y0)) =
381 // ( d0 * d1 ) - ( d2 * d3 ) =
382 // ( m0 ) - ( m1 ) = gradient
384 // This is split to do 12 elements at a time over three sets: a, b, and c.
385 // Technically we only need to do 10 elements (uvrgb_x and uvrgb_y), so
386 // two of the slots are unused.
388 // Inputs are all 16-bit signed. The m0/m1 results are 32-bit signed, as
391 // First type is: uvrg bxxx xxxx
392 // Second type is: yyyy ybyy uvrg
393 // Since x_a and y_c are the same the same variable is used for both.
417 gvreg uvrg_dx2, uvrg_dx3;
419 gvhreg zero, tmp_lo, tmp_hi;
421 gvld1q_u8(v0, (u8 *)a); // v0 = { uvrg0, b0, x0, y0 }
422 gvld1q_u8(v1, (u8 *)b); // v1 = { uvrg1, b1, x1, y1 }
423 gvld1q_u8(v2, (u8 *)c); // v2 = { uvrg2, b2, x2, y2 }
425 gvmovl_u8(uvrg_xxxx0, gvlo(v0)); // uvrg_xxxx0 = { uv0, rg0, b0-, -- }
426 gvmovl_u8(uvrg_xxxx1, gvlo(v1)); // uvrg_xxxx1 = { uv1, rg1, b1-, -- }
427 gvmovl_u8(uvrg_xxxx2, gvlo(v2)); // uvrg_xxxx2 = { uv2, rg2, b2-, -- }
429 gvdup_l_u16(tmp_lo, gvhi(v0), 1); // yyyy0 = { yy0, yy0 }
430 gvcombine_u16(y0_ab, tmp_lo, gvlo(uvrg_xxxx0));
432 gvdup_l_u16(tmp_lo, gvhi(v0), 0); // xxxx0 = { xx0, xx0 }
433 gvset_hi(uvrg_xxxx0, tmp_lo);
435 u32 x1_x2 = (u16)b->x | (c->x << 16); // x1_x2 = { x1, x2 }
436 u32 x0_x1 = (u16)a->x | (b->x << 16); // x0_x1 = { x0, x1 }
438 gvdup_l_u16(tmp_lo, gvhi(v1), 1); // yyyy1 = { yy1, yy1 }
439 gvcombine_u16(y1_ab, tmp_lo, gvlo(uvrg_xxxx1));
441 gvdup_l_u16(tmp_lo, gvhi(v1), 0); // xxxx1 = { xx1, xx1 }
442 gvset_hi(uvrg_xxxx1, tmp_lo);
444 gvdup_l_u16(tmp_lo, gvhi(v2), 1); // yyyy2 = { yy2, yy2 }
445 gvcombine_u16(y2_ab, tmp_lo, gvlo(uvrg_xxxx2));
447 gvdup_l_u16(tmp_lo, gvhi(v2), 0); // xxxx2 = { xx2, xx2 }
448 gvset_hi(uvrg_xxxx2, tmp_lo);
450 u32 y0_y1 = (u16)a->y | (b->y << 16); // y0_y1 = { y0, y1 }
451 u32 y1_y2 = (u16)b->y | (c->y << 16); // y1_y2 = { y1, y2 }
453 gvsubq_u16(d0_ab, uvrg_xxxx1, uvrg_xxxx0);
455 u32 b1_b2 = b->b | (c->b << 16); // b1_b2 = { b1, b2 }
457 gvsubq_u16(d2_ab, uvrg_xxxx2, uvrg_xxxx1);
459 gvsubq_u16(d1_ab, y2_ab, y1_ab);
461 u32 b0_b1 = a->b | (b->b << 16); // b0_b1 = { b0, b1 }
464 gssub16(dx, x1_x2, x0_x1); // dx = { x1 - x0, x2 - x1 }
465 gssub16(dy, y1_y2, y0_y1); // dy = { y1 - y0, y2 - y1 }
466 gssub16(db, b1_b2, b0_b1); // db = { b1 - b0, b2 - b1 }
469 gvsubq_u16(d3_ab, y1_ab, y0_ab);
470 gsmusdx(ga_by, dx, db); // ga_by = ((x1 - x0) * (b2 - b1)) -
471 // ((x2 - X1) * (b1 - b0))
472 gvmull_s16(ga_uvrg_x, gvlo(d0_ab), gvlo(d1_ab));
473 gsmusdx(ga_bx, db, dy); // ga_bx = ((b1 - b0) * (y2 - y1)) -
474 // ((b2 - b1) * (y1 - y0))
475 gvmlsl_s16(ga_uvrg_x, gvlo(d2_ab), gvlo(d3_ab));
476 u32 gs_bx = (s32)ga_bx >> 31; // movs
478 gvmull_s16(ga_uvrg_y, gvhi(d0_ab), gvhi(d1_ab));
479 if ((s32)gs_bx < 0) ga_bx = -ga_bx; // rsbmi
481 gvmlsl_s16(ga_uvrg_y, gvhi(d2_ab), gvhi(d3_ab));
482 u32 gs_by = (s32)ga_by >> 31; // movs
485 gvshr_n_u64(d0, d30.v, 22); // note: on "d30 >> 22" gcc generates junk code
487 gvdupq_n_u32(uvrgb_phase, psx_gpu->uvrgb_phase);
488 u32 b_base = psx_gpu->uvrgb_phase + (a->b << 16);
490 if ((s32)gs_by < 0) ga_by = -ga_by; // rsbmi
491 gvreg gs_uvrg_x, gs_uvrg_y;
492 gs_uvrg_x.s32 = ga_uvrg_x.s32 < 0; // gs_uvrg_x = ga_uvrg_x < 0
493 gs_uvrg_y.s32 = ga_uvrg_y.s32 < 0; // gs_uvrg_y = ga_uvrg_y < 0
495 gvdupq_n_u32(w_mask, -psx_gpu->triangle_winding); // w_mask = { -w, -w, -w, -w }
496 shift -= 62 - 12; // shift -= (62 - FIXED_BITS)
499 gvshll_n_u16(uvrg_base, gvlo(uvrg_xxxx0), 16); // uvrg_base = uvrg0 << 16
500 gvdupq_n_u32(r_shift, shift); // r_shift = { shift, shift, shift, shift }
502 gvaddq_u32(uvrg_base, uvrg_base, uvrgb_phase);
503 gvabsq_s32(ga_uvrg_x, ga_uvrg_x); // ga_uvrg_x = abs(ga_uvrg_x)
505 u32 area_r_s = d0.u32[0]; // area_r_s = triangle_reciprocal
506 gvabsq_s32(ga_uvrg_y, ga_uvrg_y); // ga_uvrg_y = abs(ga_uvrg_y)
508 gvmull_l_u32(gw_rg_x, gvhi(ga_uvrg_x), d0, 0);
509 gvmull_l_u32(ga_uvrg_x, gvlo(ga_uvrg_x), d0, 0);
510 gvmull_l_u32(gw_rg_y, gvhi(ga_uvrg_y), d0, 0);
511 gvmull_l_u32(ga_uvrg_y, gvlo(ga_uvrg_y), d0, 0);
513 gvshlq_u64(gw_rg_x, gw_rg_x, r_shift);
514 gvshlq_u64(ga_uvrg_x, ga_uvrg_x, r_shift);
515 gvshlq_u64(gw_rg_y, gw_rg_y, r_shift);
516 gvshlq_u64(ga_uvrg_y, ga_uvrg_y, r_shift);
518 gveorq(gs_uvrg_x, gs_uvrg_x, w_mask);
519 gvmovn_u64(tmp_lo, ga_uvrg_x);
521 gveorq(gs_uvrg_y, gs_uvrg_y, w_mask);
522 gvmovn_u64(tmp_hi, gw_rg_x);
524 gvcombine_u32(ga_uvrg_x, tmp_lo, tmp_hi);
526 gveorq(ga_uvrg_x, ga_uvrg_x, gs_uvrg_x);
527 gvmovn_u64(tmp_lo, ga_uvrg_y);
529 gvsubq_u32(ga_uvrg_x, ga_uvrg_x, gs_uvrg_x);
530 gvmovn_u64(tmp_hi, gw_rg_y);
532 gvcombine_u32(ga_uvrg_y, tmp_lo, tmp_hi);
534 gveorq(ga_uvrg_y, ga_uvrg_y, gs_uvrg_y);
537 gvsubq_u32(ga_uvrg_y, ga_uvrg_y, gs_uvrg_y);
540 u32 gw_bx_h, gw_by_h;
541 gw_bx_h = (u64)ga_bx * area_r_s >> 32;
543 gvshlq_n_u32(ga_uvrg_x, ga_uvrg_x, 4);
544 gvshlq_n_u32(ga_uvrg_y, ga_uvrg_y, 4);
546 gw_by_h = (u64)ga_by * area_r_s >> 32;
547 gvdup_n_u32(tmp_lo, a->x);
548 gvmlsq_l_s32(uvrg_base, ga_uvrg_x, tmp_lo, 0);
550 gs_bx = gs_bx ^ -psx_gpu->triangle_winding;
551 gvaddq_u32(uvrg_dx2, ga_uvrg_x, ga_uvrg_x);
553 gs_by = gs_by ^ -psx_gpu->triangle_winding;
555 u32 r11 = -shift; // r11 = negative shift for scalar lsr
556 u32 *store_a = psx_gpu->uvrg.e;
557 r11 = r11 - (32 - 13);
558 u32 *store_b = store_a + 16 / sizeof(u32);
560 gvaddq_u32(uvrg_dx3, uvrg_dx2, ga_uvrg_x);
561 gvst1q_inc_u32(uvrg_base, store_a, 32);
563 gvst1q_inc_u32(ga_uvrg_x, store_b, 32);
564 u32 g_bx = (u32)gw_bx_h >> r11;
566 gvst1q_inc_u32(ga_uvrg_y, store_a, 32);
567 u32 g_by = (u32)gw_by_h >> r11;
569 gvdup_n_u32(zero, 0);
571 gvst4_4_inc_u32(zero, gvlo(ga_uvrg_x), gvlo(uvrg_dx2), gvlo(uvrg_dx3), store_b, 32);
574 gvst4_4_inc_u32(zero, gvhi(ga_uvrg_x), gvhi(uvrg_dx2), gvhi(uvrg_dx3), store_b, 32);
580 b_base -= g_bx * a->x;
585 u32 g_bx2 = g_bx + g_bx;
586 u32 g_bx3 = g_bx + g_bx2;
594 store_b[5] = g_by; // 132
597 #define setup_spans_debug_check(span_edge_data_element) \
599 #define setup_spans_prologue_alternate_yes() \
600 vec_2x64s alternate_x; \
601 vec_2x64s alternate_dx_dy; \
602 vec_4x32s alternate_x_32; \
603 vec_2x32s alternate_x_16; \
605 vec_4x16u alternate_select; \
606 vec_4x16s y_mid_point; \
610 s32 edge_dx_dy_alt; \
613 #define setup_spans_prologue_alternate_no() \
615 #define setup_spans_prologue(alternate_active) \
616 edge_data_struct *span_edge_data; \
617 vec_4x32u *span_uvrg_offset; \
618 u32 *span_b_offset; \
623 union { vec_2x64s full; vec_1x64s h[2]; } edges_xy; \
624 vec_2x32s edges_dx_dy; \
625 vec_2x32u edge_shifts; \
627 vec_2x64s left_x, right_x; \
628 vec_2x64s left_dx_dy, right_dx_dy; \
629 vec_4x32s left_x_32, right_x_32; \
630 vec_2x32s left_x_32_lo, right_x_32_lo; \
631 vec_2x32s left_x_32_hi, right_x_32_hi; \
632 vec_4x16s left_right_x_16_lo, left_right_x_16_hi; \
634 vec_8x16s left_edge; \
635 vec_8x16s right_edge; \
636 vec_4x16u span_shift; \
640 vec_4x16u c_0xFFFE; \
643 vec_2x32s x_starts; \
654 u32 b = psx_gpu->b; \
655 u32 b_dy = psx_gpu->b_dy; \
656 const u32 *reciprocal_table = psx_gpu->reciprocal_table_ptr; \
658 gvld1q_u32(uvrg, psx_gpu->uvrg.e); \
659 gvld1q_u32(uvrg_dy, psx_gpu->uvrg_dy.e); \
660 gvdup_n_u32(c_0x01, 0x01); \
661 setup_spans_prologue_alternate_##alternate_active() \
663 #define setup_spans_prologue_b() \
664 span_edge_data = psx_gpu->span_edge_data; \
665 span_uvrg_offset = (vec_4x32u *)psx_gpu->span_uvrg_offset; \
666 span_b_offset = psx_gpu->span_b_offset; \
668 vec_8x16u c_0x0001; \
670 gvdupq_n_u16(c_0x0001, 0x0001); \
671 gvdupq_n_u16(left_edge, psx_gpu->viewport_start_x); \
672 gvdupq_n_u16(right_edge, psx_gpu->viewport_end_x); \
673 gvaddq_u16(right_edge, right_edge, c_0x0001); \
674 gvdup_n_u16(c_0x04, 0x04); \
675 gvdup_n_u16(c_0x07, 0x07); \
676 gvdup_n_u16(c_0xFFFE, 0xFFFE); \
679 #define compute_edge_delta_x2() \
682 vec_2x32s height_reciprocals; \
683 vec_2x32s heights_b; \
686 u32 edge_shift = reciprocal_table[height]; \
688 gvdup_n_u32(heights, height); \
689 gvsub_u32(widths, x_ends, x_starts); \
691 gvdup_n_u32(edge_shifts, edge_shift); \
692 gvsub_u32(heights_b, heights, c_0x01); \
693 gvshr_n_u32(height_reciprocals, edge_shifts, 10); \
695 gvmla_s32(heights_b, x_starts, heights); \
696 gvbic_n_u16(edge_shifts, 0xE0); \
697 gvmul_s32(edges_dx_dy, widths, height_reciprocals); \
698 gvmull_s32(edges_xy.full, heights_b, height_reciprocals); \
701 #define compute_edge_delta_x3(start_c, height_a, height_b) \
704 vec_2x32s height_reciprocals; \
705 vec_2x32s heights_b; \
710 u32 height_reciprocal_alt; \
712 gvcreate_u32(heights, height_a, height_b); \
713 gvcreate_u32(edge_shifts, reciprocal_table[height_a], reciprocal_table[height_b]); \
715 edge_shift_alt = reciprocal_table[height_minor_b]; \
717 gvsub_u32(widths, x_ends, x_starts); \
718 width_alt = x_c - start_c; \
720 gvshr_n_u32(height_reciprocals, edge_shifts, 10); \
721 height_reciprocal_alt = edge_shift_alt >> 10; \
723 gvbic_n_u16(edge_shifts, 0xE0); \
724 edge_shift_alt &= 0x1F; \
726 gvsub_u32(heights_b, heights, c_0x01); \
727 height_b_alt = height_minor_b - 1; \
729 gvmla_s32(heights_b, x_starts, heights); \
730 height_b_alt += height_minor_b * start_c; \
732 gvmull_s32(edges_xy.full, heights_b, height_reciprocals); \
733 edge_alt = (s64)height_b_alt * height_reciprocal_alt; \
735 gvmul_s32(edges_dx_dy, widths, height_reciprocals); \
736 edge_dx_dy_alt = width_alt * height_reciprocal_alt; \
740 #define setup_spans_adjust_y_up() \
741 gvsub_u32(y_x4, y_x4, c_0x04) \
743 #define setup_spans_adjust_y_down() \
744 gvadd_u32(y_x4, y_x4, c_0x04) \
746 #define setup_spans_adjust_interpolants_up() \
747 gvsubq_u32(uvrg, uvrg, uvrg_dy); \
750 #define setup_spans_adjust_interpolants_down() \
751 gvaddq_u32(uvrg, uvrg, uvrg_dy); \
755 #define setup_spans_clip_interpolants_increment() \
756 gvmlaq_s32(uvrg, uvrg_dy, v_clip); \
759 #define setup_spans_clip_interpolants_decrement() \
760 gvmlsq_s32(uvrg, uvrg_dy, v_clip); \
763 #define setup_spans_clip_alternate_yes() \
764 edge_alt += edge_dx_dy_alt * (s64)(clip) \
766 #define setup_spans_clip_alternate_no() \
768 #define setup_spans_clip(direction, alternate_active) \
770 gvdupq_n_u32(v_clip, clip); \
771 gvmlal_s32(edges_xy.full, edges_dx_dy, gvlo(v_clip)); \
772 setup_spans_clip_alternate_##alternate_active(); \
773 setup_spans_clip_interpolants_##direction(); \
777 #define setup_spans_adjust_edges_alternate_no(left_index, right_index) \
779 vec_2x64s edge_shifts_64; \
780 union { vec_2x64s full; vec_1x64s h[2]; } edges_dx_dy_64; \
781 vec_1x64s left_x_hi, right_x_hi; \
783 gvmovl_s32(edge_shifts_64, edge_shifts); \
784 gvshlq_s64(edges_xy.full, edges_xy.full, edge_shifts_64); \
786 gvmovl_s32(edges_dx_dy_64.full, edges_dx_dy); \
787 gvshlq_s64(edges_dx_dy_64.full, edges_dx_dy_64.full, edge_shifts_64); \
789 gvdupq_l_s64(left_x, edges_xy.h[left_index], 0); \
790 gvdupq_l_s64(right_x, edges_xy.h[right_index], 0); \
792 gvdupq_l_s64(left_dx_dy, edges_dx_dy_64.h[left_index], 0); \
793 gvdupq_l_s64(right_dx_dy, edges_dx_dy_64.h[right_index], 0); \
795 gvadd_s64(left_x_hi, gvlo(left_x), gvlo(left_dx_dy)); \
796 gvadd_s64(right_x_hi, gvlo(right_x), gvlo(right_dx_dy)); \
798 gvset_hi(left_x, left_x_hi); \
799 gvset_hi(right_x, right_x_hi); \
801 gvaddq_s64(left_dx_dy, left_dx_dy, left_dx_dy); \
802 gvaddq_s64(right_dx_dy, right_dx_dy, right_dx_dy); \
805 #define setup_spans_adjust_edges_alternate_yes(left_index, right_index) \
807 setup_spans_adjust_edges_alternate_no(left_index, right_index); \
808 s64 edge_dx_dy_alt_64; \
809 vec_1x64s alternate_x_hi; \
811 gvdup_n_u16(y_mid_point, y_b); \
813 edge_alt <<= edge_shift_alt; \
814 edge_dx_dy_alt_64 = (s64)edge_dx_dy_alt << edge_shift_alt; \
816 gvdupq_n_s64(alternate_x, edge_alt); \
817 gvdupq_n_s64(alternate_dx_dy, edge_dx_dy_alt_64); \
819 gvadd_s64(alternate_x_hi, gvlo(alternate_x), gvlo(alternate_dx_dy)); \
820 gvaddq_s64(alternate_dx_dy, alternate_dx_dy, alternate_dx_dy); \
821 gvset_hi(alternate_x, alternate_x_hi); \
825 #define setup_spans_y_select_up() \
826 gvclt_s16(alternate_select, y_x4, y_mid_point) \
828 #define setup_spans_y_select_down() \
829 gvcgt_s16(alternate_select, y_x4, y_mid_point) \
831 #define setup_spans_y_select_alternate_yes(direction) \
832 setup_spans_y_select_##direction() \
834 #define setup_spans_y_select_alternate_no(direction) \
836 #define setup_spans_alternate_select_left() \
837 gvbit(left_right_x_16_lo, alternate_x_16, alternate_select); \
839 #define setup_spans_alternate_select_right() \
840 gvbit(left_right_x_16_hi, alternate_x_16, alternate_select); \
842 #define setup_spans_alternate_select_none() \
844 #define setup_spans_increment_alternate_yes() \
846 vec_2x32s alternate_x_32_lo, alternate_x_32_hi; \
847 gvshrn_n_s64(alternate_x_32_lo, alternate_x, 32); \
848 gvaddq_s64(alternate_x, alternate_x, alternate_dx_dy); \
849 gvshrn_n_s64(alternate_x_32_hi, alternate_x, 32); \
850 gvaddq_s64(alternate_x, alternate_x, alternate_dx_dy); \
851 gvcombine_u32(alternate_x_32, alternate_x_32_lo, alternate_x_32_hi); \
852 gvmovn_u32(alternate_x_16, alternate_x_32); \
855 #define setup_spans_increment_alternate_no() \
857 #define setup_spans_set_x4(alternate, direction, alternate_active) \
859 gvst1q_pi_u32(uvrg, span_uvrg_offset); \
860 *span_b_offset++ = b; \
861 setup_spans_adjust_interpolants_##direction(); \
863 gvst1q_pi_u32(uvrg, span_uvrg_offset); \
864 *span_b_offset++ = b; \
865 setup_spans_adjust_interpolants_##direction(); \
867 gvst1q_pi_u32(uvrg, span_uvrg_offset); \
868 *span_b_offset++ = b; \
869 setup_spans_adjust_interpolants_##direction(); \
871 gvst1q_pi_u32(uvrg, span_uvrg_offset); \
872 *span_b_offset++ = b; \
873 setup_spans_adjust_interpolants_##direction(); \
875 gvshrn_n_s64(left_x_32_lo, left_x, 32); \
876 gvshrn_n_s64(right_x_32_lo, right_x, 32); \
878 gvaddq_s64(left_x, left_x, left_dx_dy); \
879 gvaddq_s64(right_x, right_x, right_dx_dy); \
881 gvshrn_n_s64(left_x_32_hi, left_x, 32); \
882 gvshrn_n_s64(right_x_32_hi, right_x, 32); \
884 gvaddq_s64(left_x, left_x, left_dx_dy); \
885 gvaddq_s64(right_x, right_x, right_dx_dy); \
887 gvcombine_s64(left_x_32, left_x_32_lo, left_x_32_hi); \
888 gvcombine_s64(right_x_32, right_x_32_lo, right_x_32_hi); \
890 gvmovn_u32(left_right_x_16_lo, left_x_32); \
891 gvmovn_u32(left_right_x_16_hi, right_x_32); \
893 setup_spans_increment_alternate_##alternate_active(); \
894 setup_spans_y_select_alternate_##alternate_active(direction); \
895 setup_spans_alternate_select_##alternate(); \
897 gvmax_s16(left_right_x_16_lo, left_right_x_16_lo, gvlo(left_edge)); \
898 gvmax_s16(left_right_x_16_hi, left_right_x_16_hi, gvhi(left_edge)); \
899 gvmin_s16(left_right_x_16_lo, left_right_x_16_lo, gvlo(right_edge)); \
900 gvmin_s16(left_right_x_16_hi, left_right_x_16_hi, gvhi(right_edge)); \
902 gvsub_u16(left_right_x_16_hi, left_right_x_16_hi, left_right_x_16_lo); \
903 gvadd_u16(left_right_x_16_hi, left_right_x_16_hi, c_0x07); \
904 gvand(span_shift, left_right_x_16_hi, c_0x07); \
905 gvshl_u16(span_shift, c_0xFFFE, span_shift); \
906 gvshr_n_u16(left_right_x_16_hi, left_right_x_16_hi, 3); \
908 gvst4_pi_u16(left_right_x_16_lo, left_right_x_16_hi, span_shift, y_x4, \
911 setup_spans_adjust_y_##direction(); \
915 #define setup_spans_alternate_adjust_yes() \
916 edge_alt -= edge_dx_dy_alt * (s64)height_minor_a \
918 #define setup_spans_alternate_adjust_no() \
921 #define setup_spans_down(left_index, right_index, alternate, alternate_active) \
922 setup_spans_alternate_adjust_##alternate_active(); \
923 if(y_c > psx_gpu->viewport_end_y) \
924 height -= y_c - psx_gpu->viewport_end_y - 1; \
926 clip = psx_gpu->viewport_start_y - y_a; \
931 setup_spans_clip(increment, alternate_active); \
934 setup_spans_prologue_b(); \
938 u64 y_x4_ = ((u64)(y_a + 3) << 48) | ((u64)(u16)(y_a + 2) << 32) \
939 | (u32)((y_a + 1) << 16) | (u16)y_a; \
940 gvcreate_u64(y_x4, y_x4_); \
941 setup_spans_adjust_edges_alternate_##alternate_active(left_index, \
944 psx_gpu->num_spans = height; \
947 setup_spans_set_x4(alternate, down, alternate_active); \
949 } while(height > 0); \
953 #define setup_spans_alternate_pre_increment_yes() \
954 edge_alt += edge_dx_dy_alt \
956 #define setup_spans_alternate_pre_increment_no() \
958 #define setup_spans_up_decrement_height_yes() \
961 #define setup_spans_up_decrement_height_no() \
964 #define setup_spans_up(left_index, right_index, alternate, alternate_active) \
965 setup_spans_alternate_adjust_##alternate_active(); \
968 if(y_c < psx_gpu->viewport_start_y) \
969 height -= psx_gpu->viewport_start_y - y_c; \
971 setup_spans_up_decrement_height_##alternate_active(); \
973 clip = y_a - psx_gpu->viewport_end_y; \
978 setup_spans_clip(decrement, alternate_active); \
981 setup_spans_prologue_b(); \
985 u64 y_x4_ = ((u64)(y_a - 3) << 48) | ((u64)(u16)(y_a - 2) << 32) \
986 | (u32)((y_a - 1) << 16) | (u16)y_a; \
987 gvcreate_u64(y_x4, y_x4_); \
988 gvaddw_s32(edges_xy.full, edges_xy.full, edges_dx_dy); \
989 setup_spans_alternate_pre_increment_##alternate_active(); \
990 setup_spans_adjust_edges_alternate_##alternate_active(left_index, \
992 setup_spans_adjust_interpolants_up(); \
994 psx_gpu->num_spans = height; \
997 setup_spans_set_x4(alternate, up, alternate_active); \
1002 #define index_left 0
1003 #define index_right 1
1005 #define setup_spans_up_up(minor, major) \
1006 setup_spans_prologue(yes); \
1007 s32 height_minor_a = y_a - y_b; \
1008 s32 height_minor_b = y_b - y_c; \
1009 s32 height = y_a - y_c; \
1011 gvdup_n_u32(x_starts, x_a); \
1012 gvcreate_u32(x_ends, x_c, x_b); \
1014 compute_edge_delta_x3(x_b, height, height_minor_a); \
1015 setup_spans_up(index_##major, index_##minor, minor, yes) \
1017 void setup_spans_up_left(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1018 vertex_struct *v_b, vertex_struct *v_c)
1021 setup_spans_up_left_(psx_gpu, v_a, v_b, v_c);
1024 setup_spans_up_up(left, right)
1027 void setup_spans_up_right(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1028 vertex_struct *v_b, vertex_struct *v_c)
1031 setup_spans_up_right_(psx_gpu, v_a, v_b, v_c);
1034 setup_spans_up_up(right, left)
1037 #define setup_spans_down_down(minor, major) \
1038 setup_spans_prologue(yes); \
1039 s32 height_minor_a = y_b - y_a; \
1040 s32 height_minor_b = y_c - y_b; \
1041 s32 height = y_c - y_a; \
1043 gvdup_n_u32(x_starts, x_a); \
1044 gvcreate_u32(x_ends, x_c, x_b); \
1046 compute_edge_delta_x3(x_b, height, height_minor_a); \
1047 setup_spans_down(index_##major, index_##minor, minor, yes) \
1049 void setup_spans_down_left(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1050 vertex_struct *v_b, vertex_struct *v_c)
1053 setup_spans_down_left_(psx_gpu, v_a, v_b, v_c);
1056 setup_spans_down_down(left, right)
1059 void setup_spans_down_right(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1060 vertex_struct *v_b, vertex_struct *v_c)
1063 setup_spans_down_right_(psx_gpu, v_a, v_b, v_c);
1066 setup_spans_down_down(right, left)
1069 #define setup_spans_up_flat() \
1070 s32 height = y_a - y_c; \
1072 compute_edge_delta_x2(); \
1073 setup_spans_up(index_left, index_right, none, no) \
1075 void setup_spans_up_a(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1076 vertex_struct *v_b, vertex_struct *v_c)
1079 setup_spans_up_a_(psx_gpu, v_a, v_b, v_c);
1082 setup_spans_prologue(no);
1084 gvcreate_u32(x_starts, x_a, x_b);
1085 gvdup_n_u32(x_ends, x_c);
1087 setup_spans_up_flat()
1090 void setup_spans_up_b(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1091 vertex_struct *v_b, vertex_struct *v_c)
1094 setup_spans_up_b_(psx_gpu, v_a, v_b, v_c);
1097 setup_spans_prologue(no);
1099 gvdup_n_u32(x_starts, x_a);
1100 gvcreate_u32(x_ends, x_b, x_c);
1102 setup_spans_up_flat()
1105 #define setup_spans_down_flat() \
1106 s32 height = y_c - y_a; \
1108 compute_edge_delta_x2(); \
1109 setup_spans_down(index_left, index_right, none, no) \
1111 void setup_spans_down_a(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1112 vertex_struct *v_b, vertex_struct *v_c)
1115 setup_spans_down_a_(psx_gpu, v_a, v_b, v_c);
1118 setup_spans_prologue(no);
1120 gvcreate_u32(x_starts, x_a, x_b);
1121 gvdup_n_u32(x_ends, x_c);
1123 setup_spans_down_flat()
1126 void setup_spans_down_b(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1127 vertex_struct *v_b, vertex_struct *v_c)
1130 setup_spans_down_b_(psx_gpu, v_a, v_b, v_c);
1133 setup_spans_prologue(no)
1135 gvdup_n_u32(x_starts, x_a);
1136 gvcreate_u32(x_ends, x_b, x_c);
1138 setup_spans_down_flat()
1141 void setup_spans_up_down(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1142 vertex_struct *v_b, vertex_struct *v_c)
1145 setup_spans_up_down_(psx_gpu, v_a, v_b, v_c);
1148 setup_spans_prologue(no);
1156 s32 height_minor_a = y_a - y_b;
1157 s32 height_minor_b = y_c - y_a;
1158 s32 height_major = y_c - y_b;
1160 vec_2x64s edges_xy_b;
1161 vec_1x64s edges_xy_b_left;
1162 vec_2x32s edges_dx_dy_b;
1163 vec_2x32u edge_shifts_b;
1165 vec_2x32s height_increment;
1167 gvcreate_u32(x_starts, x_a, x_c);
1168 gvdup_n_u32(x_ends, x_b);
1170 compute_edge_delta_x3(x_a, height_minor_a, height_major);
1172 gvcreate_s32(height_increment, 0, height_minor_b);
1174 gvmlal_s32(edges_xy.full, edges_dx_dy, height_increment);
1176 gvcreate_s64(edges_xy_b_left, edge_alt);
1177 gvcombine_s64(edges_xy_b, edges_xy_b_left, gvhi(edges_xy.full));
1179 edge_shifts_b = edge_shifts;
1180 gvmov_l_u32(edge_shifts_b, edge_shift_alt, 0);
1182 gvneg_s32(edges_dx_dy_b, edges_dx_dy);
1183 gvmov_l_s32(edges_dx_dy_b, edge_dx_dy_alt, 0);
1187 if(y_b < psx_gpu->viewport_start_y)
1188 height_minor_a -= psx_gpu->viewport_start_y - y_b;
1190 clip = y_a - psx_gpu->viewport_end_y;
1193 height_minor_a -= clip;
1195 setup_spans_clip(decrement, no);
1198 setup_spans_prologue_b();
1200 if(height_minor_a > 0)
1202 u64 y_x4_ = ((u64)(y_a - 3) << 48) | ((u64)(u16)(y_a - 2) << 32)
1203 | (u32)((y_a - 1) << 16) | (u16)y_a;
1204 gvcreate_u64(y_x4, y_x4_);
1205 gvaddw_s32(edges_xy.full, edges_xy.full, edges_dx_dy);
1206 setup_spans_adjust_edges_alternate_no(index_left, index_right);
1207 setup_spans_adjust_interpolants_up();
1209 psx_gpu->num_spans = height_minor_a;
1210 while(height_minor_a > 0)
1212 setup_spans_set_x4(none, up, no);
1213 height_minor_a -= 4;
1216 span_edge_data += height_minor_a;
1217 span_uvrg_offset += height_minor_a;
1218 span_b_offset += height_minor_a;
1221 edges_xy.full = edges_xy_b;
1222 edges_dx_dy = edges_dx_dy_b;
1223 edge_shifts = edge_shifts_b;
1225 gvld1q_u32(uvrg, psx_gpu->uvrg.e);
1230 if(y_c > psx_gpu->viewport_end_y)
1231 height_minor_b -= y_c - psx_gpu->viewport_end_y - 1;
1233 clip = psx_gpu->viewport_start_y - y_a;
1236 height_minor_b -= clip;
1238 setup_spans_clip(increment, no);
1241 if(height_minor_b > 0)
1243 u64 y_x4_ = ((u64)(y_a + 3) << 48) | ((u64)(u16)(y_a + 2) << 32)
1244 | (u32)((y_a + 1) << 16) | (u16)y_a;
1245 gvcreate_u64(y_x4, y_x4_);
1246 setup_spans_adjust_edges_alternate_no(index_left, index_right);
1248 // FIXME: overflow corner case
1249 if(psx_gpu->num_spans + height_minor_b == MAX_SPANS)
1250 height_minor_b &= ~3;
1252 psx_gpu->num_spans += height_minor_b;
1253 while(height_minor_b > 0)
1255 setup_spans_set_x4(none, down, no);
1256 height_minor_b -= 4;
1262 #define dither_table_entry_normal(value) \
1265 #define setup_blocks_load_msb_mask_indirect() \
1267 #define setup_blocks_load_msb_mask_direct() \
1268 vec_8x16u msb_mask; \
1269 gvdupq_n_u16(msb_mask, psx_gpu->mask_msb); \
1271 #define setup_blocks_variables_shaded_textured(target) \
1272 vec_4x32u u_block; \
1273 vec_4x32u v_block; \
1274 vec_4x32u r_block; \
1275 vec_4x32u g_block; \
1276 vec_4x32u b_block; \
1277 vec_4x32u uvrg_dx; \
1278 vec_4x32u uvrg_dx4; \
1279 vec_4x32u uvrg_dx8; \
1281 vec_16x8u texture_mask; \
1282 vec_8x8u texture_mask_lo, texture_mask_hi; \
1283 u32 b_dx = psx_gpu->b_block_span.e[1]; \
1284 u32 b_dx4 = b_dx << 2; \
1285 u32 b_dx8 = b_dx << 3; \
1288 gvld1q_u32(uvrg_dx, psx_gpu->uvrg_dx.e); \
1289 gvshlq_n_u32(uvrg_dx4, uvrg_dx, 2); \
1290 gvshlq_n_u32(uvrg_dx8, uvrg_dx, 3); \
1291 gvld2_dup(texture_mask_lo, texture_mask_hi, &psx_gpu->texture_mask_width); \
1292 gvcombine_u16(texture_mask, texture_mask_lo, texture_mask_hi) \
1294 #define setup_blocks_variables_shaded_untextured(target) \
1295 vec_4x32u r_block; \
1296 vec_4x32u g_block; \
1297 vec_4x32u b_block; \
1299 vec_2x32u rgb_dx_lo, rgb_dx_hi; \
1300 vec_4x32u rgb_dx4; \
1301 vec_4x32u rgb_dx8; \
1303 vec_2x32u rgb_lo, rgb_hi; \
1305 vec_8x8u d64_0x07; \
1310 gvdup_n_u8(d64_0x07, 0x07); \
1311 gvdup_n_u8(d64_1, 1); \
1312 gvdup_n_u8(d64_4, 4); \
1313 gvdup_n_u8(d64_128, 128); \
1315 gvld1_u32(rgb_dx_lo, &psx_gpu->uvrg_dx.e[2]); \
1316 gvcreate_u32(rgb_dx_hi, psx_gpu->b_block_span.e[1], 0); \
1317 gvcombine_u32(rgb_dx, rgb_dx_lo, rgb_dx_hi); \
1318 gvshlq_n_u32(rgb_dx4, rgb_dx, 2); \
1319 gvshlq_n_u32(rgb_dx8, rgb_dx, 3) \
1321 #define setup_blocks_variables_unshaded_textured(target) \
1322 vec_4x32u u_block; \
1323 vec_4x32u v_block; \
1328 vec_16x8u texture_mask; \
1329 vec_8x8u texture_mask_lo, texture_mask_hi; \
1331 gvld1_u32(uv_dx, psx_gpu->uvrg_dx.e); \
1332 gvld1_u32(uv, psx_gpu->uvrg.e); \
1333 gvshl_n_u32(uv_dx4, uv_dx, 2); \
1334 gvshl_n_u32(uv_dx8, uv_dx, 3); \
1335 gvld2_dup(texture_mask_lo, texture_mask_hi, &psx_gpu->texture_mask_width); \
1336 gvcombine_u16(texture_mask, texture_mask_lo, texture_mask_hi) \
1338 #define setup_blocks_variables_unshaded_untextured_direct() \
1339 gvorrq(colors, colors, msb_mask) \
1341 #define setup_blocks_variables_unshaded_untextured_indirect() \
1343 #define setup_blocks_variables_unshaded_untextured(target) \
1344 u32 color = psx_gpu->triangle_color; \
1347 u32 color_r = color & 0xFF; \
1348 u32 color_g = (color >> 8) & 0xFF; \
1349 u32 color_b = (color >> 16) & 0xFF; \
1351 color = (color_r >> 3) | ((color_g >> 3) << 5) | \
1352 ((color_b >> 3) << 10); \
1353 gvdupq_n_u16(colors, color); \
1354 setup_blocks_variables_unshaded_untextured_##target() \
1356 #define setup_blocks_span_initialize_dithered_textured() \
1357 vec_8x16u dither_offsets; \
1358 gvshll_n_s8(dither_offsets, dither_offsets_short, 4) \
1360 #define setup_blocks_span_initialize_dithered_untextured() \
1361 vec_8x8u dither_offsets; \
1362 gvadd_u8(dither_offsets, dither_offsets_short, d64_4) \
1364 #define setup_blocks_span_initialize_dithered(texturing) \
1365 u32 dither_row = psx_gpu->dither_table[y & 0x3]; \
1366 u32 dither_shift = (span_edge_data->left_x & 0x3) * 8; \
1367 vec_8x8s dither_offsets_short; \
1370 (dither_row >> dither_shift) | (dither_row << (32 - dither_shift)); \
1371 gvdup_n_u32(dither_offsets_short, dither_row); \
1372 setup_blocks_span_initialize_dithered_##texturing() \
1374 #define setup_blocks_span_initialize_undithered(texturing) \
1376 #define setup_blocks_span_initialize_shaded_textured() \
1378 u32 left_x = span_edge_data->left_x; \
1379 vec_4x32u block_span; \
1380 vec_4x32u v_left_x; \
1382 gvld1q_u32(uvrg, span_uvrg_offset); \
1383 gvdupq_n_u32(v_left_x, left_x); \
1384 gvmlaq_u32(uvrg, uvrg_dx, v_left_x); \
1385 b = *span_b_offset; \
1386 b += b_dx * left_x; \
1388 gvdupq_l_u32(u_block, gvlo(uvrg), 0); \
1389 gvdupq_l_u32(v_block, gvlo(uvrg), 1); \
1390 gvdupq_l_u32(r_block, gvhi(uvrg), 0); \
1391 gvdupq_l_u32(g_block, gvhi(uvrg), 1); \
1392 gvdupq_n_u32(b_block, b); \
1394 gvld1q_u32(block_span, psx_gpu->u_block_span.e); \
1395 gvaddq_u32(u_block, u_block, block_span); \
1396 gvld1q_u32(block_span, psx_gpu->v_block_span.e); \
1397 gvaddq_u32(v_block, v_block, block_span); \
1398 gvld1q_u32(block_span, psx_gpu->r_block_span.e); \
1399 gvaddq_u32(r_block, r_block, block_span); \
1400 gvld1q_u32(block_span, psx_gpu->g_block_span.e); \
1401 gvaddq_u32(g_block, g_block, block_span); \
1402 gvld1q_u32(block_span, psx_gpu->b_block_span.e); \
1403 gvaddq_u32(b_block, b_block, block_span); \
1406 #define setup_blocks_span_initialize_shaded_untextured() \
1408 u32 left_x = span_edge_data->left_x; \
1409 u32 *span_uvrg_offset_high = (u32 *)span_uvrg_offset + 2; \
1410 vec_4x32u block_span; \
1411 vec_4x32u v_left_x; \
1413 gvld1_u32(rgb_lo, span_uvrg_offset_high); \
1414 gvcreate_u32(rgb_hi, *span_b_offset, 0); \
1415 gvcombine_u32(rgb, rgb_lo, rgb_hi); \
1416 gvdupq_n_u32(v_left_x, left_x); \
1417 gvmlaq_u32(rgb, rgb_dx, v_left_x); \
1419 gvdupq_l_u32(r_block, gvlo(rgb), 0); \
1420 gvdupq_l_u32(g_block, gvlo(rgb), 1); \
1421 gvdupq_l_u32(b_block, gvhi(rgb), 0); \
1423 gvld1q_u32(block_span, psx_gpu->r_block_span.e); \
1424 gvaddq_u32(r_block, r_block, block_span); \
1425 gvld1q_u32(block_span, psx_gpu->g_block_span.e); \
1426 gvaddq_u32(g_block, g_block, block_span); \
1427 gvld1q_u32(block_span, psx_gpu->b_block_span.e); \
1428 gvaddq_u32(b_block, b_block, block_span); \
1431 #define setup_blocks_span_initialize_unshaded_textured() \
1433 u32 left_x = span_edge_data->left_x; \
1434 vec_4x32u block_span; \
1435 vec_2x32u v_left_x; \
1437 gvld1_u32(uv, span_uvrg_offset); \
1438 gvdup_n_u32(v_left_x, left_x); \
1439 gvmla_u32(uv, uv_dx, v_left_x); \
1441 gvdupq_l_u32(u_block, uv, 0); \
1442 gvdupq_l_u32(v_block, uv, 1); \
1444 gvld1q_u32(block_span, psx_gpu->u_block_span.e); \
1445 gvaddq_u32(u_block, u_block, block_span); \
1446 gvld1q_u32(block_span, psx_gpu->v_block_span.e); \
1447 gvaddq_u32(v_block, v_block, block_span); \
1450 #define setup_blocks_span_initialize_unshaded_untextured() \
1452 #define setup_blocks_texture_swizzled() \
1454 vec_8x8u u_saved = u; \
1455 gvsli_n_u8(u, v, 4); \
1456 gvsri_n_u8(v, u_saved, 4); \
1459 #define setup_blocks_texture_unswizzled() \
1461 #define setup_blocks_store_shaded_textured(swizzling, dithering, target, \
1464 vec_8x16u u_whole; \
1465 vec_8x16u v_whole; \
1466 vec_8x16u r_whole; \
1467 vec_8x16u g_whole; \
1468 vec_8x16u b_whole; \
1469 vec_4x16u u_whole_lo, u_whole_hi; \
1470 vec_4x16u v_whole_lo, v_whole_hi; \
1471 vec_4x16u r_whole_lo, r_whole_hi; \
1472 vec_4x16u g_whole_lo, g_whole_hi; \
1473 vec_4x16u b_whole_lo, b_whole_hi; \
1484 gvshrn_n_u32(u_whole_lo, u_block, 16); \
1485 gvshrn_n_u32(v_whole_lo, v_block, 16); \
1486 gvshrn_n_u32(r_whole_lo, r_block, 16); \
1487 gvshrn_n_u32(g_whole_lo, g_block, 16); \
1488 gvshrn_n_u32(b_whole_lo, b_block, 16); \
1490 gvdupq_l_u32(dx4, gvlo(uvrg_dx4), 0); \
1491 gvaddhn_u32(u_whole_hi, u_block, dx4); \
1492 gvdupq_l_u32(dx4, gvlo(uvrg_dx4), 1); \
1493 gvaddhn_u32(v_whole_hi, v_block, dx4); \
1494 gvdupq_l_u32(dx4, gvhi(uvrg_dx4), 0); \
1495 gvaddhn_u32(r_whole_hi, r_block, dx4); \
1496 gvdupq_l_u32(dx4, gvhi(uvrg_dx4), 1); \
1497 gvaddhn_u32(g_whole_hi, g_block, dx4); \
1498 gvdupq_n_u32(dx4, b_dx4); \
1499 gvaddhn_u32(b_whole_hi, b_block, dx4); \
1501 gvcombine_u16(u_whole, u_whole_lo, u_whole_hi); \
1502 gvcombine_u16(v_whole, v_whole_lo, v_whole_hi); \
1503 gvcombine_u16(r_whole, r_whole_lo, r_whole_hi); \
1504 gvcombine_u16(g_whole, g_whole_lo, g_whole_hi); \
1505 gvcombine_u16(b_whole, b_whole_lo, b_whole_hi); \
1506 gvmovn_u16(u, u_whole); \
1507 gvmovn_u16(v, v_whole); \
1508 gvmovn_u16(r, r_whole); \
1509 gvmovn_u16(g, g_whole); \
1510 gvmovn_u16(b, b_whole); \
1512 gvdupq_l_u32(dx8, gvlo(uvrg_dx8), 0); \
1513 gvaddq_u32(u_block, u_block, dx8); \
1514 gvdupq_l_u32(dx8, gvlo(uvrg_dx8), 1); \
1515 gvaddq_u32(v_block, v_block, dx8); \
1516 gvdupq_l_u32(dx8, gvhi(uvrg_dx8), 0); \
1517 gvaddq_u32(r_block, r_block, dx8); \
1518 gvdupq_l_u32(dx8, gvhi(uvrg_dx8), 1); \
1519 gvaddq_u32(g_block, g_block, dx8); \
1520 gvdupq_n_u32(dx8, b_dx8); \
1521 gvaddq_u32(b_block, b_block, dx8); \
1523 gvand(u, u, gvlo(texture_mask)); \
1524 gvand(v, v, gvhi(texture_mask)); \
1525 setup_blocks_texture_##swizzling(); \
1527 gvst2_u8(u, v, (u8 *)block->uv.e); \
1528 gvst1_u8(r, block->r.e); \
1529 gvst1_u8(g, block->g.e); \
1530 gvst1_u8(b, block->b.e); \
1531 gvst1q_u16(dither_offsets, (u16 *)block->dither_offsets.e); \
1532 block->fb_ptr = fb_ptr; \
1535 #define setup_blocks_store_unshaded_textured(swizzling, dithering, target, \
1538 vec_8x16u u_whole; \
1539 vec_8x16u v_whole; \
1540 vec_4x16u u_whole_lo, u_whole_hi; \
1541 vec_4x16u v_whole_lo, v_whole_hi; \
1549 gvshrn_n_u32(u_whole_lo, u_block, 16); \
1550 gvshrn_n_u32(v_whole_lo, v_block, 16); \
1552 gvdupq_l_u32(dx4, uv_dx4, 0); \
1553 gvaddhn_u32(u_whole_hi, u_block, dx4); \
1554 gvdupq_l_u32(dx4, uv_dx4, 1); \
1555 gvaddhn_u32(v_whole_hi, v_block, dx4); \
1557 gvcombine_u16(u_whole, u_whole_lo, u_whole_hi); \
1558 gvcombine_u16(v_whole, v_whole_lo, v_whole_hi); \
1559 gvmovn_u16(u, u_whole); \
1560 gvmovn_u16(v, v_whole); \
1562 gvdupq_l_u32(dx8, uv_dx8, 0); \
1563 gvaddq_u32(u_block, u_block, dx8); \
1564 gvdupq_l_u32(dx8, uv_dx8, 1); \
1565 gvaddq_u32(v_block, v_block, dx8); \
1567 gvand(u, u, gvlo(texture_mask)); \
1568 gvand(v, v, gvhi(texture_mask)); \
1569 setup_blocks_texture_##swizzling(); \
1571 gvst2_u8(u, v, (u8 *)block->uv.e); \
1572 gvst1q_u16(dither_offsets, (u16 *)block->dither_offsets.e); \
1573 block->fb_ptr = fb_ptr; \
1576 #define setup_blocks_store_shaded_untextured_dithered() \
1577 gvqadd_u8(r, r, dither_offsets); \
1578 gvqadd_u8(g, g, dither_offsets); \
1579 gvqadd_u8(b, b, dither_offsets); \
1581 gvqsub_u8(r, r, d64_4); \
1582 gvqsub_u8(g, g, d64_4); \
1583 gvqsub_u8(b, b, d64_4) \
1585 #define setup_blocks_store_shaded_untextured_undithered() \
1587 #define setup_blocks_store_untextured_pixels_indirect_full(_pixels) \
1588 gvst1q_u16(_pixels, block->pixels.e); \
1589 block->fb_ptr = fb_ptr \
1591 #define setup_blocks_store_untextured_pixels_indirect_edge(_pixels) \
1592 gvst1q_u16(_pixels, block->pixels.e); \
1593 block->fb_ptr = fb_ptr \
1595 #define setup_blocks_store_shaded_untextured_seed_pixels_indirect() \
1596 gvmull_u8(pixels, r, d64_1) \
1598 #define setup_blocks_store_untextured_pixels_direct_full(_pixels) \
1599 gvst1q_u16(_pixels, fb_ptr) \
1601 #define setup_blocks_store_untextured_pixels_direct_edge(_pixels) \
1603 vec_8x16u fb_pixels; \
1604 vec_8x16u draw_mask; \
1605 vec_8x16u test_mask; \
1607 gvld1q_u16(test_mask, psx_gpu->test_mask.e); \
1608 gvld1q_u16(fb_pixels, fb_ptr); \
1609 gvdupq_n_u16(draw_mask, span_edge_data->right_mask); \
1610 gvtstq_u16(draw_mask, draw_mask, test_mask); \
1611 gvbifq(fb_pixels, _pixels, draw_mask); \
1612 gvst1q_u16(fb_pixels, fb_ptr); \
1615 #define setup_blocks_store_shaded_untextured_seed_pixels_direct() \
1616 pixels = msb_mask; \
1617 gvmlal_u8(pixels, r, d64_1) \
1619 #define setup_blocks_store_shaded_untextured(swizzling, dithering, target, \
1622 vec_8x16u r_whole; \
1623 vec_8x16u g_whole; \
1624 vec_8x16u b_whole; \
1625 vec_4x16u r_whole_lo, r_whole_hi; \
1626 vec_4x16u g_whole_lo, g_whole_hi; \
1627 vec_4x16u b_whole_lo, b_whole_hi; \
1638 gvshrn_n_u32(r_whole_lo, r_block, 16); \
1639 gvshrn_n_u32(g_whole_lo, g_block, 16); \
1640 gvshrn_n_u32(b_whole_lo, b_block, 16); \
1642 gvdupq_l_u32(dx4, gvlo(rgb_dx4), 0); \
1643 gvaddhn_u32(r_whole_hi, r_block, dx4); \
1644 gvdupq_l_u32(dx4, gvlo(rgb_dx4), 1); \
1645 gvaddhn_u32(g_whole_hi, g_block, dx4); \
1646 gvdupq_l_u32(dx4, gvhi(rgb_dx4), 0); \
1647 gvaddhn_u32(b_whole_hi, b_block, dx4); \
1649 gvcombine_u16(r_whole, r_whole_lo, r_whole_hi); \
1650 gvcombine_u16(g_whole, g_whole_lo, g_whole_hi); \
1651 gvcombine_u16(b_whole, b_whole_lo, b_whole_hi); \
1652 gvmovn_u16(r, r_whole); \
1653 gvmovn_u16(g, g_whole); \
1654 gvmovn_u16(b, b_whole); \
1656 gvdupq_l_u32(dx8, gvlo(rgb_dx8), 0); \
1657 gvaddq_u32(r_block, r_block, dx8); \
1658 gvdupq_l_u32(dx8, gvlo(rgb_dx8), 1); \
1659 gvaddq_u32(g_block, g_block, dx8); \
1660 gvdupq_l_u32(dx8, gvhi(rgb_dx8), 0); \
1661 gvaddq_u32(b_block, b_block, dx8); \
1663 setup_blocks_store_shaded_untextured_##dithering(); \
1665 gvshr_n_u8(r, r, 3); \
1666 gvbic(g, g, d64_0x07); \
1667 gvbic(b, b, d64_0x07); \
1669 setup_blocks_store_shaded_untextured_seed_pixels_##target(); \
1670 gvmlal_u8(pixels, g, d64_4); \
1671 gvmlal_u8(pixels, b, d64_128); \
1673 setup_blocks_store_untextured_pixels_##target##_##edge_type(pixels); \
1676 #define setup_blocks_store_unshaded_untextured(swizzling, dithering, target, \
1678 setup_blocks_store_untextured_pixels_##target##_##edge_type(colors) \
1680 #define setup_blocks_store_draw_mask_textured_indirect(_block, bits) \
1681 (_block)->draw_mask_bits = bits \
1683 #define setup_blocks_store_draw_mask_untextured_indirect(_block, bits) \
1685 vec_8x16u bits_mask; \
1686 vec_8x16u test_mask; \
1688 gvld1q_u16(test_mask, psx_gpu->test_mask.e); \
1689 gvdupq_n_u16(bits_mask, bits); \
1690 gvtstq_u16(bits_mask, bits_mask, test_mask); \
1691 gvst1q_u16(bits_mask, (_block)->draw_mask.e); \
1694 #define setup_blocks_store_draw_mask_untextured_direct(_block, bits) \
1696 #define setup_blocks_add_blocks_indirect() \
1697 num_blocks += span_num_blocks; \
1699 if(num_blocks > MAX_BLOCKS) \
1701 psx_gpu->num_blocks = num_blocks - span_num_blocks; \
1702 flush_render_block_buffer(psx_gpu); \
1703 num_blocks = span_num_blocks; \
1704 block = psx_gpu->blocks; \
1707 #define setup_blocks_add_blocks_direct() \
1709 #define setup_blocks_do(shading, texturing, dithering, sw, target) \
1710 setup_blocks_load_msb_mask_##target(); \
1711 setup_blocks_variables_##shading##_##texturing(target); \
1713 edge_data_struct *span_edge_data = psx_gpu->span_edge_data; \
1714 vec_4x32u *span_uvrg_offset = (vec_4x32u *)psx_gpu->span_uvrg_offset; \
1715 u32 *span_b_offset = psx_gpu->span_b_offset; \
1717 block_struct *block = psx_gpu->blocks + psx_gpu->num_blocks; \
1719 u32 num_spans = psx_gpu->num_spans; \
1724 u32 num_blocks = psx_gpu->num_blocks; \
1725 u32 span_num_blocks; \
1729 span_num_blocks = span_edge_data->num_blocks; \
1730 if(span_num_blocks) \
1732 y = span_edge_data->y; \
1733 fb_ptr = psx_gpu->vram_out_ptr + span_edge_data->left_x + (y * 1024); \
1735 setup_blocks_span_initialize_##shading##_##texturing(); \
1736 setup_blocks_span_initialize_##dithering(texturing); \
1738 setup_blocks_add_blocks_##target(); \
1740 s32 pixel_span = span_num_blocks * 8; \
1741 pixel_span -= __builtin_popcount(span_edge_data->right_mask & 0xFF); \
1743 span_num_blocks--; \
1744 while(span_num_blocks) \
1746 setup_blocks_store_##shading##_##texturing(sw, dithering, target, \
1748 setup_blocks_store_draw_mask_##texturing##_##target(block, 0x00); \
1752 span_num_blocks--; \
1755 setup_blocks_store_##shading##_##texturing(sw, dithering, target, edge); \
1756 setup_blocks_store_draw_mask_##texturing##_##target(block, \
1757 span_edge_data->right_mask); \
1764 span_uvrg_offset++; \
1768 psx_gpu->num_blocks = num_blocks \
1770 void setup_blocks_shaded_textured_dithered_swizzled_indirect(psx_gpu_struct
1774 setup_blocks_shaded_textured_dithered_swizzled_indirect_(psx_gpu);
1777 setup_blocks_do(shaded, textured, dithered, swizzled, indirect);
1780 void setup_blocks_shaded_textured_dithered_unswizzled_indirect(psx_gpu_struct
1784 setup_blocks_shaded_textured_dithered_unswizzled_indirect_(psx_gpu);
1787 setup_blocks_do(shaded, textured, dithered, unswizzled, indirect);
1790 void setup_blocks_unshaded_textured_dithered_swizzled_indirect(psx_gpu_struct
1794 setup_blocks_unshaded_textured_dithered_swizzled_indirect_(psx_gpu);
1797 setup_blocks_do(unshaded, textured, dithered, swizzled, indirect);
1800 void setup_blocks_unshaded_textured_dithered_unswizzled_indirect(psx_gpu_struct
1804 setup_blocks_unshaded_textured_dithered_unswizzled_indirect_(psx_gpu);
1807 setup_blocks_do(unshaded, textured, dithered, unswizzled, indirect);
1810 void setup_blocks_unshaded_untextured_undithered_unswizzled_indirect(
1811 psx_gpu_struct *psx_gpu)
1814 setup_blocks_unshaded_untextured_undithered_unswizzled_indirect_(psx_gpu);
1817 setup_blocks_do(unshaded, untextured, undithered, unswizzled, indirect);
1820 void setup_blocks_unshaded_untextured_undithered_unswizzled_direct(
1821 psx_gpu_struct *psx_gpu)
1824 setup_blocks_unshaded_untextured_undithered_unswizzled_direct_(psx_gpu);
1827 setup_blocks_do(unshaded, untextured, undithered, unswizzled, direct);
1830 void setup_blocks_shaded_untextured_undithered_unswizzled_indirect(psx_gpu_struct
1834 setup_blocks_shaded_untextured_undithered_unswizzled_indirect_(psx_gpu);
1837 setup_blocks_do(shaded, untextured, undithered, unswizzled, indirect);
1840 void setup_blocks_shaded_untextured_dithered_unswizzled_indirect(psx_gpu_struct
1844 setup_blocks_shaded_untextured_dithered_unswizzled_indirect_(psx_gpu);
1847 setup_blocks_do(shaded, untextured, dithered, unswizzled, indirect);
1850 void setup_blocks_shaded_untextured_undithered_unswizzled_direct(
1851 psx_gpu_struct *psx_gpu)
1854 setup_blocks_shaded_untextured_undithered_unswizzled_direct_(psx_gpu);
1857 setup_blocks_do(shaded, untextured, undithered, unswizzled, direct);
1860 void setup_blocks_shaded_untextured_dithered_unswizzled_direct(psx_gpu_struct
1864 setup_blocks_shaded_untextured_dithered_unswizzled_direct_(psx_gpu);
1867 setup_blocks_do(shaded, untextured, dithered, unswizzled, direct);
1870 static void update_texture_4bpp_cache(psx_gpu_struct *psx_gpu)
1872 u32 current_texture_page = psx_gpu->current_texture_page;
1873 u8 *texture_page_ptr = psx_gpu->texture_page_base;
1874 const u16 *vram_ptr = psx_gpu->vram_ptr;
1879 vram_ptr += (current_texture_page >> 4) * 256 * 1024;
1880 vram_ptr += (current_texture_page & 0xF) * 64;
1882 gvdupq_n_u16(c_0x00f0, 0x00f0);
1884 psx_gpu->dirty_textures_4bpp_mask &= ~(psx_gpu->current_texture_mask);
1886 for (tile_y = 16; tile_y; tile_y--)
1888 for (tile_x = 16; tile_x; tile_x--)
1890 for (sub_y = 8; sub_y; sub_y--)
1892 vec_8x8u texel_block_a, texel_block_b;
1893 vec_8x16u texel_block_expanded_a, texel_block_expanded_b;
1894 vec_8x16u texel_block_expanded_c, texel_block_expanded_d;
1895 vec_8x16u texel_block_expanded_ab, texel_block_expanded_cd;
1897 gvld1_u8(texel_block_a, (u8 *)vram_ptr); vram_ptr += 1024;
1898 gvld1_u8(texel_block_b, (u8 *)vram_ptr); vram_ptr += 1024;
1900 gvmovl_u8(texel_block_expanded_a, texel_block_a);
1901 gvshll_n_u8(texel_block_expanded_b, texel_block_a, 4);
1902 gvmovl_u8(texel_block_expanded_c, texel_block_b);
1903 gvshll_n_u8(texel_block_expanded_d, texel_block_b, 4);
1905 gvbicq(texel_block_expanded_a, texel_block_expanded_a, c_0x00f0);
1906 gvbicq(texel_block_expanded_b, texel_block_expanded_b, c_0x00f0);
1907 gvbicq(texel_block_expanded_c, texel_block_expanded_c, c_0x00f0);
1908 gvbicq(texel_block_expanded_d, texel_block_expanded_d, c_0x00f0);
1910 gvorrq(texel_block_expanded_ab, texel_block_expanded_a, texel_block_expanded_b);
1911 gvorrq(texel_block_expanded_cd, texel_block_expanded_c, texel_block_expanded_d);
1913 gvst1q_2_pi_u32(texel_block_expanded_ab, texel_block_expanded_cd, texture_page_ptr);
1916 vram_ptr -= (1024 * 16) - 4;
1919 vram_ptr += (16 * 1024) - (4 * 16);
1923 void update_texture_8bpp_cache_slice(psx_gpu_struct *psx_gpu,
1927 update_texture_8bpp_cache_slice_(psx_gpu, texture_page);
1930 u16 *texture_page_ptr = psx_gpu->texture_page_base;
1931 u16 *vram_ptr = psx_gpu->vram_ptr;
1936 vram_ptr += (texture_page >> 4) * 256 * 1024;
1937 vram_ptr += (texture_page & 0xF) * 64;
1939 if((texture_page ^ psx_gpu->current_texture_page) & 0x1)
1940 texture_page_ptr += (8 * 16) * 8;
1942 for (tile_y = 16; tile_y; tile_y--)
1944 for (tile_x = 8; tile_x; tile_x--)
1946 for (sub_y = 4; sub_y; sub_y--)
1948 vec_4x32u texels_a, texels_b, texels_c, texels_d = {};
1949 gvld1q_u32(texels_a, vram_ptr); vram_ptr += 1024;
1950 gvld1q_u32(texels_b, vram_ptr); vram_ptr += 1024;
1951 gvld1q_u32(texels_c, vram_ptr); vram_ptr += 1024;
1952 gvld1q_u32(texels_d, vram_ptr); vram_ptr += 1024;
1954 gvst1q_2_pi_u32(texels_a, texels_b, texture_page_ptr);
1955 gvst1q_2_pi_u32(texels_c, texels_d, texture_page_ptr);
1958 vram_ptr -= (1024 * 16) - 8;
1961 vram_ptr -= (8 * 8);
1962 vram_ptr += (16 * 1024);
1964 texture_page_ptr += (8 * 16) * 8;
1968 void texture_blocks_untextured(psx_gpu_struct *psx_gpu)
1972 void texture_blocks_4bpp(psx_gpu_struct *psx_gpu)
1975 texture_blocks_4bpp_(psx_gpu);
1978 block_struct *block = psx_gpu->blocks;
1979 u32 num_blocks = psx_gpu->num_blocks;
1981 vec_8x8u texels_low;
1982 vec_8x8u texels_high;
1985 vec_16x8u clut_high;
1987 const u8 *texture_ptr_8bpp = psx_gpu->texture_page_ptr;
1989 gvld2q_u8(clut_low, clut_high, (u8 *)psx_gpu->clut_ptr);
1991 if(psx_gpu->current_texture_mask & psx_gpu->dirty_textures_4bpp_mask)
1992 update_texture_4bpp_cache(psx_gpu);
2000 texture_ptr_8bpp[block->uv.e[0]],
2001 texture_ptr_8bpp[block->uv.e[1]],
2002 texture_ptr_8bpp[block->uv.e[2]],
2003 texture_ptr_8bpp[block->uv.e[3]],
2004 texture_ptr_8bpp[block->uv.e[4]],
2005 texture_ptr_8bpp[block->uv.e[5]],
2006 texture_ptr_8bpp[block->uv.e[6]],
2007 texture_ptr_8bpp[block->uv.e[7]]
2011 gvtbl2_u8(texels_low, clut_low, texels);
2012 gvtbl2_u8(texels_high, clut_high, texels);
2014 gvst2_u8(texels_low, texels_high, (u8 *)block->texels.e);
2021 void texture_blocks_8bpp(psx_gpu_struct *psx_gpu)
2024 texture_blocks_8bpp_(psx_gpu);
2027 u32 num_blocks = psx_gpu->num_blocks;
2029 if(psx_gpu->current_texture_mask & psx_gpu->dirty_textures_8bpp_mask)
2030 update_texture_8bpp_cache(psx_gpu);
2032 const u8 * __restrict__ texture_ptr_8bpp = psx_gpu->texture_page_ptr;
2033 const u16 * __restrict__ clut_ptr = psx_gpu->clut_ptr;
2034 block_struct * __restrict__ block = psx_gpu->blocks;
2039 #define load_one(i_) \
2040 offset = block->uv.e[i_]; u16 texel##i_ = texture_ptr_8bpp[offset]
2041 #define store_one(i_) \
2042 block->texels.e[i_] = clut_ptr[texel##i_]
2043 load_one(0); load_one(1); load_one(2); load_one(3);
2044 load_one(4); load_one(5); load_one(6); load_one(7);
2045 store_one(0); store_one(1); store_one(2); store_one(3);
2046 store_one(4); store_one(5); store_one(6); store_one(7);
2055 void texture_blocks_16bpp(psx_gpu_struct *psx_gpu)
2058 texture_blocks_16bpp_(psx_gpu);
2061 u32 num_blocks = psx_gpu->num_blocks;
2062 const u16 * __restrict__ texture_ptr_16bpp = psx_gpu->texture_page_ptr;
2063 block_struct * __restrict__ block = psx_gpu->blocks;
2068 #define load_one(i_) \
2069 offset = block->uv.e[i_]; \
2070 offset += ((offset & 0xFF00) * 3); \
2071 u16 texel##i_ = texture_ptr_16bpp[offset]
2072 #define store_one(i_) \
2073 block->texels.e[i_] = texel##i_
2074 load_one(0); load_one(1); load_one(2); load_one(3);
2075 load_one(4); load_one(5); load_one(6); load_one(7);
2076 store_one(0); store_one(1); store_one(2); store_one(3);
2077 store_one(4); store_one(5); store_one(6); store_one(7);
2086 #define shade_blocks_load_msb_mask_indirect() \
2088 #define shade_blocks_load_msb_mask_direct() \
2089 vec_8x16u msb_mask; \
2090 gvdupq_n_u16(msb_mask, psx_gpu->mask_msb); \
2092 #define shade_blocks_store_indirect(_draw_mask, _pixels) \
2093 gvst1q_u16(_draw_mask, block->draw_mask.e); \
2094 gvst1q_u16(_pixels, block->pixels.e); \
2096 #define shade_blocks_store_direct(_draw_mask, _pixels) \
2098 vec_8x16u fb_pixels; \
2099 gvorrq(_pixels, _pixels, msb_mask); \
2100 gvld1q_u16(fb_pixels, block->fb_ptr); \
2101 gvbifq(fb_pixels, _pixels, _draw_mask); \
2102 gvst1q_u16(fb_pixels, block->fb_ptr); \
2105 #define shade_blocks_textured_false_modulated_check_dithered(target) \
2107 #define shade_blocks_textured_false_modulated_check_undithered(target) \
2108 if(psx_gpu->triangle_color == 0x808080) \
2110 shade_blocks_textured_unmodulated_##target(psx_gpu); \
2114 #define shade_blocks_textured_modulated_shaded_primitive_load(dithering, \
2117 #define shade_blocks_textured_modulated_unshaded_primitive_load(dithering, \
2120 u32 color = psx_gpu->triangle_color; \
2121 gvdup_n_u8(colors_r, color); \
2122 gvdup_n_u8(colors_g, color >> 8); \
2123 gvdup_n_u8(colors_b, color >> 16); \
2124 shade_blocks_textured_false_modulated_check_##dithering(target); \
2127 #define shade_blocks_textured_modulated_shaded_block_load() \
2128 gvld1_u8(colors_r, block->r.e); \
2129 gvld1_u8(colors_g, block->g.e); \
2130 gvld1_u8(colors_b, block->b.e) \
2132 #define shade_blocks_textured_modulated_unshaded_block_load() \
2134 #define shade_blocks_textured_modulate_dithered(component) \
2135 gvld1q_u16(pixels_##component, block->dither_offsets.e); \
2136 gvmlal_u8(pixels_##component, texels_##component, colors_##component) \
2138 #define shade_blocks_textured_modulate_undithered(component) \
2139 gvmull_u8(pixels_##component, texels_##component, colors_##component) \
2141 #define shade_blocks_textured_modulated_do(shading, dithering, target) \
2142 block_struct *block = psx_gpu->blocks; \
2143 u32 num_blocks = psx_gpu->num_blocks; \
2146 vec_8x8u texels_r; \
2147 vec_8x8u texels_g; \
2148 vec_8x8u texels_b; \
2150 vec_8x8u colors_r; \
2151 vec_8x8u colors_g; \
2152 vec_8x8u colors_b; \
2154 vec_8x8u pixels_r_low; \
2155 vec_8x8u pixels_g_low; \
2156 vec_8x8u pixels_b_low; \
2159 vec_8x16u pixels_r; \
2160 vec_8x16u pixels_g; \
2161 vec_8x16u pixels_b; \
2163 vec_8x16u draw_mask; \
2164 vec_8x16u zero_mask; \
2166 vec_8x8u d64_0x07; \
2167 vec_8x8u d64_0x1F; \
2172 vec_8x16u d128_0x8000; \
2174 vec_8x16u test_mask; \
2175 u32 draw_mask_bits; \
2177 gvld1q_u16(test_mask, psx_gpu->test_mask.e); \
2178 shade_blocks_load_msb_mask_##target(); \
2180 gvdup_n_u8(d64_0x07, 0x07); \
2181 gvdup_n_u8(d64_0x1F, 0x1F); \
2182 gvdup_n_u8(d64_1, 1); \
2183 gvdup_n_u8(d64_4, 4); \
2184 gvdup_n_u8(d64_128, 128); \
2186 gvdupq_n_u16(d128_0x8000, 0x8000); \
2188 shade_blocks_textured_modulated_##shading##_primitive_load(dithering, \
2193 draw_mask_bits = block->draw_mask_bits; \
2194 gvdupq_n_u16(draw_mask, draw_mask_bits); \
2195 gvtstq_u16(draw_mask, draw_mask, test_mask); \
2197 shade_blocks_textured_modulated_##shading##_block_load(); \
2199 gvld1q_u16(texels, block->texels.e); \
2201 gvmovn_u16(texels_r, texels); \
2202 gvshrn_n_u16(texels_g, texels, 5); \
2203 gvshrn_n_u16(texels_b, texels, 7); \
2205 gvand(texels_r, texels_r, d64_0x1F); \
2206 gvand(texels_g, texels_g, d64_0x1F); \
2207 gvshr_n_u8(texels_b, texels_b, 3); \
2209 shade_blocks_textured_modulate_##dithering(r); \
2210 shade_blocks_textured_modulate_##dithering(g); \
2211 shade_blocks_textured_modulate_##dithering(b); \
2213 gvceqzq_u16(zero_mask, texels); \
2214 gvand(pixels, texels, d128_0x8000); \
2216 gvqshrun_n_s16(pixels_r_low, pixels_r, 4); \
2217 gvqshrun_n_s16(pixels_g_low, pixels_g, 4); \
2218 gvqshrun_n_s16(pixels_b_low, pixels_b, 4); \
2220 gvorrq(zero_mask, draw_mask, zero_mask); \
2222 gvshr_n_u8(pixels_r_low, pixels_r_low, 3); \
2223 gvbic(pixels_g_low, pixels_g_low, d64_0x07); \
2224 gvbic(pixels_b_low, pixels_b_low, d64_0x07); \
2226 gvmlal_u8(pixels, pixels_r_low, d64_1); \
2227 gvmlal_u8(pixels, pixels_g_low, d64_4); \
2228 gvmlal_u8(pixels, pixels_b_low, d64_128); \
2230 shade_blocks_store_##target(zero_mask, pixels); \
2236 void shade_blocks_shaded_textured_modulated_dithered_direct(psx_gpu_struct
2240 shade_blocks_shaded_textured_modulated_dithered_direct_(psx_gpu);
2243 shade_blocks_textured_modulated_do(shaded, dithered, direct);
2246 void shade_blocks_shaded_textured_modulated_undithered_direct(psx_gpu_struct
2250 shade_blocks_shaded_textured_modulated_undithered_direct_(psx_gpu);
2253 shade_blocks_textured_modulated_do(shaded, undithered, direct);
2256 void shade_blocks_unshaded_textured_modulated_dithered_direct(psx_gpu_struct
2260 shade_blocks_unshaded_textured_modulated_dithered_direct_(psx_gpu);
2263 shade_blocks_textured_modulated_do(unshaded, dithered, direct);
2266 void shade_blocks_unshaded_textured_modulated_undithered_direct(psx_gpu_struct
2270 shade_blocks_unshaded_textured_modulated_undithered_direct_(psx_gpu);
2273 shade_blocks_textured_modulated_do(unshaded, undithered, direct);
2276 void shade_blocks_shaded_textured_modulated_dithered_indirect(psx_gpu_struct
2280 shade_blocks_shaded_textured_modulated_dithered_indirect_(psx_gpu);
2283 shade_blocks_textured_modulated_do(shaded, dithered, indirect);
2286 void shade_blocks_shaded_textured_modulated_undithered_indirect(psx_gpu_struct
2290 shade_blocks_shaded_textured_modulated_undithered_indirect_(psx_gpu);
2293 shade_blocks_textured_modulated_do(shaded, undithered, indirect);
2296 void shade_blocks_unshaded_textured_modulated_dithered_indirect(psx_gpu_struct
2300 shade_blocks_unshaded_textured_modulated_dithered_indirect_(psx_gpu);
2303 shade_blocks_textured_modulated_do(unshaded, dithered, indirect);
2306 void shade_blocks_unshaded_textured_modulated_undithered_indirect(psx_gpu_struct
2310 shade_blocks_unshaded_textured_modulated_undithered_indirect_(psx_gpu);
2313 shade_blocks_textured_modulated_do(unshaded, undithered, indirect);
2316 #define shade_blocks_textured_unmodulated_do(target) \
2317 block_struct *block = psx_gpu->blocks; \
2318 u32 num_blocks = psx_gpu->num_blocks; \
2319 vec_8x16u draw_mask; \
2320 vec_8x16u test_mask; \
2321 u32 draw_mask_bits; \
2325 gvld1q_u16(test_mask, psx_gpu->test_mask.e); \
2326 shade_blocks_load_msb_mask_##target(); \
2330 vec_8x16u zero_mask; \
2332 draw_mask_bits = block->draw_mask_bits; \
2333 gvdupq_n_u16(draw_mask, draw_mask_bits); \
2334 gvtstq_u16(draw_mask, draw_mask, test_mask); \
2336 gvld1q_u16(pixels, block->texels.e); \
2338 gvceqzq_u16(zero_mask, pixels); \
2339 gvorrq(zero_mask, draw_mask, zero_mask); \
2341 shade_blocks_store_##target(zero_mask, pixels); \
2347 void shade_blocks_textured_unmodulated_indirect(psx_gpu_struct *psx_gpu)
2350 shade_blocks_textured_unmodulated_indirect_(psx_gpu);
2353 shade_blocks_textured_unmodulated_do(indirect)
2356 void shade_blocks_textured_unmodulated_direct(psx_gpu_struct *psx_gpu)
2359 shade_blocks_textured_unmodulated_direct_(psx_gpu);
2362 shade_blocks_textured_unmodulated_do(direct)
2365 void shade_blocks_unshaded_untextured_indirect(psx_gpu_struct *psx_gpu)
2369 void shade_blocks_unshaded_untextured_direct(psx_gpu_struct *psx_gpu)
2372 shade_blocks_unshaded_untextured_direct_(psx_gpu);
2375 block_struct *block = psx_gpu->blocks;
2376 u32 num_blocks = psx_gpu->num_blocks;
2379 gvld1q_u16(pixels, block->texels.e);
2380 shade_blocks_load_msb_mask_direct();
2384 vec_8x16u draw_mask;
2385 gvld1q_u16(draw_mask, block->draw_mask.e);
2386 shade_blocks_store_direct(draw_mask, pixels);
2393 #define blend_blocks_mask_evaluate_on() \
2394 vec_8x16u mask_pixels; \
2395 gvcltzq_s16(mask_pixels, framebuffer_pixels); \
2396 gvorrq(draw_mask, draw_mask, mask_pixels) \
2398 #define blend_blocks_mask_evaluate_off() \
2400 #define blend_blocks_average() \
2402 vec_8x16u pixels_no_msb; \
2403 vec_8x16u fb_pixels_no_msb; \
2405 vec_8x16u d128_0x0421; \
2407 gvdupq_n_u16(d128_0x0421, 0x0421); \
2409 gveorq(blend_pixels, pixels, framebuffer_pixels); \
2410 gvbicq(pixels_no_msb, pixels, d128_0x8000); \
2411 gvand(blend_pixels, blend_pixels, d128_0x0421); \
2412 gvsubq_u16(blend_pixels, pixels_no_msb, blend_pixels); \
2413 gvbicq(fb_pixels_no_msb, framebuffer_pixels, d128_0x8000); \
2414 gvhaddq_u16(blend_pixels, fb_pixels_no_msb, blend_pixels); \
2417 #define blend_blocks_add() \
2419 vec_8x16u pixels_rb, pixels_g; \
2420 vec_8x16u fb_rb, fb_g; \
2422 vec_8x16u d128_0x7C1F; \
2423 vec_8x16u d128_0x03E0; \
2425 gvdupq_n_u16(d128_0x7C1F, 0x7C1F); \
2426 gvdupq_n_u16(d128_0x03E0, 0x03E0); \
2428 gvand(pixels_rb, pixels, d128_0x7C1F); \
2429 gvand(pixels_g, pixels, d128_0x03E0); \
2431 gvand(fb_rb, framebuffer_pixels, d128_0x7C1F); \
2432 gvand(fb_g, framebuffer_pixels, d128_0x03E0); \
2434 gvaddq_u16(fb_rb, fb_rb, pixels_rb); \
2435 gvaddq_u16(fb_g, fb_g, pixels_g); \
2437 gvminq_u8(fb_rb, fb_rb, d128_0x7C1F); \
2438 gvminq_u16(fb_g, fb_g, d128_0x03E0); \
2440 gvorrq(blend_pixels, fb_rb, fb_g); \
2443 #define blend_blocks_subtract() \
2445 vec_8x16u pixels_rb, pixels_g; \
2446 vec_8x16u fb_rb, fb_g; \
2448 vec_8x16u d128_0x7C1F; \
2449 vec_8x16u d128_0x03E0; \
2451 gvdupq_n_u16(d128_0x7C1F, 0x7C1F); \
2452 gvdupq_n_u16(d128_0x03E0, 0x03E0); \
2454 gvand(pixels_rb, pixels, d128_0x7C1F); \
2455 gvand(pixels_g, pixels, d128_0x03E0); \
2457 gvand(fb_rb, framebuffer_pixels, d128_0x7C1F); \
2458 gvand(fb_g, framebuffer_pixels, d128_0x03E0); \
2460 gvqsubq_u8(fb_rb, fb_rb, pixels_rb); \
2461 gvqsubq_u16(fb_g, fb_g, pixels_g); \
2463 gvorrq(blend_pixels, fb_rb, fb_g); \
2466 #define blend_blocks_add_fourth() \
2468 vec_8x16u pixels_rb, pixels_g; \
2469 vec_8x16u pixels_fourth; \
2470 vec_8x16u fb_rb, fb_g; \
2472 vec_8x16u d128_0x7C1F; \
2473 vec_8x16u d128_0x1C07; \
2474 vec_8x16u d128_0x03E0; \
2475 vec_8x16u d128_0x00E0; \
2477 gvdupq_n_u16(d128_0x7C1F, 0x7C1F); \
2478 gvdupq_n_u16(d128_0x1C07, 0x1C07); \
2479 gvdupq_n_u16(d128_0x03E0, 0x03E0); \
2480 gvdupq_n_u16(d128_0x00E0, 0x00E0); \
2482 gvshrq_n_u16(pixels_fourth, pixels, 2); \
2484 gvand(fb_rb, framebuffer_pixels, d128_0x7C1F); \
2485 gvand(fb_g, framebuffer_pixels, d128_0x03E0); \
2487 gvand(pixels_rb, pixels_fourth, d128_0x1C07); \
2488 gvand(pixels_g, pixels_fourth, d128_0x00E0); \
2490 gvaddq_u16(fb_rb, fb_rb, pixels_rb); \
2491 gvaddq_u16(fb_g, fb_g, pixels_g); \
2493 gvminq_u8(fb_rb, fb_rb, d128_0x7C1F); \
2494 gvminq_u16(fb_g, fb_g, d128_0x03E0); \
2496 gvorrq(blend_pixels, fb_rb, fb_g); \
2499 #define blend_blocks_blended_combine_textured() \
2501 vec_8x16u blend_mask; \
2502 gvcltzq_s16(blend_mask, pixels); \
2504 gvorrq(blend_pixels, blend_pixels, d128_0x8000); \
2505 gvbifq(blend_pixels, pixels, blend_mask); \
2508 #define blend_blocks_blended_combine_untextured() \
2510 #define blend_blocks_body_blend(blend_mode, texturing) \
2512 blend_blocks_##blend_mode(); \
2513 blend_blocks_blended_combine_##texturing(); \
2516 #define blend_blocks_body_average(texturing) \
2517 blend_blocks_body_blend(average, texturing) \
2519 #define blend_blocks_body_add(texturing) \
2520 blend_blocks_body_blend(add, texturing) \
2522 #define blend_blocks_body_subtract(texturing) \
2523 blend_blocks_body_blend(subtract, texturing) \
2525 #define blend_blocks_body_add_fourth(texturing) \
2526 blend_blocks_body_blend(add_fourth, texturing) \
2528 #define blend_blocks_body_unblended(texturing) \
2529 blend_pixels = pixels \
2531 #define blend_blocks_do(texturing, blend_mode, mask_evaluate) \
2532 block_struct *block = psx_gpu->blocks; \
2533 u32 num_blocks = psx_gpu->num_blocks; \
2534 vec_8x16u draw_mask; \
2536 vec_8x16u blend_pixels; \
2537 vec_8x16u framebuffer_pixels; \
2538 vec_8x16u msb_mask; \
2539 vec_8x16u d128_0x8000; \
2543 gvdupq_n_u16(d128_0x8000, 0x8000); \
2544 gvdupq_n_u16(msb_mask, psx_gpu->mask_msb); \
2545 (void)d128_0x8000; /* sometimes unused */ \
2549 gvld1q_u16(pixels, block->pixels.e); \
2550 gvld1q_u16(draw_mask, block->draw_mask.e); \
2551 fb_ptr = block->fb_ptr; \
2553 gvld1q_u16(framebuffer_pixels, fb_ptr); \
2555 blend_blocks_mask_evaluate_##mask_evaluate(); \
2556 blend_blocks_body_##blend_mode(texturing); \
2558 gvorrq(blend_pixels, blend_pixels, msb_mask); \
2559 gvbifq(framebuffer_pixels, blend_pixels, draw_mask); \
2560 gvst1q_u16(framebuffer_pixels, fb_ptr); \
2567 void blend_blocks_textured_average_off(psx_gpu_struct *psx_gpu)
2570 blend_blocks_textured_average_off_(psx_gpu);
2573 blend_blocks_do(textured, average, off);
2576 void blend_blocks_untextured_average_off(psx_gpu_struct *psx_gpu)
2579 blend_blocks_untextured_average_off_(psx_gpu);
2582 blend_blocks_do(untextured, average, off);
2585 void blend_blocks_textured_average_on(psx_gpu_struct *psx_gpu)
2588 blend_blocks_textured_average_on_(psx_gpu);
2591 blend_blocks_do(textured, average, on);
2594 void blend_blocks_untextured_average_on(psx_gpu_struct *psx_gpu)
2597 blend_blocks_untextured_average_on_(psx_gpu);
2600 blend_blocks_do(untextured, average, on);
2603 void blend_blocks_textured_add_off(psx_gpu_struct *psx_gpu)
2606 blend_blocks_textured_add_off_(psx_gpu);
2609 blend_blocks_do(textured, add, off);
2612 void blend_blocks_textured_add_on(psx_gpu_struct *psx_gpu)
2615 blend_blocks_textured_add_on_(psx_gpu);
2618 blend_blocks_do(textured, add, on);
2621 void blend_blocks_untextured_add_off(psx_gpu_struct *psx_gpu)
2624 blend_blocks_untextured_add_off_(psx_gpu);
2627 blend_blocks_do(untextured, add, off);
2630 void blend_blocks_untextured_add_on(psx_gpu_struct *psx_gpu)
2633 blend_blocks_untextured_add_on_(psx_gpu);
2636 blend_blocks_do(untextured, add, on);
2639 void blend_blocks_textured_subtract_off(psx_gpu_struct *psx_gpu)
2642 blend_blocks_textured_subtract_off_(psx_gpu);
2645 blend_blocks_do(textured, subtract, off);
2648 void blend_blocks_textured_subtract_on(psx_gpu_struct *psx_gpu)
2651 blend_blocks_textured_subtract_on_(psx_gpu);
2654 blend_blocks_do(textured, subtract, on);
2657 void blend_blocks_untextured_subtract_off(psx_gpu_struct *psx_gpu)
2660 blend_blocks_untextured_subtract_off_(psx_gpu);
2663 blend_blocks_do(untextured, subtract, off);
2666 void blend_blocks_untextured_subtract_on(psx_gpu_struct *psx_gpu)
2669 blend_blocks_untextured_subtract_on_(psx_gpu);
2672 blend_blocks_do(untextured, subtract, on);
2675 void blend_blocks_textured_add_fourth_off(psx_gpu_struct *psx_gpu)
2678 blend_blocks_textured_add_fourth_off_(psx_gpu);
2681 blend_blocks_do(textured, add_fourth, off);
2684 void blend_blocks_textured_add_fourth_on(psx_gpu_struct *psx_gpu)
2687 blend_blocks_textured_add_fourth_on_(psx_gpu);
2690 blend_blocks_do(textured, add_fourth, on);
2693 void blend_blocks_untextured_add_fourth_off(psx_gpu_struct *psx_gpu)
2696 blend_blocks_untextured_add_fourth_off_(psx_gpu);
2699 blend_blocks_do(untextured, add_fourth, off);
2702 void blend_blocks_untextured_add_fourth_on(psx_gpu_struct *psx_gpu)
2705 blend_blocks_untextured_add_fourth_on_(psx_gpu);
2708 blend_blocks_do(untextured, add_fourth, on);
2711 void blend_blocks_textured_unblended_on(psx_gpu_struct *psx_gpu)
2714 blend_blocks_textured_unblended_on_(psx_gpu);
2717 blend_blocks_do(textured, unblended, on);
2720 void blend_blocks_textured_unblended_off(psx_gpu_struct *psx_gpu)
2724 void setup_sprite_untextured(psx_gpu_struct *psx_gpu, s32 x, s32 y, s32 u,
2725 s32 v, s32 width, s32 height, u32 color)
2727 if((psx_gpu->render_state & (RENDER_STATE_MASK_EVALUATE |
2728 RENDER_FLAGS_MODULATE_TEXELS | RENDER_FLAGS_BLEND)) == 0 &&
2729 (psx_gpu->render_mode & RENDER_INTERLACE_ENABLED) == 0)
2731 setup_sprite_untextured_simple(psx_gpu, x, y, u, v, width, height, color);
2736 setup_sprite_untextured_(psx_gpu, x, y, u, v, width, height, color);
2739 u32 right_width = ((width - 1) & 0x7) + 1;
2740 u32 right_mask_bits = (0xFF << right_width);
2741 u16 *fb_ptr = psx_gpu->vram_out_ptr + (y * 1024) + x;
2742 u32 block_width = (width + 7) / 8;
2743 u32 fb_ptr_pitch = 1024 - ((block_width - 1) * 8);
2744 u32 blocks_remaining;
2745 u32 num_blocks = psx_gpu->num_blocks;
2746 block_struct *block = psx_gpu->blocks + num_blocks;
2748 u32 color_r = color & 0xFF;
2749 u32 color_g = (color >> 8) & 0xFF;
2750 u32 color_b = (color >> 16) & 0xFF;
2752 vec_8x16u right_mask;
2753 vec_8x16u test_mask;
2754 vec_8x16u zero_mask;
2756 gvld1q_u16(test_mask, psx_gpu->test_mask.e);
2757 color = (color_r >> 3) | ((color_g >> 3) << 5) | ((color_b >> 3) << 10);
2759 gvdupq_n_u16(colors, color);
2760 gvdupq_n_u16(zero_mask, 0x00);
2761 gvdupq_n_u16(right_mask, right_mask_bits);
2762 gvtstq_u16(right_mask, right_mask, test_mask);
2766 blocks_remaining = block_width - 1;
2767 num_blocks += block_width;
2769 if(num_blocks > MAX_BLOCKS)
2771 flush_render_block_buffer(psx_gpu);
2772 num_blocks = block_width;
2773 block = psx_gpu->blocks;
2776 while(blocks_remaining)
2778 gvst1q_u16(colors, block->pixels.e);
2779 gvst1q_u16(zero_mask, block->draw_mask.e);
2780 block->fb_ptr = fb_ptr;
2787 gvst1q_u16(colors, block->pixels.e);
2788 gvst1q_u16(right_mask, block->draw_mask.e);
2789 block->fb_ptr = fb_ptr;
2792 fb_ptr += fb_ptr_pitch;
2795 psx_gpu->num_blocks = num_blocks;
2799 #define setup_sprite_tiled_initialize_4bpp_clut() \
2800 vec_16x8u clut_low, clut_high; \
2802 gvld2q_u8(clut_low, clut_high, (u8 *)psx_gpu->clut_ptr) \
2804 #define setup_sprite_tiled_initialize_4bpp() \
2805 setup_sprite_tiled_initialize_4bpp_clut(); \
2807 if(psx_gpu->current_texture_mask & psx_gpu->dirty_textures_4bpp_mask) \
2808 update_texture_4bpp_cache(psx_gpu) \
2810 #define setup_sprite_tiled_initialize_8bpp() \
2811 if(psx_gpu->current_texture_mask & psx_gpu->dirty_textures_8bpp_mask) \
2812 update_texture_8bpp_cache(psx_gpu) \
2814 #define setup_sprite_tile_fetch_texel_block_8bpp(offset) \
2815 texture_block_ptr = psx_gpu->texture_page_ptr + \
2816 ((texture_offset + offset) & texture_mask); \
2818 gvld1_u8(texels, (u8 *)texture_block_ptr) \
2820 #define setup_sprite_tile_add_blocks(tile_num_blocks) \
2821 num_blocks += tile_num_blocks; \
2823 if(num_blocks > MAX_BLOCKS) \
2825 flush_render_block_buffer(psx_gpu); \
2826 num_blocks = tile_num_blocks; \
2827 block = psx_gpu->blocks; \
2830 #define setup_sprite_tile_full_4bpp(edge) \
2832 vec_8x8u texels_low, texels_high; \
2833 setup_sprite_tile_add_blocks(sub_tile_height * 2); \
2835 while(sub_tile_height) \
2837 setup_sprite_tile_fetch_texel_block_8bpp(0); \
2838 gvtbl2_u8(texels_low, clut_low, texels); \
2839 gvtbl2_u8(texels_high, clut_high, texels); \
2841 gvst2_u8(texels_low, texels_high, (u8 *)block->texels.e); \
2842 block->draw_mask_bits = left_mask_bits; \
2843 block->fb_ptr = fb_ptr; \
2846 setup_sprite_tile_fetch_texel_block_8bpp(8); \
2847 gvtbl2_u8(texels_low, clut_low, texels); \
2848 gvtbl2_u8(texels_high, clut_high, texels); \
2850 gvst2_u8(texels_low, texels_high, (u8 *)block->texels.e); \
2851 block->draw_mask_bits = right_mask_bits; \
2852 block->fb_ptr = fb_ptr + 8; \
2856 texture_offset += 0x10; \
2857 sub_tile_height--; \
2859 texture_offset += 0xF00; \
2860 psx_gpu->num_blocks = num_blocks; \
2863 #define setup_sprite_tile_half_4bpp(edge) \
2865 vec_8x8u texels_low, texels_high; \
2866 setup_sprite_tile_add_blocks(sub_tile_height); \
2868 while(sub_tile_height) \
2870 setup_sprite_tile_fetch_texel_block_8bpp(0); \
2871 gvtbl2_u8(texels_low, clut_low, texels); \
2872 gvtbl2_u8(texels_high, clut_high, texels); \
2874 gvst2_u8(texels_low, texels_high, (u8 *)block->texels.e); \
2875 block->draw_mask_bits = edge##_mask_bits; \
2876 block->fb_ptr = fb_ptr; \
2880 texture_offset += 0x10; \
2881 sub_tile_height--; \
2883 texture_offset += 0xF00; \
2884 psx_gpu->num_blocks = num_blocks; \
2887 #define setup_sprite_tile_full_8bpp(edge) \
2889 setup_sprite_tile_add_blocks(sub_tile_height * 2); \
2891 while(sub_tile_height) \
2893 setup_sprite_tile_fetch_texel_block_8bpp(0); \
2894 gvst1_u8(texels, block->r.e); \
2895 block->draw_mask_bits = left_mask_bits; \
2896 block->fb_ptr = fb_ptr; \
2899 setup_sprite_tile_fetch_texel_block_8bpp(8); \
2900 gvst1_u8(texels, block->r.e); \
2901 block->draw_mask_bits = right_mask_bits; \
2902 block->fb_ptr = fb_ptr + 8; \
2906 texture_offset += 0x10; \
2907 sub_tile_height--; \
2909 texture_offset += 0xF00; \
2910 psx_gpu->num_blocks = num_blocks; \
2913 #define setup_sprite_tile_half_8bpp(edge) \
2915 setup_sprite_tile_add_blocks(sub_tile_height); \
2917 while(sub_tile_height) \
2919 setup_sprite_tile_fetch_texel_block_8bpp(0); \
2920 gvst1_u8(texels, block->r.e); \
2921 block->draw_mask_bits = edge##_mask_bits; \
2922 block->fb_ptr = fb_ptr; \
2926 texture_offset += 0x10; \
2927 sub_tile_height--; \
2929 texture_offset += 0xF00; \
2930 psx_gpu->num_blocks = num_blocks; \
2933 #define setup_sprite_tile_column_edge_pre_adjust_half_right() \
2934 texture_offset = texture_offset_base + 8; \
2937 #define setup_sprite_tile_column_edge_pre_adjust_half_left() \
2938 texture_offset = texture_offset_base \
2940 #define setup_sprite_tile_column_edge_pre_adjust_half(edge) \
2941 setup_sprite_tile_column_edge_pre_adjust_half_##edge() \
2943 #define setup_sprite_tile_column_edge_pre_adjust_full(edge) \
2944 texture_offset = texture_offset_base \
2946 #define setup_sprite_tile_column_edge_post_adjust_half_right() \
2949 #define setup_sprite_tile_column_edge_post_adjust_half_left() \
2951 #define setup_sprite_tile_column_edge_post_adjust_half(edge) \
2952 setup_sprite_tile_column_edge_post_adjust_half_##edge() \
2954 #define setup_sprite_tile_column_edge_post_adjust_full(edge) \
2957 #define setup_sprite_tile_column_height_single(edge_mode, edge, texture_mode, \
2961 sub_tile_height = column_data; \
2962 setup_sprite_tile_column_edge_pre_adjust_##edge_mode##x4mode(edge); \
2963 setup_sprite_tile_##edge_mode##_##texture_mode##x4mode(edge); \
2964 setup_sprite_tile_column_edge_post_adjust_##edge_mode##x4mode(edge); \
2967 #define setup_sprite_tile_column_height_multi(edge_mode, edge, texture_mode, \
2971 u32 tiles_remaining = column_data >> 16; \
2972 sub_tile_height = column_data & 0xFF; \
2973 setup_sprite_tile_column_edge_pre_adjust_##edge_mode##x4mode(edge); \
2974 setup_sprite_tile_##edge_mode##_##texture_mode##x4mode(edge); \
2975 tiles_remaining -= 1; \
2977 while(tiles_remaining) \
2979 sub_tile_height = 16; \
2980 setup_sprite_tile_##edge_mode##_##texture_mode##x4mode(edge); \
2981 tiles_remaining--; \
2984 sub_tile_height = (column_data >> 8) & 0xFF; \
2985 setup_sprite_tile_##edge_mode##_##texture_mode##x4mode(edge); \
2986 setup_sprite_tile_column_edge_post_adjust_##edge_mode##x4mode(edge); \
2990 #define setup_sprite_column_data_single() \
2991 column_data = height \
2993 #define setup_sprite_column_data_multi() \
2994 column_data = 16 - offset_v; \
2995 column_data |= ((height_rounded & 0xF) + 1) << 8; \
2996 column_data |= (tile_height - 1) << 16 \
2998 #define RIGHT_MASK_BIT_SHIFT 8
2999 #define RIGHT_MASK_BIT_SHIFT_4x 16
3001 #define setup_sprite_tile_column_width_single(texture_mode, multi_height, \
3002 edge_mode, edge, x4mode) \
3004 setup_sprite_column_data_##multi_height(); \
3005 left_mask_bits = left_block_mask | right_block_mask; \
3006 right_mask_bits = left_mask_bits >> RIGHT_MASK_BIT_SHIFT##x4mode; \
3008 setup_sprite_tile_column_height_##multi_height(edge_mode, edge, \
3009 texture_mode, x4mode); \
3012 #define setup_sprite_tiled_advance_column() \
3013 texture_offset_base += 0x100; \
3014 if((texture_offset_base & 0xF00) == 0) \
3015 texture_offset_base -= (0x100 + 0xF00) \
3017 #define FB_PTR_MULTIPLIER 1
3018 #define FB_PTR_MULTIPLIER_4x 2
3020 #define setup_sprite_tile_column_width_multi(texture_mode, multi_height, \
3021 left_mode, right_mode, x4mode) \
3023 setup_sprite_column_data_##multi_height(); \
3024 s32 fb_ptr_advance_column = (16 - (1024 * height)) \
3025 * FB_PTR_MULTIPLIER##x4mode; \
3028 left_mask_bits = left_block_mask; \
3029 right_mask_bits = left_mask_bits >> RIGHT_MASK_BIT_SHIFT##x4mode; \
3031 setup_sprite_tile_column_height_##multi_height(left_mode, right, \
3032 texture_mode, x4mode); \
3033 fb_ptr += fb_ptr_advance_column; \
3035 left_mask_bits = 0x00; \
3036 right_mask_bits = 0x00; \
3040 setup_sprite_tiled_advance_column(); \
3041 setup_sprite_tile_column_height_##multi_height(full, none, \
3042 texture_mode, x4mode); \
3043 fb_ptr += fb_ptr_advance_column; \
3047 left_mask_bits = right_block_mask; \
3048 right_mask_bits = left_mask_bits >> RIGHT_MASK_BIT_SHIFT##x4mode; \
3050 setup_sprite_tiled_advance_column(); \
3051 setup_sprite_tile_column_height_##multi_height(right_mode, left, \
3052 texture_mode, x4mode); \
3057 #define setup_sprite_tiled_initialize_4bpp_4x() \
3058 setup_sprite_tiled_initialize_4bpp_clut() \
3060 #define setup_sprite_tiled_initialize_8bpp_4x() \
3062 #define setup_sprite_tile_full_4bpp_4x(edge) \
3064 vec_8x8u texels_low, texels_high; \
3066 vec_4x16u pixels_half; \
3067 setup_sprite_tile_add_blocks(sub_tile_height * 2 * 4); \
3068 u32 left_mask_bits_a = left_mask_bits & 0xFF; \
3069 u32 left_mask_bits_b = left_mask_bits >> 8; \
3070 u32 right_mask_bits_a = right_mask_bits & 0xFF; \
3071 u32 right_mask_bits_b = right_mask_bits >> 8; \
3073 while(sub_tile_height) \
3075 setup_sprite_tile_fetch_texel_block_8bpp(0); \
3076 gvtbl2_u8(texels_low, clut_low, texels); \
3077 gvtbl2_u8(texels_high, clut_high, texels); \
3078 gvzip_u8(pixels, texels_low, texels_high); \
3080 gvget_lo(pixels_half, pixels); \
3081 gvst2_u16(pixels_half, pixels_half, block->texels.e); \
3082 block->draw_mask_bits = left_mask_bits_a; \
3083 block->fb_ptr = fb_ptr; \
3086 gvst2_u16(pixels_half, pixels_half, block->texels.e); \
3087 block->draw_mask_bits = left_mask_bits_a; \
3088 block->fb_ptr = fb_ptr + 1024; \
3091 gvget_hi(pixels_half, pixels); \
3092 gvst2_u16(pixels_half, pixels_half, block->texels.e); \
3093 block->draw_mask_bits = left_mask_bits_b; \
3094 block->fb_ptr = fb_ptr + 8; \
3097 gvst2_u16(pixels_half, pixels_half, block->texels.e); \
3098 block->draw_mask_bits = left_mask_bits_b; \
3099 block->fb_ptr = fb_ptr + 1024 + 8; \
3102 setup_sprite_tile_fetch_texel_block_8bpp(8); \
3103 gvtbl2_u8(texels_low, clut_low, texels); \
3104 gvtbl2_u8(texels_high, clut_high, texels); \
3105 gvzip_u8(pixels, texels_low, texels_high); \
3107 gvget_lo(pixels_half, pixels); \
3108 gvst2_u16(pixels_half, pixels_half, block->texels.e); \
3109 block->draw_mask_bits = right_mask_bits_a; \
3110 block->fb_ptr = fb_ptr + 16; \
3113 gvst2_u16(pixels_half, pixels_half, block->texels.e); \
3114 block->draw_mask_bits = right_mask_bits_a; \
3115 block->fb_ptr = fb_ptr + 1024 + 16; \
3118 gvget_hi(pixels_half, pixels); \
3119 gvst2_u16(pixels_half, pixels_half, block->texels.e); \
3120 block->draw_mask_bits = right_mask_bits_b; \
3121 block->fb_ptr = fb_ptr + 24; \
3124 gvst2_u16(pixels_half, pixels_half, block->texels.e); \
3125 block->draw_mask_bits = right_mask_bits_b; \
3126 block->fb_ptr = fb_ptr + 1024 + 24; \
3130 texture_offset += 0x10; \
3131 sub_tile_height--; \
3133 texture_offset += 0xF00; \
3134 psx_gpu->num_blocks = num_blocks; \
3137 #define setup_sprite_tile_half_4bpp_4x(edge) \
3139 vec_8x8u texels_low, texels_high; \
3141 vec_4x16u pixels_half; \
3142 setup_sprite_tile_add_blocks(sub_tile_height * 4); \
3143 u32 edge##_mask_bits_a = edge##_mask_bits & 0xFF; \
3144 u32 edge##_mask_bits_b = edge##_mask_bits >> 8; \
3146 while(sub_tile_height) \
3148 setup_sprite_tile_fetch_texel_block_8bpp(0); \
3149 gvtbl2_u8(texels_low, clut_low, texels); \
3150 gvtbl2_u8(texels_high, clut_high, texels); \
3151 gvzip_u8(pixels, texels_low, texels_high); \
3153 gvget_lo(pixels_half, pixels); \
3154 gvst2_u16(pixels_half, pixels_half, block->texels.e); \
3155 block->draw_mask_bits = edge##_mask_bits_a; \
3156 block->fb_ptr = fb_ptr; \
3159 gvst2_u16(pixels_half, pixels_half, block->texels.e); \
3160 block->draw_mask_bits = edge##_mask_bits_a; \
3161 block->fb_ptr = fb_ptr + 1024; \
3164 gvget_hi(pixels_half, pixels); \
3165 gvst2_u16(pixels_half, pixels_half, block->texels.e); \
3166 block->draw_mask_bits = edge##_mask_bits_b; \
3167 block->fb_ptr = fb_ptr + 8; \
3170 gvst2_u16(pixels_half, pixels_half, block->texels.e); \
3171 block->draw_mask_bits = edge##_mask_bits_b; \
3172 block->fb_ptr = fb_ptr + 1024 + 8; \
3176 texture_offset += 0x10; \
3177 sub_tile_height--; \
3179 texture_offset += 0xF00; \
3180 psx_gpu->num_blocks = num_blocks; \
3183 #define setup_sprite_tile_full_8bpp_4x(edge) \
3185 setup_sprite_tile_add_blocks(sub_tile_height * 2 * 4); \
3186 vec_8x16u texels_wide; \
3187 vec_4x16u texels_half; \
3188 u32 left_mask_bits_a = left_mask_bits & 0xFF; \
3189 u32 left_mask_bits_b = left_mask_bits >> 8; \
3190 u32 right_mask_bits_a = right_mask_bits & 0xFF; \
3191 u32 right_mask_bits_b = right_mask_bits >> 8; \
3193 while(sub_tile_height) \
3195 setup_sprite_tile_fetch_texel_block_8bpp(0); \
3196 gvzip_u8(texels_wide, texels, texels); \
3197 gvget_lo(texels_half, texels_wide); \
3198 gvst1_u8(texels_half, block->r.e); \
3199 block->draw_mask_bits = left_mask_bits_a; \
3200 block->fb_ptr = fb_ptr; \
3203 gvst1_u8(texels_half, block->r.e); \
3204 block->draw_mask_bits = left_mask_bits_a; \
3205 block->fb_ptr = fb_ptr + 1024; \
3208 gvget_hi(texels_half, texels_wide); \
3209 gvst1_u8(texels_half, block->r.e); \
3210 block->draw_mask_bits = left_mask_bits_b; \
3211 block->fb_ptr = fb_ptr + 8; \
3214 gvst1_u8(texels_half, block->r.e); \
3215 block->draw_mask_bits = left_mask_bits_b; \
3216 block->fb_ptr = fb_ptr + 1024 + 8; \
3219 setup_sprite_tile_fetch_texel_block_8bpp(8); \
3220 gvzip_u8(texels_wide, texels, texels); \
3221 gvget_lo(texels_half, texels_wide); \
3222 gvst1_u8(texels_half, block->r.e); \
3223 block->draw_mask_bits = right_mask_bits_a; \
3224 block->fb_ptr = fb_ptr + 16; \
3227 gvst1_u8(texels_half, block->r.e); \
3228 block->draw_mask_bits = right_mask_bits_a; \
3229 block->fb_ptr = fb_ptr + 1024 + 16; \
3232 gvget_hi(texels_half, texels_wide); \
3233 gvst1_u8(texels_half, block->r.e); \
3234 block->draw_mask_bits = right_mask_bits_b; \
3235 block->fb_ptr = fb_ptr + 24; \
3238 gvst1_u8(texels_half, block->r.e); \
3239 block->draw_mask_bits = right_mask_bits_b; \
3240 block->fb_ptr = fb_ptr + 24 + 1024; \
3244 texture_offset += 0x10; \
3245 sub_tile_height--; \
3247 texture_offset += 0xF00; \
3248 psx_gpu->num_blocks = num_blocks; \
3251 #define setup_sprite_tile_half_8bpp_4x(edge) \
3253 setup_sprite_tile_add_blocks(sub_tile_height * 4); \
3254 vec_8x16u texels_wide; \
3255 vec_4x16u texels_half; \
3256 u32 edge##_mask_bits_a = edge##_mask_bits & 0xFF; \
3257 u32 edge##_mask_bits_b = edge##_mask_bits >> 8; \
3259 while(sub_tile_height) \
3261 setup_sprite_tile_fetch_texel_block_8bpp(0); \
3262 gvzip_u8(texels_wide, texels, texels); \
3263 gvget_lo(texels_half, texels_wide); \
3264 gvst1_u8(texels_half, block->r.e); \
3265 block->draw_mask_bits = edge##_mask_bits_a; \
3266 block->fb_ptr = fb_ptr; \
3269 gvst1_u8(texels_half, block->r.e); \
3270 block->draw_mask_bits = edge##_mask_bits_a; \
3271 block->fb_ptr = fb_ptr + 1024; \
3274 gvget_hi(texels_half, texels_wide); \
3275 gvst1_u8(texels_half, block->r.e); \
3276 block->draw_mask_bits = edge##_mask_bits_b; \
3277 block->fb_ptr = fb_ptr + 8; \
3280 gvst1_u8(texels_half, block->r.e); \
3281 block->draw_mask_bits = edge##_mask_bits_b; \
3282 block->fb_ptr = fb_ptr + 8 + 1024; \
3286 texture_offset += 0x10; \
3287 sub_tile_height--; \
3289 texture_offset += 0xF00; \
3290 psx_gpu->num_blocks = num_blocks; \
3293 #define setup_sprite_tile_column_edge_pre_adjust_half_right_4x() \
3294 texture_offset = texture_offset_base + 8; \
3297 #define setup_sprite_tile_column_edge_pre_adjust_half_left_4x() \
3298 texture_offset = texture_offset_base \
3300 #define setup_sprite_tile_column_edge_pre_adjust_half_4x(edge) \
3301 setup_sprite_tile_column_edge_pre_adjust_half_##edge##_4x() \
3303 #define setup_sprite_tile_column_edge_pre_adjust_full_4x(edge) \
3304 texture_offset = texture_offset_base \
3306 #define setup_sprite_tile_column_edge_post_adjust_half_right_4x() \
3309 #define setup_sprite_tile_column_edge_post_adjust_half_left_4x() \
3311 #define setup_sprite_tile_column_edge_post_adjust_half_4x(edge) \
3312 setup_sprite_tile_column_edge_post_adjust_half_##edge##_4x() \
3314 #define setup_sprite_tile_column_edge_post_adjust_full_4x(edge) \
3316 #define setup_sprite_offset_u_adjust() \
3318 #define setup_sprite_comapre_left_block_mask() \
3319 ((left_block_mask & 0xFF) == 0xFF) \
3321 #define setup_sprite_comapre_right_block_mask() \
3322 (((right_block_mask >> 8) & 0xFF) == 0xFF) \
3324 #define setup_sprite_offset_u_adjust_4x() \
3326 offset_u_right = offset_u_right * 2 + 1 \
3328 #define setup_sprite_comapre_left_block_mask_4x() \
3329 ((left_block_mask & 0xFFFF) == 0xFFFF) \
3331 #define setup_sprite_comapre_right_block_mask_4x() \
3332 (((right_block_mask >> 16) & 0xFFFF) == 0xFFFF) \
3335 #define setup_sprite_tiled_do(texture_mode, x4mode) \
3336 s32 offset_u = u & 0xF; \
3337 s32 offset_v = v & 0xF; \
3339 s32 width_rounded = offset_u + width + 15; \
3340 s32 height_rounded = offset_v + height + 15; \
3341 s32 tile_height = height_rounded / 16; \
3342 s32 tile_width = width_rounded / 16; \
3343 u32 offset_u_right = width_rounded & 0xF; \
3345 setup_sprite_offset_u_adjust##x4mode(); \
3347 u32 left_block_mask = ~(0xFFFFFFFF << offset_u); \
3348 u32 right_block_mask = 0xFFFFFFFE << offset_u_right; \
3350 u32 left_mask_bits; \
3351 u32 right_mask_bits; \
3353 u32 sub_tile_height; \
3356 u32 texture_mask = (psx_gpu->texture_mask_width & 0xF) | \
3357 ((psx_gpu->texture_mask_height & 0xF) << 4) | \
3358 ((psx_gpu->texture_mask_width >> 4) << 8) | \
3359 ((psx_gpu->texture_mask_height >> 4) << 12); \
3360 u32 texture_offset = ((v & 0xF) << 4) | ((u & 0xF0) << 4) | \
3361 ((v & 0xF0) << 8); \
3362 u32 texture_offset_base = texture_offset; \
3365 u16 *fb_ptr = psx_gpu->vram_out_ptr + (y * 1024) + (x - offset_u); \
3366 u32 num_blocks = psx_gpu->num_blocks; \
3367 block_struct *block = psx_gpu->blocks + num_blocks; \
3369 u16 *texture_block_ptr; \
3372 setup_sprite_tiled_initialize_##texture_mode##x4mode(); \
3374 control_mask = tile_width == 1; \
3375 control_mask |= (tile_height == 1) << 1; \
3376 control_mask |= setup_sprite_comapre_left_block_mask##x4mode() << 2; \
3377 control_mask |= setup_sprite_comapre_right_block_mask##x4mode() << 3; \
3379 switch(control_mask) \
3383 setup_sprite_tile_column_width_multi(texture_mode, multi, full, full, \
3388 setup_sprite_tile_column_width_single(texture_mode, multi, full, none, \
3393 setup_sprite_tile_column_width_multi(texture_mode, single, full, full, \
3398 setup_sprite_tile_column_width_single(texture_mode, single, full, none, \
3403 setup_sprite_tile_column_width_multi(texture_mode, multi, half, full, \
3408 setup_sprite_tile_column_width_single(texture_mode, multi, half, right, \
3413 setup_sprite_tile_column_width_multi(texture_mode, single, half, full, \
3418 setup_sprite_tile_column_width_single(texture_mode, single, half, right, \
3423 setup_sprite_tile_column_width_multi(texture_mode, multi, full, half, \
3428 setup_sprite_tile_column_width_single(texture_mode, multi, half, left, \
3433 setup_sprite_tile_column_width_multi(texture_mode, single, full, half, \
3438 setup_sprite_tile_column_width_single(texture_mode, single, half, left, \
3443 setup_sprite_tile_column_width_multi(texture_mode, multi, half, half, \
3448 setup_sprite_tile_column_width_multi(texture_mode, single, half, half, \
3453 void setup_sprite_4bpp(psx_gpu_struct *psx_gpu, s32 x, s32 y, s32 u, s32 v,
3454 s32 width, s32 height, u32 color)
3457 setup_sprite_4bpp_(psx_gpu, x, y, u, v, width, height, color);
3460 setup_sprite_tiled_do(4bpp,)
3463 void setup_sprite_8bpp(psx_gpu_struct *psx_gpu, s32 x, s32 y, s32 u, s32 v,
3464 s32 width, s32 height, u32 color)
3467 setup_sprite_8bpp_(psx_gpu, x, y, u, v, width, height, color);
3470 setup_sprite_tiled_do(8bpp,)
3473 #undef draw_mask_fb_ptr_left
3474 #undef draw_mask_fb_ptr_right
3476 void setup_sprite_4bpp_4x(psx_gpu_struct *psx_gpu, s32 x, s32 y, s32 u, s32 v,
3477 s32 width, s32 height, u32 color)
3480 setup_sprite_4bpp_4x_(psx_gpu, x, y, u, v, width, height, color);
3483 setup_sprite_tiled_do(4bpp, _4x)
3486 void setup_sprite_8bpp_4x(psx_gpu_struct *psx_gpu, s32 x, s32 y, s32 u, s32 v,
3487 s32 width, s32 height, u32 color)
3490 setup_sprite_8bpp_4x_(psx_gpu, x, y, u, v, width, height, color);
3493 setup_sprite_tiled_do(8bpp, _4x)
3497 void scale2x_tiles8(void * __restrict__ dst_, const void * __restrict__ src_, int w8, int h)
3500 scale2x_tiles8_(dst_, src_, w8, h);
3503 const u16 * __restrict__ src = src_;
3504 const u16 * __restrict__ src1;
3505 u16 * __restrict__ dst = dst_;
3506 u16 * __restrict__ dst1;
3509 for (; h > 0; h--, src += 1024, dst += 1024*2)
3513 for (w = w8; w > 0; w--, src1 += 8, dst1 += 8*2)
3515 gvld1q_u16(a, src1);
3516 gvzipq_u16(a, b, a, a);
3517 gvst1q_u16(a, dst1);
3518 gvst1q_u16(b, dst1 + 8);
3519 gvst1q_u16(a, dst1 + 1024);
3520 gvst1q_u16(b, dst1 + 1024 + 8);
3525 // vim:ts=2:sw=2:expandtab