| 1 | /* |
| 2 | * (C) GraÅžvydas "notaz" Ignotas, 2014 |
| 3 | * |
| 4 | * This work is licensed under the terms of GNU GPL version 2 or later. |
| 5 | * See the COPYING file in the top-level directory. |
| 6 | */ |
| 7 | |
| 8 | #include "arm_features.h" |
| 9 | |
| 10 | .syntax unified |
| 11 | .text |
| 12 | .align 3 |
| 13 | |
| 14 | .macro do_mac_flags rr1 rr2 rr3 |
| 15 | cmp \rr1, #1 |
| 16 | .endm |
| 17 | |
| 18 | |
| 19 | /* |
| 20 | * ProcessVertexData register map: |
| 21 | * |
| 22 | * q | d | c code |
| 23 | * 0 0 |
| 24 | * 1 |
| 25 | * 1 2 |
| 26 | * 3 |
| 27 | * 2 4 |
| 28 | * 5 |
| 29 | * 3 6 |
| 30 | * 7 |
| 31 | * 4 8 |
| 32 | * 9 |
| 33 | * 5 10 |
| 34 | * 11 |
| 35 | * 6 12 |
| 36 | * 13 |
| 37 | * 7 14 g_vecProjected |
| 38 | * 15 |
| 39 | * 8 16 |
| 40 | * 17 |
| 41 | * ... |
| 42 | * 12 24 gRSPworldProject _11,_21,_31,_41 |
| 43 | * 25 |
| 44 | * 13 26 gRSPworldProject _12,_22,_32,_42 |
| 45 | * 27 |
| 46 | * 14 28 gRSPworldProject _13,_23,_33,_43 |
| 47 | * 29 |
| 48 | * 15 30 gRSPworldProject _14,_24,_34,_44 |
| 49 | * 31 |
| 50 | * |
| 51 | * r4 vtx[], 16 bytes: |
| 52 | * short y, x, flag, z, tv, tu; |
| 53 | * / uint8 a, b, g, r; |
| 54 | * \ char a, z, y, x; |
| 55 | * |
| 56 | * outputs: |
| 57 | * r0 - XVECTOR4 *g_vtxTransformed |
| 58 | * r1 - XVECTOR4 *g_vecProjected |
| 59 | * r2 - uint32 *g_dwVtxDifColor |
| 60 | * r3 - VECTOR2 *g_fVtxTxtCoords |
| 61 | * sp+00 - float *g_fFogCoord |
| 62 | * sp+04 - uint32 *g_clipFlag2 |
| 63 | * inputs: |
| 64 | * r11 sp+08 - uint32 dwNum |
| 65 | * sp+0c - int neon_flags |
| 66 | * r4 sp+10 - FiddledVtx vtx[] |
| 67 | * sp+14 - Light *gRSPlights |
| 68 | * sp+18 - float *fRSPAmbientLightRGBA |
| 69 | * sp+1c - XMATRIX *gRSPworldProjectTransported |
| 70 | * sp+20 - XMATRIX *gRSPmodelViewTop |
| 71 | * sp+24 - uint32 gRSPnumLights |
| 72 | * sp+28 - float gRSPfFogMin |
| 73 | */ |
| 74 | FUNCTION(pv_neon): @ |
| 75 | ldr r12, [sp, #0x10] |
| 76 | pld [r12] |
| 77 | |
| 78 | push {r4-r11,lr} |
| 79 | vpush {q4-q7} |
| 80 | |
| 81 | mov r4, r12 @ vtx |
| 82 | ldr r12, [sp, #0x64+0x1c] |
| 83 | ldr r11, [sp, #0x64+0x08] |
| 84 | vld1.32 {q12,q13}, [r12, :128]! @ load gRSPworldProject |
| 85 | vld1.32 {q14,q15}, [r12, :128] |
| 86 | |
| 87 | 0: |
| 88 | vld1.16 d4[1], [r4]! @ y |
| 89 | vld1.16 d4[0], [r4]! @ x |
| 90 | vld1.16 d4[3], [r4]! @ flag |
| 91 | vld1.16 d4[2], [r4]! @ z |
| 92 | vld1.16 d5[1], [r4]! @ v |
| 93 | vld1.16 d5[0], [r4]! @ u |
| 94 | vmovl.s16 q0, d4 |
| 95 | vmovl.s16 q1, d5 |
| 96 | mov r12, #0x3f800000 @ 1.0f |
| 97 | vcvt.f32.s32 q2, q0 @ q2 = vtx_raw |
| 98 | vcvt.f32.s32 q3, q1 @ d6 = float u, v |
| 99 | vmov.32 d5[1], r12 @ q2 = { x, y, x, 1.0f } |
| 100 | vmul.f32 q4, q2, q12 |
| 101 | vmul.f32 q5, q2, q13 |
| 102 | vmul.f32 q6, q2, q14 |
| 103 | vmul.f32 q7, q2, q15 |
| 104 | /* wrt u,v to g_fVtxTxtCoords */ vst1.32 {d6}, [r3]! |
| 105 | vpadd.f32 d7, d14, d15 |
| 106 | vpadd.f32 d6, d12, d13 |
| 107 | vpadd.f32 d5, d10, d11 |
| 108 | vpadd.f32 d4, d8, d9 |
| 109 | vpadd.f32 d1, d6, d7 @ g_vtxTransformed .z .w |
| 110 | vpadd.f32 d0, d4, d5 @ g_vtxTransformed .x .y |
| 111 | |
| 112 | #if 1 |
| 113 | vrecpe.f32 d2, d1 @ inv .z(unused) .w |
| 114 | vrecps.f32 d3, d1, d2 @ step |
| 115 | /* wrt g_vtxTransformed */ vst1.32 {q0}, [r0]! |
| 116 | vmul.f32 d2, d3, d2 @ better inv |
| 117 | vrecps.f32 d3, d1, d2 @ step |
| 118 | vmul.f32 d2, d3, d2 @ better inv |
| 119 | #if 0 |
| 120 | vrecps.f32 d3, d1, d2 @ step |
| 121 | vmul.f32 d2, d3, d2 @ better inv |
| 122 | #endif |
| 123 | #else |
| 124 | vmov.f32 s5, r12 |
| 125 | /* wrt g_vtxTransformed */ vst1.32 {q0}, [r0]! |
| 126 | vdiv.f32 s5, s5, s3 |
| 127 | #endif |
| 128 | |
| 129 | vmul.f32 q7, q0, d2[1] |
| 130 | vshr.u64 d2, #32 |
| 131 | vsli.64 d15, d2, #32 |
| 132 | /* wrt g_vecProjected */ vst1.32 {q7}, [r1]! |
| 133 | |
| 134 | vpop {q4-q7} |
| 135 | pop {r4-r11,pc} |
| 136 | .size pv_neon, .-pv_neon |
| 137 | |
| 138 | |
| 139 | @ vim:filetype=armasm |