• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 /*
2  * Copyright 2012 Red Hat Inc.
3  *
4  * Permission is hereby granted, free of charge, to any person obtaining a
5  * copy of this software and associated documentation files (the "Software"),
6  * to deal in the Software without restriction, including without limitation
7  * the rights to use, copy, modify, merge, publish, distribute, sublicense,
8  * and/or sell copies of the Software, and to permit persons to whom the
9  * Software is furnished to do so, subject to the following conditions:
10  *
11  * The above copyright notice and this permission notice shall be included in
12  * all copies or substantial portions of the Software.
13  *
14  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
15  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
16  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
17  * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
18  * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
19  * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
20  * OTHER DEALINGS IN THE SOFTWARE.
21  *
22  * Authors: Ben Skeggs
23  *
24  */
25 
26 #include <xf86drm.h>
27 #include <nouveau_drm.h>
28 #include "util/format/u_format.h"
29 #include "util/format/u_format_s3tc.h"
30 #include "util/u_screen.h"
31 
32 #include "nv_object.xml.h"
33 #include "nv_m2mf.xml.h"
34 #include "nv30/nv30-40_3d.xml.h"
35 #include "nv30/nv01_2d.xml.h"
36 
37 #include "nouveau_fence.h"
38 #include "nv30/nv30_screen.h"
39 #include "nv30/nv30_context.h"
40 #include "nv30/nv30_resource.h"
41 #include "nv30/nv30_format.h"
42 
43 #define RANKINE_0397_CHIPSET 0x00000003
44 #define RANKINE_0497_CHIPSET 0x000001e0
45 #define RANKINE_0697_CHIPSET 0x00000010
46 #define CURIE_4097_CHIPSET   0x00000baf
47 #define CURIE_4497_CHIPSET   0x00005450
48 #define CURIE_4497_CHIPSET6X 0x00000088
49 
50 static int
nv30_screen_get_param(struct pipe_screen * pscreen,enum pipe_cap param)51 nv30_screen_get_param(struct pipe_screen *pscreen, enum pipe_cap param)
52 {
53    struct nv30_screen *screen = nv30_screen(pscreen);
54    struct nouveau_object *eng3d = screen->eng3d;
55    struct nouveau_device *dev = nouveau_screen(pscreen)->device;
56 
57    switch (param) {
58    /* non-boolean capabilities */
59    case PIPE_CAP_MAX_RENDER_TARGETS:
60       return (eng3d->oclass >= NV40_3D_CLASS) ? 4 : 1;
61    case PIPE_CAP_MAX_TEXTURE_2D_SIZE:
62       return 4096;
63    case PIPE_CAP_MAX_TEXTURE_3D_LEVELS:
64       return 10;
65    case PIPE_CAP_MAX_TEXTURE_CUBE_LEVELS:
66       return 13;
67    case PIPE_CAP_GLSL_FEATURE_LEVEL:
68    case PIPE_CAP_GLSL_FEATURE_LEVEL_COMPATIBILITY:
69       return 120;
70    case PIPE_CAP_ENDIANNESS:
71       return PIPE_ENDIAN_LITTLE;
72    case PIPE_CAP_CONSTANT_BUFFER_OFFSET_ALIGNMENT:
73       return 16;
74    case PIPE_CAP_MIN_MAP_BUFFER_ALIGNMENT:
75       return NOUVEAU_MIN_BUFFER_MAP_ALIGN;
76    case PIPE_CAP_MAX_VIEWPORTS:
77       return 1;
78    case PIPE_CAP_MAX_VERTEX_ATTRIB_STRIDE:
79       return 2048;
80    case PIPE_CAP_MAX_TEXTURE_UPLOAD_MEMORY_BUDGET:
81       return 8 * 1024 * 1024;
82    case PIPE_CAP_MAX_VARYINGS:
83       return 8;
84 
85    /* supported capabilities */
86    case PIPE_CAP_ANISOTROPIC_FILTER:
87    case PIPE_CAP_POINT_SPRITE:
88    case PIPE_CAP_OCCLUSION_QUERY:
89    case PIPE_CAP_QUERY_TIME_ELAPSED:
90    case PIPE_CAP_QUERY_TIMESTAMP:
91    case PIPE_CAP_TEXTURE_SWIZZLE:
92    case PIPE_CAP_DEPTH_CLIP_DISABLE:
93    case PIPE_CAP_FS_COORD_ORIGIN_UPPER_LEFT:
94    case PIPE_CAP_FS_COORD_ORIGIN_LOWER_LEFT:
95    case PIPE_CAP_FS_COORD_PIXEL_CENTER_HALF_INTEGER:
96    case PIPE_CAP_FS_COORD_PIXEL_CENTER_INTEGER:
97    case PIPE_CAP_TGSI_TEXCOORD:
98    case PIPE_CAP_BUFFER_MAP_PERSISTENT_COHERENT:
99    case PIPE_CAP_CLEAR_SCISSORED:
100    case PIPE_CAP_VERTEX_BUFFER_OFFSET_4BYTE_ALIGNED_ONLY:
101    case PIPE_CAP_VERTEX_BUFFER_STRIDE_4BYTE_ALIGNED_ONLY:
102    case PIPE_CAP_VERTEX_ELEMENT_SRC_OFFSET_4BYTE_ALIGNED_ONLY:
103    case PIPE_CAP_ALLOW_MAPPED_BUFFERS_DURING_EXECUTION:
104       return 1;
105    case PIPE_CAP_TEXTURE_TRANSFER_MODES:
106       return PIPE_TEXTURE_TRANSFER_BLIT;
107    /* nv35 capabilities */
108    case PIPE_CAP_DEPTH_BOUNDS_TEST:
109       return eng3d->oclass == NV35_3D_CLASS || eng3d->oclass >= NV40_3D_CLASS;
110    case PIPE_CAP_SUPPORTED_PRIM_MODES_WITH_RESTART:
111    case PIPE_CAP_SUPPORTED_PRIM_MODES:
112       return BITFIELD_MASK(PIPE_PRIM_MAX);
113    /* nv4x capabilities */
114    case PIPE_CAP_BLEND_EQUATION_SEPARATE:
115    case PIPE_CAP_NPOT_TEXTURES:
116    case PIPE_CAP_CONDITIONAL_RENDER:
117    case PIPE_CAP_TEXTURE_MIRROR_CLAMP:
118    case PIPE_CAP_TEXTURE_MIRROR_CLAMP_TO_EDGE:
119    case PIPE_CAP_PRIMITIVE_RESTART:
120    case PIPE_CAP_PRIMITIVE_RESTART_FIXED_INDEX:
121       return (eng3d->oclass >= NV40_3D_CLASS) ? 1 : 0;
122    /* unsupported */
123    case PIPE_CAP_EMULATE_NONFIXED_PRIMITIVE_RESTART:
124    case PIPE_CAP_DEPTH_CLIP_DISABLE_SEPARATE:
125    case PIPE_CAP_MAX_DUAL_SOURCE_RENDER_TARGETS:
126    case PIPE_CAP_FRAGMENT_SHADER_TEXTURE_LOD:
127    case PIPE_CAP_FRAGMENT_SHADER_DERIVATIVES:
128    case PIPE_CAP_INDEP_BLEND_ENABLE:
129    case PIPE_CAP_INDEP_BLEND_FUNC:
130    case PIPE_CAP_MAX_TEXTURE_ARRAY_LAYERS:
131    case PIPE_CAP_SHADER_STENCIL_EXPORT:
132    case PIPE_CAP_VS_INSTANCEID:
133    case PIPE_CAP_VERTEX_ELEMENT_INSTANCE_DIVISOR: /* XXX: yes? */
134    case PIPE_CAP_MAX_STREAM_OUTPUT_BUFFERS:
135    case PIPE_CAP_STREAM_OUTPUT_PAUSE_RESUME:
136    case PIPE_CAP_STREAM_OUTPUT_INTERLEAVE_BUFFERS:
137    case PIPE_CAP_MIN_TEXEL_OFFSET:
138    case PIPE_CAP_MAX_TEXEL_OFFSET:
139    case PIPE_CAP_MIN_TEXTURE_GATHER_OFFSET:
140    case PIPE_CAP_MAX_TEXTURE_GATHER_OFFSET:
141    case PIPE_CAP_MAX_STREAM_OUTPUT_SEPARATE_COMPONENTS:
142    case PIPE_CAP_MAX_STREAM_OUTPUT_INTERLEAVED_COMPONENTS:
143    case PIPE_CAP_MAX_GEOMETRY_OUTPUT_VERTICES:
144    case PIPE_CAP_MAX_GEOMETRY_TOTAL_OUTPUT_COMPONENTS:
145    case PIPE_CAP_MAX_VERTEX_STREAMS:
146    case PIPE_CAP_TGSI_CAN_COMPACT_CONSTANTS:
147    case PIPE_CAP_TEXTURE_BARRIER:
148    case PIPE_CAP_SEAMLESS_CUBE_MAP:
149    case PIPE_CAP_SEAMLESS_CUBE_MAP_PER_TEXTURE:
150    case PIPE_CAP_CUBE_MAP_ARRAY:
151    case PIPE_CAP_VERTEX_COLOR_UNCLAMPED:
152    case PIPE_CAP_FRAGMENT_COLOR_CLAMPED:
153    case PIPE_CAP_VERTEX_COLOR_CLAMPED:
154    case PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION:
155    case PIPE_CAP_MIXED_COLORBUFFER_FORMATS:
156    case PIPE_CAP_START_INSTANCE:
157    case PIPE_CAP_TEXTURE_MULTISAMPLE:
158    case PIPE_CAP_TEXTURE_BUFFER_OBJECTS:
159    case PIPE_CAP_TEXTURE_BUFFER_OFFSET_ALIGNMENT:
160    case PIPE_CAP_QUERY_PIPELINE_STATISTICS:
161    case PIPE_CAP_TEXTURE_BORDER_COLOR_QUIRK:
162    case PIPE_CAP_MAX_TEXEL_BUFFER_ELEMENTS_UINT:
163    case PIPE_CAP_MIXED_FRAMEBUFFER_SIZES:
164    case PIPE_CAP_VS_LAYER_VIEWPORT:
165    case PIPE_CAP_MAX_TEXTURE_GATHER_COMPONENTS:
166    case PIPE_CAP_TEXTURE_GATHER_SM5:
167    case PIPE_CAP_FAKE_SW_MSAA:
168    case PIPE_CAP_TEXTURE_QUERY_LOD:
169    case PIPE_CAP_SAMPLE_SHADING:
170    case PIPE_CAP_TEXTURE_GATHER_OFFSETS:
171    case PIPE_CAP_VS_WINDOW_SPACE_POSITION:
172    case PIPE_CAP_USER_VERTEX_BUFFERS:
173    case PIPE_CAP_COMPUTE:
174    case PIPE_CAP_DRAW_INDIRECT:
175    case PIPE_CAP_MULTI_DRAW_INDIRECT:
176    case PIPE_CAP_MULTI_DRAW_INDIRECT_PARAMS:
177    case PIPE_CAP_FS_FINE_DERIVATIVE:
178    case PIPE_CAP_CONDITIONAL_RENDER_INVERTED:
179    case PIPE_CAP_SAMPLER_VIEW_TARGET:
180    case PIPE_CAP_CLIP_HALFZ:
181    case PIPE_CAP_VERTEXID_NOBASE:
182    case PIPE_CAP_POLYGON_OFFSET_CLAMP:
183    case PIPE_CAP_MULTISAMPLE_Z_RESOLVE:
184    case PIPE_CAP_RESOURCE_FROM_USER_MEMORY:
185    case PIPE_CAP_DEVICE_RESET_STATUS_QUERY:
186    case PIPE_CAP_MAX_SHADER_PATCH_VARYINGS:
187    case PIPE_CAP_TEXTURE_FLOAT_LINEAR:
188    case PIPE_CAP_TEXTURE_HALF_FLOAT_LINEAR:
189    case PIPE_CAP_TEXTURE_QUERY_SAMPLES:
190    case PIPE_CAP_FORCE_PERSAMPLE_INTERP:
191    case PIPE_CAP_COPY_BETWEEN_COMPRESSED_AND_PLAIN_FORMATS:
192    case PIPE_CAP_SHAREABLE_SHADERS:
193    case PIPE_CAP_CLEAR_TEXTURE:
194    case PIPE_CAP_DRAW_PARAMETERS:
195    case PIPE_CAP_SHADER_PACK_HALF_FLOAT:
196    case PIPE_CAP_FS_POSITION_IS_SYSVAL:
197    case PIPE_CAP_FS_FACE_IS_INTEGER_SYSVAL:
198    case PIPE_CAP_SHADER_BUFFER_OFFSET_ALIGNMENT:
199    case PIPE_CAP_INVALIDATE_BUFFER:
200    case PIPE_CAP_GENERATE_MIPMAP:
201    case PIPE_CAP_STRING_MARKER:
202    case PIPE_CAP_BUFFER_SAMPLER_VIEW_RGBA_ONLY:
203    case PIPE_CAP_SURFACE_REINTERPRET_BLOCKS:
204    case PIPE_CAP_QUERY_BUFFER_OBJECT:
205    case PIPE_CAP_QUERY_MEMORY_INFO:
206    case PIPE_CAP_PCI_GROUP:
207    case PIPE_CAP_PCI_BUS:
208    case PIPE_CAP_PCI_DEVICE:
209    case PIPE_CAP_PCI_FUNCTION:
210    case PIPE_CAP_FRAMEBUFFER_NO_ATTACHMENT:
211    case PIPE_CAP_ROBUST_BUFFER_ACCESS_BEHAVIOR:
212    case PIPE_CAP_CULL_DISTANCE:
213    case PIPE_CAP_SHADER_GROUP_VOTE:
214    case PIPE_CAP_MAX_WINDOW_RECTANGLES:
215    case PIPE_CAP_POLYGON_OFFSET_UNITS_UNSCALED:
216    case PIPE_CAP_VIEWPORT_SUBPIXEL_BITS:
217    case PIPE_CAP_MIXED_COLOR_DEPTH_BITS:
218    case PIPE_CAP_SHADER_ARRAY_COMPONENTS:
219    case PIPE_CAP_SHADER_CAN_READ_OUTPUTS:
220    case PIPE_CAP_NATIVE_FENCE_FD:
221    case PIPE_CAP_FBFETCH:
222    case PIPE_CAP_LEGACY_MATH_RULES:
223    case PIPE_CAP_DOUBLES:
224    case PIPE_CAP_INT64:
225    case PIPE_CAP_INT64_DIVMOD:
226    case PIPE_CAP_TGSI_TEX_TXF_LZ:
227    case PIPE_CAP_SHADER_CLOCK:
228    case PIPE_CAP_POLYGON_MODE_FILL_RECTANGLE:
229    case PIPE_CAP_SPARSE_BUFFER_PAGE_SIZE:
230    case PIPE_CAP_SHADER_BALLOT:
231    case PIPE_CAP_TES_LAYER_VIEWPORT:
232    case PIPE_CAP_CAN_BIND_CONST_BUFFER_AS_VERTEX:
233    case PIPE_CAP_POST_DEPTH_COVERAGE:
234    case PIPE_CAP_BINDLESS_TEXTURE:
235    case PIPE_CAP_NIR_SAMPLERS_AS_DEREF:
236    case PIPE_CAP_QUERY_SO_OVERFLOW:
237    case PIPE_CAP_MEMOBJ:
238    case PIPE_CAP_LOAD_CONSTBUF:
239    case PIPE_CAP_TILE_RASTER_ORDER:
240    case PIPE_CAP_MAX_COMBINED_SHADER_OUTPUT_RESOURCES:
241    case PIPE_CAP_FRAMEBUFFER_MSAA_CONSTRAINTS:
242    case PIPE_CAP_SIGNED_VERTEX_BUFFER_OFFSET:
243    case PIPE_CAP_CONTEXT_PRIORITY_MASK:
244    case PIPE_CAP_FENCE_SIGNAL:
245    case PIPE_CAP_CONSTBUF0_FLAGS:
246    case PIPE_CAP_PACKED_UNIFORMS:
247    case PIPE_CAP_CONSERVATIVE_RASTER_POST_SNAP_TRIANGLES:
248    case PIPE_CAP_CONSERVATIVE_RASTER_POST_SNAP_POINTS_LINES:
249    case PIPE_CAP_CONSERVATIVE_RASTER_PRE_SNAP_TRIANGLES:
250    case PIPE_CAP_CONSERVATIVE_RASTER_PRE_SNAP_POINTS_LINES:
251    case PIPE_CAP_CONSERVATIVE_RASTER_POST_DEPTH_COVERAGE:
252    case PIPE_CAP_MAX_CONSERVATIVE_RASTER_SUBPIXEL_PRECISION_BIAS:
253    case PIPE_CAP_PROGRAMMABLE_SAMPLE_LOCATIONS:
254    case PIPE_CAP_IMAGE_LOAD_FORMATTED:
255    case PIPE_CAP_TGSI_DIV:
256    case PIPE_CAP_IMAGE_ATOMIC_INC_WRAP:
257    case PIPE_CAP_IMAGE_STORE_FORMATTED:
258       return 0;
259 
260    case PIPE_CAP_MAX_GS_INVOCATIONS:
261       return 32;
262    case PIPE_CAP_MAX_SHADER_BUFFER_SIZE_UINT:
263       return 1 << 27;
264    case PIPE_CAP_VENDOR_ID:
265       return 0x10de;
266    case PIPE_CAP_DEVICE_ID: {
267       uint64_t device_id;
268       if (nouveau_getparam(dev, NOUVEAU_GETPARAM_PCI_DEVICE, &device_id)) {
269          NOUVEAU_ERR("NOUVEAU_GETPARAM_PCI_DEVICE failed.\n");
270          return -1;
271       }
272       return device_id;
273    }
274    case PIPE_CAP_ACCELERATED:
275       return 1;
276    case PIPE_CAP_VIDEO_MEMORY:
277       return dev->vram_size >> 20;
278    case PIPE_CAP_UMA:
279       return 0;
280    default:
281       return u_pipe_screen_get_param_defaults(pscreen, param);
282    }
283 }
284 
285 static float
nv30_screen_get_paramf(struct pipe_screen * pscreen,enum pipe_capf param)286 nv30_screen_get_paramf(struct pipe_screen *pscreen, enum pipe_capf param)
287 {
288    struct nv30_screen *screen = nv30_screen(pscreen);
289    struct nouveau_object *eng3d = screen->eng3d;
290 
291    switch (param) {
292    case PIPE_CAPF_MIN_LINE_WIDTH:
293    case PIPE_CAPF_MIN_LINE_WIDTH_AA:
294    case PIPE_CAPF_MIN_POINT_SIZE:
295    case PIPE_CAPF_MIN_POINT_SIZE_AA:
296       return 1;
297    case PIPE_CAPF_POINT_SIZE_GRANULARITY:
298    case PIPE_CAPF_LINE_WIDTH_GRANULARITY:
299       return 0.1;
300    case PIPE_CAPF_MAX_LINE_WIDTH:
301    case PIPE_CAPF_MAX_LINE_WIDTH_AA:
302       return 10.0;
303    case PIPE_CAPF_MAX_POINT_SIZE:
304    case PIPE_CAPF_MAX_POINT_SIZE_AA:
305       return 64.0;
306    case PIPE_CAPF_MAX_TEXTURE_ANISOTROPY:
307       return (eng3d->oclass >= NV40_3D_CLASS) ? 16.0 : 8.0;
308    case PIPE_CAPF_MAX_TEXTURE_LOD_BIAS:
309       return 15.0;
310    case PIPE_CAPF_MIN_CONSERVATIVE_RASTER_DILATE:
311    case PIPE_CAPF_MAX_CONSERVATIVE_RASTER_DILATE:
312    case PIPE_CAPF_CONSERVATIVE_RASTER_DILATE_GRANULARITY:
313       return 0.0;
314    default:
315       debug_printf("unknown paramf %d\n", param);
316       return 0;
317    }
318 }
319 
320 static int
nv30_screen_get_shader_param(struct pipe_screen * pscreen,enum pipe_shader_type shader,enum pipe_shader_cap param)321 nv30_screen_get_shader_param(struct pipe_screen *pscreen,
322                              enum pipe_shader_type shader,
323                              enum pipe_shader_cap param)
324 {
325    struct nv30_screen *screen = nv30_screen(pscreen);
326    struct nouveau_object *eng3d = screen->eng3d;
327 
328    switch (shader) {
329    case PIPE_SHADER_VERTEX:
330       switch (param) {
331       case PIPE_SHADER_CAP_MAX_INSTRUCTIONS:
332       case PIPE_SHADER_CAP_MAX_ALU_INSTRUCTIONS:
333          return (eng3d->oclass >= NV40_3D_CLASS) ? 512 : 256;
334       case PIPE_SHADER_CAP_MAX_TEX_INSTRUCTIONS:
335       case PIPE_SHADER_CAP_MAX_TEX_INDIRECTIONS:
336          return (eng3d->oclass >= NV40_3D_CLASS) ? 512 : 0;
337       case PIPE_SHADER_CAP_MAX_CONTROL_FLOW_DEPTH:
338          return 0;
339       case PIPE_SHADER_CAP_MAX_INPUTS:
340       case PIPE_SHADER_CAP_MAX_OUTPUTS:
341          return 16;
342       case PIPE_SHADER_CAP_MAX_CONST_BUFFER0_SIZE:
343          return ((eng3d->oclass >= NV40_3D_CLASS) ? (468 - 6): (256 - 6)) * sizeof(float[4]);
344       case PIPE_SHADER_CAP_MAX_CONST_BUFFERS:
345          return 1;
346       case PIPE_SHADER_CAP_MAX_TEMPS:
347          return (eng3d->oclass >= NV40_3D_CLASS) ? 32 : 13;
348       case PIPE_SHADER_CAP_PREFERRED_IR:
349          return (NOUVEAU_DEBUG & NOUVEAU_DEBUG_USE_TGSI) ? PIPE_SHADER_IR_TGSI : PIPE_SHADER_IR_NIR;
350       case PIPE_SHADER_CAP_MAX_TEXTURE_SAMPLERS:
351       case PIPE_SHADER_CAP_MAX_SAMPLER_VIEWS:
352          return 0;
353       case PIPE_SHADER_CAP_CONT_SUPPORTED:
354       case PIPE_SHADER_CAP_TGSI_SQRT_SUPPORTED:
355       case PIPE_SHADER_CAP_INDIRECT_INPUT_ADDR:
356       case PIPE_SHADER_CAP_INDIRECT_OUTPUT_ADDR:
357       case PIPE_SHADER_CAP_INDIRECT_TEMP_ADDR:
358       case PIPE_SHADER_CAP_INDIRECT_CONST_ADDR:
359       case PIPE_SHADER_CAP_SUBROUTINES:
360       case PIPE_SHADER_CAP_INTEGERS:
361       case PIPE_SHADER_CAP_INT64_ATOMICS:
362       case PIPE_SHADER_CAP_FP16:
363       case PIPE_SHADER_CAP_FP16_DERIVATIVES:
364       case PIPE_SHADER_CAP_FP16_CONST_BUFFERS:
365       case PIPE_SHADER_CAP_INT16:
366       case PIPE_SHADER_CAP_GLSL_16BIT_CONSTS:
367       case PIPE_SHADER_CAP_DROUND_SUPPORTED:
368       case PIPE_SHADER_CAP_DFRACEXP_DLDEXP_SUPPORTED:
369       case PIPE_SHADER_CAP_LDEXP_SUPPORTED:
370       case PIPE_SHADER_CAP_TGSI_ANY_INOUT_DECL_RANGE:
371       case PIPE_SHADER_CAP_MAX_SHADER_BUFFERS:
372       case PIPE_SHADER_CAP_MAX_SHADER_IMAGES:
373       case PIPE_SHADER_CAP_MAX_HW_ATOMIC_COUNTERS:
374       case PIPE_SHADER_CAP_MAX_HW_ATOMIC_COUNTER_BUFFERS:
375          return 0;
376       case PIPE_SHADER_CAP_SUPPORTED_IRS:
377          return (1 << PIPE_SHADER_IR_NIR) | (1 << PIPE_SHADER_IR_TGSI);
378       default:
379          debug_printf("unknown vertex shader param %d\n", param);
380          return 0;
381       }
382       break;
383    case PIPE_SHADER_FRAGMENT:
384       switch (param) {
385       case PIPE_SHADER_CAP_MAX_INSTRUCTIONS:
386       case PIPE_SHADER_CAP_MAX_ALU_INSTRUCTIONS:
387       case PIPE_SHADER_CAP_MAX_TEX_INSTRUCTIONS:
388       case PIPE_SHADER_CAP_MAX_TEX_INDIRECTIONS:
389          return 4096;
390       case PIPE_SHADER_CAP_MAX_CONTROL_FLOW_DEPTH:
391          return 0;
392       case PIPE_SHADER_CAP_MAX_INPUTS:
393          return 8; /* should be possible to do 10 with nv4x */
394       case PIPE_SHADER_CAP_MAX_OUTPUTS:
395          return 4;
396       case PIPE_SHADER_CAP_MAX_CONST_BUFFER0_SIZE:
397          return ((eng3d->oclass >= NV40_3D_CLASS) ? 224 : 32) * sizeof(float[4]);
398       case PIPE_SHADER_CAP_MAX_CONST_BUFFERS:
399          return 1;
400       case PIPE_SHADER_CAP_MAX_TEMPS:
401          return 32;
402       case PIPE_SHADER_CAP_MAX_TEXTURE_SAMPLERS:
403       case PIPE_SHADER_CAP_MAX_SAMPLER_VIEWS:
404          return 16;
405       case PIPE_SHADER_CAP_PREFERRED_IR:
406          return (NOUVEAU_DEBUG & NOUVEAU_DEBUG_USE_TGSI) ? PIPE_SHADER_IR_TGSI : PIPE_SHADER_IR_NIR;
407       case PIPE_SHADER_CAP_CONT_SUPPORTED:
408       case PIPE_SHADER_CAP_TGSI_SQRT_SUPPORTED:
409       case PIPE_SHADER_CAP_INDIRECT_INPUT_ADDR:
410       case PIPE_SHADER_CAP_INDIRECT_OUTPUT_ADDR:
411       case PIPE_SHADER_CAP_INDIRECT_TEMP_ADDR:
412       case PIPE_SHADER_CAP_INDIRECT_CONST_ADDR:
413       case PIPE_SHADER_CAP_SUBROUTINES:
414       case PIPE_SHADER_CAP_INTEGERS:
415       case PIPE_SHADER_CAP_FP16:
416       case PIPE_SHADER_CAP_FP16_DERIVATIVES:
417       case PIPE_SHADER_CAP_FP16_CONST_BUFFERS:
418       case PIPE_SHADER_CAP_INT16:
419       case PIPE_SHADER_CAP_GLSL_16BIT_CONSTS:
420       case PIPE_SHADER_CAP_DROUND_SUPPORTED:
421       case PIPE_SHADER_CAP_DFRACEXP_DLDEXP_SUPPORTED:
422       case PIPE_SHADER_CAP_LDEXP_SUPPORTED:
423       case PIPE_SHADER_CAP_TGSI_ANY_INOUT_DECL_RANGE:
424       case PIPE_SHADER_CAP_MAX_SHADER_BUFFERS:
425       case PIPE_SHADER_CAP_MAX_SHADER_IMAGES:
426       case PIPE_SHADER_CAP_MAX_HW_ATOMIC_COUNTERS:
427       case PIPE_SHADER_CAP_MAX_HW_ATOMIC_COUNTER_BUFFERS:
428          return 0;
429       case PIPE_SHADER_CAP_SUPPORTED_IRS:
430          return (1 << PIPE_SHADER_IR_NIR) | (1 << PIPE_SHADER_IR_TGSI);
431       default:
432          debug_printf("unknown fragment shader param %d\n", param);
433          return 0;
434       }
435       break;
436    default:
437       return 0;
438    }
439 }
440 
441 static bool
nv30_screen_is_format_supported(struct pipe_screen * pscreen,enum pipe_format format,enum pipe_texture_target target,unsigned sample_count,unsigned storage_sample_count,unsigned bindings)442 nv30_screen_is_format_supported(struct pipe_screen *pscreen,
443                                 enum pipe_format format,
444                                 enum pipe_texture_target target,
445                                 unsigned sample_count,
446                                 unsigned storage_sample_count,
447                                 unsigned bindings)
448 {
449    if (sample_count > nv30_screen(pscreen)->max_sample_count)
450       return false;
451 
452    if (!(0x00000017 & (1 << sample_count)))
453       return false;
454 
455    if (MAX2(1, sample_count) != MAX2(1, storage_sample_count))
456       return false;
457 
458    /* No way to render to a swizzled 3d texture. We don't necessarily know if
459     * it's swizzled or not here, but we have to assume anyways.
460     */
461    if (target == PIPE_TEXTURE_3D && (bindings & PIPE_BIND_RENDER_TARGET))
462       return false;
463 
464    /* shared is always supported */
465    bindings &= ~PIPE_BIND_SHARED;
466 
467    if (bindings & PIPE_BIND_INDEX_BUFFER) {
468       if (format != PIPE_FORMAT_R8_UINT &&
469           format != PIPE_FORMAT_R16_UINT &&
470           format != PIPE_FORMAT_R32_UINT)
471          return false;
472       bindings &= ~PIPE_BIND_INDEX_BUFFER;
473    }
474 
475    return (nv30_format_info(pscreen, format)->bindings & bindings) == bindings;
476 }
477 
478 static const nir_shader_compiler_options nv30_base_compiler_options = {
479    .fuse_ffma32 = true,
480    .fuse_ffma64 = true,
481    .lower_bitops = true,
482    .lower_extract_byte = true,
483    .lower_extract_word = true,
484    .lower_fdiv = true,
485    .lower_fsat = true,
486    .lower_insert_byte = true,
487    .lower_insert_word = true,
488    .lower_fdph = true,
489    .lower_flrp32 = true,
490    .lower_flrp64 = true,
491    .lower_fmod = true,
492    .lower_fpow = true, /* In hardware as of nv40 FS */
493    .lower_rotate = true,
494    .lower_uniforms_to_ubo = true,
495    .lower_vector_cmp = true,
496    .force_indirect_unrolling = nir_var_all,
497    .force_indirect_unrolling_sampler = true,
498    .max_unroll_iterations = 32,
499    .no_integers = true,
500 
501    .use_interpolated_input_intrinsics = true,
502 };
503 
504 static const void *
nv30_screen_get_compiler_options(struct pipe_screen * pscreen,enum pipe_shader_ir ir,enum pipe_shader_type shader)505 nv30_screen_get_compiler_options(struct pipe_screen *pscreen,
506                                  enum pipe_shader_ir ir,
507                                  enum pipe_shader_type shader)
508 {
509    struct nv30_screen *screen = nv30_screen(pscreen);
510    assert(ir == PIPE_SHADER_IR_NIR);
511 
512    /* The FS compiler options are different between nv30 and nv40, and are set
513     * up at screen creation time.
514     */
515    if (shader == PIPE_SHADER_FRAGMENT)
516       return &screen->fs_compiler_options;
517 
518    return &nv30_base_compiler_options;
519 }
520 
521 static void
nv30_screen_fence_emit(struct pipe_screen * pscreen,uint32_t * sequence)522 nv30_screen_fence_emit(struct pipe_screen *pscreen, uint32_t *sequence)
523 {
524    struct nv30_screen *screen = nv30_screen(pscreen);
525    struct nouveau_pushbuf *push = screen->base.pushbuf;
526 
527    *sequence = ++screen->base.fence.sequence;
528 
529    assert(PUSH_AVAIL(push) + push->rsvd_kick >= 3);
530    PUSH_DATA (push, NV30_3D_FENCE_OFFSET |
531               (2 /* size */ << 18) | (7 /* subchan */ << 13));
532    PUSH_DATA (push, 0);
533    PUSH_DATA (push, *sequence);
534 }
535 
536 static uint32_t
nv30_screen_fence_update(struct pipe_screen * pscreen)537 nv30_screen_fence_update(struct pipe_screen *pscreen)
538 {
539    struct nv30_screen *screen = nv30_screen(pscreen);
540    struct nv04_notify *fence = screen->fence->data;
541    return *(uint32_t *)((char *)screen->notify->map + fence->offset);
542 }
543 
544 static void
nv30_screen_destroy(struct pipe_screen * pscreen)545 nv30_screen_destroy(struct pipe_screen *pscreen)
546 {
547    struct nv30_screen *screen = nv30_screen(pscreen);
548 
549    if (!nouveau_drm_screen_unref(&screen->base))
550       return;
551 
552    nouveau_fence_cleanup(&screen->base);
553 
554    nouveau_bo_ref(NULL, &screen->notify);
555 
556    nouveau_heap_destroy(&screen->query_heap);
557    nouveau_heap_destroy(&screen->vp_exec_heap);
558    nouveau_heap_destroy(&screen->vp_data_heap);
559 
560    nouveau_object_del(&screen->query);
561    nouveau_object_del(&screen->fence);
562    nouveau_object_del(&screen->ntfy);
563 
564    nouveau_object_del(&screen->sifm);
565    nouveau_object_del(&screen->swzsurf);
566    nouveau_object_del(&screen->surf2d);
567    nouveau_object_del(&screen->m2mf);
568    nouveau_object_del(&screen->eng3d);
569    nouveau_object_del(&screen->null);
570 
571    nouveau_screen_fini(&screen->base);
572    FREE(screen);
573 }
574 
575 #define FAIL_SCREEN_INIT(str, err)                    \
576    do {                                               \
577       NOUVEAU_ERR(str, err);                          \
578       screen->base.base.context_create = NULL;        \
579       return &screen->base;                           \
580    } while(0)
581 
582 struct nouveau_screen *
nv30_screen_create(struct nouveau_device * dev)583 nv30_screen_create(struct nouveau_device *dev)
584 {
585    struct nv30_screen *screen;
586    struct pipe_screen *pscreen;
587    struct nouveau_pushbuf *push;
588    struct nv04_fifo *fifo;
589    unsigned oclass = 0;
590    int ret, i;
591 
592    switch (dev->chipset & 0xf0) {
593    case 0x30:
594       if (RANKINE_0397_CHIPSET & (1 << (dev->chipset & 0x0f)))
595          oclass = NV30_3D_CLASS;
596       else
597       if (RANKINE_0697_CHIPSET & (1 << (dev->chipset & 0x0f)))
598          oclass = NV34_3D_CLASS;
599       else
600       if (RANKINE_0497_CHIPSET & (1 << (dev->chipset & 0x0f)))
601          oclass = NV35_3D_CLASS;
602       break;
603    case 0x40:
604       if (CURIE_4097_CHIPSET & (1 << (dev->chipset & 0x0f)))
605          oclass = NV40_3D_CLASS;
606       else
607       if (CURIE_4497_CHIPSET & (1 << (dev->chipset & 0x0f)))
608          oclass = NV44_3D_CLASS;
609       break;
610    case 0x60:
611       if (CURIE_4497_CHIPSET6X & (1 << (dev->chipset & 0x0f)))
612          oclass = NV44_3D_CLASS;
613       break;
614    default:
615       break;
616    }
617 
618    if (!oclass) {
619       NOUVEAU_ERR("unknown 3d class for 0x%02x\n", dev->chipset);
620       return NULL;
621    }
622 
623    screen = CALLOC_STRUCT(nv30_screen);
624    if (!screen)
625       return NULL;
626 
627    pscreen = &screen->base.base;
628    pscreen->destroy = nv30_screen_destroy;
629 
630    /*
631     * Some modern apps try to use msaa without keeping in mind the
632     * restrictions on videomem of older cards. Resulting in dmesg saying:
633     * [ 1197.850642] nouveau E[soffice.bin[3785]] fail ttm_validate
634     * [ 1197.850648] nouveau E[soffice.bin[3785]] validating bo list
635     * [ 1197.850654] nouveau E[soffice.bin[3785]] validate: -12
636     *
637     * Because we are running out of video memory, after which the program
638     * using the msaa visual freezes, and eventually the entire system freezes.
639     *
640     * To work around this we do not allow msaa visauls by default and allow
641     * the user to override this via NV30_MAX_MSAA.
642     */
643    screen->max_sample_count = debug_get_num_option("NV30_MAX_MSAA", 0);
644    if (screen->max_sample_count > 4)
645       screen->max_sample_count = 4;
646 
647    pscreen->get_param = nv30_screen_get_param;
648    pscreen->get_paramf = nv30_screen_get_paramf;
649    pscreen->get_shader_param = nv30_screen_get_shader_param;
650    pscreen->context_create = nv30_context_create;
651    pscreen->is_format_supported = nv30_screen_is_format_supported;
652    pscreen->get_compiler_options = nv30_screen_get_compiler_options;
653 
654    nv30_resource_screen_init(pscreen);
655    nouveau_screen_init_vdec(&screen->base);
656 
657    screen->base.fence.emit = nv30_screen_fence_emit;
658    screen->base.fence.update = nv30_screen_fence_update;
659 
660    ret = nouveau_screen_init(&screen->base, dev);
661    if (ret)
662       FAIL_SCREEN_INIT("nv30_screen_init failed: %d\n", ret);
663 
664    screen->base.vidmem_bindings |= PIPE_BIND_VERTEX_BUFFER;
665    screen->base.sysmem_bindings |= PIPE_BIND_VERTEX_BUFFER;
666    if (oclass == NV40_3D_CLASS) {
667       screen->base.vidmem_bindings |= PIPE_BIND_INDEX_BUFFER;
668       screen->base.sysmem_bindings |= PIPE_BIND_INDEX_BUFFER;
669    }
670 
671    screen->fs_compiler_options = nv30_base_compiler_options;
672    screen->fs_compiler_options.lower_fsat = false;
673    if (oclass >= NV40_3D_CLASS)
674       screen->fs_compiler_options.lower_fpow = false;
675 
676    fifo = screen->base.channel->data;
677    push = screen->base.pushbuf;
678    push->rsvd_kick = 16;
679 
680    ret = nouveau_object_new(screen->base.channel, 0x00000000, NV01_NULL_CLASS,
681                             NULL, 0, &screen->null);
682    if (ret)
683       FAIL_SCREEN_INIT("error allocating null object: %d\n", ret);
684 
685    /* DMA_FENCE refuses to accept DMA objects with "adjust" filled in,
686     * this means that the address pointed at by the DMA object must
687     * be 4KiB aligned, which means this object needs to be the first
688     * one allocated on the channel.
689     */
690    ret = nouveau_object_new(screen->base.channel, 0xbeef1e00,
691                             NOUVEAU_NOTIFIER_CLASS, &(struct nv04_notify) {
692                             .length = 32 }, sizeof(struct nv04_notify),
693                             &screen->fence);
694    if (ret)
695       FAIL_SCREEN_INIT("error allocating fence notifier: %d\n", ret);
696 
697    /* DMA_NOTIFY object, we don't actually use this but M2MF fails without */
698    ret = nouveau_object_new(screen->base.channel, 0xbeef0301,
699                             NOUVEAU_NOTIFIER_CLASS, &(struct nv04_notify) {
700                             .length = 32 }, sizeof(struct nv04_notify),
701                             &screen->ntfy);
702    if (ret)
703       FAIL_SCREEN_INIT("error allocating sync notifier: %d\n", ret);
704 
705    /* DMA_QUERY, used to implement occlusion queries, we attempt to allocate
706     * the remainder of the "notifier block" assigned by the kernel for
707     * use as query objects
708     */
709    ret = nouveau_object_new(screen->base.channel, 0xbeef0351,
710                             NOUVEAU_NOTIFIER_CLASS, &(struct nv04_notify) {
711                             .length = 4096 - 128 }, sizeof(struct nv04_notify),
712                             &screen->query);
713    if (ret)
714       FAIL_SCREEN_INIT("error allocating query notifier: %d\n", ret);
715 
716    ret = nouveau_heap_init(&screen->query_heap, 0, 4096 - 128);
717    if (ret)
718       FAIL_SCREEN_INIT("error creating query heap: %d\n", ret);
719 
720    list_inithead(&screen->queries);
721 
722    /* Vertex program resources (code/data), currently 6 of the constant
723     * slots are reserved to implement user clipping planes
724     */
725    if (oclass < NV40_3D_CLASS) {
726       nouveau_heap_init(&screen->vp_exec_heap, 0, 256);
727       nouveau_heap_init(&screen->vp_data_heap, 6, 256 - 6);
728    } else {
729       nouveau_heap_init(&screen->vp_exec_heap, 0, 512);
730       nouveau_heap_init(&screen->vp_data_heap, 6, 468 - 6);
731    }
732 
733    ret = nouveau_bo_wrap(screen->base.device, fifo->notify, &screen->notify);
734    if (ret == 0)
735       ret = nouveau_bo_map(screen->notify, 0, screen->base.client);
736    if (ret)
737       FAIL_SCREEN_INIT("error mapping notifier memory: %d\n", ret);
738 
739    ret = nouveau_object_new(screen->base.channel, 0xbeef3097, oclass,
740                             NULL, 0, &screen->eng3d);
741    if (ret)
742       FAIL_SCREEN_INIT("error allocating 3d object: %d\n", ret);
743 
744    BEGIN_NV04(push, NV01_SUBC(3D, OBJECT), 1);
745    PUSH_DATA (push, screen->eng3d->handle);
746    BEGIN_NV04(push, NV30_3D(DMA_NOTIFY), 13);
747    PUSH_DATA (push, screen->ntfy->handle);
748    PUSH_DATA (push, fifo->vram);     /* TEXTURE0 */
749    PUSH_DATA (push, fifo->gart);     /* TEXTURE1 */
750    PUSH_DATA (push, fifo->vram);     /* COLOR1 */
751    PUSH_DATA (push, screen->null->handle);  /* UNK190 */
752    PUSH_DATA (push, fifo->vram);     /* COLOR0 */
753    PUSH_DATA (push, fifo->vram);     /* ZETA */
754    PUSH_DATA (push, fifo->vram);     /* VTXBUF0 */
755    PUSH_DATA (push, fifo->gart);     /* VTXBUF1 */
756    PUSH_DATA (push, screen->fence->handle);  /* FENCE */
757    PUSH_DATA (push, screen->query->handle);  /* QUERY - intr 0x80 if nullobj */
758    PUSH_DATA (push, screen->null->handle);  /* UNK1AC */
759    PUSH_DATA (push, screen->null->handle);  /* UNK1B0 */
760    if (screen->eng3d->oclass < NV40_3D_CLASS) {
761       BEGIN_NV04(push, SUBC_3D(0x03b0), 1);
762       PUSH_DATA (push, 0x00100000);
763       BEGIN_NV04(push, SUBC_3D(0x1d80), 1);
764       PUSH_DATA (push, 3);
765 
766       BEGIN_NV04(push, SUBC_3D(0x1e98), 1);
767       PUSH_DATA (push, 0);
768       BEGIN_NV04(push, SUBC_3D(0x17e0), 3);
769       PUSH_DATA (push, fui(0.0));
770       PUSH_DATA (push, fui(0.0));
771       PUSH_DATA (push, fui(1.0));
772       BEGIN_NV04(push, SUBC_3D(0x1f80), 16);
773       for (i = 0; i < 16; i++)
774          PUSH_DATA (push, (i == 8) ? 0x0000ffff : 0);
775 
776       BEGIN_NV04(push, NV30_3D(RC_ENABLE), 1);
777       PUSH_DATA (push, 0);
778    } else {
779       BEGIN_NV04(push, NV40_3D(DMA_COLOR2), 2);
780       PUSH_DATA (push, fifo->vram);
781       PUSH_DATA (push, fifo->vram);  /* COLOR3 */
782 
783       BEGIN_NV04(push, SUBC_3D(0x1450), 1);
784       PUSH_DATA (push, 0x00000004);
785 
786       BEGIN_NV04(push, SUBC_3D(0x1ea4), 3); /* ZCULL */
787       PUSH_DATA (push, 0x00000010);
788       PUSH_DATA (push, 0x01000100);
789       PUSH_DATA (push, 0xff800006);
790 
791       /* vtxprog output routing */
792       BEGIN_NV04(push, SUBC_3D(0x1fc4), 1);
793       PUSH_DATA (push, 0x06144321);
794       BEGIN_NV04(push, SUBC_3D(0x1fc8), 2);
795       PUSH_DATA (push, 0xedcba987);
796       PUSH_DATA (push, 0x0000006f);
797       BEGIN_NV04(push, SUBC_3D(0x1fd0), 1);
798       PUSH_DATA (push, 0x00171615);
799       BEGIN_NV04(push, SUBC_3D(0x1fd4), 1);
800       PUSH_DATA (push, 0x001b1a19);
801 
802       BEGIN_NV04(push, SUBC_3D(0x1ef8), 1);
803       PUSH_DATA (push, 0x0020ffff);
804       BEGIN_NV04(push, SUBC_3D(0x1d64), 1);
805       PUSH_DATA (push, 0x01d300d4);
806 
807       BEGIN_NV04(push, NV40_3D(MIPMAP_ROUNDING), 1);
808       PUSH_DATA (push, NV40_3D_MIPMAP_ROUNDING_MODE_DOWN);
809    }
810 
811    ret = nouveau_object_new(screen->base.channel, 0xbeef3901, NV03_M2MF_CLASS,
812                             NULL, 0, &screen->m2mf);
813    if (ret)
814       FAIL_SCREEN_INIT("error allocating m2mf object: %d\n", ret);
815 
816    BEGIN_NV04(push, NV01_SUBC(M2MF, OBJECT), 1);
817    PUSH_DATA (push, screen->m2mf->handle);
818    BEGIN_NV04(push, NV03_M2MF(DMA_NOTIFY), 1);
819    PUSH_DATA (push, screen->ntfy->handle);
820 
821    ret = nouveau_object_new(screen->base.channel, 0xbeef6201,
822                             NV10_SURFACE_2D_CLASS, NULL, 0, &screen->surf2d);
823    if (ret)
824       FAIL_SCREEN_INIT("error allocating surf2d object: %d\n", ret);
825 
826    BEGIN_NV04(push, NV01_SUBC(SF2D, OBJECT), 1);
827    PUSH_DATA (push, screen->surf2d->handle);
828    BEGIN_NV04(push, NV04_SF2D(DMA_NOTIFY), 1);
829    PUSH_DATA (push, screen->ntfy->handle);
830 
831    if (dev->chipset < 0x40)
832       oclass = NV30_SURFACE_SWZ_CLASS;
833    else
834       oclass = NV40_SURFACE_SWZ_CLASS;
835 
836    ret = nouveau_object_new(screen->base.channel, 0xbeef5201, oclass,
837                             NULL, 0, &screen->swzsurf);
838    if (ret)
839       FAIL_SCREEN_INIT("error allocating swizzled surface object: %d\n", ret);
840 
841    BEGIN_NV04(push, NV01_SUBC(SSWZ, OBJECT), 1);
842    PUSH_DATA (push, screen->swzsurf->handle);
843    BEGIN_NV04(push, NV04_SSWZ(DMA_NOTIFY), 1);
844    PUSH_DATA (push, screen->ntfy->handle);
845 
846    if (dev->chipset < 0x40)
847       oclass = NV30_SIFM_CLASS;
848    else
849       oclass = NV40_SIFM_CLASS;
850 
851    ret = nouveau_object_new(screen->base.channel, 0xbeef7701, oclass,
852                             NULL, 0, &screen->sifm);
853    if (ret)
854       FAIL_SCREEN_INIT("error allocating scaled image object: %d\n", ret);
855 
856    BEGIN_NV04(push, NV01_SUBC(SIFM, OBJECT), 1);
857    PUSH_DATA (push, screen->sifm->handle);
858    BEGIN_NV04(push, NV03_SIFM(DMA_NOTIFY), 1);
859    PUSH_DATA (push, screen->ntfy->handle);
860    BEGIN_NV04(push, NV05_SIFM(COLOR_CONVERSION), 1);
861    PUSH_DATA (push, NV05_SIFM_COLOR_CONVERSION_TRUNCATE);
862 
863    nouveau_pushbuf_kick(push, push->channel);
864 
865    nouveau_fence_new(&screen->base, &screen->base.fence.current);
866    return &screen->base;
867 }
868