2 * Copyright 2016 Intel Corporation
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:
11 * The above copyright notice and this permission notice (including the next
12 * paragraph) shall be included in all copies or substantial portions of the
15 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
16 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
17 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
18 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
19 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
20 * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
26 #define __gen_address_type uint64_t
27 #define __gen_user_data void
29 static inline uint64_t
30 __gen_combine_address(void *data
, void *loc
, uint64_t addr
, uint32_t delta
)
35 #include "genxml/gen_macros.h"
36 #include "genxml/genX_pack.h"
40 #define __PASTE2(x, y) x ## y
41 #define __PASTE(x, y) __PASTE2(x, y)
42 #define isl_genX(x) __PASTE(isl_, genX(x))
45 static const uint8_t isl_to_gen_halign
[] = {
51 static const uint8_t isl_to_gen_halign
[] = {
58 static const uint8_t isl_to_gen_valign
[] = {
64 static const uint8_t isl_to_gen_valign
[] = {
71 static const uint8_t isl_to_gen_tiling
[] = {
72 [ISL_TILING_LINEAR
] = LINEAR
,
73 [ISL_TILING_X
] = XMAJOR
,
74 [ISL_TILING_Y0
] = YMAJOR
,
75 [ISL_TILING_Yf
] = YMAJOR
,
76 [ISL_TILING_Ys
] = YMAJOR
,
77 [ISL_TILING_W
] = WMAJOR
,
82 static const uint32_t isl_to_gen_multisample_layout
[] = {
83 [ISL_MSAA_LAYOUT_NONE
] = MSFMT_MSS
,
84 [ISL_MSAA_LAYOUT_INTERLEAVED
] = MSFMT_DEPTH_STENCIL
,
85 [ISL_MSAA_LAYOUT_ARRAY
] = MSFMT_MSS
,
90 static const uint32_t isl_to_gen_aux_mode
[] = {
91 [ISL_AUX_USAGE_NONE
] = AUX_NONE
,
92 [ISL_AUX_USAGE_HIZ
] = AUX_HIZ
,
93 [ISL_AUX_USAGE_MCS
] = AUX_CCS_D
,
94 [ISL_AUX_USAGE_CCS_D
] = AUX_CCS_D
,
95 [ISL_AUX_USAGE_CCS_E
] = AUX_CCS_E
,
98 static const uint32_t isl_to_gen_aux_mode
[] = {
99 [ISL_AUX_USAGE_NONE
] = AUX_NONE
,
100 [ISL_AUX_USAGE_HIZ
] = AUX_HIZ
,
101 [ISL_AUX_USAGE_MCS
] = AUX_MCS
,
102 [ISL_AUX_USAGE_CCS_D
] = AUX_MCS
,
107 get_surftype(enum isl_surf_dim dim
, isl_surf_usage_flags_t usage
)
111 unreachable("bad isl_surf_dim");
112 case ISL_SURF_DIM_1D
:
113 assert(!(usage
& ISL_SURF_USAGE_CUBE_BIT
));
115 case ISL_SURF_DIM_2D
:
116 if ((usage
& ISL_SURF_USAGE_CUBE_BIT
) &&
117 (usage
& ISL_SURF_USAGE_TEXTURE_BIT
)) {
118 /* We need SURFTYPE_CUBE to make cube sampling work */
119 return SURFTYPE_CUBE
;
121 /* Everything else (render and storage) treat cubes as plain
126 case ISL_SURF_DIM_3D
:
127 assert(!(usage
& ISL_SURF_USAGE_CUBE_BIT
));
133 * Get the horizontal and vertical alignment in the units expected by the
134 * hardware. Note that this does NOT give you the actual hardware enum values
135 * but an index into the isl_to_gen_[hv]align arrays above.
137 static inline struct isl_extent3d
138 get_image_alignment(const struct isl_surf
*surf
)
141 if (isl_tiling_is_std_y(surf
->tiling
) ||
142 surf
->dim_layout
== ISL_DIM_LAYOUT_GEN9_1D
) {
143 /* The hardware ignores the alignment values. Anyway, the surface's
144 * true alignment is likely outside the enum range of HALIGN* and
147 return isl_extent3d(4, 4, 1);
149 /* In Skylake, RENDER_SUFFACE_STATE.SurfaceVerticalAlignment is in units
150 * of surface elements (not pixels nor samples). For compressed formats,
151 * a "surface element" is defined as a compression block. For example,
152 * if SurfaceVerticalAlignment is VALIGN_4 and SurfaceFormat is an ETC2
153 * format (ETC2 has a block height of 4), then the vertical alignment is
154 * 4 compression blocks or, equivalently, 16 pixels.
156 return isl_surf_get_image_alignment_el(surf
);
159 /* Pre-Skylake, RENDER_SUFFACE_STATE.SurfaceVerticalAlignment is in
160 * units of surface samples. For example, if SurfaceVerticalAlignment
161 * is VALIGN_4 and the surface is singlesampled, then for any surface
162 * format (compressed or not) the vertical alignment is
165 return isl_surf_get_image_alignment_sa(surf
);
171 get_qpitch(const struct isl_surf
*surf
)
173 switch (surf
->dim_layout
) {
175 unreachable("Bad isl_surf_dim");
176 case ISL_DIM_LAYOUT_GEN4_2D
:
178 if (surf
->dim
== ISL_SURF_DIM_3D
&& surf
->tiling
== ISL_TILING_W
) {
179 /* This is rather annoying and completely undocumented. It
180 * appears that the hardware has a bug (or undocumented feature)
181 * regarding stencil buffers most likely related to the way
182 * W-tiling is handled as modified Y-tiling. If you bind a 3-D
183 * stencil buffer normally, and use texelFetch on it, the z or
184 * array index will get implicitly multiplied by 2 for no obvious
185 * reason. The fix appears to be to divide qpitch by 2 for
188 return isl_surf_get_array_pitch_el_rows(surf
) / 2;
190 return isl_surf_get_array_pitch_el_rows(surf
);
193 /* From the Broadwell PRM for RENDER_SURFACE_STATE.QPitch
195 * "This field must be set to an integer multiple of the Surface
196 * Vertical Alignment. For compressed textures (BC*, FXT1,
197 * ETC*, and EAC* Surface Formats), this field is in units of
198 * rows in the uncompressed surface, and must be set to an
199 * integer multiple of the vertical alignment parameter "j"
200 * defined in the Common Surface Formats section."
202 return isl_surf_get_array_pitch_sa_rows(surf
);
204 case ISL_DIM_LAYOUT_GEN9_1D
:
205 /* QPitch is usually expressed as rows of surface elements (where
206 * a surface element is an compression block or a single surface
207 * sample). Skylake 1D is an outlier.
209 * From the Skylake BSpec >> Memory Views >> Common Surface
210 * Formats >> Surface Layout and Tiling >> 1D Surfaces:
212 * Surface QPitch specifies the distance in pixels between array
215 return isl_surf_get_array_pitch_el(surf
);
216 case ISL_DIM_LAYOUT_GEN4_3D
:
217 /* QPitch doesn't make sense for ISL_DIM_LAYOUT_GEN4_3D since it uses a
218 * different pitch at each LOD. Also, the QPitch field is ignored for
219 * these surfaces. From the Broadwell PRM documentation for QPitch:
221 * This field specifies the distance in rows between array slices. It
222 * is used only in the following cases:
223 * - Surface Array is enabled OR
224 * - Number of Mulitsamples is not NUMSAMPLES_1 and Multisampled
225 * Surface Storage Format set to MSFMT_MSS OR
226 * - Surface Type is SURFTYPE_CUBE
228 * None of the three conditions above can possibly apply to a 3D surface
229 * so it is safe to just set QPitch to 0.
234 #endif /* GEN_GEN >= 8 */
237 isl_genX(surf_fill_state_s
)(const struct isl_device
*dev
, void *state
,
238 const struct isl_surf_fill_state_info
*restrict info
)
240 struct GENX(RENDER_SURFACE_STATE
) s
= { 0 };
242 s
.SurfaceType
= get_surftype(info
->surf
->dim
, info
->view
->usage
);
244 if (info
->view
->usage
& ISL_SURF_USAGE_RENDER_TARGET_BIT
)
245 assert(isl_format_supports_rendering(dev
->info
, info
->view
->format
));
246 else if (info
->view
->usage
& ISL_SURF_USAGE_TEXTURE_BIT
)
247 assert(isl_format_supports_sampling(dev
->info
, info
->view
->format
));
249 /* From the Sky Lake PRM Vol. 2d, RENDER_SURFACE_STATE::SurfaceFormat
251 * This field cannot be a compressed (BC*, DXT*, FXT*, ETC*, EAC*)
252 * format if the Surface Type is SURFTYPE_1D
254 if (info
->surf
->dim
== ISL_SURF_DIM_1D
)
255 assert(!isl_format_is_compressed(info
->view
->format
));
257 s
.SurfaceFormat
= info
->view
->format
;
260 s
.ColorBufferComponentWriteDisables
= info
->write_disables
;
262 assert(info
->write_disables
== 0);
266 s
.IntegerSurfaceFormat
= isl_format_has_int_channel(s
.SurfaceFormat
);
269 assert(info
->surf
->logical_level0_px
.width
> 0 &&
270 info
->surf
->logical_level0_px
.height
> 0);
272 s
.Width
= info
->surf
->logical_level0_px
.width
- 1;
273 s
.Height
= info
->surf
->logical_level0_px
.height
- 1;
275 /* In the gen6 PRM Volume 1 Part 1: Graphics Core, Section 7.18.3.7.1
276 * (Surface Arrays For all surfaces other than separate stencil buffer):
278 * "[DevSNB] Errata: Sampler MSAA Qpitch will be 4 greater than the value
279 * calculated in the equation above , for every other odd Surface Height
280 * starting from 1 i.e. 1,5,9,13"
282 * Since this Qpitch errata only impacts the sampler, we have to adjust the
283 * input for the rendering surface to achieve the same qpitch. For the
284 * affected heights, we increment the height by 1 for the rendering
287 if (GEN_GEN
== 6 && (info
->view
->usage
& ISL_SURF_USAGE_RENDER_TARGET_BIT
) &&
288 info
->surf
->samples
> 1 &&
289 (info
->surf
->logical_level0_px
.height
% 4) == 1)
292 switch (s
.SurfaceType
) {
295 /* From the Ivy Bridge PRM >> RENDER_SURFACE_STATE::MinimumArrayElement:
297 * "If Number of Multisamples is not MULTISAMPLECOUNT_1, this field
298 * must be set to zero if this surface is used with sampling engine
301 * This restriction appears to exist only on Ivy Bridge.
303 if (GEN_GEN
== 7 && !GEN_IS_HASWELL
&& !ISL_DEV_IS_BAYTRAIL(dev
) &&
304 (info
->view
->usage
& ISL_SURF_USAGE_TEXTURE_BIT
) &&
305 info
->surf
->samples
> 1)
306 assert(info
->view
->base_array_layer
== 0);
308 s
.MinimumArrayElement
= info
->view
->base_array_layer
;
310 /* From the Broadwell PRM >> RENDER_SURFACE_STATE::Depth:
312 * For SURFTYPE_1D, 2D, and CUBE: The range of this field is reduced
313 * by one for each increase from zero of Minimum Array Element. For
314 * example, if Minimum Array Element is set to 1024 on a 2D surface,
315 * the range of this field is reduced to [0,1023].
317 * In other words, 'Depth' is the number of array layers.
319 s
.Depth
= info
->view
->array_len
- 1;
321 /* From the Broadwell PRM >> RENDER_SURFACE_STATE::RenderTargetViewExtent:
323 * For Render Target and Typed Dataport 1D and 2D Surfaces:
324 * This field must be set to the same value as the Depth field.
326 if (info
->view
->usage
& (ISL_SURF_USAGE_RENDER_TARGET_BIT
|
327 ISL_SURF_USAGE_STORAGE_BIT
))
328 s
.RenderTargetViewExtent
= s
.Depth
;
331 s
.MinimumArrayElement
= info
->view
->base_array_layer
;
332 /* Same as SURFTYPE_2D, but divided by 6 */
333 s
.Depth
= info
->view
->array_len
/ 6 - 1;
334 if (info
->view
->usage
& (ISL_SURF_USAGE_RENDER_TARGET_BIT
|
335 ISL_SURF_USAGE_STORAGE_BIT
))
336 s
.RenderTargetViewExtent
= s
.Depth
;
339 /* From the Broadwell PRM >> RENDER_SURFACE_STATE::Depth:
341 * If the volume texture is MIP-mapped, this field specifies the
342 * depth of the base MIP level.
344 s
.Depth
= info
->surf
->logical_level0_px
.depth
- 1;
346 /* From the Broadwell PRM >> RENDER_SURFACE_STATE::RenderTargetViewExtent:
348 * For Render Target and Typed Dataport 3D Surfaces: This field
349 * indicates the extent of the accessible 'R' coordinates minus 1 on
350 * the LOD currently being rendered to.
352 * The docs specify that this only matters for render targets and
353 * surfaces used with typed dataport messages. Prior to Ivy Bridge, the
354 * Depth field has more bits than RenderTargetViewExtent so we can have
355 * textures with more levels than we can render to. In order to prevent
356 * assert-failures in the packing function below, we only set the field
357 * when it's actually going to be used by the hardware.
359 * Similaraly, the MinimumArrayElement field is ignored by all hardware
360 * prior to Sky Lake when texturing and we want it set to 0 anyway.
361 * Since it's already initialized to 0, we can just leave it alone for
364 if (info
->view
->usage
& (ISL_SURF_USAGE_RENDER_TARGET_BIT
|
365 ISL_SURF_USAGE_STORAGE_BIT
)) {
366 s
.MinimumArrayElement
= info
->view
->base_array_layer
;
367 s
.RenderTargetViewExtent
= info
->view
->array_len
- 1;
371 unreachable("bad SurfaceType");
375 s
.SurfaceArray
= info
->surf
->dim
!= ISL_SURF_DIM_3D
;
378 if (info
->view
->usage
& ISL_SURF_USAGE_RENDER_TARGET_BIT
) {
379 /* For render target surfaces, the hardware interprets field
380 * MIPCount/LOD as LOD. The Broadwell PRM says:
382 * MIPCountLOD defines the LOD that will be rendered into.
383 * SurfaceMinLOD is ignored.
385 s
.MIPCountLOD
= info
->view
->base_level
;
388 /* For non render target surfaces, the hardware interprets field
389 * MIPCount/LOD as MIPCount. The range of levels accessible by the
390 * sampler engine is [SurfaceMinLOD, SurfaceMinLOD + MIPCountLOD].
392 s
.SurfaceMinLOD
= info
->view
->base_level
;
393 s
.MIPCountLOD
= MAX(info
->view
->levels
, 1) - 1;
397 /* We don't use miptails yet. The PRM recommends that you set "Mip Tail
398 * Start LOD" to 15 to prevent the hardware from trying to use them.
400 s
.TiledResourceMode
= NONE
;
401 s
.MipTailStartLOD
= 15;
405 const struct isl_extent3d image_align
= get_image_alignment(info
->surf
);
406 s
.SurfaceVerticalAlignment
= isl_to_gen_valign
[image_align
.height
];
408 s
.SurfaceHorizontalAlignment
= isl_to_gen_halign
[image_align
.width
];
412 if (info
->surf
->dim_layout
== ISL_DIM_LAYOUT_GEN9_1D
) {
413 /* For gen9 1-D textures, surface pitch is ignored */
416 s
.SurfacePitch
= info
->surf
->row_pitch
- 1;
420 s
.SurfaceQPitch
= get_qpitch(info
->surf
) >> 2;
422 s
.SurfaceArraySpacing
= info
->surf
->array_pitch_span
==
423 ISL_ARRAY_PITCH_SPAN_COMPACT
;
427 s
.TileMode
= isl_to_gen_tiling
[info
->surf
->tiling
];
429 s
.TiledSurface
= info
->surf
->tiling
!= ISL_TILING_LINEAR
,
430 s
.TileWalk
= info
->surf
->tiling
== ISL_TILING_Y0
? TILEWALK_YMAJOR
:
435 s
.RenderCacheReadWriteMode
= WriteOnlyCache
;
437 s
.RenderCacheReadWriteMode
= 0;
440 if (info
->view
->usage
& ISL_SURF_USAGE_CUBE_BIT
) {
442 s
.CubeFaceEnablePositiveZ
= 1;
443 s
.CubeFaceEnableNegativeZ
= 1;
444 s
.CubeFaceEnablePositiveY
= 1;
445 s
.CubeFaceEnableNegativeY
= 1;
446 s
.CubeFaceEnablePositiveX
= 1;
447 s
.CubeFaceEnableNegativeX
= 1;
449 s
.CubeFaceEnables
= 0x3f;
454 s
.NumberofMultisamples
= ffs(info
->surf
->samples
) - 1;
456 s
.MultisampledSurfaceStorageFormat
=
457 isl_to_gen_multisample_layout
[info
->surf
->msaa_layout
];
461 #if (GEN_GEN >= 8 || GEN_IS_HASWELL)
462 if (info
->view
->usage
& ISL_SURF_USAGE_RENDER_TARGET_BIT
) {
463 /* From the Sky Lake PRM Vol. 2d,
464 * RENDER_SURFACE_STATE::Shader Channel Select Red
466 * "For Render Target, Red, Green and Blue Shader Channel Selects
467 * MUST be such that only valid components can be swapped i.e. only
468 * change the order of components in the pixel. Any other values for
469 * these Shader Channel Select fields are not valid for Render
470 * Targets. This also means that there MUST not be multiple shader
471 * channels mapped to the same RT channel."
473 assert(info
->view
->swizzle
.r
== ISL_CHANNEL_SELECT_RED
||
474 info
->view
->swizzle
.r
== ISL_CHANNEL_SELECT_GREEN
||
475 info
->view
->swizzle
.r
== ISL_CHANNEL_SELECT_BLUE
);
476 assert(info
->view
->swizzle
.g
== ISL_CHANNEL_SELECT_RED
||
477 info
->view
->swizzle
.g
== ISL_CHANNEL_SELECT_GREEN
||
478 info
->view
->swizzle
.g
== ISL_CHANNEL_SELECT_BLUE
);
479 assert(info
->view
->swizzle
.b
== ISL_CHANNEL_SELECT_RED
||
480 info
->view
->swizzle
.b
== ISL_CHANNEL_SELECT_GREEN
||
481 info
->view
->swizzle
.b
== ISL_CHANNEL_SELECT_BLUE
);
482 assert(info
->view
->swizzle
.r
!= info
->view
->swizzle
.g
);
483 assert(info
->view
->swizzle
.r
!= info
->view
->swizzle
.b
);
484 assert(info
->view
->swizzle
.g
!= info
->view
->swizzle
.b
);
486 /* From the Sky Lake PRM Vol. 2d,
487 * RENDER_SURFACE_STATE::Shader Channel Select Alpha
489 * "For Render Target, this field MUST be programmed to
490 * value = SCS_ALPHA."
492 assert(info
->view
->swizzle
.a
== ISL_CHANNEL_SELECT_ALPHA
);
494 s
.ShaderChannelSelectRed
= info
->view
->swizzle
.r
;
495 s
.ShaderChannelSelectGreen
= info
->view
->swizzle
.g
;
496 s
.ShaderChannelSelectBlue
= info
->view
->swizzle
.b
;
497 s
.ShaderChannelSelectAlpha
= info
->view
->swizzle
.a
;
500 s
.SurfaceBaseAddress
= info
->address
;
506 #if GEN_GEN > 4 || GEN_IS_G4X
507 if (info
->x_offset_sa
!= 0 || info
->y_offset_sa
!= 0) {
508 /* There are fairly strict rules about when the offsets can be used.
509 * These are mostly taken from the Sky Lake PRM documentation for
510 * RENDER_SURFACE_STATE.
512 assert(info
->surf
->tiling
!= ISL_TILING_LINEAR
);
513 assert(info
->surf
->dim
== ISL_SURF_DIM_2D
);
514 assert(isl_is_pow2(isl_format_get_layout(info
->view
->format
)->bpb
));
515 assert(info
->surf
->levels
== 1);
516 assert(info
->surf
->logical_level0_px
.array_len
== 1);
517 assert(info
->aux_usage
== ISL_AUX_USAGE_NONE
);
520 /* Broadwell added more rules. */
521 assert(info
->surf
->samples
== 1);
522 if (isl_format_get_layout(info
->view
->format
)->bpb
== 8)
523 assert(info
->x_offset_sa
% 16 == 0);
524 if (isl_format_get_layout(info
->view
->format
)->bpb
== 16)
525 assert(info
->x_offset_sa
% 8 == 0);
529 s
.SurfaceArray
= false;
533 const unsigned x_div
= 4;
534 const unsigned y_div
= GEN_GEN
>= 8 ? 4 : 2;
535 assert(info
->x_offset_sa
% x_div
== 0);
536 assert(info
->y_offset_sa
% y_div
== 0);
537 s
.XOffset
= info
->x_offset_sa
/ x_div
;
538 s
.YOffset
= info
->y_offset_sa
/ y_div
;
540 assert(info
->x_offset_sa
== 0);
541 assert(info
->y_offset_sa
== 0);
545 if (info
->aux_surf
&& info
->aux_usage
!= ISL_AUX_USAGE_NONE
) {
546 /* The docs don't appear to say anything whatsoever about compression
547 * and the data port. Testing seems to indicate that the data port
548 * completely ignores the AuxiliarySurfaceMode field.
550 assert(!(info
->view
->usage
& ISL_SURF_USAGE_STORAGE_BIT
));
552 struct isl_tile_info tile_info
;
553 isl_surf_get_tile_info(info
->aux_surf
, &tile_info
);
554 uint32_t pitch_in_tiles
=
555 info
->aux_surf
->row_pitch
/ tile_info
.phys_extent_B
.width
;
557 s
.AuxiliarySurfaceBaseAddress
= info
->aux_address
;
558 s
.AuxiliarySurfacePitch
= pitch_in_tiles
- 1;
561 assert(GEN_GEN
>= 9 || info
->aux_usage
!= ISL_AUX_USAGE_CCS_E
);
562 /* Auxiliary surfaces in ISL have compressed formats but the hardware
563 * doesn't expect our definition of the compression, it expects qpitch
564 * in units of samples on the main surface.
566 s
.AuxiliarySurfaceQPitch
=
567 isl_surf_get_array_pitch_sa_rows(info
->aux_surf
) >> 2;
569 if (info
->aux_usage
== ISL_AUX_USAGE_HIZ
) {
570 /* The number of samples must be 1 */
571 assert(info
->surf
->samples
== 1);
573 /* The dimension must not be 3D */
574 assert(info
->surf
->dim
!= ISL_SURF_DIM_3D
);
576 /* The format must be one of the following: */
577 switch (info
->view
->format
) {
578 case ISL_FORMAT_R32_FLOAT
:
579 case ISL_FORMAT_R24_UNORM_X8_TYPELESS
:
580 case ISL_FORMAT_R16_UNORM
:
583 assert(!"Incompatible HiZ Sampling format");
588 s
.AuxiliarySurfaceMode
= isl_to_gen_aux_mode
[info
->aux_usage
];
590 assert(info
->aux_usage
== ISL_AUX_USAGE_MCS
||
591 info
->aux_usage
== ISL_AUX_USAGE_CCS_D
);
598 /* From the CHV PRM, Volume 2d, page 321 (RENDER_SURFACE_STATE dword 0
599 * bit 9 "Sampler L2 Bypass Mode Disable" Programming Notes):
601 * This bit must be set for the following surface types: BC2_UNORM
602 * BC3_UNORM BC5_UNORM BC5_SNORM BC7_UNORM
604 if (GEN_GEN
>= 9 || dev
->info
->is_cherryview
) {
605 switch (info
->view
->format
) {
606 case ISL_FORMAT_BC2_UNORM
:
607 case ISL_FORMAT_BC3_UNORM
:
608 case ISL_FORMAT_BC5_UNORM
:
609 case ISL_FORMAT_BC5_SNORM
:
610 case ISL_FORMAT_BC7_UNORM
:
611 s
.SamplerL2BypassModeDisable
= true;
614 /* From the SKL PRM, Programming Note under Sampler Output Channel
617 * If a surface has an associated HiZ Auxilliary surface, the
618 * Sampler L2 Bypass Mode Disable field in the RENDER_SURFACE_STATE
621 if (GEN_GEN
>= 9 && info
->aux_usage
== ISL_AUX_USAGE_HIZ
)
622 s
.SamplerL2BypassModeDisable
= true;
628 if (info
->aux_usage
!= ISL_AUX_USAGE_NONE
) {
630 s
.RedClearColor
= info
->clear_color
.u32
[0];
631 s
.GreenClearColor
= info
->clear_color
.u32
[1];
632 s
.BlueClearColor
= info
->clear_color
.u32
[2];
633 s
.AlphaClearColor
= info
->clear_color
.u32
[3];
635 /* Prior to Sky Lake, we only have one bit for the clear color which
636 * gives us 0 or 1 in whatever the surface's format happens to be.
638 if (isl_format_has_int_channel(info
->view
->format
)) {
639 for (unsigned i
= 0; i
< 4; i
++) {
640 assert(info
->clear_color
.u32
[i
] == 0 ||
641 info
->clear_color
.u32
[i
] == 1);
643 s
.RedClearColor
= info
->clear_color
.u32
[0] != 0;
644 s
.GreenClearColor
= info
->clear_color
.u32
[1] != 0;
645 s
.BlueClearColor
= info
->clear_color
.u32
[2] != 0;
646 s
.AlphaClearColor
= info
->clear_color
.u32
[3] != 0;
648 for (unsigned i
= 0; i
< 4; i
++) {
649 assert(info
->clear_color
.f32
[i
] == 0.0f
||
650 info
->clear_color
.f32
[i
] == 1.0f
);
652 s
.RedClearColor
= info
->clear_color
.f32
[0] != 0.0f
;
653 s
.GreenClearColor
= info
->clear_color
.f32
[1] != 0.0f
;
654 s
.BlueClearColor
= info
->clear_color
.f32
[2] != 0.0f
;
655 s
.AlphaClearColor
= info
->clear_color
.f32
[3] != 0.0f
;
660 GENX(RENDER_SURFACE_STATE_pack
)(NULL
, state
, &s
);
664 isl_genX(buffer_fill_state_s
)(void *state
,
665 const struct isl_buffer_fill_state_info
*restrict info
)
667 uint32_t num_elements
= info
->size
/ info
->stride
;
670 /* From the IVB PRM, SURFACE_STATE::Height,
672 * For typed buffer and structured buffer surfaces, the number
673 * of entries in the buffer ranges from 1 to 2^27. For raw buffer
674 * surfaces, the number of entries in the buffer is the number of bytes
675 * which can range from 1 to 2^30.
677 if (info
->format
== ISL_FORMAT_RAW
) {
678 assert(num_elements
<= (1ull << 30));
679 assert(num_elements
> 0);
681 assert(num_elements
<= (1ull << 27));
684 assert(num_elements
<= (1ull << 27));
687 struct GENX(RENDER_SURFACE_STATE
) s
= { 0, };
689 s
.SurfaceType
= SURFTYPE_BUFFER
;
690 s
.SurfaceFormat
= info
->format
;
693 s
.SurfaceVerticalAlignment
= isl_to_gen_valign
[4];
695 s
.SurfaceHorizontalAlignment
= isl_to_gen_halign
[4];
696 s
.SurfaceArray
= false;
701 s
.Height
= ((num_elements
- 1) >> 7) & 0x3fff;
702 s
.Width
= (num_elements
- 1) & 0x7f;
703 s
.Depth
= ((num_elements
- 1) >> 21) & 0x3ff;
705 s
.Height
= ((num_elements
- 1) >> 7) & 0x1fff;
706 s
.Width
= (num_elements
- 1) & 0x7f;
707 s
.Depth
= ((num_elements
- 1) >> 20) & 0x7f;
710 s
.SurfacePitch
= info
->stride
- 1;
713 s
.NumberofMultisamples
= MULTISAMPLECOUNT_1
;
719 s
.TiledSurface
= false;
723 s
.RenderCacheReadWriteMode
= WriteOnlyCache
;
725 s
.RenderCacheReadWriteMode
= 0;
728 s
.SurfaceBaseAddress
= info
->address
;
733 #if (GEN_GEN >= 8 || GEN_IS_HASWELL)
734 s
.ShaderChannelSelectRed
= SCS_RED
;
735 s
.ShaderChannelSelectGreen
= SCS_GREEN
;
736 s
.ShaderChannelSelectBlue
= SCS_BLUE
;
737 s
.ShaderChannelSelectAlpha
= SCS_ALPHA
;
740 GENX(RENDER_SURFACE_STATE_pack
)(NULL
, state
, &s
);