/* * Copyright 2016 Intel Corporation * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the "Software"), * to deal in the Software without restriction, including without limitation * the rights to use, copy, modify, merge, publish, distribute, sublicense, * and/or sell copies of the Software, and to permit persons to whom the * Software is furnished to do so, subject to the following conditions: * * The above copyright notice and this permission notice (including the next * paragraph) shall be included in all copies or substantial portions of the * Software. * * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS * IN THE SOFTWARE. */ #include #define __gen_address_type uint64_t #define __gen_user_data void static uint64_t __gen_combine_address(__attribute__((unused)) void *data, __attribute__((unused)) void *loc, uint64_t addr, uint32_t delta) { return addr + delta; } #include "genxml/gen_macros.h" #include "genxml/genX_pack.h" #include "isl_priv.h" #if GFX_VER >= 7 static const uint8_t isl_encode_halign(uint8_t halign) { switch (halign) { #if GFX_VERx10 >= 125 case 16: return HALIGN_16; case 32: return HALIGN_32; case 64: return HALIGN_64; case 128: return HALIGN_128; #elif GFX_VER >= 8 case 4: return HALIGN_4; case 8: return HALIGN_8; case 16: return HALIGN_16; #else case 4: return HALIGN_4; case 8: return HALIGN_8; #endif default: unreachable("Invalid halign"); } } #endif #if GFX_VER >= 6 static const uint8_t isl_encode_valign(uint8_t valign) { switch (valign) { #if GFX_VER >= 8 case 4: return VALIGN_4; case 8: return VALIGN_8; case 16: return VALIGN_16; #else case 2: return VALIGN_2; case 4: return VALIGN_4; #endif default: unreachable("Invalid valign"); } } #endif #if GFX_VER >= 8 static const uint8_t isl_encode_tiling[] = { [ISL_TILING_LINEAR] = LINEAR, [ISL_TILING_X] = XMAJOR, #if GFX_VERx10 >= 125 [ISL_TILING_4] = TILE4, [ISL_TILING_64] = TILE64, #else [ISL_TILING_Y0] = YMAJOR, [ISL_TILING_Yf] = YMAJOR, [ISL_TILING_Ys] = YMAJOR, #endif #if GFX_VER <= 11 [ISL_TILING_W] = WMAJOR, #endif }; #endif #if GFX_VER >= 7 static const uint32_t isl_encode_multisample_layout[] = { [ISL_MSAA_LAYOUT_NONE] = MSFMT_MSS, [ISL_MSAA_LAYOUT_INTERLEAVED] = MSFMT_DEPTH_STENCIL, [ISL_MSAA_LAYOUT_ARRAY] = MSFMT_MSS, }; #endif #if GFX_VER >= 12 static const uint32_t isl_encode_aux_mode[] = { [ISL_AUX_USAGE_NONE] = AUX_NONE, [ISL_AUX_USAGE_MC] = AUX_NONE, [ISL_AUX_USAGE_MCS] = AUX_CCS_E, [ISL_AUX_USAGE_GFX12_CCS_E] = AUX_CCS_E, [ISL_AUX_USAGE_CCS_E] = AUX_CCS_E, [ISL_AUX_USAGE_HIZ_CCS_WT] = AUX_CCS_E, [ISL_AUX_USAGE_MCS_CCS] = AUX_MCS_LCE, [ISL_AUX_USAGE_STC_CCS] = AUX_CCS_E, }; #elif GFX_VER >= 9 static const uint32_t isl_encode_aux_mode[] = { [ISL_AUX_USAGE_NONE] = AUX_NONE, [ISL_AUX_USAGE_HIZ] = AUX_HIZ, [ISL_AUX_USAGE_MCS] = AUX_CCS_D, [ISL_AUX_USAGE_CCS_D] = AUX_CCS_D, [ISL_AUX_USAGE_CCS_E] = AUX_CCS_E, }; #elif GFX_VER >= 8 static const uint32_t isl_encode_aux_mode[] = { [ISL_AUX_USAGE_NONE] = AUX_NONE, [ISL_AUX_USAGE_HIZ] = AUX_HIZ, [ISL_AUX_USAGE_MCS] = AUX_MCS, [ISL_AUX_USAGE_CCS_D] = AUX_MCS, }; #endif static uint8_t get_surftype(enum isl_surf_dim dim, isl_surf_usage_flags_t usage) { switch (dim) { default: unreachable("bad isl_surf_dim"); case ISL_SURF_DIM_1D: assert(!(usage & ISL_SURF_USAGE_CUBE_BIT)); return SURFTYPE_1D; case ISL_SURF_DIM_2D: if ((usage & ISL_SURF_USAGE_CUBE_BIT) && (usage & ISL_SURF_USAGE_TEXTURE_BIT)) { /* We need SURFTYPE_CUBE to make cube sampling work */ return SURFTYPE_CUBE; } else { /* Everything else (render and storage) treat cubes as plain * 2D array textures */ return SURFTYPE_2D; } case ISL_SURF_DIM_3D: assert(!(usage & ISL_SURF_USAGE_CUBE_BIT)); return SURFTYPE_3D; } } /** * Get the horizontal and vertical alignment in the units expected by the * hardware. Note that this does NOT give you the actual hardware enum values * but an index into the isl_encode_[hv]align arrays above. */ UNUSED static struct isl_extent3d get_image_alignment(const struct isl_surf *surf) { if (GFX_VERx10 >= 125) { if (surf->tiling == ISL_TILING_64) { /* The hardware ignores the alignment values. Anyway, the surface's * true alignment is likely outside the enum range of HALIGN* and * VALIGN*. */ return isl_extent3d(128, 4, 1); } else if (isl_format_get_layout(surf->format)->bpb % 3 == 0) { /* On XeHP, RENDER_SURFACE_STATE.SurfaceHorizontalAlignment is in * units of elements for 24, 48, and 96 bpb formats. */ return isl_surf_get_image_alignment_el(surf); } else { /* On XeHP, RENDER_SURFACE_STATE.SurfaceHorizontalAlignment is in * units of bytes for formats that are powers of two. */ const uint32_t bs = isl_format_get_layout(surf->format)->bpb / 8; return isl_extent3d(surf->image_alignment_el.w * bs, surf->image_alignment_el.h, surf->image_alignment_el.d); } } else if (GFX_VER >= 9) { if (isl_tiling_is_std_y(surf->tiling) || surf->dim_layout == ISL_DIM_LAYOUT_GFX9_1D) { /* The hardware ignores the alignment values. Anyway, the surface's * true alignment is likely outside the enum range of HALIGN* and * VALIGN*. */ return isl_extent3d(4, 4, 1); } else { /* In Skylake, RENDER_SUFFACE_STATE.SurfaceVerticalAlignment is in units * of surface elements (not pixels nor samples). For compressed formats, * a "surface element" is defined as a compression block. For example, * if SurfaceVerticalAlignment is VALIGN_4 and SurfaceFormat is an ETC2 * format (ETC2 has a block height of 4), then the vertical alignment is * 4 compression blocks or, equivalently, 16 pixels. */ return isl_surf_get_image_alignment_el(surf); } } else { /* Pre-Skylake, RENDER_SUFFACE_STATE.SurfaceVerticalAlignment is in * units of surface samples. For example, if SurfaceVerticalAlignment * is VALIGN_4 and the surface is singlesampled, then for any surface * format (compressed or not) the vertical alignment is * 4 pixels. */ return isl_surf_get_image_alignment_sa(surf); } } #if GFX_VER >= 8 static uint32_t get_qpitch(const struct isl_surf *surf) { switch (surf->dim_layout) { default: unreachable("Bad isl_surf_dim"); case ISL_DIM_LAYOUT_GFX4_2D: if (GFX_VER >= 9) { if (surf->dim == ISL_SURF_DIM_3D && surf->tiling == ISL_TILING_W) { /* This is rather annoying and completely undocumented. It * appears that the hardware has a bug (or undocumented feature) * regarding stencil buffers most likely related to the way * W-tiling is handled as modified Y-tiling. If you bind a 3-D * stencil buffer normally, and use texelFetch on it, the z or * array index will get implicitly multiplied by 2 for no obvious * reason. The fix appears to be to divide qpitch by 2 for * W-tiled surfaces. */ return isl_surf_get_array_pitch_el_rows(surf) / 2; } else { return isl_surf_get_array_pitch_el_rows(surf); } } else { /* From the Broadwell PRM for RENDER_SURFACE_STATE.QPitch * * "This field must be set to an integer multiple of the Surface * Vertical Alignment. For compressed textures (BC*, FXT1, * ETC*, and EAC* Surface Formats), this field is in units of * rows in the uncompressed surface, and must be set to an * integer multiple of the vertical alignment parameter "j" * defined in the Common Surface Formats section." */ return isl_surf_get_array_pitch_sa_rows(surf); } case ISL_DIM_LAYOUT_GFX9_1D: /* QPitch is usually expressed as rows of surface elements (where * a surface element is an compression block or a single surface * sample). Skylake 1D is an outlier. * * From the Skylake BSpec >> Memory Views >> Common Surface * Formats >> Surface Layout and Tiling >> 1D Surfaces: * * Surface QPitch specifies the distance in pixels between array * slices. */ return isl_surf_get_array_pitch_el(surf); case ISL_DIM_LAYOUT_GFX4_3D: /* QPitch doesn't make sense for ISL_DIM_LAYOUT_GFX4_3D since it uses a * different pitch at each LOD. Also, the QPitch field is ignored for * these surfaces. From the Broadwell PRM documentation for QPitch: * * This field specifies the distance in rows between array slices. It * is used only in the following cases: * - Surface Array is enabled OR * - Number of Mulitsamples is not NUMSAMPLES_1 and Multisampled * Surface Storage Format set to MSFMT_MSS OR * - Surface Type is SURFTYPE_CUBE * * None of the three conditions above can possibly apply to a 3D surface * so it is safe to just set QPitch to 0. */ return 0; } } #endif /* GFX_VER >= 8 */ void isl_genX(surf_fill_state_s)(const struct isl_device *dev, void *state, const struct isl_surf_fill_state_info *restrict info) { struct GENX(RENDER_SURFACE_STATE) s = { 0 }; s.SurfaceType = get_surftype(info->surf->dim, info->view->usage); if (info->view->usage & ISL_SURF_USAGE_RENDER_TARGET_BIT) assert(isl_format_supports_rendering(dev->info, info->view->format)); else if (info->view->usage & ISL_SURF_USAGE_TEXTURE_BIT) assert(isl_format_supports_sampling(dev->info, info->view->format)); /* From the Sky Lake PRM Vol. 2d, RENDER_SURFACE_STATE::SurfaceFormat * * This field cannot be a compressed (BC*, DXT*, FXT*, ETC*, EAC*) * format if the Surface Type is SURFTYPE_1D */ if (info->surf->dim == ISL_SURF_DIM_1D) assert(!isl_format_is_compressed(info->view->format)); if (isl_format_is_compressed(info->surf->format)) { /* You're not allowed to make a view of a compressed format with any * format other than the surface format. None of the userspace APIs * allow for this directly and doing so would mess up a number of * surface parameters such as Width, Height, and alignments. Ideally, * we'd like to assert that the two formats match. However, we have an * S3TC workaround that requires us to do reinterpretation. So assert * that they're at least the same bpb and block size. */ ASSERTED const struct isl_format_layout *surf_fmtl = isl_format_get_layout(info->surf->format); ASSERTED const struct isl_format_layout *view_fmtl = isl_format_get_layout(info->surf->format); assert(surf_fmtl->bpb == view_fmtl->bpb); assert(surf_fmtl->bw == view_fmtl->bw); assert(surf_fmtl->bh == view_fmtl->bh); } s.SurfaceFormat = info->view->format; #if GFX_VER >= 12 /* The BSpec description of this field says: * * "This bit field, when set, indicates if the resource is created as * Depth/Stencil resource." * * "SW must set this bit for any resource that was created with * Depth/Stencil resource flag. Setting this bit allows HW to properly * interpret the data-layout for various cases. For any resource that's * created without Depth/Stencil resource flag, it must be reset." * * Even though the docs for this bit seem to imply that it's required for * anything which might have been used for depth/stencil, empirical * evidence suggests that it only affects CCS compression usage. There are * a few things which back this up: * * 1. The docs are also pretty clear that this bit was added as part * of enabling Gfx12 depth/stencil lossless compression. * * 2. The only new difference between depth/stencil and color images on * Gfx12 (where the bit was added) is how they treat CCS compression. * All other differences such as alignment requirements and MSAA layout * are already covered by other bits. * * Under these assumptions, it makes sense for ISL to model this bit as * being an extension of AuxiliarySurfaceMode where STC_CCS and HIZ_CCS_WT * are indicated by AuxiliarySurfaceMode == CCS_E and DepthStencilResource * == true. */ s.DepthStencilResource = info->aux_usage == ISL_AUX_USAGE_HIZ_CCS_WT || info->aux_usage == ISL_AUX_USAGE_STC_CCS; #endif #if GFX_VER <= 5 s.ColorBufferComponentWriteDisables = info->write_disables; s.ColorBlendEnable = info->blend_enable; #else assert(info->write_disables == 0); #endif #if GFX_VERx10 == 75 s.IntegerSurfaceFormat = isl_format_has_int_channel((enum isl_format) s.SurfaceFormat); #endif assert(info->surf->logical_level0_px.width > 0 && info->surf->logical_level0_px.height > 0); s.Width = info->surf->logical_level0_px.width - 1; s.Height = info->surf->logical_level0_px.height - 1; /* In the gfx6 PRM Volume 1 Part 1: Graphics Core, Section 7.18.3.7.1 * (Surface Arrays For all surfaces other than separate stencil buffer): * * "[DevSNB] Errata: Sampler MSAA Qpitch will be 4 greater than the value * calculated in the equation above , for every other odd Surface Height * starting from 1 i.e. 1,5,9,13" * * Since this Qpitch errata only impacts the sampler, we have to adjust the * input for the rendering surface to achieve the same qpitch. For the * affected heights, we increment the height by 1 for the rendering * surface. */ if (GFX_VER == 6 && (info->view->usage & ISL_SURF_USAGE_RENDER_TARGET_BIT) && info->surf->samples > 1 && (info->surf->logical_level0_px.height % 4) == 1) s.Height++; switch (s.SurfaceType) { case SURFTYPE_1D: case SURFTYPE_2D: /* From the Ivy Bridge PRM >> RENDER_SURFACE_STATE::MinimumArrayElement: * * "If Number of Multisamples is not MULTISAMPLECOUNT_1, this field * must be set to zero if this surface is used with sampling engine * messages." * * This restriction appears to exist only on Ivy Bridge. */ if (GFX_VERx10 == 70 && !ISL_DEV_IS_BAYTRAIL(dev) && (info->view->usage & ISL_SURF_USAGE_TEXTURE_BIT) && info->surf->samples > 1) assert(info->view->base_array_layer == 0); s.MinimumArrayElement = info->view->base_array_layer; /* From the Broadwell PRM >> RENDER_SURFACE_STATE::Depth: * * For SURFTYPE_1D, 2D, and CUBE: The range of this field is reduced * by one for each increase from zero of Minimum Array Element. For * example, if Minimum Array Element is set to 1024 on a 2D surface, * the range of this field is reduced to [0,1023]. * * In other words, 'Depth' is the number of array layers. */ s.Depth = info->view->array_len - 1; /* From the Broadwell PRM >> RENDER_SURFACE_STATE::RenderTargetViewExtent: * * For Render Target and Typed Dataport 1D and 2D Surfaces: * This field must be set to the same value as the Depth field. */ if (info->view->usage & (ISL_SURF_USAGE_RENDER_TARGET_BIT | ISL_SURF_USAGE_STORAGE_BIT)) s.RenderTargetViewExtent = s.Depth; break; case SURFTYPE_CUBE: s.MinimumArrayElement = info->view->base_array_layer; /* Same as SURFTYPE_2D, but divided by 6 */ s.Depth = info->view->array_len / 6 - 1; if (info->view->usage & (ISL_SURF_USAGE_RENDER_TARGET_BIT | ISL_SURF_USAGE_STORAGE_BIT)) s.RenderTargetViewExtent = s.Depth; break; case SURFTYPE_3D: /* From the Broadwell PRM >> RENDER_SURFACE_STATE::Depth: * * If the volume texture is MIP-mapped, this field specifies the * depth of the base MIP level. */ s.Depth = info->surf->logical_level0_px.depth - 1; /* From the Broadwell PRM >> RENDER_SURFACE_STATE::RenderTargetViewExtent: * * For Render Target and Typed Dataport 3D Surfaces: This field * indicates the extent of the accessible 'R' coordinates minus 1 on * the LOD currently being rendered to. * * The docs specify that this only matters for render targets and * surfaces used with typed dataport messages. Prior to Ivy Bridge, the * Depth field has more bits than RenderTargetViewExtent so we can have * textures with more levels than we can render to. In order to prevent * assert-failures in the packing function below, we only set the field * when it's actually going to be used by the hardware. * * Similaraly, the MinimumArrayElement field is ignored by all hardware * prior to Sky Lake when texturing and we want it set to 0 anyway. * Since it's already initialized to 0, we can just leave it alone for * texture surfaces. */ if (info->view->usage & (ISL_SURF_USAGE_RENDER_TARGET_BIT | ISL_SURF_USAGE_STORAGE_BIT)) { s.MinimumArrayElement = info->view->base_array_layer; s.RenderTargetViewExtent = info->view->array_len - 1; } break; default: unreachable("bad SurfaceType"); } #if GFX_VER >= 12 /* Wa_1806565034: Only set SurfaceArray if arrayed surface is > 1. */ s.SurfaceArray = info->surf->dim != ISL_SURF_DIM_3D && info->view->array_len > 1; #elif GFX_VER >= 7 s.SurfaceArray = info->surf->dim != ISL_SURF_DIM_3D; #endif if (info->view->usage & ISL_SURF_USAGE_RENDER_TARGET_BIT) { /* For render target surfaces, the hardware interprets field * MIPCount/LOD as LOD. The Broadwell PRM says: * * MIPCountLOD defines the LOD that will be rendered into. * SurfaceMinLOD is ignored. */ s.MIPCountLOD = info->view->base_level; s.SurfaceMinLOD = 0; } else { /* For non render target surfaces, the hardware interprets field * MIPCount/LOD as MIPCount. The range of levels accessible by the * sampler engine is [SurfaceMinLOD, SurfaceMinLOD + MIPCountLOD]. */ s.SurfaceMinLOD = info->view->base_level; s.MIPCountLOD = MAX(info->view->levels, 1) - 1; } #if GFX_VER >= 9 /* We don't use miptails yet. The PRM recommends that you set "Mip Tail * Start LOD" to 15 to prevent the hardware from trying to use them. */ s.TiledResourceMode = NONE; s.MipTailStartLOD = 15; #endif #if GFX_VER >= 6 const struct isl_extent3d image_align = get_image_alignment(info->surf); s.SurfaceVerticalAlignment = isl_encode_valign(image_align.height); #if GFX_VER >= 7 s.SurfaceHorizontalAlignment = isl_encode_halign(image_align.width); #endif #endif if (info->surf->dim_layout == ISL_DIM_LAYOUT_GFX9_1D) { /* For gfx9 1-D textures, surface pitch is ignored */ s.SurfacePitch = 0; } else { s.SurfacePitch = info->surf->row_pitch_B - 1; } #if GFX_VER >= 8 s.SurfaceQPitch = get_qpitch(info->surf) >> 2; #elif GFX_VER == 7 s.SurfaceArraySpacing = info->surf->array_pitch_span == ISL_ARRAY_PITCH_SPAN_COMPACT; #endif #if GFX_VER >= 8 assert(GFX_VER < 12 || info->surf->tiling != ISL_TILING_W); s.TileMode = isl_encode_tiling[info->surf->tiling]; #else s.TiledSurface = info->surf->tiling != ISL_TILING_LINEAR, s.TileWalk = info->surf->tiling == ISL_TILING_Y0 ? TILEWALK_YMAJOR : TILEWALK_XMAJOR, #endif #if GFX_VER >= 8 s.RenderCacheReadWriteMode = WriteOnlyCache; #else s.RenderCacheReadWriteMode = 0; #endif #if GFX_VER >= 11 /* We've seen dEQP failures when enabling this bit with UINT formats, * which particularly affects blorp_copy() operations. It shouldn't * have any effect on UINT textures anyway, so disable it for them. */ s.EnableUnormPathInColorPipe = !isl_format_has_int_channel(info->view->format); #endif s.CubeFaceEnablePositiveZ = 1; s.CubeFaceEnableNegativeZ = 1; s.CubeFaceEnablePositiveY = 1; s.CubeFaceEnableNegativeY = 1; s.CubeFaceEnablePositiveX = 1; s.CubeFaceEnableNegativeX = 1; #if GFX_VER >= 6 s.NumberofMultisamples = ffs(info->surf->samples) - 1; #if GFX_VER >= 7 s.MultisampledSurfaceStorageFormat = isl_encode_multisample_layout[info->surf->msaa_layout]; #endif #endif #if (GFX_VERx10 >= 75) if (info->view->usage & ISL_SURF_USAGE_RENDER_TARGET_BIT) assert(isl_swizzle_supports_rendering(dev->info, info->view->swizzle)); s.ShaderChannelSelectRed = (enum GENX(ShaderChannelSelect)) info->view->swizzle.r; s.ShaderChannelSelectGreen = (enum GENX(ShaderChannelSelect)) info->view->swizzle.g; s.ShaderChannelSelectBlue = (enum GENX(ShaderChannelSelect)) info->view->swizzle.b; s.ShaderChannelSelectAlpha = (enum GENX(ShaderChannelSelect)) info->view->swizzle.a; #else assert(isl_swizzle_is_identity(info->view->swizzle)); #endif s.SurfaceBaseAddress = info->address; #if GFX_VER >= 6 s.MOCS = info->mocs; #endif #if GFX_VERx10 >= 45 if (info->x_offset_sa != 0 || info->y_offset_sa != 0) { /* There are fairly strict rules about when the offsets can be used. * These are mostly taken from the Sky Lake PRM documentation for * RENDER_SURFACE_STATE. */ assert(info->surf->tiling != ISL_TILING_LINEAR); assert(info->surf->dim == ISL_SURF_DIM_2D); assert(isl_is_pow2(isl_format_get_layout(info->view->format)->bpb)); assert(info->surf->levels == 1); assert(info->surf->logical_level0_px.array_len == 1); assert(info->aux_usage == ISL_AUX_USAGE_NONE); if (GFX_VER >= 8) { /* Broadwell added more rules. */ assert(info->surf->samples == 1); if (isl_format_get_layout(info->view->format)->bpb == 8) assert(info->x_offset_sa % 16 == 0); if (isl_format_get_layout(info->view->format)->bpb == 16) assert(info->x_offset_sa % 8 == 0); } #if GFX_VER >= 7 s.SurfaceArray = false; #endif } const unsigned x_div = 4; const unsigned y_div = GFX_VER >= 8 ? 4 : 2; assert(info->x_offset_sa % x_div == 0); assert(info->y_offset_sa % y_div == 0); s.XOffset = info->x_offset_sa / x_div; s.YOffset = info->y_offset_sa / y_div; #else assert(info->x_offset_sa == 0); assert(info->y_offset_sa == 0); #endif #if GFX_VER >= 7 if (info->aux_usage != ISL_AUX_USAGE_NONE) { /* Check valid aux usages per-gen */ if (GFX_VER >= 12) { assert(info->aux_usage == ISL_AUX_USAGE_MCS || info->aux_usage == ISL_AUX_USAGE_CCS_E || info->aux_usage == ISL_AUX_USAGE_GFX12_CCS_E || info->aux_usage == ISL_AUX_USAGE_MC || info->aux_usage == ISL_AUX_USAGE_HIZ_CCS_WT || info->aux_usage == ISL_AUX_USAGE_MCS_CCS || info->aux_usage == ISL_AUX_USAGE_STC_CCS); } else if (GFX_VER >= 9) { assert(info->aux_usage == ISL_AUX_USAGE_HIZ || info->aux_usage == ISL_AUX_USAGE_MCS || info->aux_usage == ISL_AUX_USAGE_CCS_D || info->aux_usage == ISL_AUX_USAGE_CCS_E); } else if (GFX_VER >= 8) { assert(info->aux_usage == ISL_AUX_USAGE_HIZ || info->aux_usage == ISL_AUX_USAGE_MCS || info->aux_usage == ISL_AUX_USAGE_CCS_D); } else if (GFX_VER >= 7) { assert(info->aux_usage == ISL_AUX_USAGE_MCS || info->aux_usage == ISL_AUX_USAGE_CCS_D); } /* The docs don't appear to say anything whatsoever about compression * and the data port. Testing seems to indicate that the data port * completely ignores the AuxiliarySurfaceMode field. * * On gfx12 HDC supports compression. */ if (GFX_VER < 12) assert(!(info->view->usage & ISL_SURF_USAGE_STORAGE_BIT)); if (isl_surf_usage_is_depth(info->surf->usage)) assert(isl_aux_usage_has_hiz(info->aux_usage)); if (isl_surf_usage_is_stencil(info->surf->usage)) assert(info->aux_usage == ISL_AUX_USAGE_STC_CCS); if (isl_aux_usage_has_hiz(info->aux_usage)) { /* For Gfx8-10, there are some restrictions around sampling from HiZ. * The Skylake PRM docs for RENDER_SURFACE_STATE::AuxiliarySurfaceMode * say: * * "If this field is set to AUX_HIZ, Number of Multisamples must * be MULTISAMPLECOUNT_1, and Surface Type cannot be SURFTYPE_3D." * * On Gfx12, the docs are a bit less obvious but the restriction is * the same. The limitation isn't called out explicitly but the docs * for the CCS_E value of RENDER_SURFACE_STATE::AuxiliarySurfaceMode * say: * * "If Number of multisamples > 1, programming this value means * MSAA compression is enabled for that surface. Auxillary surface * is MSC with tile y." * * Since this interpretation ignores whether the surface is * depth/stencil or not and since multisampled depth buffers use * ISL_MSAA_LAYOUT_INTERLEAVED which is incompatible with MCS * compression, this means that we can't even specify MSAA depth CCS * in RENDER_SURFACE_STATE::AuxiliarySurfaceMode. */ assert(info->surf->samples == 1); /* The dimension must not be 3D */ assert(info->surf->dim != ISL_SURF_DIM_3D); /* The format must be one of the following: */ switch (info->view->format) { case ISL_FORMAT_R32_FLOAT: case ISL_FORMAT_R24_UNORM_X8_TYPELESS: case ISL_FORMAT_R16_UNORM: break; default: assert(!"Incompatible HiZ Sampling format"); break; } } #if GFX_VERx10 >= 125 s.RenderCompressionFormat = isl_get_render_compression_format(info->surf->format); #endif #if GFX_VER >= 12 s.MemoryCompressionEnable = info->aux_usage == ISL_AUX_USAGE_MC; #endif #if GFX_VER >= 8 s.AuxiliarySurfaceMode = isl_encode_aux_mode[info->aux_usage]; #else s.MCSEnable = true; #endif } /* The auxiliary buffer info is filled when it's useable by the HW. * * Starting with Gfx12, the only form of compression that can be used * with RENDER_SURFACE_STATE which requires an aux surface is MCS. * HiZ still requires a surface but the HiZ surface can only be * accessed through 3DSTATE_HIER_DEPTH_BUFFER. * * On all earlier hardware, an aux surface is required for all forms * of compression. */ if ((GFX_VER < 12 && info->aux_usage != ISL_AUX_USAGE_NONE) || (GFX_VER >= 12 && isl_aux_usage_has_mcs(info->aux_usage))) { assert(info->aux_surf != NULL); struct isl_tile_info tile_info; isl_surf_get_tile_info(info->aux_surf, &tile_info); uint32_t pitch_in_tiles = info->aux_surf->row_pitch_B / tile_info.phys_extent_B.width; s.AuxiliarySurfaceBaseAddress = info->aux_address; s.AuxiliarySurfacePitch = pitch_in_tiles - 1; #if GFX_VER >= 8 /* Auxiliary surfaces in ISL have compressed formats but the hardware * doesn't expect our definition of the compression, it expects qpitch * in units of samples on the main surface. */ s.AuxiliarySurfaceQPitch = isl_surf_get_array_pitch_sa_rows(info->aux_surf) >> 2; #endif } #endif #if GFX_VER >= 8 && GFX_VER < 11 /* From the CHV PRM, Volume 2d, page 321 (RENDER_SURFACE_STATE dword 0 * bit 9 "Sampler L2 Bypass Mode Disable" Programming Notes): * * This bit must be set for the following surface types: BC2_UNORM * BC3_UNORM BC5_UNORM BC5_SNORM BC7_UNORM */ if (GFX_VER >= 9 || dev->info->is_cherryview) { switch (info->view->format) { case ISL_FORMAT_BC2_UNORM: case ISL_FORMAT_BC3_UNORM: case ISL_FORMAT_BC5_UNORM: case ISL_FORMAT_BC5_SNORM: case ISL_FORMAT_BC7_UNORM: s.SamplerL2BypassModeDisable = true; break; default: /* From the SKL PRM, Programming Note under Sampler Output Channel * Mapping: * * If a surface has an associated HiZ Auxilliary surface, the * Sampler L2 Bypass Mode Disable field in the RENDER_SURFACE_STATE * must be set. */ if (GFX_VER >= 9 && info->aux_usage == ISL_AUX_USAGE_HIZ) s.SamplerL2BypassModeDisable = true; break; } } #endif if (isl_aux_usage_has_fast_clears(info->aux_usage)) { if (info->use_clear_address) { #if GFX_VER >= 10 s.ClearValueAddressEnable = true; s.ClearValueAddress = info->clear_address; #else unreachable("Gfx9 and earlier do not support indirect clear colors"); #endif } #if GFX_VER == 11 /* * From BXML > GT > Shared Functions > vol5c Shared Functions > * [Structure] RENDER_SURFACE_STATE [BDW+] > ClearColorConversionEnable: * * Project: Gfx11 * * "Enables Pixel backend hw to convert clear values into native format * and write back to clear address, so that display and sampler can use * the converted value for resolving fast cleared RTs." * * Summary: * Clear color conversion must be enabled if the clear color is stored * indirectly and fast color clears are enabled. */ if (info->use_clear_address) { s.ClearColorConversionEnable = true; } #endif #if GFX_VER >= 12 assert(info->use_clear_address); #elif GFX_VER >= 9 if (!info->use_clear_address) { s.RedClearColor = info->clear_color.u32[0]; s.GreenClearColor = info->clear_color.u32[1]; s.BlueClearColor = info->clear_color.u32[2]; s.AlphaClearColor = info->clear_color.u32[3]; } #elif GFX_VER >= 7 /* Prior to Sky Lake, we only have one bit for the clear color which * gives us 0 or 1 in whatever the surface's format happens to be. */ if (isl_format_has_int_channel(info->view->format)) { for (unsigned i = 0; i < 4; i++) { assert(info->clear_color.u32[i] == 0 || info->clear_color.u32[i] == 1); } s.RedClearColor = info->clear_color.u32[0] != 0; s.GreenClearColor = info->clear_color.u32[1] != 0; s.BlueClearColor = info->clear_color.u32[2] != 0; s.AlphaClearColor = info->clear_color.u32[3] != 0; } else { for (unsigned i = 0; i < 4; i++) { assert(info->clear_color.f32[i] == 0.0f || info->clear_color.f32[i] == 1.0f); } s.RedClearColor = info->clear_color.f32[0] != 0.0f; s.GreenClearColor = info->clear_color.f32[1] != 0.0f; s.BlueClearColor = info->clear_color.f32[2] != 0.0f; s.AlphaClearColor = info->clear_color.f32[3] != 0.0f; } #endif } GENX(RENDER_SURFACE_STATE_pack)(NULL, state, &s); } void isl_genX(buffer_fill_state_s)(const struct isl_device *dev, void *state, const struct isl_buffer_fill_state_info *restrict info) { uint64_t buffer_size = info->size_B; /* Uniform and Storage buffers need to have surface size not less that the * aligned 32-bit size of the buffer. To calculate the array lenght on * unsized arrays in StorageBuffer the last 2 bits store the padding size * added to the surface, so we can calculate latter the original buffer * size to know the number of elements. * * surface_size = isl_align(buffer_size, 4) + * (isl_align(buffer_size) - buffer_size) * * buffer_size = (surface_size & ~3) - (surface_size & 3) */ if ((info->format == ISL_FORMAT_RAW || info->stride_B < isl_format_get_layout(info->format)->bpb / 8) && !info->is_scratch) { assert(info->stride_B == 1); uint64_t aligned_size = isl_align(buffer_size, 4); buffer_size = aligned_size + (aligned_size - buffer_size); } uint32_t num_elements = buffer_size / info->stride_B; assert(num_elements > 0); if (info->format == ISL_FORMAT_RAW) { assert(num_elements <= dev->max_buffer_size); } else { /* From the IVB PRM, SURFACE_STATE::Height, * * For typed buffer and structured buffer surfaces, the number * of entries in the buffer ranges from 1 to 2^27. */ assert(num_elements <= (1ull << 27)); } struct GENX(RENDER_SURFACE_STATE) s = { 0, }; s.SurfaceFormat = info->format; s.SurfaceType = SURFTYPE_BUFFER; #if GFX_VERx10 >= 125 if (info->is_scratch) { /* From the BSpec: * * "For surfaces of type SURFTYPE_SCRATCH, valid range of pitch is: * [63,262143] -> [64B, 256KB]. Also, for SURFTYPE_SCRATCH, the * pitch must be a multiple of 64bytes." */ assert(info->format == ISL_FORMAT_RAW); assert(info->stride_B % 64 == 0); assert(info->stride_B <= 256 * 1024); s.SurfaceType = SURFTYPE_SCRATCH; } #else assert(!info->is_scratch); #endif s.SurfacePitch = info->stride_B - 1; #if GFX_VER >= 6 s.SurfaceVerticalAlignment = isl_encode_valign(4); #if GFX_VERx10 >= 125 s.SurfaceHorizontalAlignment = isl_encode_halign(128); #elif GFX_VER >= 7 s.SurfaceHorizontalAlignment = isl_encode_halign(4); s.SurfaceArray = false; #endif #endif #if GFX_VER >= 7 s.Height = ((num_elements - 1) >> 7) & 0x3fff; s.Width = (num_elements - 1) & 0x7f; s.Depth = ((num_elements - 1) >> 21) & 0x3ff; #else s.Height = ((num_elements - 1) >> 7) & 0x1fff; s.Width = (num_elements - 1) & 0x7f; s.Depth = ((num_elements - 1) >> 20) & 0x7f; #endif if (GFX_VER == 12 && dev->info->revision == 0) { /* TGL-LP A0 has a HW bug (fixed in later HW) which causes buffer * textures with very close base addresses (delta < 64B) to corrupt each * other. We can sort-of work around this by making small buffer * textures 1D textures instead. This doesn't fix the problem for large * buffer textures but the liklihood of large, overlapping, and very * close buffer textures is fairly low and the point is to hack around * the bug so we can run apps and tests. */ if (info->format != ISL_FORMAT_RAW && info->stride_B == isl_format_get_layout(info->format)->bpb / 8 && num_elements <= (1 << 14)) { s.SurfaceType = SURFTYPE_1D; s.Width = num_elements - 1; s.Height = 0; s.Depth = 0; } } #if GFX_VER >= 6 s.NumberofMultisamples = MULTISAMPLECOUNT_1; #endif #if (GFX_VER >= 8) s.TileMode = LINEAR; #else s.TiledSurface = false; #endif #if (GFX_VER >= 8) s.RenderCacheReadWriteMode = WriteOnlyCache; #else s.RenderCacheReadWriteMode = 0; #endif s.SurfaceBaseAddress = info->address; #if GFX_VER >= 6 s.MOCS = info->mocs; #endif #if (GFX_VERx10 >= 75) s.ShaderChannelSelectRed = (enum GENX(ShaderChannelSelect)) info->swizzle.r; s.ShaderChannelSelectGreen = (enum GENX(ShaderChannelSelect)) info->swizzle.g; s.ShaderChannelSelectBlue = (enum GENX(ShaderChannelSelect)) info->swizzle.b; s.ShaderChannelSelectAlpha = (enum GENX(ShaderChannelSelect)) info->swizzle.a; #endif GENX(RENDER_SURFACE_STATE_pack)(NULL, state, &s); } void isl_genX(null_fill_state)(void *state, const struct isl_null_fill_state_info *restrict info) { struct GENX(RENDER_SURFACE_STATE) s = { .SurfaceType = SURFTYPE_NULL, /* We previously had this format set to B8G8R8A8_UNORM but ran into * hangs on IVB. R32_UINT seems to work for everybody. * * https://gitlab.freedesktop.org/mesa/mesa/-/issues/1872 */ .SurfaceFormat = ISL_FORMAT_R32_UINT, #if GFX_VER >= 7 .SurfaceArray = info->size.depth > 1, #endif #if GFX_VERx10 >= 125 .TileMode = TILE4, #elif GFX_VER >= 8 .TileMode = YMAJOR, #else .TiledSurface = true, .TileWalk = TILEWALK_YMAJOR, #endif #if GFX_VER == 7 /* According to PRMs: "Volume 4 Part 1: Subsystem and Cores – Shared * Functions" * * RENDER_SURFACE_STATE::Surface Vertical Alignment * * "This field must be set to VALIGN_4 for all tiled Y Render Target * surfaces." * * Affect IVB, HSW. */ .SurfaceVerticalAlignment = VALIGN_4, #endif .MIPCountLOD = info->levels, .Width = info->size.width - 1, .Height = info->size.height - 1, .Depth = info->size.depth - 1, .RenderTargetViewExtent = info->size.depth - 1, #if GFX_VER <= 5 .MinimumArrayElement = info->minimum_array_element, .ColorBufferComponentWriteDisables = 0xf, #endif }; GENX(RENDER_SURFACE_STATE_pack)(NULL, state, &s); }