diff options
author | Jordan Justen <[email protected]> | 2018-02-26 15:39:59 -0800 |
---|---|---|
committer | Jordan Justen <[email protected]> | 2018-03-05 09:47:37 -0800 |
commit | 272bef0601a1bdb5292771aefc8d62fcbdf4c47f (patch) | |
tree | c94ae513a916ed81151067da24260bbb424e5deb /src/intel/dev | |
parent | 9a0d7bb48c93e7d0109751469a8b32c94e85bc24 (diff) |
intel: Split gen_device_info out into libintel_dev
Split out the device info so isl doesn't depend on intel/common. Now
it will depend on the new intel/dev device info lib.
This will allow the decoder in intel/common to use isl, allowing us to
apply Ken's patch that removes the genxml duplication of surface
formats.
Signed-off-by: Jordan Justen <[email protected]>
Reviewed-by: Samuel Iglesias Gonsálvez <[email protected]>
Diffstat (limited to 'src/intel/dev')
-rw-r--r-- | src/intel/dev/gen_device_info.c | 902 | ||||
-rw-r--r-- | src/intel/dev/gen_device_info.h | 213 | ||||
-rw-r--r-- | src/intel/dev/meson.build | 33 |
3 files changed, 1148 insertions, 0 deletions
diff --git a/src/intel/dev/gen_device_info.c b/src/intel/dev/gen_device_info.c new file mode 100644 index 00000000000..1773009d33c --- /dev/null +++ b/src/intel/dev/gen_device_info.c @@ -0,0 +1,902 @@ +/* + * Copyright © 2013 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 <assert.h> +#include <stdio.h> +#include <stdlib.h> +#include <string.h> +#include <unistd.h> +#include "gen_device_info.h" +#include "compiler/shader_enums.h" +#include "util/macros.h" + +/** + * Get the PCI ID for the device name. + * + * Returns -1 if the device is not known. + */ +int +gen_device_name_to_pci_device_id(const char *name) +{ + static const struct { + const char *name; + int pci_id; + } name_map[] = { + { "brw", 0x2a02 }, + { "g4x", 0x2a42 }, + { "ilk", 0x0042 }, + { "snb", 0x0126 }, + { "ivb", 0x016a }, + { "hsw", 0x0d2e }, + { "byt", 0x0f33 }, + { "bdw", 0x162e }, + { "chv", 0x22B3 }, + { "skl", 0x1912 }, + { "bxt", 0x5A85 }, + { "kbl", 0x5912 }, + { "glk", 0x3185 }, + { "cnl", 0x5a52 }, + { "icl", 0x8a52 }, + }; + + for (unsigned i = 0; i < ARRAY_SIZE(name_map); i++) { + if (!strcmp(name_map[i].name, name)) + return name_map[i].pci_id; + } + + return -1; +} + +/** + * Get the overridden PCI ID for the device. This is set with the + * INTEL_DEVID_OVERRIDE environment variable. + * + * Returns -1 if the override is not set. + */ +int +gen_get_pci_device_id_override(void) +{ + if (geteuid() == getuid()) { + const char *devid_override = getenv("INTEL_DEVID_OVERRIDE"); + if (devid_override) { + const int id = gen_device_name_to_pci_device_id(devid_override); + return id >= 0 ? id : strtol(devid_override, NULL, 0); + } + } + + return -1; +} + +static const struct gen_device_info gen_device_info_i965 = { + .gen = 4, + .has_negative_rhw_bug = true, + .num_slices = 1, + .num_subslices = { 1, }, + .num_thread_per_eu = 4, + .max_vs_threads = 16, + .max_gs_threads = 2, + .max_wm_threads = 8 * 4, + .urb = { + .size = 256, + }, + .timestamp_frequency = 12500000, +}; + +static const struct gen_device_info gen_device_info_g4x = { + .gen = 4, + .has_pln = true, + .has_compr4 = true, + .has_surface_tile_offset = true, + .is_g4x = true, + .num_slices = 1, + .num_subslices = { 1, }, + .num_thread_per_eu = 5, + .max_vs_threads = 32, + .max_gs_threads = 2, + .max_wm_threads = 10 * 5, + .urb = { + .size = 384, + }, + .timestamp_frequency = 12500000, +}; + +static const struct gen_device_info gen_device_info_ilk = { + .gen = 5, + .has_pln = true, + .has_compr4 = true, + .has_surface_tile_offset = true, + .num_slices = 1, + .num_subslices = { 1, }, + .num_thread_per_eu = 6, + .max_vs_threads = 72, + .max_gs_threads = 32, + .max_wm_threads = 12 * 6, + .urb = { + .size = 1024, + }, + .timestamp_frequency = 12500000, +}; + +static const struct gen_device_info gen_device_info_snb_gt1 = { + .gen = 6, + .gt = 1, + .has_hiz_and_separate_stencil = true, + .has_llc = true, + .has_pln = true, + .has_surface_tile_offset = true, + .needs_unlit_centroid_workaround = true, + .num_slices = 1, + .num_subslices = { 1, }, + .num_thread_per_eu = 6, /* Not confirmed */ + .max_vs_threads = 24, + .max_gs_threads = 21, /* conservative; 24 if rendering disabled. */ + .max_wm_threads = 40, + .urb = { + .size = 32, + .min_entries = { + [MESA_SHADER_VERTEX] = 24, + }, + .max_entries = { + [MESA_SHADER_VERTEX] = 256, + [MESA_SHADER_GEOMETRY] = 256, + }, + }, + .timestamp_frequency = 12500000, +}; + +static const struct gen_device_info gen_device_info_snb_gt2 = { + .gen = 6, + .gt = 2, + .has_hiz_and_separate_stencil = true, + .has_llc = true, + .has_pln = true, + .has_surface_tile_offset = true, + .needs_unlit_centroid_workaround = true, + .num_slices = 1, + .num_subslices = { 1, }, + .num_thread_per_eu = 6, /* Not confirmed */ + .max_vs_threads = 60, + .max_gs_threads = 60, + .max_wm_threads = 80, + .urb = { + .size = 64, + .min_entries = { + [MESA_SHADER_VERTEX] = 24, + }, + .max_entries = { + [MESA_SHADER_VERTEX] = 256, + [MESA_SHADER_GEOMETRY] = 256, + }, + }, + .timestamp_frequency = 12500000, +}; + +#define GEN7_FEATURES \ + .gen = 7, \ + .has_hiz_and_separate_stencil = true, \ + .must_use_separate_stencil = true, \ + .has_llc = true, \ + .has_pln = true, \ + .has_64bit_types = true, \ + .has_surface_tile_offset = true, \ + .timestamp_frequency = 12500000 + +static const struct gen_device_info gen_device_info_ivb_gt1 = { + GEN7_FEATURES, .is_ivybridge = true, .gt = 1, + .num_slices = 1, + .num_subslices = { 1, }, + .num_thread_per_eu = 6, + .l3_banks = 2, + .max_vs_threads = 36, + .max_tcs_threads = 36, + .max_tes_threads = 36, + .max_gs_threads = 36, + .max_wm_threads = 48, + .max_cs_threads = 36, + .urb = { + .size = 128, + .min_entries = { + [MESA_SHADER_VERTEX] = 32, + [MESA_SHADER_TESS_EVAL] = 10, + }, + .max_entries = { + [MESA_SHADER_VERTEX] = 512, + [MESA_SHADER_TESS_CTRL] = 32, + [MESA_SHADER_TESS_EVAL] = 288, + [MESA_SHADER_GEOMETRY] = 192, + }, + }, +}; + +static const struct gen_device_info gen_device_info_ivb_gt2 = { + GEN7_FEATURES, .is_ivybridge = true, .gt = 2, + .num_slices = 1, + .num_subslices = { 1, }, + .num_thread_per_eu = 8, /* Not sure why this isn't a multiple of + * @max_wm_threads ... */ + .l3_banks = 4, + .max_vs_threads = 128, + .max_tcs_threads = 128, + .max_tes_threads = 128, + .max_gs_threads = 128, + .max_wm_threads = 172, + .max_cs_threads = 64, + .urb = { + .size = 256, + .min_entries = { + [MESA_SHADER_VERTEX] = 32, + [MESA_SHADER_TESS_EVAL] = 10, + }, + .max_entries = { + [MESA_SHADER_VERTEX] = 704, + [MESA_SHADER_TESS_CTRL] = 64, + [MESA_SHADER_TESS_EVAL] = 448, + [MESA_SHADER_GEOMETRY] = 320, + }, + }, +}; + +static const struct gen_device_info gen_device_info_byt = { + GEN7_FEATURES, .is_baytrail = true, .gt = 1, + .num_slices = 1, + .num_subslices = { 1, }, + .num_thread_per_eu = 8, + .l3_banks = 1, + .has_llc = false, + .max_vs_threads = 36, + .max_tcs_threads = 36, + .max_tes_threads = 36, + .max_gs_threads = 36, + .max_wm_threads = 48, + .max_cs_threads = 32, + .urb = { + .size = 128, + .min_entries = { + [MESA_SHADER_VERTEX] = 32, + [MESA_SHADER_TESS_EVAL] = 10, + }, + .max_entries = { + [MESA_SHADER_VERTEX] = 512, + [MESA_SHADER_TESS_CTRL] = 32, + [MESA_SHADER_TESS_EVAL] = 288, + [MESA_SHADER_GEOMETRY] = 192, + }, + }, +}; + +#define HSW_FEATURES \ + GEN7_FEATURES, \ + .is_haswell = true, \ + .supports_simd16_3src = true, \ + .has_resource_streamer = true + +static const struct gen_device_info gen_device_info_hsw_gt1 = { + HSW_FEATURES, .gt = 1, + .num_slices = 1, + .num_subslices = { 1, }, + .num_thread_per_eu = 7, + .l3_banks = 2, + .max_vs_threads = 70, + .max_tcs_threads = 70, + .max_tes_threads = 70, + .max_gs_threads = 70, + .max_wm_threads = 102, + .max_cs_threads = 70, + .urb = { + .size = 128, + .min_entries = { + [MESA_SHADER_VERTEX] = 32, + [MESA_SHADER_TESS_EVAL] = 10, + }, + .max_entries = { + [MESA_SHADER_VERTEX] = 640, + [MESA_SHADER_TESS_CTRL] = 64, + [MESA_SHADER_TESS_EVAL] = 384, + [MESA_SHADER_GEOMETRY] = 256, + }, + }, +}; + +static const struct gen_device_info gen_device_info_hsw_gt2 = { + HSW_FEATURES, .gt = 2, + .num_slices = 1, + .num_subslices = { 2, }, + .num_thread_per_eu = 7, + .l3_banks = 4, + .max_vs_threads = 280, + .max_tcs_threads = 256, + .max_tes_threads = 280, + .max_gs_threads = 256, + .max_wm_threads = 204, + .max_cs_threads = 70, + .urb = { + .size = 256, + .min_entries = { + [MESA_SHADER_VERTEX] = 64, + [MESA_SHADER_TESS_EVAL] = 10, + }, + .max_entries = { + [MESA_SHADER_VERTEX] = 1664, + [MESA_SHADER_TESS_CTRL] = 128, + [MESA_SHADER_TESS_EVAL] = 960, + [MESA_SHADER_GEOMETRY] = 640, + }, + }, +}; + +static const struct gen_device_info gen_device_info_hsw_gt3 = { + HSW_FEATURES, .gt = 3, + .num_slices = 2, + .num_subslices = { 2, }, + .num_thread_per_eu = 7, + .l3_banks = 8, + .max_vs_threads = 280, + .max_tcs_threads = 256, + .max_tes_threads = 280, + .max_gs_threads = 256, + .max_wm_threads = 408, + .max_cs_threads = 70, + .urb = { + .size = 512, + .min_entries = { + [MESA_SHADER_VERTEX] = 64, + [MESA_SHADER_TESS_EVAL] = 10, + }, + .max_entries = { + [MESA_SHADER_VERTEX] = 1664, + [MESA_SHADER_TESS_CTRL] = 128, + [MESA_SHADER_TESS_EVAL] = 960, + [MESA_SHADER_GEOMETRY] = 640, + }, + }, +}; + +/* It's unclear how well supported sampling from the hiz buffer is on GEN8, + * so keep things conservative for now and set has_sample_with_hiz = false. + */ +#define GEN8_FEATURES \ + .gen = 8, \ + .has_hiz_and_separate_stencil = true, \ + .has_resource_streamer = true, \ + .must_use_separate_stencil = true, \ + .has_llc = true, \ + .has_sample_with_hiz = false, \ + .has_pln = true, \ + .has_integer_dword_mul = true, \ + .has_64bit_types = true, \ + .supports_simd16_3src = true, \ + .has_surface_tile_offset = true, \ + .max_vs_threads = 504, \ + .max_tcs_threads = 504, \ + .max_tes_threads = 504, \ + .max_gs_threads = 504, \ + .max_wm_threads = 384, \ + .timestamp_frequency = 12500000 + +static const struct gen_device_info gen_device_info_bdw_gt1 = { + GEN8_FEATURES, .gt = 1, + .is_broadwell = true, + .num_slices = 1, + .num_subslices = { 2, }, + .num_thread_per_eu = 7, + .l3_banks = 2, + .max_cs_threads = 42, + .urb = { + .size = 192, + .min_entries = { + [MESA_SHADER_VERTEX] = 64, + [MESA_SHADER_TESS_EVAL] = 34, + }, + .max_entries = { + [MESA_SHADER_VERTEX] = 2560, + [MESA_SHADER_TESS_CTRL] = 504, + [MESA_SHADER_TESS_EVAL] = 1536, + [MESA_SHADER_GEOMETRY] = 960, + }, + } +}; + +static const struct gen_device_info gen_device_info_bdw_gt2 = { + GEN8_FEATURES, .gt = 2, + .is_broadwell = true, + .num_slices = 1, + .num_subslices = { 3, }, + .num_thread_per_eu = 7, + .l3_banks = 4, + .max_cs_threads = 56, + .urb = { + .size = 384, + .min_entries = { + [MESA_SHADER_VERTEX] = 64, + [MESA_SHADER_TESS_EVAL] = 34, + }, + .max_entries = { + [MESA_SHADER_VERTEX] = 2560, + [MESA_SHADER_TESS_CTRL] = 504, + [MESA_SHADER_TESS_EVAL] = 1536, + [MESA_SHADER_GEOMETRY] = 960, + }, + } +}; + +static const struct gen_device_info gen_device_info_bdw_gt3 = { + GEN8_FEATURES, .gt = 3, + .is_broadwell = true, + .num_slices = 2, + .num_subslices = { 3, 3, }, + .num_thread_per_eu = 7, + .l3_banks = 8, + .max_cs_threads = 56, + .urb = { + .size = 384, + .min_entries = { + [MESA_SHADER_VERTEX] = 64, + [MESA_SHADER_TESS_EVAL] = 34, + }, + .max_entries = { + [MESA_SHADER_VERTEX] = 2560, + [MESA_SHADER_TESS_CTRL] = 504, + [MESA_SHADER_TESS_EVAL] = 1536, + [MESA_SHADER_GEOMETRY] = 960, + }, + } +}; + +static const struct gen_device_info gen_device_info_chv = { + GEN8_FEATURES, .is_cherryview = 1, .gt = 1, + .has_llc = false, + .has_integer_dword_mul = false, + .num_slices = 1, + .num_subslices = { 2, }, + .num_thread_per_eu = 7, + .l3_banks = 2, + .max_vs_threads = 80, + .max_tcs_threads = 80, + .max_tes_threads = 80, + .max_gs_threads = 80, + .max_wm_threads = 128, + .max_cs_threads = 6 * 7, + .urb = { + .size = 192, + .min_entries = { + [MESA_SHADER_VERTEX] = 34, + [MESA_SHADER_TESS_EVAL] = 34, + }, + .max_entries = { + [MESA_SHADER_VERTEX] = 640, + [MESA_SHADER_TESS_CTRL] = 80, + [MESA_SHADER_TESS_EVAL] = 384, + [MESA_SHADER_GEOMETRY] = 256, + }, + } +}; + +#define GEN9_HW_INFO \ + .gen = 9, \ + .max_vs_threads = 336, \ + .max_gs_threads = 336, \ + .max_tcs_threads = 336, \ + .max_tes_threads = 336, \ + .max_cs_threads = 56, \ + .timestamp_frequency = 12000000, \ + .urb = { \ + .size = 384, \ + .min_entries = { \ + [MESA_SHADER_VERTEX] = 64, \ + [MESA_SHADER_TESS_EVAL] = 34, \ + }, \ + .max_entries = { \ + [MESA_SHADER_VERTEX] = 1856, \ + [MESA_SHADER_TESS_CTRL] = 672, \ + [MESA_SHADER_TESS_EVAL] = 1120, \ + [MESA_SHADER_GEOMETRY] = 640, \ + }, \ + } + +#define GEN9_LP_FEATURES \ + GEN8_FEATURES, \ + GEN9_HW_INFO, \ + .has_integer_dword_mul = false, \ + .gt = 1, \ + .has_llc = false, \ + .has_sample_with_hiz = true, \ + .num_slices = 1, \ + .num_thread_per_eu = 6, \ + .max_vs_threads = 112, \ + .max_tcs_threads = 112, \ + .max_tes_threads = 112, \ + .max_gs_threads = 112, \ + .max_cs_threads = 6 * 6, \ + .timestamp_frequency = 19200000, \ + .urb = { \ + .size = 192, \ + .min_entries = { \ + [MESA_SHADER_VERTEX] = 34, \ + [MESA_SHADER_TESS_EVAL] = 34, \ + }, \ + .max_entries = { \ + [MESA_SHADER_VERTEX] = 704, \ + [MESA_SHADER_TESS_CTRL] = 256, \ + [MESA_SHADER_TESS_EVAL] = 416, \ + [MESA_SHADER_GEOMETRY] = 256, \ + }, \ + } + +#define GEN9_LP_FEATURES_3X6 \ + GEN9_LP_FEATURES, \ + .num_subslices = { 3, } + +#define GEN9_LP_FEATURES_2X6 \ + GEN9_LP_FEATURES, \ + .num_subslices = { 2, }, \ + .max_vs_threads = 56, \ + .max_tcs_threads = 56, \ + .max_tes_threads = 56, \ + .max_gs_threads = 56, \ + .max_cs_threads = 6 * 6, \ + .urb = { \ + .size = 128, \ + .min_entries = { \ + [MESA_SHADER_VERTEX] = 34, \ + [MESA_SHADER_TESS_EVAL] = 34, \ + }, \ + .max_entries = { \ + [MESA_SHADER_VERTEX] = 352, \ + [MESA_SHADER_TESS_CTRL] = 128, \ + [MESA_SHADER_TESS_EVAL] = 208, \ + [MESA_SHADER_GEOMETRY] = 128, \ + }, \ + } + +#define GEN9_FEATURES \ + GEN8_FEATURES, \ + GEN9_HW_INFO, \ + .has_sample_with_hiz = true, \ + .num_thread_per_eu = 7 + +static const struct gen_device_info gen_device_info_skl_gt1 = { + GEN9_FEATURES, .gt = 1, + .is_skylake = true, + .num_slices = 1, + .num_subslices = { 2, }, + .l3_banks = 2, + .urb.size = 192, +}; + +static const struct gen_device_info gen_device_info_skl_gt2 = { + GEN9_FEATURES, .gt = 2, + .is_skylake = true, + .num_slices = 1, + .num_subslices = { 3, }, + .l3_banks = 4, +}; + +static const struct gen_device_info gen_device_info_skl_gt3 = { + GEN9_FEATURES, .gt = 3, + .is_skylake = true, + .num_slices = 2, + .num_subslices = { 3, 3, }, + .l3_banks = 8, +}; + +static const struct gen_device_info gen_device_info_skl_gt4 = { + GEN9_FEATURES, .gt = 4, + .is_skylake = true, + .num_slices = 3, + .num_subslices = { 3, 3, 3, }, + .l3_banks = 12, + /* From the "L3 Allocation and Programming" documentation: + * + * "URB is limited to 1008KB due to programming restrictions. This is not a + * restriction of the L3 implementation, but of the FF and other clients. + * Therefore, in a GT4 implementation it is possible for the programmed + * allocation of the L3 data array to provide 3*384KB=1152KB for URB, but + * only 1008KB of this will be used." + */ + .urb.size = 1008 / 3, +}; + +static const struct gen_device_info gen_device_info_bxt = { + GEN9_LP_FEATURES_3X6, + .is_broxton = true, + .l3_banks = 2, +}; + +static const struct gen_device_info gen_device_info_bxt_2x6 = { + GEN9_LP_FEATURES_2X6, + .is_broxton = true, + .l3_banks = 1, +}; +/* + * Note: for all KBL SKUs, the PRM says SKL for GS entries, not SKL+. + * There's no KBL entry. Using the default SKL (GEN9) GS entries value. + */ + +static const struct gen_device_info gen_device_info_kbl_gt1 = { + GEN9_FEATURES, + .is_kabylake = true, + .gt = 1, + + .max_cs_threads = 7 * 6, + .urb.size = 192, + .num_slices = 1, + .num_subslices = { 2, }, + .l3_banks = 2, +}; + +static const struct gen_device_info gen_device_info_kbl_gt1_5 = { + GEN9_FEATURES, + .is_kabylake = true, + .gt = 1, + + .max_cs_threads = 7 * 6, + .num_slices = 1, + .num_subslices = { 3, }, + .l3_banks = 4, +}; + +static const struct gen_device_info gen_device_info_kbl_gt2 = { + GEN9_FEATURES, + .is_kabylake = true, + .gt = 2, + + .num_slices = 1, + .num_subslices = { 3, }, + .l3_banks = 4, +}; + +static const struct gen_device_info gen_device_info_kbl_gt3 = { + GEN9_FEATURES, + .is_kabylake = true, + .gt = 3, + + .num_slices = 2, + .num_subslices = { 3, 3, }, + .l3_banks = 8, +}; + +static const struct gen_device_info gen_device_info_kbl_gt4 = { + GEN9_FEATURES, + .is_kabylake = true, + .gt = 4, + + /* + * From the "L3 Allocation and Programming" documentation: + * + * "URB is limited to 1008KB due to programming restrictions. This + * is not a restriction of the L3 implementation, but of the FF and + * other clients. Therefore, in a GT4 implementation it is + * possible for the programmed allocation of the L3 data array to + * provide 3*384KB=1152KB for URB, but only 1008KB of this + * will be used." + */ + .urb.size = 1008 / 3, + .num_slices = 3, + .num_subslices = { 3, 3, 3, }, + .l3_banks = 12, +}; + +static const struct gen_device_info gen_device_info_glk = { + GEN9_LP_FEATURES_3X6, + .is_geminilake = true, + .l3_banks = 2, +}; + +/*TODO: Initialize l3_banks when we know the number. */ +static const struct gen_device_info gen_device_info_glk_2x6 = { + GEN9_LP_FEATURES_2X6, + .is_geminilake = true, +}; + +static const struct gen_device_info gen_device_info_cfl_gt1 = { + GEN9_FEATURES, + .is_coffeelake = true, + .gt = 1, + + .num_slices = 1, + .num_subslices = { 2, }, + .l3_banks = 2, +}; +static const struct gen_device_info gen_device_info_cfl_gt2 = { + GEN9_FEATURES, + .is_coffeelake = true, + .gt = 2, + + .num_slices = 1, + .num_subslices = { 3, }, + .l3_banks = 4, +}; + +static const struct gen_device_info gen_device_info_cfl_gt3 = { + GEN9_FEATURES, + .is_coffeelake = true, + .gt = 3, + + .num_slices = 2, + .num_subslices = { 3, 3, }, + .l3_banks = 8, +}; + +#define GEN10_HW_INFO \ + .gen = 10, \ + .num_thread_per_eu = 7, \ + .max_vs_threads = 728, \ + .max_gs_threads = 432, \ + .max_tcs_threads = 432, \ + .max_tes_threads = 624, \ + .max_cs_threads = 56, \ + .timestamp_frequency = 19200000, \ + .urb = { \ + .size = 256, \ + .min_entries = { \ + [MESA_SHADER_VERTEX] = 64, \ + [MESA_SHADER_TESS_EVAL] = 34, \ + }, \ + .max_entries = { \ + [MESA_SHADER_VERTEX] = 3936, \ + [MESA_SHADER_TESS_CTRL] = 896, \ + [MESA_SHADER_TESS_EVAL] = 2064, \ + [MESA_SHADER_GEOMETRY] = 832, \ + }, \ + } + +#define subslices(args...) { args, } + +#define GEN10_FEATURES(_gt, _slices, _subslices, _l3) \ + GEN8_FEATURES, \ + GEN10_HW_INFO, \ + .has_sample_with_hiz = true, \ + .gt = _gt, \ + .num_slices = _slices, \ + .num_subslices = _subslices, \ + .l3_banks = _l3 + +static const struct gen_device_info gen_device_info_cnl_2x8 = { + /* GT0.5 */ + GEN10_FEATURES(1, 1, subslices(2), 2), + .is_cannonlake = true, +}; + +static const struct gen_device_info gen_device_info_cnl_3x8 = { + /* GT1 */ + GEN10_FEATURES(1, 1, subslices(3), 3), + .is_cannonlake = true, +}; + +static const struct gen_device_info gen_device_info_cnl_4x8 = { + /* GT 1.5 */ + GEN10_FEATURES(1, 2, subslices(2, 2), 6), + .is_cannonlake = true, +}; + +static const struct gen_device_info gen_device_info_cnl_5x8 = { + /* GT2 */ + GEN10_FEATURES(2, 2, subslices(3, 2), 6), + .is_cannonlake = true, +}; + +#define GEN11_HW_INFO \ + .gen = 11, \ + .has_pln = false, \ + .max_vs_threads = 364, \ + .max_gs_threads = 224, \ + .max_tcs_threads = 224, \ + .max_tes_threads = 364, \ + .max_cs_threads = 56, \ + .urb = { \ + .size = 1024, \ + .min_entries = { \ + [MESA_SHADER_VERTEX] = 64, \ + [MESA_SHADER_TESS_EVAL] = 34, \ + }, \ + .max_entries = { \ + [MESA_SHADER_VERTEX] = 2384, \ + [MESA_SHADER_TESS_CTRL] = 1032, \ + [MESA_SHADER_TESS_EVAL] = 2384, \ + [MESA_SHADER_GEOMETRY] = 1032, \ + }, \ + } + +#define GEN11_FEATURES(_gt, _slices, _subslices, _l3) \ + GEN8_FEATURES, \ + GEN11_HW_INFO, \ + .has_64bit_types = false, \ + .has_integer_dword_mul = false, \ + .gt = _gt, .num_slices = _slices, .l3_banks = _l3, \ + .num_subslices = _subslices + +static const struct gen_device_info gen_device_info_icl_8x8 = { + GEN11_FEATURES(2, 1, subslices(8), 8), +}; + +static const struct gen_device_info gen_device_info_icl_6x8 = { + GEN11_FEATURES(1, 1, subslices(6), 6), +}; + +static const struct gen_device_info gen_device_info_icl_4x8 = { + GEN11_FEATURES(1, 1, subslices(4), 6), +}; + +static const struct gen_device_info gen_device_info_icl_1x8 = { + GEN11_FEATURES(1, 1, subslices(1), 6), +}; + +bool +gen_get_device_info(int devid, struct gen_device_info *devinfo) +{ + switch (devid) { +#undef CHIPSET +#define CHIPSET(id, family, name) \ + case id: *devinfo = gen_device_info_##family; break; +#include "pci_ids/i965_pci_ids.h" + default: + fprintf(stderr, "i965_dri.so does not support the 0x%x PCI ID.\n", devid); + return false; + } + + /* From the Skylake PRM, 3DSTATE_PS::Scratch Space Base Pointer: + * + * "Scratch Space per slice is computed based on 4 sub-slices. SW must + * allocate scratch space enough so that each slice has 4 slices allowed." + * + * The equivalent internal documentation says that this programming note + * applies to all Gen9+ platforms. + * + * The hardware typically calculates the scratch space pointer by taking + * the base address, and adding per-thread-scratch-space * thread ID. + * Extra padding can be necessary depending how the thread IDs are + * calculated for a particular shader stage. + */ + + switch(devinfo->gen) { + case 9: + case 10: + devinfo->max_wm_threads = 64 /* threads-per-PSD */ + * devinfo->num_slices + * 4; /* effective subslices per slice */ + break; + case 11: + devinfo->max_wm_threads = 128 /* threads-per-PSD */ + * devinfo->num_slices + * 8; /* subslices per slice */ + break; + default: + break; + } + + assert(devinfo->num_slices <= ARRAY_SIZE(devinfo->num_subslices)); + + return true; +} + +const char * +gen_get_device_name(int devid) +{ + switch (devid) { +#undef CHIPSET +#define CHIPSET(id, family, name) case id: return name; +#include "pci_ids/i965_pci_ids.h" + default: + return NULL; + } +} diff --git a/src/intel/dev/gen_device_info.h b/src/intel/dev/gen_device_info.h new file mode 100644 index 00000000000..b8044d00032 --- /dev/null +++ b/src/intel/dev/gen_device_info.h @@ -0,0 +1,213 @@ + /* + * Copyright © 2013 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. + * + */ + +#ifndef GEN_DEVICE_INFO_H +#define GEN_DEVICE_INFO_H + +#include <stdbool.h> +#include <stdint.h> + +#ifdef __cplusplus +extern "C" { +#endif + +/** + * Intel hardware information and quirks + */ +struct gen_device_info +{ + int gen; /**< Generation number: 4, 5, 6, 7, ... */ + int gt; + + bool is_g4x; + bool is_ivybridge; + bool is_baytrail; + bool is_haswell; + bool is_broadwell; + bool is_cherryview; + bool is_skylake; + bool is_broxton; + bool is_kabylake; + bool is_geminilake; + bool is_coffeelake; + bool is_cannonlake; + + bool has_hiz_and_separate_stencil; + bool must_use_separate_stencil; + bool has_sample_with_hiz; + bool has_llc; + + bool has_pln; + bool has_64bit_types; + bool has_integer_dword_mul; + bool has_compr4; + bool has_surface_tile_offset; + bool supports_simd16_3src; + bool has_resource_streamer; + + /** + * \name Intel hardware quirks + * @{ + */ + bool has_negative_rhw_bug; + + /** + * Some versions of Gen hardware don't do centroid interpolation correctly + * on unlit pixels, causing incorrect values for derivatives near triangle + * edges. Enabling this flag causes the fragment shader to use + * non-centroid interpolation for unlit pixels, at the expense of two extra + * fragment shader instructions. + */ + bool needs_unlit_centroid_workaround; + /** @} */ + + /** + * \name GPU hardware limits + * + * In general, you can find shader thread maximums by looking at the "Maximum + * Number of Threads" field in the Intel PRM description of the 3DSTATE_VS, + * 3DSTATE_GS, 3DSTATE_HS, 3DSTATE_DS, and 3DSTATE_PS commands. URB entry + * limits come from the "Number of URB Entries" field in the + * 3DSTATE_URB_VS command and friends. + * + * These fields are used to calculate the scratch space to allocate. The + * amount of scratch space can be larger without being harmful on modern + * GPUs, however, prior to Haswell, programming the maximum number of threads + * to greater than the hardware maximum would cause GPU performance to tank. + * + * @{ + */ + /** + * Total number of slices present on the device whether or not they've been + * fused off. + * + * XXX: CS thread counts are limited by the inability to do cross subslice + * communication. It is the effectively the number of logical threads which + * can be executed in a subslice. Fuse configurations may cause this number + * to change, so we program @max_cs_threads as the lower maximum. + */ + unsigned num_slices; + + /** + * Number of subslices for each slice (used to be uniform until CNL). + */ + unsigned num_subslices[3]; + + /** + * Number of threads per eu, varies between 4 and 8 between generations. + */ + unsigned num_thread_per_eu; + + unsigned l3_banks; + unsigned max_vs_threads; /**< Maximum Vertex Shader threads */ + unsigned max_tcs_threads; /**< Maximum Hull Shader threads */ + unsigned max_tes_threads; /**< Maximum Domain Shader threads */ + unsigned max_gs_threads; /**< Maximum Geometry Shader threads. */ + /** + * Theoretical maximum number of Pixel Shader threads. + * + * PSD means Pixel Shader Dispatcher. On modern Intel GPUs, hardware will + * automatically scale pixel shader thread count, based on a single value + * programmed into 3DSTATE_PS. + * + * To calculate the maximum number of threads for Gen8 beyond (which have + * multiple Pixel Shader Dispatchers): + * + * - Look up 3DSTATE_PS and find "Maximum Number of Threads Per PSD" + * - Usually there's only one PSD per subslice, so use the number of + * subslices for number of PSDs. + * - For max_wm_threads, the total should be PSD threads * #PSDs. + */ + unsigned max_wm_threads; + + /** + * Maximum Compute Shader threads. + * + * Thread count * number of EUs per subslice + */ + unsigned max_cs_threads; + + struct { + /** + * Hardware default URB size. + * + * The units this is expressed in are somewhat inconsistent: 512b units + * on Gen4-5, KB on Gen6-7, and KB times the slice count on Gen8+. + * + * Look up "URB Size" in the "Device Attributes" page, and take the + * maximum. Look up the slice count for each GT SKU on the same page. + * urb.size = URB Size (kbytes) / slice count + */ + unsigned size; + + /** + * The minimum number of URB entries. See the 3DSTATE_URB_<XS> docs. + */ + unsigned min_entries[4]; + + /** + * The maximum number of URB entries. See the 3DSTATE_URB_<XS> docs. + */ + unsigned max_entries[4]; + } urb; + + /** + * For the longest time the timestamp frequency for Gen's timestamp counter + * could be assumed to be 12.5MHz, where the least significant bit neatly + * corresponded to 80 nanoseconds. + * + * Since Gen9 the numbers aren't so round, with a a frequency of 12MHz for + * SKL (or scale factor of 83.33333333) and a frequency of 19200000Hz for + * BXT. + * + * For simplicty to fit with the current code scaling by a single constant + * to map from raw timestamps to nanoseconds we now do the conversion in + * floating point instead of integer arithmetic. + * + * In general it's probably worth noting that the documented constants we + * have for the per-platform timestamp frequencies aren't perfect and + * shouldn't be trusted for scaling and comparing timestamps with a large + * delta. + * + * E.g. with crude testing on my system using the 'correct' scale factor I'm + * seeing a drift of ~2 milliseconds per second. + */ + uint64_t timestamp_frequency; + + /** @} */ +}; + +#define gen_device_info_is_9lp(devinfo) \ + ((devinfo)->is_broxton || (devinfo)->is_geminilake) + +int gen_get_pci_device_id_override(void); +int gen_device_name_to_pci_device_id(const char *name); +bool gen_get_device_info(int devid, struct gen_device_info *devinfo); +const char *gen_get_device_name(int devid); + +#ifdef __cplusplus +} +#endif + +#endif /* GEN_DEVICE_INFO_H */ diff --git a/src/intel/dev/meson.build b/src/intel/dev/meson.build new file mode 100644 index 00000000000..3346fe60c07 --- /dev/null +++ b/src/intel/dev/meson.build @@ -0,0 +1,33 @@ +# Copyright © 2017 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 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. + +# TODO: android? + +files_libintel_dev = files( + 'gen_device_info.c', + 'gen_device_info.h', +) + +libintel_dev = static_library( + ['intel_dev'], + files_libintel_dev, + include_directories : [inc_common, inc_intel], + c_args : [c_vis_args, no_override_init_args], +) |