1 /* 2 * Copyright 2013 Intel Corporation 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 (including the next 12 * paragraph) shall be included in all copies or substantial portions of the 13 * Software. 14 * 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 21 * IN THE SOFTWARE. 22 * 23 */ 24 25 #ifndef GEN_DEVICE_INFO_H 26 #define GEN_DEVICE_INFO_H 27 28 #include <stdbool.h> 29 30 /** 31 * Intel hardware information and quirks 32 */ 33 struct gen_device_info 34 { 35 int gen; /**< Generation number: 4, 5, 6, 7, ... */ 36 int gt; 37 38 bool is_g4x; 39 bool is_ivybridge; 40 bool is_baytrail; 41 bool is_haswell; 42 bool is_cherryview; 43 bool is_broxton; 44 bool is_kabylake; 45 46 bool has_hiz_and_separate_stencil; 47 bool must_use_separate_stencil; 48 49 bool has_llc; 50 51 bool has_pln; 52 bool has_compr4; 53 bool has_surface_tile_offset; 54 bool supports_simd16_3src; 55 bool has_resource_streamer; 56 57 /** 58 * \name Intel hardware quirks 59 * @{ 60 */ 61 bool has_negative_rhw_bug; 62 63 /** 64 * Some versions of Gen hardware don't do centroid interpolation correctly 65 * on unlit pixels, causing incorrect values for derivatives near triangle 66 * edges. Enabling this flag causes the fragment shader to use 67 * non-centroid interpolation for unlit pixels, at the expense of two extra 68 * fragment shader instructions. 69 */ 70 bool needs_unlit_centroid_workaround; 71 /** @} */ 72 73 /** 74 * \name GPU hardware limits 75 * 76 * In general, you can find shader thread maximums by looking at the "Maximum 77 * Number of Threads" field in the Intel PRM description of the 3DSTATE_VS, 78 * 3DSTATE_GS, 3DSTATE_HS, 3DSTATE_DS, and 3DSTATE_PS commands. URB entry 79 * limits come from the "Number of URB Entries" field in the 80 * 3DSTATE_URB_VS command and friends. 81 * 82 * These fields are used to calculate the scratch space to allocate. The 83 * amount of scratch space can be larger without being harmful on modern 84 * GPUs, however, prior to Haswell, programming the maximum number of threads 85 * to greater than the hardware maximum would cause GPU performance to tank. 86 * 87 * @{ 88 */ 89 /** 90 * Total number of slices present on the device whether or not they've been 91 * fused off. 92 * 93 * XXX: CS thread counts are limited by the inability to do cross subslice 94 * communication. It is the effectively the number of logical threads which 95 * can be executed in a subslice. Fuse configurations may cause this number 96 * to change, so we program @max_cs_threads as the lower maximum. 97 */ 98 unsigned num_slices; 99 unsigned max_vs_threads; /**< Maximum Vertex Shader threads */ 100 unsigned max_tcs_threads; /**< Maximum Hull Shader threads */ 101 unsigned max_tes_threads; /**< Maximum Domain Shader threads */ 102 unsigned max_gs_threads; /**< Maximum Geometry Shader threads. */ 103 /** 104 * Theoretical maximum number of Pixel Shader threads. 105 * 106 * PSD means Pixel Shader Dispatcher. On modern Intel GPUs, hardware will 107 * automatically scale pixel shader thread count, based on a single value 108 * programmed into 3DSTATE_PS. 109 * 110 * To calculate the maximum number of threads for Gen8 beyond (which have 111 * multiple Pixel Shader Dispatchers): 112 * 113 * - Look up 3DSTATE_PS and find "Maximum Number of Threads Per PSD" 114 * - Usually there's only one PSD per subslice, so use the number of 115 * subslices for number of PSDs. 116 * - For max_wm_threads, the total should be PSD threads * #PSDs. 117 */ 118 unsigned max_wm_threads; 119 120 /** 121 * Maximum Compute Shader threads. 122 * 123 * Thread count * number of EUs per subslice 124 */ 125 unsigned max_cs_threads; 126 127 struct { 128 /** 129 * Hardware default URB size. 130 * 131 * The units this is expressed in are somewhat inconsistent: 512b units 132 * on Gen4-5, KB on Gen6-7, and KB times the slice count on Gen8+. 133 * 134 * Look up "URB Size" in the "Device Attributes" page, and take the 135 * maximum. Look up the slice count for each GT SKU on the same page. 136 * urb.size = URB Size (kbytes) / slice count 137 */ 138 unsigned size; 139 140 /** 141 * The minimum number of URB entries. See the 3DSTATE_URB_<XS> docs. 142 */ 143 unsigned min_entries[4]; 144 145 /** 146 * The maximum number of URB entries. See the 3DSTATE_URB_<XS> docs. 147 */ 148 unsigned max_entries[4]; 149 } urb; 150 /** @} */ 151 }; 152 153 bool gen_get_device_info(int devid, struct gen_device_info *devinfo); 154 const char *gen_get_device_name(int devid); 155 156 #endif /* GEN_DEVICE_INFO_H */ 157
