2 * Copyright © 2013 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
25 #ifndef GEN_DEVICE_INFO_H
26 #define GEN_DEVICE_INFO_H
32 * Intel hardware information and quirks
34 struct gen_device_info
36 int gen
; /**< Generation number: 4, 5, 6, 7, ... */
52 bool has_hiz_and_separate_stencil
;
53 bool must_use_separate_stencil
;
59 bool has_surface_tile_offset
;
60 bool supports_simd16_3src
;
61 bool has_resource_streamer
;
64 * \name Intel hardware quirks
67 bool has_negative_rhw_bug
;
70 * Some versions of Gen hardware don't do centroid interpolation correctly
71 * on unlit pixels, causing incorrect values for derivatives near triangle
72 * edges. Enabling this flag causes the fragment shader to use
73 * non-centroid interpolation for unlit pixels, at the expense of two extra
74 * fragment shader instructions.
76 bool needs_unlit_centroid_workaround
;
80 * \name GPU hardware limits
82 * In general, you can find shader thread maximums by looking at the "Maximum
83 * Number of Threads" field in the Intel PRM description of the 3DSTATE_VS,
84 * 3DSTATE_GS, 3DSTATE_HS, 3DSTATE_DS, and 3DSTATE_PS commands. URB entry
85 * limits come from the "Number of URB Entries" field in the
86 * 3DSTATE_URB_VS command and friends.
88 * These fields are used to calculate the scratch space to allocate. The
89 * amount of scratch space can be larger without being harmful on modern
90 * GPUs, however, prior to Haswell, programming the maximum number of threads
91 * to greater than the hardware maximum would cause GPU performance to tank.
96 * Total number of slices present on the device whether or not they've been
99 * XXX: CS thread counts are limited by the inability to do cross subslice
100 * communication. It is the effectively the number of logical threads which
101 * can be executed in a subslice. Fuse configurations may cause this number
102 * to change, so we program @max_cs_threads as the lower maximum.
107 * Number of subslices for each slice (used to be uniform until CNL).
109 unsigned num_subslices
[3];
112 * Number of threads per eu, varies between 4 and 8 between generations.
114 unsigned num_thread_per_eu
;
117 unsigned max_vs_threads
; /**< Maximum Vertex Shader threads */
118 unsigned max_tcs_threads
; /**< Maximum Hull Shader threads */
119 unsigned max_tes_threads
; /**< Maximum Domain Shader threads */
120 unsigned max_gs_threads
; /**< Maximum Geometry Shader threads. */
122 * Theoretical maximum number of Pixel Shader threads.
124 * PSD means Pixel Shader Dispatcher. On modern Intel GPUs, hardware will
125 * automatically scale pixel shader thread count, based on a single value
126 * programmed into 3DSTATE_PS.
128 * To calculate the maximum number of threads for Gen8 beyond (which have
129 * multiple Pixel Shader Dispatchers):
131 * - Look up 3DSTATE_PS and find "Maximum Number of Threads Per PSD"
132 * - Usually there's only one PSD per subslice, so use the number of
133 * subslices for number of PSDs.
134 * - For max_wm_threads, the total should be PSD threads * #PSDs.
136 unsigned max_wm_threads
;
139 * Maximum Compute Shader threads.
141 * Thread count * number of EUs per subslice
143 unsigned max_cs_threads
;
147 * Hardware default URB size.
149 * The units this is expressed in are somewhat inconsistent: 512b units
150 * on Gen4-5, KB on Gen6-7, and KB times the slice count on Gen8+.
152 * Look up "URB Size" in the "Device Attributes" page, and take the
153 * maximum. Look up the slice count for each GT SKU on the same page.
154 * urb.size = URB Size (kbytes) / slice count
159 * The minimum number of URB entries. See the 3DSTATE_URB_<XS> docs.
161 unsigned min_entries
[4];
164 * The maximum number of URB entries. See the 3DSTATE_URB_<XS> docs.
166 unsigned max_entries
[4];
170 * For the longest time the timestamp frequency for Gen's timestamp counter
171 * could be assumed to be 12.5MHz, where the least significant bit neatly
172 * corresponded to 80 nanoseconds.
174 * Since Gen9 the numbers aren't so round, with a a frequency of 12MHz for
175 * SKL (or scale factor of 83.33333333) and a frequency of 19200000Hz for
178 * For simplicty to fit with the current code scaling by a single constant
179 * to map from raw timestamps to nanoseconds we now do the conversion in
180 * floating point instead of integer arithmetic.
182 * In general it's probably worth noting that the documented constants we
183 * have for the per-platform timestamp frequencies aren't perfect and
184 * shouldn't be trusted for scaling and comparing timestamps with a large
187 * E.g. with crude testing on my system using the 'correct' scale factor I'm
188 * seeing a drift of ~2 milliseconds per second.
190 uint64_t timestamp_frequency
;
195 #define gen_device_info_is_9lp(devinfo) \
196 ((devinfo)->is_broxton || (devinfo)->is_geminilake)
198 bool gen_get_device_info(int devid
, struct gen_device_info
*devinfo
);
199 const char *gen_get_device_name(int devid
);
201 #endif /* GEN_DEVICE_INFO_H */