2 * Copyright (C) 2019 Collabora, Ltd.
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 FROM,
20 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
24 * Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
27 #include "util/u_math.h"
28 #include "util/macros.h"
29 #include "pan_encoder.h"
31 /* Midgard has a small register file, so shaders with high register pressure
32 * need to spill from the register file onto the stack. In addition to
33 * spilling, it is desireable to allocate temporary arrays on the stack (for
34 * instance because the register file does not support indirect access but the
37 * The stack is located in "Thread Local Storage", sometimes abbreviated TLS in
38 * the kernel source code. Thread local storage is allocated per-thread,
39 * per-core, so threads executing concurrently do not interfere with each
40 * other's stacks. On modern kernels, we may query
41 * DRM_PANFROST_PARAM_THREAD_TLS_ALLOC for the number of threads per core we
42 * must allocate for, and DRM_PANFROST_PARAM_SHADER_PRESENT for a bitmask of
43 * shader cores (so take a popcount of that mask for the number of shader
44 * cores). On older kernels that do not support querying these values,
45 * following kbase, we may use the worst-case value of 256 threads for
46 * THREAD_TLS_ALLOC, and the worst-case value of 16 cores for Midgard per the
47 * "shader core count" column of the implementations table in
48 * https://en.wikipedia.org/wiki/Mali_%28GPU% [citation needed]
50 * Within a particular thread, there is stack allocated. If it is present, its
51 * size is a power-of-two, and it is at least 16 bytes. Stack is allocated
52 * with the shared memory descriptor used for all shaders within a frame (note
53 * that they don't execute concurrently so it's fine). So, consider the maximum
54 * stack size used by any shader within a job, and then compute (where npot
55 * denotes the next power of two):
57 * bytes/thread = npot(max(size, 16))
58 * allocated = (# of bytes/thread) * (# of threads/core) * (# of cores)
60 * The size of Thread Local Storage is signaled to the GPU in a dedicated
61 * log_stack_size field. Since stack sizes are powers of two, it follows that
62 * stack_size is logarithmic. Consider some sample values:
64 * stack size | log_stack_size
65 * ---------------------------
70 * Noting that log2(256) = 8, we have the relation:
72 * stack_size <= 2^(log_stack_size + 4)
74 * Given the constraints about powers-of-two and the minimum of 256, we thus
75 * derive a formula for log_stack_size in terms of stack size (s), where s is
78 * log_stack_size = ceil(log2(max(s, 16))) - 4
80 * There are other valid characterisations of this formula, of course, but this
81 * is computationally simple, so good enough for our purposes. If s=0, since
82 * there is no spilling used whatsoever, we may set log_stack_size to 0 to
86 /* Computes log_stack_size = ceil(log2(max(s, 16))) - 4 */
89 panfrost_get_stack_shift(unsigned stack_size
)
92 return util_logbase2_ceil(MAX2(stack_size
, 16)) - 4;
97 /* Computes the aligned stack size given the shift and thread count. */
100 panfrost_get_total_stack_size(
101 unsigned thread_size
,
102 unsigned threads_per_core
,
105 unsigned size_per_thread
= (thread_size
== 0) ? 0 :
106 util_next_power_of_two(ALIGN_POT(thread_size
, 16));
108 return size_per_thread
* threads_per_core
* core_count
;