1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
|
/*
* Copyright (C) 2019 Collabora, Ltd.
*
* 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.
*
* Authors:
* Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
*/
#include "util/u_math.h"
#include "pan_encoder.h"
/* Midgard has a small register file, so shaders with high register pressure
* need to spill from the register file onto the stack. In addition to
* spilling, it is desireable to allocate temporary arrays on the stack (for
* instance because the register file does not support indirect access but the
* stack does).
*
* The stack is located in "Thread Local Storage", sometimes abbreviated TLS in
* the kernel source code. Thread local storage is allocated per-thread,
* per-core, so threads executing concurrently do not interfere with each
* other's stacks. On modern kernels, we may query
* DRM_PANFROST_PARAM_THREAD_TLS_ALLOC for the number of threads per core we
* must allocate for, and DRM_PANFROST_PARAM_SHADER_PRESENT for a bitmask of
* shader cores (so take a popcount of that mask for the number of shader
* cores). On older kernels that do not support querying these values,
* following kbase, we may use the worst-case value of 256 threads for
* THREAD_TLS_ALLOC, and the worst-case value of 16 cores for Midgard per the
* "shader core count" column of the implementations table in
* https://en.wikipedia.org/wiki/Mali_%28GPU% [citation needed]
*
* Within a particular thread, there is stack allocated. If it is present, its
* size is a power-of-two, and it is at least 16 bytes. Stack is allocated
* with the shared memory descriptor used for all shaders within a frame (note
* that they don't execute concurrently so it's fine). So, consider the maximum
* stack size used by any shader within a job, and then compute (where npot
* denotes the next power of two):
*
* bytes/thread = npot(max(size, 16))
* allocated = (# of bytes/thread) * (# of threads/core) * (# of cores)
*
* The size of Thread Local Storage is signaled to the GPU in a dedicated
* log_stack_size field. Since stack sizes are powers of two, it follows that
* stack_size is logarithmic. Consider some sample values:
*
* stack size | log_stack_size
* ---------------------------
* 256 | 4
* 512 | 5
* 1024 | 6
*
* Noting that log2(256) = 8, we have the relation:
*
* stack_size <= 2^(log_stack_size + 4)
*
* Given the constraints about powers-of-two and the minimum of 256, we thus
* derive a formula for log_stack_size in terms of stack size (s), where s is
* positive:
*
* log_stack_size = ceil(log2(max(s, 16))) - 4
*
* There are other valid characterisations of this formula, of course, but this
* is computationally simple, so good enough for our purposes. If s=0, since
* there is no spilling used whatsoever, we may set log_stack_size to 0 to
* disable the stack.
*/
/* Computes log_stack_size = ceil(log2(max(s, 16))) - 4 */
unsigned
panfrost_get_stack_shift(unsigned stack_size)
{
if (stack_size)
return util_logbase2_ceil(MAX2(stack_size, 16)) - 4;
else
return 0;
}
/* Computes the aligned stack size given the shift and thread count. The blob
* reserves an extra page, and since this is hardware-internal, we do too. */
unsigned
panfrost_get_total_stack_size(
unsigned stack_shift,
unsigned threads_per_core,
unsigned core_count)
{
unsigned size_per_thread = MAX2(1 << (stack_shift + 4), 32);
unsigned size = size_per_thread * threads_per_core * core_count;
return size + 4096;
}
|
