1 /* fuc microcode for nve0 PGRAPH/GPC
3 * Copyright 2011 Red Hat Inc.
5 * Permission is hereby granted, free of charge, to any person obtaining a
6 * copy of this software and associated documentation files (the "Software"),
7 * to deal in the Software without restriction, including without limitation
8 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
9 * and/or sell copies of the Software, and to permit persons to whom the
10 * Software is furnished to do so, subject to the following conditions:
12 * The above copyright notice and this permission notice shall be included in
13 * all copies or substantial portions of the Software.
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 COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
19 * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
20 * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
21 * OTHER DEALINGS IN THE SOFTWARE.
27 * m4 nve0_grgpc.fuc | envyas -a -w -m fuc -V nva3 -o nve0_grgpc.fuc.h
31 * - bracket certain functions with scratch writes, useful for debugging
32 * - watchdog timer around ctx operations
35 .section #nve0_grgpc_data
38 gpc_mmio_list_head: .b32 0
39 gpc_mmio_list_tail: .b32 0
43 tpc_mmio_list_head: .b32 0
44 tpc_mmio_list_tail: .b32 0
48 // chipset descriptions
51 .b16 #nve4_gpc_mmio_head
52 .b16 #nve4_gpc_mmio_tail
53 .b16 #nve4_tpc_mmio_head
54 .b16 #nve4_tpc_mmio_tail
56 .b16 #nve4_gpc_mmio_head
57 .b16 #nve4_gpc_mmio_tail
58 .b16 #nve6_tpc_mmio_head
59 .b16 #nve6_tpc_mmio_tail
61 .b16 #nve4_gpc_mmio_head
62 .b16 #nve4_gpc_mmio_tail
63 .b16 #nve6_tpc_mmio_head
64 .b16 #nve6_tpc_mmio_tail
69 mmctx_data(0x000380, 1)
70 mmctx_data(0x000400, 2)
71 mmctx_data(0x00040c, 3)
72 mmctx_data(0x000450, 9)
73 mmctx_data(0x000600, 1)
74 mmctx_data(0x000684, 1)
75 mmctx_data(0x000700, 5)
76 mmctx_data(0x000800, 1)
77 mmctx_data(0x000808, 3)
78 mmctx_data(0x000828, 1)
79 mmctx_data(0x000830, 1)
80 mmctx_data(0x0008d8, 1)
81 mmctx_data(0x0008e0, 1)
82 mmctx_data(0x0008e8, 6)
83 mmctx_data(0x00091c, 1)
84 mmctx_data(0x000924, 3)
85 mmctx_data(0x000b00, 1)
86 mmctx_data(0x000b08, 6)
87 mmctx_data(0x000bb8, 1)
88 mmctx_data(0x000c08, 1)
89 mmctx_data(0x000c10, 8)
90 mmctx_data(0x000c40, 1)
91 mmctx_data(0x000c6c, 1)
92 mmctx_data(0x000c80, 1)
93 mmctx_data(0x000c8c, 1)
94 mmctx_data(0x001000, 3)
95 mmctx_data(0x001014, 1)
96 mmctx_data(0x003024, 1)
97 mmctx_data(0x0030c0, 2)
98 mmctx_data(0x0030e4, 1)
99 mmctx_data(0x003100, 6)
100 mmctx_data(0x0031d0, 1)
101 mmctx_data(0x0031e0, 2)
106 mmctx_data(0x000048, 1)
107 mmctx_data(0x000064, 1)
108 mmctx_data(0x000088, 1)
109 mmctx_data(0x000200, 6)
110 mmctx_data(0x00021c, 2)
111 mmctx_data(0x000230, 1)
112 mmctx_data(0x0002c4, 1)
113 mmctx_data(0x000400, 3)
114 mmctx_data(0x000420, 3)
115 mmctx_data(0x0004e8, 1)
116 mmctx_data(0x0004f4, 1)
117 mmctx_data(0x000604, 4)
118 mmctx_data(0x000644, 22)
119 mmctx_data(0x0006ac, 2)
120 mmctx_data(0x0006c8, 1)
121 mmctx_data(0x000730, 8)
122 mmctx_data(0x000758, 1)
123 mmctx_data(0x000778, 1)
127 mmctx_data(0x000048, 1)
128 mmctx_data(0x000064, 1)
129 mmctx_data(0x000088, 1)
130 mmctx_data(0x000200, 6)
131 mmctx_data(0x00021c, 2)
132 mmctx_data(0x000230, 1)
133 mmctx_data(0x0002c4, 1)
134 mmctx_data(0x000400, 3)
135 mmctx_data(0x000420, 3)
136 mmctx_data(0x0004e8, 1)
137 mmctx_data(0x0004f4, 1)
138 mmctx_data(0x000604, 4)
139 mmctx_data(0x000644, 22)
140 mmctx_data(0x0006ac, 2)
141 mmctx_data(0x0006c8, 1)
142 mmctx_data(0x000730, 8)
143 mmctx_data(0x000758, 1)
144 mmctx_data(0x000770, 1)
145 mmctx_data(0x000778, 2)
148 .section #nve0_grgpc_code
150 define(`include_code')
153 // reports an exception to the host
155 // In: $r15 error code (see nve0.fuc)
159 mov $r14 -0x67ec // 0x9814
161 call #nv_wr32 // HUB_CTXCTL_CC_SCRATCH[5] = error code
164 call #nv_wr32 // HUB_CTXCTL_INTR_UP_SET
168 // GPC fuc initialisation, executed by triggering ucode start, will
169 // fall through to main loop after completion.
172 // CC_SCRATCH[0]: chipset (PMC_BOOT_0 read returns 0x0bad0bad... sigh)
173 // CC_SCRATCH[1]: context base
177 // 31:31: set to signal completion
179 // 31:0: GPC context size
185 // enable fifo access
188 iowr I[$r1 + 0x000] $r2 // FIFO_ENABLE
190 // setup i0 handler, and route all interrupts to it
194 iowr I[$r1 + 0x300] $r0 // INTR_DISPATCH
196 // enable fifo interrupt
198 iowr I[$r1 + 0x000] $r2 // INTR_EN_SET
203 // figure out which GPC we are, and how many TPCs we have
206 iord $r2 I[$r1 + 0x000] // UNITS
211 st b32 D[$r0 + #tpc_count] $r2
212 st b32 D[$r0 + #tpc_mask] $r3
214 iord $r2 I[$r1 + 0x000] // MYINDEX
215 st b32 D[$r0 + #gpc_id] $r2
217 // find context data for this chipset
220 iord $r2 I[$r2 + 0x000] // CC_SCRATCH[0]
221 mov $r1 #chipsets - 12
224 ld b32 $r3 D[$r1 + 0x00]
228 bra ne #init_find_chipset
232 // initialise context base, and size tracking
236 iord $r2 I[$r2 + 0x100] // CC_SCRATCH[1], initial base
237 clear b32 $r3 // track GPC context size here
239 // set mmctx base addresses now so we don't have to do it later,
240 // they don't currently ever change
244 iowr I[$r4 + 0x000] $r5 // MMCTX_SAVE_SWBASE
245 iowr I[$r4 + 0x100] $r5 // MMCTX_LOAD_SWBASE
247 // calculate GPC mmio context size, store the chipset-specific
248 // mmio list pointers somewhere we can get at them later without
249 // re-parsing the chipset list
252 ld b16 $r14 D[$r1 + 4]
253 ld b16 $r15 D[$r1 + 6]
254 st b16 D[$r0 + #gpc_mmio_list_head] $r14
255 st b16 D[$r0 + #gpc_mmio_list_tail] $r15
260 // calculate per-TPC mmio context size, store the list pointers
261 ld b16 $r14 D[$r1 + 8]
262 ld b16 $r15 D[$r1 + 10]
263 st b16 D[$r0 + #tpc_mmio_list_head] $r14
264 st b16 D[$r0 + #tpc_mmio_list_tail] $r15
266 ld b32 $r14 D[$r0 + #tpc_count]
271 // round up base/size to 256 byte boundary (for strand SWBASE)
274 iowr I[$r4 + 0x000] $r3 // MMCTX_LOAD_COUNT, wtf for?!?
282 // calculate size of strand context data
284 call #strand_ctx_init
287 // save context size, and tell HUB we're done
290 iowr I[$r1 + 0x100] $r3 // CC_SCRATCH[1] = context size
294 iowr I[$r1 + 0x000] $r2 // CC_SCRATCH[0] |= 0x80000000
296 // Main program loop, very simple, sleeps until woken up by the interrupt
297 // handler, pulls a command from the queue and executes its handler
306 // 0x0000-0x0003 are all context transfers
308 bra nc #main_not_ctx_xfer
309 // fetch $flags and mask off $p1/$p2
314 // set $p1/$p2 according to transfer type
318 // transfer context data
324 or $r15 E_BAD_COMMAND
340 // incoming fifo command?
341 iord $r10 I[$r0 + 0x200] // INTR
342 and $r11 $r10 0x00000004
344 // queue incoming fifo command for later processing
347 iord $r14 I[$r11 + 0x100] // FIFO_CMD
348 iord $r15 I[$r11 + 0x000] // FIFO_DATA
352 iowr I[$r11 + 0x000] $r14 // FIFO_ACK
354 // ack, and wake up main()
356 iowr I[$r0 + 0x100] $r10 // INTR_ACK
370 // Set this GPC's bit in HUB_BAR, used to signal completion of various
371 // activities to the HUB fuc
375 ld b32 $r14 D[$r0 + #gpc_id]
377 mov $r14 -0x6be8 // 0x409418 - HUB_BAR_SET
382 // Disables various things, waits a bit, and re-enables them..
384 // Not sure how exactly this helps, perhaps "ENABLE" is not such a
385 // good description for the bits we turn off? Anyways, without this,
386 // funny things happen.
392 iowr I[$r14] $r15 // GPC_RED_SWITCH = POWER
396 bra ne #ctx_redswitch_delay
398 iowr I[$r14] $r15 // GPC_RED_SWITCH = UNK11, ENABLE, POWER
401 // Transfer GPC context data between GPU and storage area
403 // In: $r15 context base address
404 // $p1 clear on save, set on load
405 // $p2 set if opposite direction done/will be done, so:
406 // on save it means: "a load will follow this save"
407 // on load it means: "a save preceeded this load"
410 // set context base address
413 iowr I[$r1 + 0x000] $r15// MEM_BASE
414 bra not $p1 #ctx_xfer_not_load
422 iowr I[$r1] $r2 // STRAND_CMD(0x3f) = 0x0c
426 iowr I[$r2] $r0 // STRAND_FIRST_GENE(0x3f) = 0x00
429 iowr I[$r1] $r2 // STRAND_CMD(0x3f) = 0x03/0x04 (SAVE/LOAD)
432 xbit $r10 $flags $p1 // direction
436 ld b32 $r12 D[$r0 + #gpc_id]
438 add b32 $r11 $r12 // base = NV_PGRAPH_GPCn
439 ld b32 $r12 D[$r0 + #gpc_mmio_list_head]
440 ld b32 $r13 D[$r0 + #gpc_mmio_list_tail]
441 mov $r14 0 // not multi
444 // per-TPC mmio context
445 xbit $r10 $flags $p1 // direction
448 sethi $r11 0x500000 // base = NV_PGRAPH_GPC0_TPC0
449 ld b32 $r12 D[$r0 + #gpc_id]
451 add b32 $r11 $r12 // base = NV_PGRAPH_GPCn_TPC0
452 ld b32 $r12 D[$r0 + #tpc_mmio_list_head]
453 ld b32 $r13 D[$r0 + #tpc_mmio_list_tail]
454 ld b32 $r15 D[$r0 + #tpc_mask]
455 mov $r14 0x800 // stride = 0x800
458 // wait for strands to finish
461 // if load, or a save without a load following, do some
462 // unknown stuff that's done after finishing a block of
464 bra $p1 #ctx_xfer_post
465 bra not $p2 #ctx_xfer_done
470 iowr I[$r1] $r2 // STRAND_CMD(0x3f) = 0x0d
473 // mark completion in HUB's barrier
475 call #hub_barrier_done