Daniel Schürmann
b7af10449b
aco/ra: create nested affinities for loop header phis
...
Totals from 875 (0.58% of 150170) affected shaders: (GFX10.3)
CodeSize: 6084528 -> 6066628 (-0.29%); split: -0.32%, +0.02%
Instrs: 1136497 -> 1133565 (-0.26%); split: -0.28%, +0.02%
Latency: 23355051 -> 22952592 (-1.72%); split: -1.83%, +0.10%
InvThroughput: 13028151 -> 12859628 (-1.29%); split: -1.38%, +0.09%
Copies: 85673 -> 82790 (-3.37%); split: -3.62%, +0.26%
Branches: 25049 -> 25098 (+0.20%); split: -0.08%, +0.28%
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12836 >
2021-09-30 09:10:07 +00:00
Daniel Schürmann
d6bfc95732
aco/ra: create affinities between nested phis
...
Totals from 17143 (11.42% of 150170) affected shaders: (GFX10.3)
VGPRs: 1138112 -> 1138440 (+0.03%); split: -0.00%, +0.03%
CodeSize: 131235532 -> 131147080 (-0.07%); split: -0.14%, +0.07%
Instrs: 24848044 -> 24775419 (-0.29%); split: -0.32%, +0.02%
Latency: 599031816 -> 596005601 (-0.51%); split: -0.52%, +0.01%
InvThroughput: 152059329 -> 151054105 (-0.66%); split: -0.66%, +0.00%
VClause: 410951 -> 410958 (+0.00%); split: -0.01%, +0.01%
Copies: 1696885 -> 1621908 (-4.42%); split: -4.64%, +0.22%
Branches: 846710 -> 851052 (+0.51%); split: -0.29%, +0.80%
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12836 >
2021-09-30 09:10:07 +00:00
Daniel Schürmann
a2ffdca26a
aco/ra: don't set affinities for ssa-repair phis
...
These have no effect anymore.
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12836 >
2021-09-30 09:10:07 +00:00
Daniel Schürmann
ae5cbd8f3a
aco/ra: for phis try to find an operand-matching register earlier
...
Totals from 3557 (2.37% of 150170) affected shaders: (GFX10.3)
VGPRs: 257976 -> 257984 (+0.00%)
CodeSize: 34296232 -> 34270552 (-0.07%); split: -0.09%, +0.01%
Instrs: 6512289 -> 6506900 (-0.08%); split: -0.10%, +0.01%
Latency: 136376181 -> 136262553 (-0.08%); split: -0.10%, +0.02%
InvThroughput: 33042816 -> 32992849 (-0.15%); split: -0.18%, +0.03%
VClause: 104687 -> 104686 (-0.00%)
SClause: 238657 -> 238663 (+0.00%); split: -0.00%, +0.00%
Copies: 477690 -> 471058 (-1.39%); split: -1.52%, +0.13%
Branches: 223058 -> 224326 (+0.57%); split: -0.02%, +0.59%
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12836 >
2021-09-30 09:10:07 +00:00
Daniel Schürmann
8995599c80
aco/ra: try more aggressive to assign phi defs the same register
...
Totals from 4158 (2.77% of 150170) affected shaders: (GFX10.3)
VGPRs: 312008 -> 312000 (-0.00%)
CodeSize: 42902064 -> 42892200 (-0.02%); split: -0.06%, +0.04%
Instrs: 8086443 -> 8084532 (-0.02%); split: -0.07%, +0.05%
Latency: 138551153 -> 138215222 (-0.24%); split: -0.28%, +0.03%
InvThroughput: 39676773 -> 39570850 (-0.27%); split: -0.29%, +0.02%
SClause: 306299 -> 306284 (-0.00%); split: -0.01%, +0.00%
Copies: 552481 -> 553353 (+0.16%); split: -0.75%, +0.91%
Branches: 284381 -> 282409 (-0.69%); split: -0.74%, +0.04%
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12836 >
2021-09-30 09:10:07 +00:00
Daniel Schürmann
28fe49c35f
aco/ra: split register assignment for phis into separate function
...
No fossil-db changes.
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12836 >
2021-09-30 09:10:07 +00:00
Daniel Schürmann
302cb5c900
aco/ra: remove some redundant code
...
No fossil-db changes.
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12836 >
2021-09-30 09:10:07 +00:00
Daniel Schürmann
d7cb169819
aco/ra: refactor affinities into assignment struct
...
This lets us get rid of an unordered_map<>.
No fossil-db changes.
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12836 >
2021-09-30 09:10:07 +00:00
Daniel Schürmann
5afcc17f50
aco/ra: fix intersects()
...
The previous implementation failed when a contained b.
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12836 >
2021-09-30 09:10:07 +00:00
Daniel Schürmann
40a93e271c
aco: clang-format
...
No changes, just formatting.
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13087 >
2021-09-28 19:48:00 +00:00
Daniel Schürmann
23452f9eb0
aco/ra: don't copy linear VGPRs within CF in get_reg_create_vector()
...
Fixes: 6ed18749de ('aco: allow live-range splits of linear vgprs in top-level blocks')
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13058 >
2021-09-27 11:38:56 +00:00
Rhys Perry
6ed18749de
aco: allow live-range splits of linear vgprs in top-level blocks
...
Fixes dEQP-VK.ssbo.phys.layout.random.8bit.all_per_block_buffers.46 on
GFX8.
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12172 >
2021-09-17 14:36:03 +00:00
Rhys Perry
207b2cd434
aco: find scratch reg for sub-dword psuedo instructions which read sgprs
...
If there's a sgpr operand before a sub-dword operand, a scratch register
will not be found on GFX6/7.
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12172 >
2021-09-17 14:36:03 +00:00
Rhys Perry
f2a48f4eae
aco: find a scratch register for sub-dword copies on GFX7 if scc is empty
...
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12172 >
2021-09-17 14:36:03 +00:00
Rhys Perry
783609a849
aco: add RegClass::is_linear_vgpr helper
...
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12172 >
2021-09-17 14:36:03 +00:00
Daniel Schürmann
eb8ec12b23
aco/ra: Fix potential out-of-bounds array accesses.
...
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12748 >
2021-09-10 19:39:18 +00:00
Rhys Perry
c1e668d5d1
aco/ra: don't use ds_write_b8_d16_hi/ds_write_b16_d16_hi on GFX8
...
GFX8 doesn't support these opcodes.
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Fixes: c75138ed64 ("aco/ra: refactor subdword definition info")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12721 >
2021-09-06 15:10:26 +00:00
Rhys Perry
8037b21573
aco/ra: allow v1b operands with 16-bit instructions
...
Instruction selection can create these.
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Fixes: ec1bbfa608 ("aco/ra: refactor subdword operand stride")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12617 >
2021-09-03 13:21:28 +00:00
Rhys Perry
2a7fa132be
aco: implement udot_4x8/sdot_4x8/udot_2x16/sdot_2x16 opcodes
...
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12617 >
2021-09-03 13:21:28 +00:00
Daniel Schürmann
ec1bbfa608
aco/ra: refactor subdword operand stride
...
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12364 >
2021-08-23 10:31:40 +00:00
Daniel Schürmann
c75138ed64
aco/ra: refactor subdword definition info
...
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12364 >
2021-08-23 10:31:40 +00:00
Daniel Schürmann
3870c52159
aco/ra: don't allocate vector space for MIMG NSA operands
...
In this case, the MIMG vaddr components are not vector-aligned anymore, anyway.
Totals from 11866 (7.90% of 150170) affected shaders: (GFX10.3)
VGPRs: 733064 -> 728408 (-0.64%); split: -0.66%, +0.02%
CodeSize: 67968356 -> 67968440 (+0.00%); split: -0.02%, +0.02%
MaxWaves: 214022 -> 214014 (-0.00%)
Instrs: 12798200 -> 12797232 (-0.01%); split: -0.02%, +0.01%
Latency: 196427665 -> 196418706 (-0.00%); split: -0.02%, +0.01%
InvThroughput: 37082037 -> 37080799 (-0.00%); split: -0.02%, +0.02%
VClause: 246097 -> 246031 (-0.03%); split: -0.16%, +0.13%
Copies: 494852 -> 493923 (-0.19%); split: -0.52%, +0.34%
Branches: 220323 -> 220294 (-0.01%); split: -0.03%, +0.02%
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11980 >
2021-07-21 09:38:15 +00:00
Daniel Schürmann
1e2639026f
aco: Format.
...
Manually adjusted some comments for more intuitive line breaks.
Reviewed-by: Tony Wasserka <tony.wasserka@gmx.de>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11258 >
2021-07-12 21:27:31 +00:00
Daniel Schürmann
3f9e986d33
aco: add missing Licenses and remove Authors from files
...
Reviewed-by: Tony Wasserka <tony.wasserka@gmx.de>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11271 >
2021-07-12 12:09:31 +00:00
Daniel Schürmann
59fdaa1985
aco: reorder and cleanup #includes
...
Reviewed-by: Tony Wasserka <tony.wasserka@gmx.de>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11271 >
2021-07-12 12:09:31 +00:00
Daniel Schürmann
b14bd285f8
aco/ra: handle copies of copies better
...
Instead of adding a second copy, just redirect
the existing copy.
No fossil-db changes.
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11571 >
2021-06-24 16:53:10 +00:00
Daniel Schürmann
995e218993
aco/ra: handle copies of definition registers
...
Previously, it could happen that a parallelcopy of
a definition was inserted before the instruction.
Fixes Rage 2 with GFX7.
No fossil-db changes.
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11571 >
2021-06-24 16:53:10 +00:00
Rhys Perry
f8bf6b9e0a
aco/ra: use adjust_max_used_regs() in compact_relocate_vars()
...
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6489 >
2021-06-14 15:47:37 +00:00
Daniel Schürmann
bb1c06343d
aco/ra: refactor register assignment for vector operands
...
No functional changes.
Reviewed-by: Tony Wasserka <tony.wasserka@gmx.de>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8764 >
2021-06-11 12:35:46 +02:00
Daniel Schürmann
09b99f1b7c
aco/ra: refactor affinity coalescing
...
Also adds v_interp_p2_f32 to the list of
affinity-related instructions.
Totals from 68 (0.05% of 149839) affected shaders (GFX10.3):
CodeSize: 792928 -> 792056 (-0.11%)
Instrs: 152843 -> 152625 (-0.14%)
Latency: 1235353 -> 1235278 (-0.01%)
InvThroughput: 224087 -> 224049 (-0.02%)
Copies: 9218 -> 9000 (-2.36%)
Reviewed-by: Tony Wasserka <tony.wasserka@gmx.de>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8764 >
2021-06-11 12:35:31 +02:00
Daniel Schürmann
3a98f484d1
aco/ra: only create phi-affinities for killed operands
...
If a phi-operand is not killed, it must be copied anyway.
The additional affinity would only overwrite any potential
better affinity that was already created
Totals from 1067 (0.71% of 149839) affected shaders (GFX10.3):
VGPRs: 68072 -> 68064 (-0.01%)
CodeSize: 8252588 -> 8245220 (-0.09%); split: -0.12%, +0.03%
Instrs: 1596146 -> 1593941 (-0.14%); split: -0.16%, +0.02%
Latency: 18828176 -> 18823914 (-0.02%); split: -0.08%, +0.06%
InvThroughput: 3575063 -> 3574787 (-0.01%); split: -0.05%, +0.04%
VClause: 24345 -> 24325 (-0.08%); split: -0.16%, +0.07%
Copies: 88712 -> 87398 (-1.48%); split: -1.77%, +0.29%
Branches: 52067 -> 51364 (-1.35%); split: -1.38%, +0.03%
Reviewed-by: Tony Wasserka <tony.wasserka@gmx.de>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8764 >
2021-06-11 12:35:12 +02:00
Tony Wasserka
3b81f53e34
aco/ra: Split print_regs by lines of 64 registers
...
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10517 >
2021-06-08 17:03:08 +02:00
Tony Wasserka
69584478c9
aco/ra: Clean up print_regs output and support byte-allocated variables
...
Example output:
00 03 06 09 12 15 18 21 24 27 30 33 36 39 42
sgprs: ·▉█▉███▉▉█··████···········▉████············
00 03 06 09 12 15 18 21 24 27 30 33 36 39 42
vgprs: ▉▉··▉▉▉▉▘▀▉▉▉···▉▘▘▉▉▉▉···▉▉▉▀▀▉············
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10517 >
2021-06-08 17:03:08 +02:00
Tony Wasserka
5bfef2de66
aco/ra: Fix off-by-one-error in print_regs
...
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Fixes: 3675aefa84 ("aco/ra: Fix build with print_regs enabled")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10517 >
2021-06-08 17:03:08 +02:00
Rhys Perry
54292e99c7
aco: optimize 32-bit extracts and inserts using SDWA
...
Still need to use dst_u=preserve field to optimize packs
fossil-db (Sienna Cichlid):
Totals from 15974 (10.66% of 149839) affected shaders:
VGPRs: 1009064 -> 1008968 (-0.01%); split: -0.03%, +0.02%
SpillSGPRs: 7959 -> 7964 (+0.06%)
CodeSize: 101716436 -> 101159568 (-0.55%); split: -0.55%, +0.01%
MaxWaves: 284464 -> 284490 (+0.01%); split: +0.02%, -0.01%
Instrs: 19334216 -> 19224241 (-0.57%); split: -0.57%, +0.00%
Latency: 375465295 -> 375230478 (-0.06%); split: -0.14%, +0.08%
InvThroughput: 79006105 -> 78860705 (-0.18%); split: -0.25%, +0.07%
fossil-db (Polaris):
Totals from 11369 (7.51% of 151365) affected shaders:
SGPRs: 787920 -> 787680 (-0.03%); split: -0.04%, +0.01%
VGPRs: 681056 -> 681040 (-0.00%); split: -0.01%, +0.00%
CodeSize: 68127288 -> 67664120 (-0.68%); split: -0.69%, +0.01%
MaxWaves: 54370 -> 54371 (+0.00%)
Instrs: 13294638 -> 13214109 (-0.61%); split: -0.62%, +0.01%
Latency: 373515759 -> 373214571 (-0.08%); split: -0.11%, +0.03%
InvThroughput: 166529524 -> 166275291 (-0.15%); split: -0.20%, +0.05%
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/3151 >
2021-06-08 08:57:43 +00:00
Rhys Perry
fb31dda909
aco/ra: use flags instead of booleans for update_renames()
...
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10459 >
2021-05-17 13:31:07 +00:00
Rhys Perry
6fd6374e27
aco/ra: fix get_reg_for_operand() with vector operands
...
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10459 >
2021-05-17 13:31:07 +00:00
Rhys Perry
c08bfa110c
aco/ra: fix get_reg_for_operand() when the blocking var is a vector
...
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10459 >
2021-05-17 13:31:07 +00:00
Rhys Perry
bc95d55e1f
aco/ra: fix get_reg_for_operand() with no free registers
...
fossil-db (Sienna Cichlid):
Totals from 195 (0.13% of 149839) affected shaders:
CodeSize: 2352160 -> 2356720 (+0.19%); split: -0.00%, +0.20%
Instrs: 431976 -> 433124 (+0.27%); split: -0.00%, +0.27%
Latency: 10174434 -> 10174897 (+0.00%); split: -0.00%, +0.00%
InvThroughput: 4044388 -> 4044425 (+0.00%); split: -0.00%, +0.00%
Copies: 67634 -> 68762 (+1.67%); split: -0.00%, +1.67%
fossil-db (Polaris):
Totals from 186 (0.12% of 151365) affected shaders:
CodeSize: 2272356 -> 2276848 (+0.20%); split: -0.00%, +0.20%
Instrs: 432390 -> 433513 (+0.26%); split: -0.00%, +0.26%
Latency: 13153394 -> 13160194 (+0.05%); split: -0.00%, +0.05%
InvThroughput: 10889509 -> 10889967 (+0.00%); split: -0.00%, +0.00%
SClause: 12745 -> 12747 (+0.02%)
Copies: 74832 -> 75945 (+1.49%); split: -0.01%, +1.50%
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10459 >
2021-05-17 13:31:07 +00:00
Rhys Perry
4e459df0fc
aco/ra: initialize temp_in_scc earlier
...
We need to know if there's a temporary in SCC before the instruction, not
after.
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Fixes: 93c8ebfa78 ("aco: Initial commit of independent AMD compiler")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10459 >
2021-05-17 13:31:07 +00:00
Daniel Schürmann
b960169257
aco/ra: also prevent overflow register for p_create_vector operands
...
Fixes: d659ce0d6c ('aco/ra: prevent underflow register for p_create_vector operands')
Reviewed-by: Tony Wasserka <tony.wasserka@gmx.de>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10832 >
2021-05-17 11:18:25 +00:00
Daniel Schürmann
d659ce0d6c
aco/ra: prevent underflow register for p_create_vector operands
...
It could happen that we tested negative out-of-range
registers for p_create_vector operands resulting in a crash.
Fixes: 8962510e38 ('aco/ra: Conservatively refactor get_reg_specified to use PhysRegInterval')
Closes : #4697
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Tony Wasserka <tony.wasserka@gmx.de>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10799 >
2021-05-14 17:26:41 +00:00
Rhys Perry
776ba40115
aco: add and use Program::progress
...
This is used when printing the program and to avoid updating register
demand during post-RA liveness analysis.
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10315 >
2021-04-21 11:09:33 +00:00
Rhys Perry
5b8a4516e6
aco/ra: remove live-in temporary from live_out_per_block when moving it
...
Otherwise, handle_loop_phis() might pass it to handle_live_in() and then
we could have two phis for this variable.
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Fixes: 7c64623e94 ("aco/ra: refactor SSA repairing during register allocation")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10236 >
2021-04-14 19:04:08 +00:00
Rhys Perry
11fde1247c
aco/ra: use original names when renaming loop carried phi operands
...
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Fixes: 7c64623e94 ("aco/ra: refactor SSA repairing during register allocation")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10236 >
2021-04-14 19:04:08 +00:00
Daniel Schürmann
8e43abcd2c
aco/ra: remove exec handling for phis
...
These are not temporaries anymore.
Reviewed-by: Tony Wasserka <tony.wasserka@gmx.de>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9763 >
2021-03-25 15:12:19 +00:00
Daniel Schürmann
3284f755a3
aco/ra: allow m0 in get_reg_specified()
...
Totals from 1 (0.00% of 136546) affected shaders (Navi10):
CodeSize: 12788 -> 12776 (-0.09%)
Instrs: 2441 -> 2438 (-0.12%)
Latency: 29713 -> 29731 (+0.06%)
InvThroughput: 14857 -> 14866 (+0.06%)
Copies: 354 -> 353 (-0.28%)
Branches: 66 -> 65 (-1.52%)
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Tony Wasserka <tony.wasserka@gmx.de>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9763 >
2021-03-25 15:12:19 +00:00
Daniel Schürmann
4bfbd4de84
aco/ra: iterate backwards when coalescing phis
...
Aligning the phi definition with the operand from
the else- block can reduce the number of branches
if the else- block is otherwise empty.
Totals from 16 (0.01% of 136546) affected shaders (Navi10:
CodeSize: 707848 -> 707312 (-0.08%); split: -0.09%, +0.01%
Instrs: 126534 -> 126400 (-0.11%); split: -0.13%, +0.02%
Latency: 6399306 -> 6395082 (-0.07%)
InvThroughput: 6134374 -> 6132119 (-0.04%); split: -0.04%, +0.00%
SClause: 1879 -> 1871 (-0.43%)
Copies: 36316 -> 36219 (-0.27%); split: -0.37%, +0.10%
Branches: 4154 -> 4127 (-0.65%); split: -0.67%, +0.02%
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Tony Wasserka <tony.wasserka@gmx.de>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9763 >
2021-03-25 15:12:19 +00:00
Daniel Schürmann
7c64623e94
aco/ra: refactor SSA repairing during register allocation
...
The previous approach attempted to construct phi nodes
on-demand and on-the-fly. Due to several bugs, it became
necessary to always create incomplete phis for all live-in
variables on loop headers, which is highly inefficient.
The new approach assumes that live-in variables on loop-
headers don't get renamed, and afterwards does one renaming
pass per loop nest. This greatly simplifies the code and
reduces the memory footprint.
Totals from 37 (0.03% of 136546) affected shaders (Navi10):
CodeSize: 588148 -> 588020 (-0.02%); split: -0.03%, +0.01%
Instrs: 111793 -> 111761 (-0.03%); split: -0.04%, +0.01%
Latency: 4546013 -> 4545611 (-0.01%); split: -0.02%, +0.01%
InvThroughput: 2806217 -> 2805730 (-0.02%); split: -0.03%, +0.01%
VClause: 2044 -> 2046 (+0.10%)
SClause: 3889 -> 3884 (-0.13%)
Copies: 17730 -> 17700 (-0.17%); split: -0.23%, +0.06%
Branches: 3282 -> 3280 (-0.06%)
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Acked-by: Tony Wasserka <tony.wasserka@gmx.de>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9763 >
2021-03-25 15:12:19 +00:00
Daniel Schürmann
3ea2c05b32
aco/ra: split register_file initialization into separate function
...
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Tony Wasserka <tony.wasserka@gmx.de>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9763 >
2021-03-25 15:12:19 +00:00