OpenHarmony/third_party_mesa3d
2
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
161,155 Commits 9 Branches 1 Tag
bcb2cf96880c65de20f856b73298bc994dc9606c
Commit Graph

36 Commits

Author SHA1 Message Date
Alyssa Rosenzweig
bcb2cf9688 agx: Write to r0l with a "nesting" instruction 
This avoids modeling the r0l register explicitly in the IR, which would
complicate RA for little benefit at this stage. Do the simplest thing
that could possibly work in SSA.

glmark2 subset.

total instructions in shared programs: 6442 -> 6442 (0.00%)
instructions in affected programs: 701 -> 701 (0.00%)
helped: 4
HURT: 5
helped stats (abs) min: 1.0 max: 3.0 x̄: 2.00 x̃: 2
helped stats (rel) min: 1.46% max: 7.69% x̄: 4.03% x̃: 3.48%
HURT stats (abs)   min: 1.0 max: 3.0 x̄: 1.60 x̃: 1
HURT stats (rel)   min: 0.81% max: 7.41% x̄: 2.67% x̃: 1.14%
95% mean confidence interval for instructions value: -1.58 1.58
95% mean confidence interval for instructions %-change: -3.70% 3.08%
Inconclusive result (value mean confidence interval includes 0).

total bytes in shared programs: 42196 -> 42186 (-0.02%)
bytes in affected programs: 7768 -> 7758 (-0.13%)
helped: 8
HURT: 5
helped stats (abs) min: 2.0 max: 18.0 x̄: 7.25 x̃: 4
helped stats (rel) min: 0.13% max: 7.26% x̄: 2.02% x̃: 0.97%
HURT stats (abs)   min: 6.0 max: 18.0 x̄: 9.60 x̃: 6
HURT stats (rel)   min: 0.82% max: 6.32% x̄: 2.37% x̃: 1.02%
95% mean confidence interval for bytes value: -7.02 5.48
95% mean confidence interval for bytes %-change: -2.30% 1.63%
Inconclusive result (value mean confidence interval includes 0).

total halfregs in shared programs: 1926 -> 1769 (-8.15%)
halfregs in affected programs: 1395 -> 1238 (-11.25%)
helped: 71
HURT: 0
helped stats (abs) min: 1.0 max: 10.0 x̄: 2.21 x̃: 2
helped stats (rel) min: 1.92% max: 52.63% x̄: 15.33% x̃: 11.76%
95% mean confidence interval for halfregs value: -2.69 -1.73
95% mean confidence interval for halfregs %-change: -17.98% -12.68%
Halfregs are helped.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18804>
 2022-10-14 01:37:39 +00:00
Alyssa Rosenzweig
c9a96d4615 agx: Preload vertex/instance ID only at start 
This means we don't reserve the registers, which improves RA
considerably. Using a special preload psuedo-op instead of a regular
move allows us to constrain semantics and gaurantee coalescing.

shader-db on glmark2 subset:

total instructions in shared programs: 6448 -> 6442 (-0.09%)
instructions in affected programs: 230 -> 224 (-2.61%)
helped: 4
HURT: 0

total bytes in shared programs: 42232 -> 42196 (-0.09%)
bytes in affected programs: 1530 -> 1494 (-2.35%)
helped: 4
HURT: 0

total halfregs in shared programs: 2291 -> 1926 (-15.93%)
halfregs in affected programs: 2185 -> 1820 (-16.70%)
helped: 75
HURT: 0

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18804>
 2022-10-14 01:37:39 +00:00
Alyssa Rosenzweig
f665229d77 agx: Print agx_dim appropriately 
Easier to read, and gets us closer to proper disasm in Mesa.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18804>
 2022-10-14 01:37:39 +00:00
Alyssa Rosenzweig
1dcaade3e2 agx: Rename "combine" to "collect" 
For consistency with ir3 and bifrost.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18804>
 2022-10-14 01:37:39 +00:00
Alyssa Rosenzweig
82e8e709cb agx: Dynamically size split instruction 
This is more flexible.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18804>
 2022-10-14 01:37:39 +00:00
Alyssa Rosenzweig
7c9fba34bc agx: Switch to dynamic allocation of srcs/dests 
So we can handle parallel copies later.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18804>
 2022-10-14 01:37:39 +00:00
Alyssa Rosenzweig
c2bc8c1384 agx: Don't prefix pseudo-ops 
It's not really buying us anything and it clutters the IR.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18804>
 2022-10-14 01:37:39 +00:00
Alyssa Rosenzweig
5cd2371318 agx: Pass mask into ld/st_tile instructions 
Properly handle render target formats with <4 components.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18687>
 2022-09-22 03:23:36 +00:00
Alyssa Rosenzweig
640fd089a2 agx: Ensure that the optimizer sees legitimate SSA 
Expecting it to keep around unused definitions around is wishful. Add an
"anchoring" unit_test instruction to consume the results so they don't
have to be precoloured registers.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18687>
 2022-09-22 03:23:36 +00:00
Alyssa Rosenzweig
7a4e0a4d35 agx: Implement texture offsets and comparators 
Texture offsets and shadow comparison values get grouped into a vector
passed by register. Comparison values are provided as-is (fp32). Texture
offsets are packed into nibbles, but we can do this on the CPU, as
nonconstant offsets are forbidden in GLSL at least. They're also
forbidden in Vulkan/SPIR-V without ImageGatherExtended/
shaderImageGatherExtended. I'm happy kicking the NIR lowering can down
the line, this commit is complicated enough already.

Passes dEQP-GLES3.functional.shaders.texture_functions.texture.* and
dEQP-GLES3.functional.shaders.texture_functions.textureoffset.*

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18525>
 2022-09-13 16:04:29 +00:00
Alyssa Rosenzweig
4f85a7be8c agx: Make p_combine take a dynamic src count 
For larger vectors.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18525>
 2022-09-13 16:04:28 +00:00
Alyssa Rosenzweig
10c9aceadf agx: Plumb through nir_op_txf 
nir_op_txf maps to AGX's texture_load, which is encoded like
texture_sample. The one outstanding question with this commit is the
rules for the sample required ... in NIR and APIs (including Metal), txf
is samplerless, but the AGX instruction consumes a sampler. Hopefully
any sampler will do, though?

Passes dEQP-GLES3.functional.shaders.texture_functions.texelfetch.*

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18380>
 2022-09-04 18:05:31 +00:00
Alyssa Rosenzweig
17168162fb agx: Remove p_extract 
It's now unused. We didn't have coalescing for it anyway, splits are the
preferred alternative.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17198>
 2022-08-07 20:43:54 -04:00
Alyssa Rosenzweig
744b4a0f7f agx: Rename varying load instructions 
Unlike Mali (where I borrowed the old names from), these are not loads
in the memory sense. They are simply register loads and arithmetic.
Rename accordingly, using PowerVR names and public Apple names as a
guide.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17198>
 2022-08-07 20:43:53 -04:00
Alyssa Rosenzweig
14bf020d75 agx: Model perspective coefficient reg in the IR 
For perspective-correct interpolation, the W coefficient register is
needed. Instead of hardcoding this to cf0 and special casing, model this
in the IR and let the general handling kick in.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17198>
 2022-08-07 20:43:53 -04:00
Alyssa Rosenzweig
a2a947031a agx: Mark the logical ends of blocks 
We need to insert parallel copies at the logical end of blocks, before branches.
Add a pseudo instruction signaling that. Cribbed from ACO.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16268>
 2022-05-01 22:00:00 -04:00
Alyssa Rosenzweig
d285c63417 agx: Add phi pseudo instruction 
For SSA.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16268>
 2022-05-01 21:58:29 -04:00
Alyssa Rosenzweig
2a4a8a8902 agx: Add p_split pseudoinstruction 
Easier on RA for extracts.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16268>
 2022-05-01 21:58:29 -04:00
Alyssa Rosenzweig
7d38bcb7ee agx: Use pseudo ops for mov/not/and/xor/or 
Rather than using builder magic (implicitly lowered on emit), add actual pseudo
operations (explicitly lowered before encoding). In theory this is slower, I
doubt it matters. This makes the instruction aliases first-class for IR prining
and machine inspection, which will make optimization passes easier to write.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16268>
 2022-05-01 21:58:29 -04:00
Alyssa Rosenzweig
f0a973081f agx: Use correct types for some IR enums 
Otherwise there are implicit int->enum casts which prevent us from building as
C++ (with -fpermissive).

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16268>
 2022-05-01 21:58:29 -04:00
Alyssa Rosenzweig
f248f6623c agx: Add sample_mask instruction 
Sets the output sample mask to a given 8-bit immediate or 16-bit
register. Also used to implement discards, which is my ES2 interest.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/14219>
 2022-01-16 18:23:28 +00:00
Alyssa Rosenzweig
3d3e4928b2 agx: Add ld_vary_flat opcode 
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11718>
 2021-07-05 20:56:03 +00:00
Alyssa Rosenzweig
10b8563966 agx: Update ld_vary encoding mask 
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11718>
 2021-07-05 20:56:03 +00:00
Alyssa Rosenzweig
99b67ab355 agx: Add perspective bit to ld_var 
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11718>
 2021-07-05 20:56:03 +00:00
Alyssa Rosenzweig
307b8f1b2f agx: List sr enum in Python 
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11199>
 2021-06-05 20:38:22 +00:00
Alyssa Rosenzweig
cc8fec8b74 agx: Generate enums from Python 
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11199>
 2021-06-05 20:38:22 +00:00
Alyssa Rosenzweig
223476aff3 agx: Model get_sr 
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11199>
 2021-06-05 20:38:22 +00:00
Alyssa Rosenzweig
4fb964ccb8 agx: Model jump instructions 
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11023>
 2021-05-31 01:23:35 +05:30
Alyssa Rosenzweig
4eb8fbf780 agx: Model pop_exec 
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11023>
 2021-05-31 01:23:35 +05:30
Alyssa Rosenzweig
a270665a9e agx: Model control flow instructions 
Thankfully the encoding is regular, if a bit complicated.

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11023>
 2021-05-31 01:23:35 +05:30
Alyssa Rosenzweig
8454d08aa3 agx: Add branch target to IR 
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11023>
 2021-05-31 01:23:35 +05:30
Alyssa Rosenzweig
ad4dfb3321 agx: Add invert_cond (ccn) to IR 
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11023>
 2021-05-31 01:23:35 +05:30
Alyssa Rosenzweig
21cf528e76 agx: Add nest field to IR 
Needed to model control flow instructions.

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11023>
 2021-05-31 01:23:35 +05:30
Alyssa Rosenzweig
1164c992cf agx: Add ld_tile opcode 
Variant of st_tile.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10720>
 2021-05-09 23:29:45 -04:00
Alyssa Rosenzweig
07fdc0015e agx: Rename blend -> st_tile 
For symmetry.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10720>
 2021-05-09 23:29:45 -04:00
Alyssa Rosenzweig
50b5c94885 agx: Add opcode descriptions as Python 
Pattern lifted from NIR.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Acked-by: Jason Ekstrand <jason@jlekstrand.net>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10582>
 2021-05-02 17:41:09 -04:00