Commit graph

91452 commits

Author SHA1 Message Date
Marek Olšák
4e50062028 radeonsi: pass tessellation ring addresses via user SGPRs
This removes s_load_dword latency for tess rings.

We need just 1 SGPR for the address if we use 64K alignment. The final asm
for recreating the descriptor is:

    // s2 is (address >> 16)
    s_mov_b32 s3, 0
    s_lshl_b64 s[4:5], s[2:3], 16
    s_mov_b32 s6, -1
    s_mov_b32 s7, 0x27fac

v2: bitcast the descriptor type from v2i64 to v4i32

Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
2017-04-28 21:47:35 +02:00
Marek Olšák
2823e15f60 radeonsi: use si_insert_input_ret in si_llvm_emit_tcs_epilogue
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
2017-04-28 21:47:35 +02:00
Marek Olšák
9fd9a7d0ba radeonsi: remove VS epilog code, compile VS with PrimID export on demand
The use of PrimID in the pixel shader is too rare to deserve such
a sizable support code.

The initial idea of the VS epilog was to move the clipping code there and
remove it based on states, but optimized variants are now used to do that
and are easier to support, so the VS epilog has turned out to be not so
useful.

Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
2017-04-28 21:47:35 +02:00
Marek Olšák
3b2e93e472 radeonsi: get InstanceID from VGPR1 (or VGPR2 for tess) instead of VGPR3
VGPR1 = InstanceID / StepRate0; // StepRate0 can be set to 1

Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
2017-04-28 21:47:35 +02:00
Marek Olšák
678d568c7b radeonsi: don't load PrimID in TES if it's not used
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
2017-04-28 21:47:35 +02:00
Marek Olšák
808c33f6f0 radeonsi: explain (non-)monolithic shaders
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
2017-04-28 21:47:35 +02:00
Marek Olšák
fc478248f3 radeonsi/gfx9: enable OpenGL 4.5
Tentatively enable it, expecting the scratch buffer support to be done before
the next Mesa release.

Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
2017-04-28 21:47:35 +02:00
Marek Olšák
ed9a51cd3b radeonsi/gfx9: 2nd shader of merged shaders should hold a reference of the 1st
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
2017-04-28 21:47:35 +02:00
Marek Olšák
ef40937854 radeonsi: add reference counting for shader selectors
The 2nd shader of merged shaders should take a reference of the 1st shader.
The next commit will do that.

Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
2017-04-28 21:47:35 +02:00
Marek Olšák
6c15e15af4 radeonsi/gfx9: set VGT_VERTEX_REUSE for ES in ES-GS
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
2017-04-28 21:47:35 +02:00
Marek Olšák
887ef1de34 radeonsi/gfx9: set TES registers for merged ES-GS
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
2017-04-28 21:47:35 +02:00
Marek Olšák
49cd0cbfd5 radeonsi/gfx9: disallow scratch buffer for LS-HS and ES-GS
not implemented yet

Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
2017-04-28 21:47:35 +02:00
Marek Olšák
2857b14bba radeonsi/gfx9: always compile monolithic ES-GS (asynchronously)
In addition to the non-monolithic variant.

Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
2017-04-28 21:47:35 +02:00
Marek Olšák
a82398a8f5 radeonsi/gfx9: add support for monolithic ES-GS
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
2017-04-28 21:47:35 +02:00
Marek Olšák
6a9c20fdd5 radeonsi/gfx9: make sure the 1st shader's main part exists for merged shaders
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
2017-04-28 21:47:35 +02:00
Marek Olšák
7df682c291 radeonsi/gfx9: select shader parts for non-monolithic ES-GS
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
2017-04-28 21:47:35 +02:00
Marek Olšák
cd99c442c4 radeonsi/gfx9: add GS prolog support for merged ES-GS
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
2017-04-28 21:47:35 +02:00
Marek Olšák
e0570bc283 radeonsi/gfx9: add VS prolog support for merged ES-GS
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
2017-04-28 21:47:35 +02:00
Marek Olšák
6b93452b24 radeonsi/gfx9: pass GS input SGPRs and VGPRs from the ES part to GS
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
2017-04-28 21:47:35 +02:00
Marek Olšák
37e22ab65e radeonsi/gfx9: store ES outputs to LDS
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
2017-04-28 21:47:35 +02:00
Marek Olšák
d616c57342 radeonsi/gfx9: load GS inputs from LDS
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
2017-04-28 21:47:35 +02:00
Marek Olšák
fc781fa0ab radeonsi/gfx9: get GS wave ID from the correct input
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
2017-04-28 21:47:35 +02:00
Marek Olšák
bcaf905129 radeonsi/gfx9: add the function signature of merged ES-GS
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
2017-04-28 21:47:35 +02:00
Marek Olšák
8b220877ad radeonsi/gfx9: set registers and shader key for merged ES-GS
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
2017-04-28 21:47:35 +02:00
Marek Olšák
ab197ad8d1 radeonsi/gfx9: add GS user SGPRs
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
2017-04-28 21:47:35 +02:00
Marek Olšák
b2f5d03152 radeonsi: rename declare_tess_lds -> declare_lds_as_pointer
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
2017-04-28 21:47:35 +02:00
Marek Olšák
e3caa1cd36 radeonsi: simplify some shader type conditions
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
2017-04-28 21:47:35 +02:00
Marek Olšák
021e65640e radeonsi: rename the swizzle parameter of lds_store
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
2017-04-28 21:47:35 +02:00
Marek Olšák
dcea7e5d19 radeonsi: add si_shader::prolog2
For a GS prolog in merged ES-GS.

Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
2017-04-28 21:47:35 +02:00
Marek Olšák
eb35238ffe radeonsi/gfx9: move RW_BUFFERS to s[0:1] for merged shaders
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
2017-04-28 21:47:35 +02:00
Marek Olšák
0af00f179e radeonsi/gfx9: add support for monolithic merged LS-HS
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
2017-04-28 21:47:35 +02:00
Marek Olšák
0d6d25475d radeonsi/gfx9: set EXEC for non-mono merged shaders, add a barrier between them
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
2017-04-28 21:47:35 +02:00
Marek Olšák
a84a6feac9 radeonsi/gfx9: don't store the HS control word
GFX9 doesn't have it.

Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
2017-04-28 21:47:35 +02:00
Marek Olšák
1d90ecd3a5 radeonsi/gfx9: pass inputs from LS to TCS
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
2017-04-28 21:47:35 +02:00
Marek Olšák
cbd1bc2e3e radeonsi/gfx9: add TCS epilog support for merged LS-HS
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
2017-04-28 21:47:35 +02:00
Marek Olšák
f11ced475e radeonsi/gfx9: add VS prolog support for merged LS-HS
HS input VGPRs must be reserved.

Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
2017-04-28 21:47:35 +02:00
Marek Olšák
82a0e4f658 radeonsi/gfx9: merged shaders have scratch offset at the beginning
also, screen wasn't initialized for compute shaders

Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
2017-04-28 21:47:35 +02:00
Marek Olšák
0c253557b2 radeonsi/gfx9: define LS-HS main shader function prototype
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
2017-04-28 21:47:35 +02:00
Marek Olšák
852ea69a2d radeonsi: assign VS/TCS/TES/GS shader input parameter locations dynamically
They will vary with merged stages.

Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
2017-04-28 21:47:35 +02:00
Marek Olšák
067dacd1b1 radeonsi/gfx9: define and set LS-HS user SGPRs
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
2017-04-28 21:47:35 +02:00
Marek Olšák
0588146cb0 radeonsi/gfx9: set up shader registers for merged LS-HS
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
2017-04-28 21:47:35 +02:00
Marek Olšák
62abdb17bb radeonsi/gfx9: add initial code generation for non-monolithic merged LS-HS
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
2017-04-28 21:47:35 +02:00
Marek Olšák
c73d9bd643 radeonsi: separate out code for selecting the VS prolog
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
2017-04-28 21:47:35 +02:00
Marek Olšák
a98c9ba580 radeonsi/gfx9: add si_shader::previous_stage for merged shaders
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
2017-04-28 21:47:35 +02:00
Marek Olšák
cfb0798bb3 radeonsi/gfx9: enlarge num_input_sgprs in shader keys due to higher hw limit
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
2017-04-28 21:47:35 +02:00
Marek Olšák
4ab36e0ebc radeonsi/gfx9: update the summary of shader stage configs
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
2017-04-28 21:47:35 +02:00
Marek Olšák
9d6ed572d9 radeonsi: adjust the signature of si_get_vs_prolog_key
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
2017-04-28 21:47:35 +02:00
Marek Olšák
b1ed3ffc56 radeonsi: separate out VS prolog key generation
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
2017-04-28 21:47:35 +02:00
Marek Olšák
e4542f00ce radeonsi: separate out VS prolog key printing
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
2017-04-28 21:47:35 +02:00
Marek Olšák
983d7e743e radeonsi: code shuffling in si_emit_derived_tess_state
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
2017-04-28 21:47:35 +02:00