Commit Graph

225 Commits

Author SHA1 Message Date
Nicolai Hähnle 227c29a80d amd/common/gfx10: implement scan & reduce operations
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
2019-07-03 15:51:12 -04:00
Marek Olšák ac4b1e2f0a radeonsi: set the calling convention for inlined function calls
otherwise the behavior is undefined

Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Tested-by: Dieter Nützel <Dieter@nuetzel-hh.de>
2019-06-24 21:04:10 -04:00
Daniel Schürmann 0daeb1d127 amd/common: lower bitfield_extract to ubfe/ibfe.
Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
2019-06-24 18:42:20 +02:00
Connor Abbott 3bf8981c51 ac,radeonsi: Always mark buffer stores as inaccessiblememonly
inaccessiblememonly means that it doesn't modify memory accesible via
normal LLVM pointers. This lets LLVM's dead store elimination, memcpy
forwarding, etc. ignore functions with this attribute. We don't
represent descriptors as pointers, so this property is always true of
buffer and image stores. There are plans to represent descriptors via
pointers, but this just means that now nothing is inaccessiblememonly,
as LLVM will then understand loads/stores via its usual alias analysis.

Radeonsi was mistakenly only setting it if the driver could prove that
there were no reads, and then it was cargo-culted into ac_llvm_build
and ac_llvm_to_nir. Rip it out of everything.

statistics with nir enabled:

Totals from affected shaders:
SGPRS: 152 -> 152 (0.00 %)
VGPRS: 128 -> 132 (3.12 %)
Spilled SGPRs: 0 -> 0 (0.00 %)
Spilled VGPRs: 0 -> 0 (0.00 %)
Private memory VGPRs: 0 -> 0 (0.00 %)
Scratch size: 0 -> 0 (0.00 %) dwords per thread
Code Size: 9324 -> 9244 (-0.86 %) bytes
LDS: 2 -> 2 (0.00 %) blocks
Max Waves: 17 -> 17 (0.00 %)
Wait states: 0 -> 0 (0.00 %)

The only difference was a manhattan31 shader.

Acked-by: Timothy Arceri <tarceri@itsqueeze.com>
Acked-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
2019-06-19 14:08:27 +02:00
Samuel Pitoiset b5012a0518 ac: update llvm.amdgcn.icmp intrinsic name for LLVM 9+
LLVM r363339 changed llvm.amdgcn.icmp.i* to llvm.amdgcn.icmp.i64.i*.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-By: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
2019-06-17 08:58:33 +02:00
Marek Olšák 4773f5a293 radeonsi: use the ac helper for index buffer stores in the culling shader 2019-06-11 20:05:21 -04:00
Marek Olšák ff63b99531 ac: rename LLVM <= 7 helpers for readability
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
2019-06-04 18:53:46 -04:00
Marek Olšák c9b64b58de ac: fix a typo in ac_build_wg_scan_bottom
Cc: 19.1 <mesa-stable@lists.freedesktop.org>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
2019-06-04 18:53:46 -04:00
Samuel Pitoiset 33f4e04d5a ac,radv: do not emit vec3 for raw load/store on SI
It's unsupported, only load/store format with vec3 are supported.

Fixes: 6970a9a6ca ("ac,radv: remove the vec3 restriction with LLVM 9+")"
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
2019-06-04 08:47:26 +02:00
Samuel Pitoiset 6970a9a6ca ac,radv: remove the vec3 restriction with LLVM 9+
This changes requires LLVM r356755.

32706 shaders in 16744 tests
Totals:
SGPRS: 1448848 -> 1455984 (0.49 %)
VGPRS: 1016684 -> 1016220 (-0.05 %)
Spilled SGPRs: 25871 -> 25815 (-0.22 %)
Spilled VGPRs: 122 -> 122 (0.00 %)
Scratch size: 11964 -> 11956 (-0.07 %) dwords per thread
Code Size: 55324500 -> 55301152 (-0.04 %) bytes
Max Waves: 235660 -> 235586 (-0.03 %)

Totals from affected shaders:
SGPRS: 293704 -> 300840 (2.43 %)
VGPRS: 246716 -> 246252 (-0.19 %)
Spilled SGPRs: 159 -> 103 (-35.22 %)
Scratch size: 188 -> 180 (-4.26 %) dwords per thread
Code Size: 8653664 -> 8630316 (-0.27 %) bytes
Max Waves: 60811 -> 60737 (-0.12 %)

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
2019-06-03 11:30:08 +02:00
Marek Olšák ccfcb9d818 ac: rename SI-CIK-VI to GFX6-GFX7-GFX8
Acked-by: Dave Airlie <airlied@redhat.com>

We already use GFX9 and I don't want us to have confusing naming
in the driver. GFXn naming is better from the driver perspective,
because it's the real version of the gfx portion of the hw. Also,
CIK means Bonaire-Kaveri-Kabini, it doesn't mean CI.

It shouldn't confuse our SDMA, UVD, VCE etc. code much. Those have
nothing to do with GFXn and they have their own version numbers.
2019-05-15 20:54:10 -04:00
Marek Olšák 6b0b8f132a ac: use 1D GEPs for descriptors and constants
just a cleanup

Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Tested-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
2019-05-14 15:15:11 -04:00
Nicolai Hähnle 81fe33735a amd/common: add ac_build_opencoded_fetch_format
Implement software emulation of buffer_load_format for all types required
by vertex buffer fetches.

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
2019-05-13 17:07:23 +02:00
Samuel Pitoiset 492e828848 ac: tidy up ac_build_llvm8_tbuffer_{load,store}
For consistency with ac_build_llvm8_buffer_{load,store}_common
helpers and that will help a bit for removing the vec3 restriction.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
2019-05-02 09:24:05 +02:00
Marek Olšák 35cd57df2e ac: add ac_get_i1_sgpr_mask
Tested-by: Dieter Nützel <Dieter@nuetzel-hh.de>
Acked-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
2019-04-23 11:28:56 -04:00
Samuel Pitoiset ad6dc13fc7 ac: use struct/raw store intrinsics for 8-bit/16-bit int with LLVM 9+
This changes requires LLVM r356465.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
2019-04-17 22:10:30 +02:00
Samuel Pitoiset 26ea506235 ac: use struct/raw load intrinsics for 8-bit/16-bit int with LLVM 9+
This changes requires LLVM r356465.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
2019-04-17 22:10:28 +02:00
Samuel Pitoiset 6fd5e39b60 ac: add support for more types with struct/raw LLVM intrinsics
LLVM 9+ now supports 8-bit and 16-bit types.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
2019-04-17 22:10:25 +02:00
Marek Olšák 971bc10177 radeonsi: set AC_FUNC_ATTR_READNONE for image opcodes where it was missing
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
2019-04-12 11:34:39 -04:00
Samuel Pitoiset fd4041987b ac: add ac_build_load_helper_invocation() helper
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
2019-04-12 17:30:55 +02:00
Samuel Pitoiset 590a4c8981 ac: add ac_build_ddxy_interp() helper
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
2019-04-12 17:30:55 +02:00
Samuel Pitoiset 4cb13e9462 ac: add ac_build_umax() and use it where possible
This changes the predicate from LessThan to Equal.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
2019-04-12 17:30:55 +02:00
Rhys Perry fd1fc255d9 ac: add 16-bit support to ac_build_ddxy()
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
2019-04-10 09:05:58 +02:00
Samuel Pitoiset d099bc5829 ac: add 8-bit and 64-bit support to ac_build_bitfield_reverse()
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
2019-04-01 18:53:57 +02:00
Samuel Pitoiset 2cecf6c5cc ac: add 8-bit support to ac_build_umsb()
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
2019-04-01 18:53:55 +02:00
Samuel Pitoiset a45d9e3e8d ac: add 8-bit support to ac_find_lsb()
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
2019-04-01 18:53:53 +02:00
Samuel Pitoiset 89cf8ca0ae ac: add 8-bit support to ac_build_bit_count()
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
2019-04-01 18:53:52 +02:00
Samuel Pitoiset 4d5fce29c3 ac: fix ac_build_umsb() for 16-bit integer type
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
2019-04-01 09:51:56 +02:00
Samuel Pitoiset 7a088d1ac8 ac: fix ac_find_lsb() for 16-bit integer type
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
2019-04-01 09:51:54 +02:00
Samuel Pitoiset b16dffff23 ac: fix ac_build_bitfield_reverse() for 16-bit integer type
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
2019-04-01 09:51:52 +02:00
Samuel Pitoiset 9d13b9e53e ac: fix ac_build_bit_count() for 16-bit integer type
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
2019-04-01 09:51:49 +02:00
Samuel Pitoiset cc752dea61 ac: fix return type for llvm.amdgcn.frexp.exp.i32.64
This fixes the following piglit with RadeonSI
tests/spec/arb_gpu_shader_fp64/execution/built-in-functions/fs-frexp-dvec4.shader_test

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
2019-03-29 09:18:24 +01:00
Samuel Pitoiset 52c02d921f ac: add ac_build_frex_exp() helper ans 16-bit/32-bit support
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
2019-03-28 13:02:48 +01:00
Samuel Pitoiset 1bf9311c59 ac: add ac_build_frexp_mant() helper and 16-bit/32-bit support
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
2019-03-28 13:02:46 +01:00
Samuel Pitoiset d6a07732c9 ac: use llvm.amdgcn.fmed3 intrinsic for nir_op_fmed3
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
2019-03-27 14:45:52 +01:00
Samuel Pitoiset 00327f827f ac: fix incorrect argument type for tbuffer.{load,store} with LLVM 7
GLC/SLC are boolean.

This fixes the following LLVM error when checkir is set:
Intrinsic has incorrect argument type!
void (i32, <4 x i32>, i32, i32, i32, i32, i32, i32, i32, i32)* @llvm.amdgcn.tbuffer.store.i32

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl
2019-03-21 14:02:00 +01:00
Samuel Pitoiset 2ac5c5c1b5 ac: add 16-bit support to fract
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
2019-03-21 12:13:09 +01:00
Samuel Pitoiset 0eb1478ac2 ac: add 16-bit support fo fsign
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
2019-03-21 12:13:07 +01:00
Samuel Pitoiset ff11c9dcc7 ac: add f16_0 and f16_1 constants
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
2019-03-21 12:13:05 +01:00
Samuel Pitoiset b235d77e18 ac: add ac_build_tbuffer_store_byte() helper
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
2019-03-21 09:02:18 +01:00
Samuel Pitoiset 104dbc64a5 ac: add ac_build_tbuffer_load_byte() helper
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
2019-03-21 09:02:14 +01:00
Samuel Pitoiset 6e632eb24b ac: add various int8 definitions
Original patch by Rhys Perry.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
2019-03-21 09:02:10 +01:00
Samuel Pitoiset 72e366b4c2 ac: use new LLVM 8 intrinsics in ac_build_buffer_store_dword()
New buffer intrinsics have a separate soffset parameter.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
2019-03-20 22:19:19 +01:00
Samuel Pitoiset 9d960c17a8 ac: use new LLVM 8 intrinsic when storing 16-bit values
vindex is always 0.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
2019-03-20 22:19:14 +01:00
Samuel Pitoiset 2a9d331898 ac: add ac_build_{struct,raw}_tbuffer_store() helpers
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
2019-03-20 22:19:12 +01:00
Samuel Pitoiset 30c2aca67f ac: use new LLVM 8 intrinsics in ac_build_buffer_load()
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
2019-03-20 22:19:08 +01:00
Samuel Pitoiset a2073f49f1 ac: add ac_build_buffer_store_format() helper
Similar to ac_build_buffer_load_format().

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
2019-03-20 22:18:50 +01:00
Samuel Pitoiset 1b553dd47f ac: make use of ac_get_store_intr_attribs() where possible
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
2019-03-20 22:18:35 +01:00
Samuel Pitoiset f4f0e3a395 ac: use llvm.amdgcn.fract intrinsic for nir_op_ffract
Noticed with a Doom shader.

29077 shaders in 15096 tests
Totals:
SGPRS: 1282125 -> 1282133 (0.00 %)
VGPRS: 908716 -> 908616 (-0.01 %)
Spilled SGPRs: 24811 -> 24779 (-0.13 %)
Code Size: 49048176 -> 48936488 (-0.23 %) bytes
Max Waves: 244232 -> 244226 (-0.00 %)

Totals from affected shaders:
SGPRS: 229584 -> 229592 (0.00 %)
VGPRS: 163268 -> 163168 (-0.06 %)
Spilled SGPRs: 8682 -> 8650 (-0.37 %)
Code Size: 12819572 -> 12707884 (-0.87 %) bytes
Max Waves: 24398 -> 24392 (-0.02 %)

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
2019-03-20 09:06:35 +01:00
Samuel Pitoiset cbf022cb31 ac: use the raw tbuffer version for 16-bit SSBO loads
vindex is always 0.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
2019-03-13 14:16:14 +01:00