Searched refs:NGG (Results 1 - 25 of 31) sorted by relevance

12

/xsrc/external/mit/MesaLib/dist/src/gallium/drivers/radeonsi/
H A Dsi_state_draw.cpp108 template <chip_class GFX_VERSION, si_has_tess HAS_TESS, si_has_gs HAS_GS, si_has_ngg NGG>
158 } else if (NGG) {
179 if (!NGG) {
189 if (!NGG) {
206 if (NGG) {
236 if (NGG)
254 (!NGG && si_pm4_state_changed(sctx, vs)) ||
255 (NGG && si_pm4_state_changed(sctx, gs))) {
271 /* NGG cull state uses smoothing_enabled. */
351 if ((HAS_GS || NGG)
[all...]
/xsrc/external/mit/MesaLib/dist/docs/relnotes/
H A D20.3.1.rst141 - radv: fix exporting multiviews with NGG
146 - radv: don't count unusable vertices to the NGG LDS size
147 - radv: fix applying the NGG minimum vertex count requirement
148 - radv: do VGT_FLUSH when switching NGG -\> legacy on Sienna Cichlid
H A D21.1.4.rst154 - radv/llvm: Emit s_barrier at the beginning of NGG non-GS shaders.
155 - aco/gfx10: NGG zero output workaround for conservative rasterization.
156 - aco/gfx10: Emit barrier at the start of NGG VS and TES.
H A D20.2.5.rst143 - radv: fix exporting multiviews with NGG
147 - radv: do VGT_FLUSH when switching NGG -> legacy on Sienna Cichlid
148 - radv: fix applying the NGG minimum vertex count requirement
149 - radv: don't count unusable vertices to the NGG LDS size
H A D20.0.2.rst120 - radeonsi: add a bug workaround for NGG - LATE_ALLOC_GS
121 - ac: add a bug workaround for the 100% NGG culling case
H A D21.2.3.rst100 - radeonsi: fix clearing index_size for NGG fast launch
118 - radv: don't require a GS copy shader to use the cache with NGG VS+GS
H A D20.1.10.rst111 - aco/isel: Always export position data from VS/NGG
H A D20.1.6.rst110 - radeonsi: fix applying the NGG minimum vertex count requirement
H A D20.2.2.rst147 - aco/isel: Always export position data from VS/NGG
H A D20.2.3.rst121 - radeonsi: do VGT_FLUSH when switching NGG -> legacy on Sienna Cichlid
H A D19.3.0.rst478 - radv: Disable NGG for geometry shaders.
2796 - radv: never kill a NGG GS shader
2828 - radv/aco: disable NGG when ACO is used
3044 - radv/gfx10: do not use NGG with NAVI14
3089 - radv/gfx10: account for the subpass view for the NGG GS storage
3090 - radv/gfx10: make use the output usage mask when exporting NGG GS
3096 - radv: store GFX10 NGG state as part of the shader info
3098 - radv: calculate GFX9 GS and GFX10 NGG states before compiling shader
3100 - radv/gfx10: declare a LDS symbol for the NGG emit space
3103 - radv/gfx10: add an option to switch from legacy to NGG streamou
[all...]
H A D19.2.0.rst149 - radv: Disable NGG for geometry shaders.
387 - radv/gfx10: do not use NGG with NAVI14
392 - radv/gfx10: fix VK_KHR_pipeline_executable_properties with NGG GS
H A D21.3.0.rst49 - NGG shader based culling is now enabled by default on GFX10.3 on RADV.
80 - RADV: Resident Evil Village needs invariantgeom when NGG culling is enabled
2634 - ac/llvm: implement a bunch of NIR AMD intrinsics for NGG
2656 - radeonsi: don't use NGG passthrough if culling is possible for better perf
2681 - radeonsi: precompute si_vgt_stages_key for NGG in si_shader
2702 - radeonsi: enable NGG passthrough when LDS is used, document the real constraints
2716 - radeonsi: reduce NGG culling on/off transitions by keeping it enabled
2719 - ac/gpu_info: fix the comment for the NGG->legacy transition bug
2721 - radeonsi: fix clearing index_size for NGG fast launch
2722 - radeonsi: disallow NGG fas
[all...]
H A D20.3.0.rst52 - NGG GS support in ACO
859 - radv: Disable NGG on APUs.
3312 - radeonsi: increase minimum NGG vertex count requirement per workgroup on gfx 10.3
3313 - radeonsi: fix applying the NGG minimum vertex count requirement
3314 - radeonsi: don't count unusable vertices to the NGG LDS size
3316 - radeonsi: remove the NGG hack decreasing LDS usage to deal with overflows
3318 - radeonsi: disable NGG culling on gfx10.3 because of hangs
3417 - radeonsi: eliminate unused shader outputs for separate NGG geometry shaders
3490 - radeonsi: simplify NGG culling enablement and add radeonsi_shader_culling option
3527 - radeonsi: don't disable NGG cullin
[all...]
H A D20.0.0.rst2289 - radeonsi/gfx10: simplify some duplicated NGG GS code
2321 - radeonsi/gfx10: don't insert NGG streamout atomics if they are never
2330 - radeonsi/gfx10: don't declare any LDS for NGG if it's not used
2331 - radeonsi/gfx10: enable NGG passthrough for eligible shaders
2404 for NGG
2427 - radeonsi/gfx10: correct VS PrimitiveID implementation for NGG
2433 - radeonsi/gfx10: don't initialize VGPRs not used by NGG passthrough
2435 - radeonsi/gfx10: implement NGG culling for 4x wave32 subgroups
2439 NGG culling
3037 - radv: declare NGG scratc
[all...]
H A D21.2.0.rst64 - NGG shader based primitive culling is now supported by RADV.
3455 - radeonsi: simplify the NGG culling vertex count heuristic
3483 - radeonsi: allow changing the NGG subgroup size to 256 but don't change it yet
3523 - radeonsi: move the accepting code into the bbox cull branch in NGG cull code
3528 - radeonsi: set more precise max_waves in NGG code
3542 - radeonsi: don't use NGG culling on 1 RB chips
4742 - radv/llvm: adjust NGG if provoking vertex mode is last
4743 - aco: adjust NGG if provoking vertex mode is last
4746 - radv: fix computation of the number of user SGPRS for NGG GS state
4974 - radv: Ignore GS copy shader when determining NGG G
[all...]
H A D20.3.5.rst245 - radv: correctly enable WGP_MODE for NGG and GS
H A D21.0.0.rst2054 - radeonsi: do VGT_FLUSH when switching NGG -\> legacy on Sienna Cichlid
2069 - radeonsi: fix NGG streamout regression
2157 - radeonsi: enable NGG and NGG culling on gfx10.3 APUs by default
2226 - radeonsi: disable NGG fast launch with indexed triangle strips to fix a hang
2813 - radv: correctly enable WGP_MODE for NGG and GS
2964 - radv: do VGT_FLUSH when switching NGG -\> legacy on Sienna Cichlid
2965 - radv: fix applying the NGG minimum vertex count requirement
2966 - radv: don't count unusable vertices to the NGG LDS size
2995 - radv: enable NGG o
[all...]
H A D20.1.0.rst3031 - radeonsi: add a bug workaround for NGG - LATE_ALLOC_GS
3032 - ac: add a bug workaround for the 100% NGG culling case
3115 - radeonsi/gfx10: don't use NGG culling if compute-based culling is
3994 - radv: set missing SHARED_VGPR_CNT for NGG VS and ACO
4223 - aco/ngg: Initialize exec mask for NGG VS and TES.
4224 - aco/ngg: Fix exports for NGG VS and TES.
4225 - aco/ngg: Setup NGG VS and TES stages.
4226 - aco/ngg: Implement NGG VS and TES.
4227 - aco/ngg: Schedule position exports of NGG VS/TES.
4228 - aco/ngg: Run GS_ALLOC_REQ on priority 3 for NGG V
[all...]
/xsrc/external/mit/MesaLib/dist/src/amd/compiler/
H A DREADME.md139 * NGG = Next Generation Geometry, a new hardware stage that replaces legacy HW GS and HW VS on RDNA GPUs
146 However, GS store their output in VRAM (except GFX10/NGG).
153 On GFX10/NGG this limitation no longer exists, as the HW NGG GS can now export directly where it needs to.
157 The merged stages on GFX9 (and GFX10/legacy) are: LSHS and ESGS. On GFX10/NGG the ESGS is merged with HW VS into NGG GS.
196 ##### NGG (GFX10+ only):
198 * HW GS and VS stages are now merged, and NGG GS can export directly
201 | GFX10/NGG HW stages: | LSHS | NGG G
[all...]
H A Daco_instruction_selection_setup.cpp260 * When there are no param exports in an NGG (or legacy VS) shader,
263 * before the NGG (or VS) waves finish.
274 /* TODO: NGG streamout */
275 if (ctx->stage.hw == HWStage::NGG)
318 /* TODO: NGG streamout */
319 if (ctx->stage.hw == HWStage::NGG)
846 hw_stage = HWStage::NGG; /* GFX10/NGG: VS without GS uses the HW GS stage */
858 hw_stage = HWStage::NGG; /* GFX10+: VS+GS merged into an NGG G
[all...]
H A DREADME-ISA.md162 What the doc doesn't say is that in case of NGG (and legacy VS) when there
166 for the NGG (or VS) to finish.
168 It can even launch PS waves before NGG (or VS) ends.
H A Daco_assembler.cpp779 if (program->stage.hw == HWStage::VS || program->stage.hw == HWStage::NGG) {
800 (program->stage.hw == HWStage::VS || program->stage.hw == HWStage::NGG);
802 "Missing export in %s shader:", is_vertex_or_ngg ? "vertex or NGG" : "fragment");
972 program->stage.hw == HWStage::NGG)
H A Daco_ir.h1940 NGG, /* Primitive shader, used to implement VS, TES, GS. */ enumerator in enum:aco::HWStage
1981 /* GFX10/NGG */
1982 static constexpr Stage vertex_ngg(HWStage::NGG, SWStage::VS);
1983 static constexpr Stage vertex_geometry_ngg(HWStage::NGG, SWStage::VS_GS);
1984 static constexpr Stage tess_eval_ngg(HWStage::NGG, SWStage::TES);
1985 static constexpr Stage tess_eval_geometry_ngg(HWStage::NGG, SWStage::TES_GS);
1986 /* GFX9 (and GFX10 if NGG isn't used) */
/xsrc/external/mit/MesaLib/dist/docs/
H A Denvvars.rst659 disable NGG for GFX10+
661 disable NGG culling on GPUs where it's enabled by default (GFX10.3+ only).
719 enable NGG culling on GPUs where it's not enabled by default (GFX10.1 only).
832 Disable NGG and use the legacy pipeline.
834 Always use NGG culling even when it can hurt.
836 Disable NGG culling.

Completed in 81 milliseconds

12