On Oct 2, 2013 9:29 PM, "Francisco Jerez" <<a href="mailto:currojerez@riseup.net">currojerez@riseup.net</a>> wrote: > > The maximum number of atomic buffer objects is somewhat arbitrary, we > can change it in the future easily if it turns out it's not enough... > > v2: Add comments with the relevant mesa dirty bits. Fix usage of > BRW_NEW_UNIFORM_BUFFER in the GS ABO state atom. > v3: Update binding table layout diagrams. > > Reviewed-by: Paul Berry <<a href="mailto:stereotype441@gmail.com">stereotype441@gmail.com</a>> Hi, I guess this patch also depends on pending patches, right? Aiaiai-mesa fails to built it: On Oct 2, 2013 9:39 PM, "Aiaiai" <aiaiai@aiaiai.ku> wrote: Hi, I have tested your changes: [Mesa-dev] [PATCH] i965: Implement ABO surface state emission. Project: mesa (Mesa build tests) Configurations: android linux ================================================================================ Tested the patch(es) on top of the following commits: 4e4c32b r600/llvm: Adds support for MSAA 8edbd76 r600g/llvm: Undef z and w component of 2D TXP inst 9f183eb r600g/llvm: fix txq for texture buffer 848c0e7 i965: compute DDX in a subspan based only on top row 72edba1 i965/blorp: Use passed in framebuffer rather than ctx->DrawBuffer ef8cc3e ralloc: Remove the rzalloc-based new/delete operator definition macro. fcbbecb st/mesa: Switch glsl_to_tgsi_instruction to the non-zeroing allocator. ================================================================================ Failed to build for "android"" 4e4c32b r600/llvm: Adds support for MSAA 8edbd76 r600g/llvm: Undef z and w component of 2D TXP inst 9f183eb r600g/llvm: fix txq for texture buffer 848c0e7 i965: compute DDX in a subspan based only on top row 72edba1 i965/blorp: Use passed in framebuffer rather than ctx->DrawBuffer ef8cc3e ralloc: Remove the rzalloc-based new/delete operator definition macro. fcbbecb st/mesa: Switch glsl_to_tgsi_instruction to the non-zeroing allocator. src/mesa/drivers/dri/i965/brw_state_dump.c:423:63: warning: pointer of type 'void *' used in arithmetic [-Wpointer-arith] src/mesa/drivers/dri/i965/brw_state_dump.c: In function 'dump_vs_constants': src/mesa/drivers/dri/i965/brw_state_dump.c:435:47: warning: pointer of type 'void *' used in arithmetic [-Wpointer-arith] src/mesa/drivers/dri/i965/brw_state_dump.c:436:45: warning: pointer of type 'void *' used in arithmetic [-Wpointer-arith] src/mesa/drivers/dri/i965/brw_state_dump.c: In function 'dump_wm_constants': src/mesa/drivers/dri/i965/brw_state_dump.c:451:47: warning: pointer of type 'void *' used in arithmetic [-Wpointer-arith] src/mesa/drivers/dri/i965/brw_state_dump.c:452:45: warning: pointer of type 'void *' used in arithmetic [-Wpointer-arith] src/mesa/drivers/dri/i965/brw_state_dump.c: In function 'dump_binding_table': src/mesa/drivers/dri/i965/brw_state_dump.c:468:44: warning: pointer of type 'void *' used in arithmetic [-Wpointer-arith] src/mesa/drivers/dri/i965/brw_state_dump.c: In function 'dump_prog_cache': src/mesa/drivers/dri/i965/brw_state_dump.c:495:33: warning: pointer of type 'void *' used in arithmetic [-Wpointer-arith] src/mesa/drivers/dri/i965/brw_vs_surface_state.c: In function 'brw_upload_vec4_pull_constants': src/mesa/drivers/dri/i965/brw_vs_surface_state.c:72:45: warning: pointer of type 'void *' used in arithmetic [-Wpointer-arith] src/mesa/drivers/dri/i965/brw_wm_surface_state.c: In function 'brw_upload_abo_surfaces': src/mesa/drivers/dri/i965/brw_wm_surface_state.c:842:28: error: 'struct gl_shader_program' has no member named 'NumAtomicBuffers' src/mesa/drivers/dri/i965/brw_wm_surface_state.c:844:14: error: 'struct gl_context' has no member named 'AtomicBufferBindings' src/mesa/drivers/dri/i965/brw_wm_surface_state.c:844:41: error: 'struct gl_shader_program' has no member named 'AtomicBuffers' src/mesa/drivers/dri/i965/brw_wm_surface_state.c:846:37: error: dereferencing pointer to incomplete type src/mesa/drivers/dri/i965/brw_wm_surface_state.c:850:16: error: 'struct <anonymous>' has no member named 'create_raw_surface' src/mesa/drivers/dri/i965/brw_wm_surface_state.c:850:52: error: dereferencing pointer to incomplete type src/mesa/drivers/dri/i965/brw_wm_surface_state.c:851:54: error: dereferencing pointer to incomplete type src/mesa/drivers/dri/i965/brw_wm_surface_state.c:855:12: error: 'struct gl_shader_program' has no member named 'NumAtomicBuffers' make: *** [out/target/product/samsungxe700t/obj/SHARED_LIBRARIES/i965_dri_intermediates/brw_wm_surface_state.o] Error 1 FAILURE ================================ > --- > src/mesa/drivers/dri/i965/brw_context.h | 27 ++++++++++++-- > src/mesa/drivers/dri/i965/brw_gs_surface_state.c | 21 +++++++++++ > src/mesa/drivers/dri/i965/brw_state.h | 3 ++ > src/mesa/drivers/dri/i965/brw_state_upload.c | 4 +++ > src/mesa/drivers/dri/i965/brw_vs_surface_state.c | 21 +++++++++++ > src/mesa/drivers/dri/i965/brw_wm_surface_state.c | 45 ++++++++++++++++++++++++ > 6 files changed, 119 insertions(+), 2 deletions(-) > > diff --git a/src/mesa/drivers/dri/i965/brw_context.h b/src/mesa/drivers/dri/i965/brw_context.h > index 3922705..1407f20 100644 > --- a/src/mesa/drivers/dri/i965/brw_context.h > +++ b/src/mesa/drivers/dri/i965/brw_context.h > @@ -158,6 +158,7 @@ enum brw_state_id { > BRW_STATE_RASTERIZER_DISCARD, > BRW_STATE_STATS_WM, > BRW_STATE_UNIFORM_BUFFER, > + BRW_STATE_ATOMIC_BUFFER, > BRW_STATE_META_IN_PROGRESS, > BRW_STATE_INTERPOLATION_MAP, > BRW_STATE_PUSH_CONSTANT_ALLOCATION, > @@ -196,6 +197,7 @@ enum brw_state_id { > #define BRW_NEW_RASTERIZER_DISCARD (1 << BRW_STATE_RASTERIZER_DISCARD) > #define BRW_NEW_STATS_WM (1 << BRW_STATE_STATS_WM) > #define BRW_NEW_UNIFORM_BUFFER (1 << BRW_STATE_UNIFORM_BUFFER) > +#define BRW_NEW_ATOMIC_BUFFER (1 << BRW_STATE_ATOMIC_BUFFER) > #define BRW_NEW_META_IN_PROGRESS (1 << BRW_STATE_META_IN_PROGRESS) > #define BRW_NEW_INTERPOLATION_MAP (1 << BRW_STATE_INTERPOLATION_MAP) > #define BRW_NEW_PUSH_CONSTANT_ALLOCATION (1 << BRW_STATE_PUSH_CONSTANT_ALLOCATION) > @@ -599,6 +601,12 @@ struct brw_gs_prog_data > /** Max number of render targets in a shader */ > #define BRW_MAX_DRAW_BUFFERS 8 > > +/** Max number of uniform buffer objects in a shader */ > +#define BRW_MAX_UBO 12 > + > +/** Max number of atomic counter buffer objects in a shader */ > +#define BRW_MAX_ABO 4 > + > /** > * Max number of binding table entries used for stream output. > * > @@ -661,6 +669,11 @@ struct brw_gs_prog_data > * | : | : | > * | 36 | UBO 11 | > * +-------------------------------+ > + * | 37 | ABO 0 | > + * | . | . | > + * | : | : | > + * | 40 | ABO 3 | > + * +-------------------------------+ > * > * Our VS (and Gen7 GS) binding tables are programmed as follows: > * > @@ -677,6 +690,11 @@ struct brw_gs_prog_data > * | : | : | > * | 28 | UBO 11 | > * +-------------------------------+ > + * | 29 | ABO 0 | > + * | . | . | > + * | : | : | > + * | 32 | ABO 3 | > + * +-------------------------------+ > * > * Our (gen6) GS binding tables are programmed as follows: > * > @@ -691,14 +709,16 @@ struct brw_gs_prog_data > #define SURF_INDEX_FRAG_CONST_BUFFER (BRW_MAX_DRAW_BUFFERS + 1) > #define SURF_INDEX_TEXTURE(t) (BRW_MAX_DRAW_BUFFERS + 2 + (t)) > #define SURF_INDEX_WM_UBO(u) (SURF_INDEX_TEXTURE(BRW_MAX_TEX_UNIT) + u) > -#define SURF_INDEX_WM_SHADER_TIME (SURF_INDEX_WM_UBO(12)) > +#define SURF_INDEX_WM_ABO(a) (SURF_INDEX_WM_UBO(BRW_MAX_UBO) + a) > +#define SURF_INDEX_WM_SHADER_TIME (SURF_INDEX_WM_ABO(BRW_MAX_ABO)) > /** Maximum size of the binding table. */ > #define BRW_MAX_WM_SURFACES (SURF_INDEX_WM_SHADER_TIME + 1) > > #define SURF_INDEX_VEC4_CONST_BUFFER (0) > #define SURF_INDEX_VEC4_TEXTURE(t) (SURF_INDEX_VEC4_CONST_BUFFER + 1 + (t)) > #define SURF_INDEX_VEC4_UBO(u) (SURF_INDEX_VEC4_TEXTURE(BRW_MAX_TEX_UNIT) + u) > -#define SURF_INDEX_VEC4_SHADER_TIME (SURF_INDEX_VEC4_UBO(12)) > +#define SURF_INDEX_VEC4_ABO(a) (SURF_INDEX_VEC4_UBO(BRW_MAX_UBO) + a) > +#define SURF_INDEX_VEC4_SHADER_TIME (SURF_INDEX_VEC4_ABO(BRW_MAX_ABO)) > #define BRW_MAX_VEC4_SURFACES (SURF_INDEX_VEC4_SHADER_TIME + 1) > > #define SURF_INDEX_GEN6_SOL_BINDING(t) (t) > @@ -1498,6 +1518,9 @@ brw_update_sol_surface(struct brw_context *brw, > void brw_upload_ubo_surfaces(struct brw_context *brw, > struct gl_shader *shader, > uint32_t *surf_offsets); > +void brw_upload_abo_surfaces(struct brw_context *brw, > + struct gl_shader_program *prog, > + uint32_t *surf_offsets); > > /* brw_surface_formats.c */ > bool brw_is_hiz_depth_format(struct brw_context *ctx, gl_format format); > diff --git a/src/mesa/drivers/dri/i965/brw_gs_surface_state.c b/src/mesa/drivers/dri/i965/brw_gs_surface_state.c > index d0ce412..05f9efa 100644 > --- a/src/mesa/drivers/dri/i965/brw_gs_surface_state.c > +++ b/src/mesa/drivers/dri/i965/brw_gs_surface_state.c > @@ -87,3 +87,24 @@ const struct brw_tracked_state brw_gs_ubo_surfaces = { > }, > .emit = brw_upload_gs_ubo_surfaces, > }; > + > +static void > +brw_upload_gs_abo_surfaces(struct brw_context *brw) > +{ > + struct gl_context *ctx = &brw->ctx; > + /* _NEW_PROGRAM */ > + struct gl_shader_program *prog = ctx->Shader.CurrentGeometryProgram; > + > + if (prog) > + brw_upload_abo_surfaces( > + brw, prog, &brw->gs.base.surf_offset[SURF_INDEX_VEC4_ABO(0)]); > +} > + > +const struct brw_tracked_state brw_gs_abo_surfaces = { > + .dirty = { > + .mesa = _NEW_PROGRAM, > + .brw = BRW_NEW_BATCH | BRW_NEW_ATOMIC_BUFFER, > + .cache = 0, > + }, > + .emit = brw_upload_gs_abo_surfaces, > +}; > diff --git a/src/mesa/drivers/dri/i965/brw_state.h b/src/mesa/drivers/dri/i965/brw_state.h > index ec64328..ed8ce33 100644 > --- a/src/mesa/drivers/dri/i965/brw_state.h > +++ b/src/mesa/drivers/dri/i965/brw_state.h > @@ -71,7 +71,9 @@ extern const struct brw_tracked_state brw_vs_prog; > extern const struct brw_tracked_state brw_vs_samplers; > extern const struct brw_tracked_state brw_gs_samplers; > extern const struct brw_tracked_state brw_vs_ubo_surfaces; > +extern const struct brw_tracked_state brw_vs_abo_surfaces; > extern const struct brw_tracked_state brw_gs_ubo_surfaces; > +extern const struct brw_tracked_state brw_gs_abo_surfaces; > extern const struct brw_tracked_state brw_vs_unit; > extern const struct brw_tracked_state brw_gs_prog; > extern const struct brw_tracked_state brw_wm_prog; > @@ -81,6 +83,7 @@ extern const struct brw_tracked_state brw_wm_binding_table; > extern const struct brw_tracked_state brw_gs_binding_table; > extern const struct brw_tracked_state brw_vs_binding_table; > extern const struct brw_tracked_state brw_wm_ubo_surfaces; > +extern const struct brw_tracked_state brw_wm_abo_surfaces; > extern const struct brw_tracked_state brw_wm_unit; > extern const struct brw_tracked_state brw_interpolation_map; > > diff --git a/src/mesa/drivers/dri/i965/brw_state_upload.c b/src/mesa/drivers/dri/i965/brw_state_upload.c > index d7fe319..64ffcd0 100644 > --- a/src/mesa/drivers/dri/i965/brw_state_upload.c > +++ b/src/mesa/drivers/dri/i965/brw_state_upload.c > @@ -204,10 +204,13 @@ static const struct brw_tracked_state *gen7_atoms[] = > */ > &brw_vs_pull_constants, > &brw_vs_ubo_surfaces, > + &brw_vs_abo_surfaces, > &brw_gs_pull_constants, > &brw_gs_ubo_surfaces, > + &brw_gs_abo_surfaces, > &brw_wm_pull_constants, > &brw_wm_ubo_surfaces, > + &brw_wm_abo_surfaces, > &gen6_renderbuffer_surfaces, > &brw_texture_surfaces, > &brw_vs_binding_table, > @@ -398,6 +401,7 @@ static struct dirty_bit_map brw_bits[] = { > DEFINE_BIT(BRW_NEW_RASTERIZER_DISCARD), > DEFINE_BIT(BRW_NEW_STATS_WM), > DEFINE_BIT(BRW_NEW_UNIFORM_BUFFER), > + DEFINE_BIT(BRW_NEW_ATOMIC_BUFFER), > DEFINE_BIT(BRW_NEW_META_IN_PROGRESS), > DEFINE_BIT(BRW_NEW_INTERPOLATION_MAP), > DEFINE_BIT(BRW_NEW_PUSH_CONSTANT_ALLOCATION), > diff --git a/src/mesa/drivers/dri/i965/brw_vs_surface_state.c b/src/mesa/drivers/dri/i965/brw_vs_surface_state.c > index 2c5d06f..c6461c7 100644 > --- a/src/mesa/drivers/dri/i965/brw_vs_surface_state.c > +++ b/src/mesa/drivers/dri/i965/brw_vs_surface_state.c > @@ -148,3 +148,24 @@ const struct brw_tracked_state brw_vs_ubo_surfaces = { > }, > .emit = brw_upload_vs_ubo_surfaces, > }; > + > +static void > +brw_upload_vs_abo_surfaces(struct brw_context *brw) > +{ > + struct gl_context *ctx = &brw->ctx; > + /* _NEW_PROGRAM */ > + struct gl_shader_program *prog = ctx->Shader.CurrentVertexProgram; > + > + if (prog) > + brw_upload_abo_surfaces( > + brw, prog, &brw->vs.base.surf_offset[SURF_INDEX_VEC4_ABO(0)]); > +} > + > +const struct brw_tracked_state brw_vs_abo_surfaces = { > + .dirty = { > + .mesa = _NEW_PROGRAM, > + .brw = BRW_NEW_BATCH | BRW_NEW_ATOMIC_BUFFER, > + .cache = 0, > + }, > + .emit = brw_upload_vs_abo_surfaces, > +}; > diff --git a/src/mesa/drivers/dri/i965/brw_wm_surface_state.c b/src/mesa/drivers/dri/i965/brw_wm_surface_state.c > index 4c3eb69..70f1391 100644 > --- a/src/mesa/drivers/dri/i965/brw_wm_surface_state.c > +++ b/src/mesa/drivers/dri/i965/brw_wm_surface_state.c > @@ -833,6 +833,51 @@ const struct brw_tracked_state brw_wm_ubo_surfaces = { > }; > > void > +brw_upload_abo_surfaces(struct brw_context *brw, > + struct gl_shader_program *prog, > + uint32_t *surf_offsets) > +{ > + struct gl_context *ctx = &brw->ctx; > + > + for (int i = 0; i < prog->NumAtomicBuffers; i++) { > + struct gl_atomic_buffer_binding *binding = > + &ctx->AtomicBufferBindings[prog->AtomicBuffers[i].Binding]; > + struct intel_buffer_object *intel_bo = > + intel_buffer_object(binding->BufferObject); > + drm_intel_bo *bo = intel_bufferobj_buffer(brw, intel_bo, > + INTEL_READ | INTEL_WRITE_PART); > + > + brw->vtbl.create_raw_surface(brw, bo, binding->Offset, > + bo->size - binding->Offset, > + &surf_offsets[i], true); > + } > + > + if (prog->NumAtomicBuffers) > + brw->state.dirty.brw |= BRW_NEW_SURFACES; > +} > + > +static void > +brw_upload_wm_abo_surfaces(struct brw_context *brw) > +{ > + struct gl_context *ctx = &brw->ctx; > + /* _NEW_PROGRAM */ > + struct gl_shader_program *prog = ctx->Shader._CurrentFragmentProgram; > + > + if (prog) > + brw_upload_abo_surfaces( > + brw, prog, &brw->wm.base.surf_offset[SURF_INDEX_WM_ABO(0)]); > +} > + > +const struct brw_tracked_state brw_wm_abo_surfaces = { > + .dirty = { > + .mesa = _NEW_PROGRAM, > + .brw = BRW_NEW_BATCH | BRW_NEW_ATOMIC_BUFFER, > + .cache = 0, > + }, > + .emit = brw_upload_wm_abo_surfaces, > +}; > + > +void > gen4_init_vtable_surface_functions(struct brw_context *brw) > { > brw->vtbl.update_texture_surface = brw_update_texture_surface; > -- > 1.8.3.4 > > _______________________________________________ > mesa-dev mailing list > <a href="mailto:mesa-dev@lists.freedesktop.org">mesa-dev@lists.freedesktop.org</a> > <a href="http://lists.freedesktop.org/mailman/listinfo/mesa-dev">http://lists.freedesktop.org/mailman/listinfo/mesa-dev</a>