[Mesa-dev] [PATCH] [v2] i965: implement ARB_pipeline_statistics_query
Ian Romanick
idr at freedesktop.org
Tue Dec 2 23:07:34 PST 2014
Since there will be a v3 anyway, nits below...
On 12/02/2014 06:33 PM, Ben Widawsky wrote:
> This patch implements ARB_pipeline_statistics_query. This addition to GL does
> not add a new API. Instead, it adds new tokens to the existing query APIs. The
> work to hook up the new tokens is trivial due to it's similarity to the previous
> work done for the query APIs. I've implemented all the new tokens to some
> degree, but have stubbed out the untested ones at the entry point for Begin().
> Doing this should allow the remainder of the code to be left in.
>
> The new tokens give GL clients a way to obtain stats about the GL pipeline.
> Generally, you get the number of things going in, invocations, and number of
> things coming out, primitives, of the various stages. There are two immediate
> uses for this, performance information, and debugging various types of
> misrendering. I doubt one can use these for debugging very complex applications,
> but for piglit tests, it should be quite useful.
>
> Tessellation shaders, and compute shaders are not addressed in this patch
> because there is no upstream implementation. I've implemented how I believe
> tessellation shader stats will work for Intel hardware (though there is a bit of
> ambiguity). Compute shaders are a bit more interesting though, and I don't yet
> know what we'll do there.
>
> For the lazy, here is a link to the relevant part of the spec:
> https://www.opengl.org/registry/specs/ARB/pipeline_statistics_query.txt
>
> Running the piglit tests
> http://lists.freedesktop.org/archives/piglit/2014-November/013321.html
> (http://cgit.freedesktop.org/~bwidawsk/piglit/log/?h=pipe_stats)
> yield the following results:
>
>> python2 ./piglit-run.py -t stats tests/all.py output/pipeline_stats
>> [5/5] pass: 5 Running Test(s): 5
>
> Previously I was seeing the adjacent vertex test failing on certain Intel
> hardware. I am currently not able to reproduce this, and therefore for now, I'll
> assume it was some transient issue which has been fixed.
>
> v2:
> - Don't allow pipeline_stats to be per stream (Ilia). This would be needed for
> AMD_transform_feedback4, which we do not support.
> > If AMD_transform_feedback4 is supported then GEOMETRY_SHADER_PRIMITIVES_-
> > EMITTED_ARB counts primitives emitted to any of the vertex streams for which
> > STREAM_RASTERIZATION_AMD is enabled.
> - Remove comment from GL3.txt because it is only used for extensions that are
> part of required versions (Ilia)
> - Move the new tokens to a new XML doc instead of using the main GL4x.xml (Ilia)
> - Add a fallthrough comment (Ilia)
> - Only divide PS invocations by 4 on HSW+ (Ben)
>
> Cc: Ilia Mirkin <imirkin at alum.mit.edu>
> Signed-off-by: Ben Widawsky <ben at bwidawsk.net>
> ---
> .../glapi/gen/ARB_pipeline_statistics_query.xml | 24 ++++
> src/mesa/drivers/dri/i965/gen6_queryobj.c | 121 +++++++++++++++++++++
> src/mesa/drivers/dri/i965/intel_extensions.c | 1 +
> src/mesa/main/config.h | 3 +
> src/mesa/main/extensions.c | 1 +
> src/mesa/main/mtypes.h | 15 +++
> src/mesa/main/queryobj.c | 77 +++++++++++++
> 7 files changed, 242 insertions(+)
> create mode 100644 src/mapi/glapi/gen/ARB_pipeline_statistics_query.xml
>
> diff --git a/src/mapi/glapi/gen/ARB_pipeline_statistics_query.xml b/src/mapi/glapi/gen/ARB_pipeline_statistics_query.xml
> new file mode 100644
> index 0000000..db37267
> --- /dev/null
> +++ b/src/mapi/glapi/gen/ARB_pipeline_statistics_query.xml
> @@ -0,0 +1,24 @@
> +<?xml version="1.0"?>
> +<!DOCTYPE OpenGLAPI SYSTEM "gl_API.dtd">
> +
> +<!-- Note: no GLX protocol info yet. -->
> +
> +<OpenGLAPI>
> +
> +<category name="GL_ARB_pipeline_statistics_query" number="100">
> +
> + <enum name="VERTICES_SUBMITTED" value="0x82EE"/>
> + <enum name="PRIMITIVES_SUBMITTED" value="0x82EF"/>
> + <enum name="VERTEX_SHADER_INVOCATIONS" value="0x82F0"/>
> + <enum name="TESS_CONTROL_SHADER_PATCHES" value="0x82F1"/>
> + <enum name="TESS_EVALUATION_SHADER_INVOCATIONS" value="0x82F2"/>
> + <!-- <enum name="GEOMETRY_SHADER_INVOCATIONS" value="0x887F"/> -->
> + <enum name="GEOMETRY_SHADER_PRIMITIVES_EMITTED" value="0x82F3"/>
> + <enum name="FRAGMENT_SHADER_INVOCATIONS" value="0x82F4"/>
> + <enum name="COMPUTE_SHADER_INVOCATIONS" value="0x82F5"/>
> + <enum name="CLIPPING_INPUT_PRIMITIVES" value="0x82F6"/>
> + <enum name="CLIPPING_OUTPUT_PRIMITIVES" value="0x82F7"/>
> +
> +</category>
> +
> +</OpenGLAPI>
> diff --git a/src/mesa/drivers/dri/i965/gen6_queryobj.c b/src/mesa/drivers/dri/i965/gen6_queryobj.c
> index 130236e..f8b9bc3 100644
> --- a/src/mesa/drivers/dri/i965/gen6_queryobj.c
> +++ b/src/mesa/drivers/dri/i965/gen6_queryobj.c
> @@ -109,6 +109,73 @@ write_xfb_primitives_written(struct brw_context *brw,
> }
> }
>
> +static inline const int
> +pipeline_target_to_index(int target)
> +{
> + if (target == GL_GEOMETRY_SHADER_INVOCATIONS)
> + return MAX_PIPELINE_STATISTICS - 1;
> + else
> + return target - GL_VERTICES_SUBMITTED_ARB;
> +}
> +
> +static void
> +emit_pipeline_stat(struct brw_context *brw, drm_intel_bo *bo,
> + int stream, int target, int idx)
> +{
> + /*
> + * There are 2 confusing parts to implementing the various target. The first is
> + * the distinction between vertices submitted and primitives submitted. The
> + * spec tries to clear this up, and luckily our hardware seems to understand
> + * it:
> + *
> + * (8) What stage the VERTICES_SUBMITTED_ARB and PRIMITIVES_SUBMITTED_ARB
> + * belong to? What do they count?
> + *
> + * DISCUSSION: There is no separate pipeline stage introduced in the
> + * specification that matches D3D's "input assembler" stage. While the
> + * latest version of the GL specification mentions a "vertex puller" stage
> + * in the pipeline diagram, this stage does not have a corresponding chapter
> + * in the specification that introduces it.
> + *
> + * RESOLVED: Introduce VERTICES_SUBMITTED_ARB and PRIMITIVES_SUBMITTED_ARB
> + * in chapter 10, Vertex Specification and Drawing Command. They count the
> + * total number of vertices and primitives processed by the GL. Including
> + * multiple instances.
> + *
> + * The second confusion is the tessellation shader statistics. Our hardware has
> + * no statistics specific to the TE unit. Ideally we could have the HS
> + * primitives for TESS_CONTROL_SHADER_PATCHES_ARB, and the DS invocations as
> + * the register for TESS_CONTROL_SHADER_PATCHES_ARB. Unfortunately we don't
> + * have HS primitives, we only have HS invocations.
> + */
> +
> + /* Everything except GEOMETRY_SHADER_INVOCATIONS can be kept in a simple
> + * lookup table */
*/ should be on it's own line (here and elsewhere).
> + const uint32_t target_to_register[] = {
static. Otherwise GCC emits code to initialize the array.
> + IA_VERTICES_COUNT, /* VERTICES_SUBMITTED */
> + IA_PRIMITIVES_COUNT, /* PRIMITIVES_SUBMITTED */
> + VS_INVOCATION_COUNT, /* VERTEX_SHADER_INVOCATIONS */
> + 0, /* HS_INVOCATION_COUNT,*/ /* TESS_CONTROL_SHADER_PATCHES */
> + 0, /* DS_INVOCATION_COUNT,*/ /* TESS_EVALUATION_SHADER_INVOCATIONS */
> + GS_PRIMITIVES_COUNT, /* GEOMETRY_SHADER_PRIMITIVES_EMITTED */
> + PS_INVOCATION_COUNT, /* FRAGMENT_SHADER_INVOCATIONS */
> + 0, /* COMPUTE_SHADER_INVOCATIONS */
> + CL_INVOCATION_COUNT, /* CLIPPING_INPUT_PRIMITIVES */
> + CL_PRIMITIVES_COUNT, /* CLIPPING_OUTPUT_PRIMITIVES */
> + GS_INVOCATION_COUNT /* This one is special... */
> + };
> + STATIC_ASSERT(ARRAY_SIZE(target_to_register) == MAX_PIPELINE_STATISTICS);
> + uint32_t reg = target_to_register[pipeline_target_to_index(target)];
> + assert(reg != 0);
> +
> + /* Emit a flush to make sure various parts of the pipeline are complete and
> + * we get an accurate value */
> + intel_batchbuffer_emit_mi_flush(brw);
> +
> + brw_store_register_mem64(brw, bo, reg, idx);
> +}
> +
> +
> /**
> * Wait on the query object's BO and calculate the final result.
> */
> @@ -182,9 +249,29 @@ gen6_queryobj_get_results(struct gl_context *ctx,
>
> case GL_PRIMITIVES_GENERATED:
> case GL_TRANSFORM_FEEDBACK_PRIMITIVES_WRITTEN:
> + case GL_VERTICES_SUBMITTED_ARB:
> + case GL_PRIMITIVES_SUBMITTED_ARB:
> + case GL_VERTEX_SHADER_INVOCATIONS_ARB:
> + case GL_GEOMETRY_SHADER_INVOCATIONS:
> + case GL_GEOMETRY_SHADER_PRIMITIVES_EMITTED_ARB:
> + case GL_CLIPPING_INPUT_PRIMITIVES_ARB:
> + case GL_CLIPPING_OUTPUT_PRIMITIVES_ARB:
> query->Base.Result = results[1] - results[0];
> break;
>
> + case GL_FRAGMENT_SHADER_INVOCATIONS_ARB:
> + query->Base.Result = (results[1] - results[0]);
> + /* HW reports this count 4X the actual value and therefore SW must divide
> + * the count by 4 for correct reporting.
> + * XXX: Not in public docs?
> + */
> + if (brw->gen >= 8 || brw->is_haswell)
> + query->Base.Result /= 4;
> + break;
> +
> + case GL_TESS_CONTROL_SHADER_PATCHES_ARB:
> + case GL_TESS_EVALUATION_SHADER_INVOCATIONS_ARB:
> + case GL_COMPUTE_SHADER_INVOCATIONS_ARB:
> default:
> unreachable("Unrecognized query target in brw_queryobj_get_results()");
> }
> @@ -251,6 +338,20 @@ gen6_begin_query(struct gl_context *ctx, struct gl_query_object *q)
> write_xfb_primitives_written(brw, query->bo, query->Base.Stream, 0);
> break;
>
> + case GL_VERTICES_SUBMITTED_ARB:
> + case GL_PRIMITIVES_SUBMITTED_ARB:
> + case GL_VERTEX_SHADER_INVOCATIONS_ARB:
> + case GL_GEOMETRY_SHADER_INVOCATIONS:
> + case GL_GEOMETRY_SHADER_PRIMITIVES_EMITTED_ARB:
> + case GL_FRAGMENT_SHADER_INVOCATIONS_ARB:
> + case GL_CLIPPING_INPUT_PRIMITIVES_ARB:
> + case GL_CLIPPING_OUTPUT_PRIMITIVES_ARB:
> + emit_pipeline_stat(brw, query->bo, query->Base.Stream, query->Base.Target, 0);
> + break;
> +
> + case GL_TESS_CONTROL_SHADER_PATCHES_ARB:
> + case GL_TESS_EVALUATION_SHADER_INVOCATIONS_ARB:
> + case GL_COMPUTE_SHADER_INVOCATIONS_ARB:
> default:
> unreachable("Unrecognized query target in brw_begin_query()");
> }
> @@ -289,6 +390,26 @@ gen6_end_query(struct gl_context *ctx, struct gl_query_object *q)
> write_xfb_primitives_written(brw, query->bo, query->Base.Stream, 1);
> break;
>
> + case GL_VERTICES_SUBMITTED_ARB:
> + case GL_PRIMITIVES_SUBMITTED_ARB:
> + case GL_VERTEX_SHADER_INVOCATIONS_ARB:
> + case GL_TESS_CONTROL_SHADER_PATCHES_ARB:
> + case GL_TESS_EVALUATION_SHADER_INVOCATIONS_ARB:
> + case GL_GEOMETRY_SHADER_PRIMITIVES_EMITTED_ARB:
> + case GL_FRAGMENT_SHADER_INVOCATIONS_ARB:
> + case GL_COMPUTE_SHADER_INVOCATIONS_ARB:
> + case GL_CLIPPING_INPUT_PRIMITIVES_ARB:
> + case GL_CLIPPING_OUTPUT_PRIMITIVES_ARB:
> + emit_pipeline_stat(brw, query->bo,
> + query->Base.Stream, query->Base.Target, 1);
> + break;
> + case GL_GEOMETRY_SHADER_INVOCATIONS:
> + /* GEOMETRY_SHADER_INVOCATIONS has a weirdly numbered target */
> + emit_pipeline_stat(brw, query->bo, query->Base.Stream,
> + GL_VERTICES_SUBMITTED_ARB + MAX_PIPELINE_STATISTICS - 1,
Is this actually necessary? It looks like this parameter is eventually
just passed to pipeline_target_to_index, which does the right thing with
GL_GEOMETRY_SHADER_INVOCATIONS.
> + 1);
> + break;
> +
> default:
> unreachable("Unrecognized query target in brw_end_query()");
> }
> diff --git a/src/mesa/drivers/dri/i965/intel_extensions.c b/src/mesa/drivers/dri/i965/intel_extensions.c
> index bbbb76f..a3645b9 100644
> --- a/src/mesa/drivers/dri/i965/intel_extensions.c
> +++ b/src/mesa/drivers/dri/i965/intel_extensions.c
> @@ -186,6 +186,7 @@ intelInitExtensions(struct gl_context *ctx)
> ctx->Extensions.ARB_map_buffer_range = true;
> ctx->Extensions.ARB_occlusion_query = true;
> ctx->Extensions.ARB_occlusion_query2 = true;
> + ctx->Extensions.ARB_pipeline_statistics_query = true;
> ctx->Extensions.ARB_point_sprite = true;
> ctx->Extensions.ARB_seamless_cube_map = true;
> ctx->Extensions.ARB_shader_bit_encoding = true;
> diff --git a/src/mesa/main/config.h b/src/mesa/main/config.h
> index 4ec4b75..50e87fa 100644
> --- a/src/mesa/main/config.h
> +++ b/src/mesa/main/config.h
> @@ -300,6 +300,9 @@
> #define MAX_COMPUTE_IMAGE_UNIFORMS 8
> /*@}*/
>
> +/** For GL_ARB_pipeline_statistics_query */
> +#define MAX_PIPELINE_STATISTICS 11
> +
> /*
> * Color channel component order
> *
> diff --git a/src/mesa/main/extensions.c b/src/mesa/main/extensions.c
> index 0df04c2..997c774 100644
> --- a/src/mesa/main/extensions.c
> +++ b/src/mesa/main/extensions.c
> @@ -133,6 +133,7 @@ static const struct extension extension_table[] = {
> { "GL_ARB_multitexture", o(dummy_true), GLL, 1998 },
> { "GL_ARB_occlusion_query2", o(ARB_occlusion_query2), GL, 2003 },
> { "GL_ARB_occlusion_query", o(ARB_occlusion_query), GLL, 2001 },
> + { "GL_ARB_pipeline_statistics_query", o(ARB_pipeline_statistics_query), GL, 2014 },
> { "GL_ARB_pixel_buffer_object", o(EXT_pixel_buffer_object), GL, 2004 },
> { "GL_ARB_point_parameters", o(EXT_point_parameters), GLL, 1997 },
> { "GL_ARB_point_sprite", o(ARB_point_sprite), GL, 2003 },
> diff --git a/src/mesa/main/mtypes.h b/src/mesa/main/mtypes.h
> index 7389baa..04a87be 100644
> --- a/src/mesa/main/mtypes.h
> +++ b/src/mesa/main/mtypes.h
> @@ -3053,6 +3053,9 @@ struct gl_query_state
> /** GL_ARB_timer_query */
> struct gl_query_object *TimeElapsed;
>
> + /** GL_ARB_pipeline_statistics_query */
> + struct gl_query_object *pipeline_stats[MAX_PIPELINE_STATISTICS];
> +
> GLenum CondRenderMode;
> };
>
> @@ -3439,6 +3442,17 @@ struct gl_constants
> GLuint Timestamp;
> GLuint PrimitivesGenerated;
> GLuint PrimitivesWritten;
> + GLuint VerticesSubmitted;
> + GLuint PrimitivesSubmitted;
> + GLuint VsInvocations;
> + GLuint TessPatches;
> + GLuint TessInvocations;
> + GLuint GsInvocations;
> + GLuint GsPrimitives;
> + GLuint FsInvocations;
> + GLuint ComputeInvocations;
> + GLuint ClInPrimitives;
> + GLuint ClOutPrimitives;
> } QueryCounterBits;
>
> GLuint MaxDrawBuffers; /**< GL_ARB_draw_buffers */
> @@ -3735,6 +3749,7 @@ struct gl_extensions
> GLboolean ARB_map_buffer_range;
> GLboolean ARB_occlusion_query;
> GLboolean ARB_occlusion_query2;
> + GLboolean ARB_pipeline_statistics_query;
> GLboolean ARB_point_sprite;
> GLboolean ARB_sample_shading;
> GLboolean ARB_seamless_cube_map;
> diff --git a/src/mesa/main/queryobj.c b/src/mesa/main/queryobj.c
> index 932359c..6cbacb3 100644
> --- a/src/mesa/main/queryobj.c
> +++ b/src/mesa/main/queryobj.c
> @@ -142,6 +142,14 @@ _mesa_init_query_object_functions(struct dd_function_table *driver)
> driver->CheckQuery = _mesa_check_query;
> }
>
> +static struct gl_query_object **
> +get_pipe_stats_binding_point(struct gl_context *ctx,
> + GLenum target)
> +{
> + const int which = target - GL_VERTICES_SUBMITTED_ARB;
> + assert(which < MAX_PIPELINE_STATISTICS);
> + return &ctx->Query.pipeline_stats[which];
> +}
>
> /**
> * Return pointer to the query object binding point for the given target and
> @@ -183,6 +191,42 @@ get_query_binding_point(struct gl_context *ctx, GLenum target, GLuint index)
> return &ctx->Query.PrimitivesWritten[index];
> else
> return NULL;
> +
> + case GL_VERTICES_SUBMITTED_ARB:
> + case GL_PRIMITIVES_SUBMITTED_ARB:
> + case GL_VERTEX_SHADER_INVOCATIONS_ARB:
> + case GL_FRAGMENT_SHADER_INVOCATIONS_ARB:
> + case GL_CLIPPING_INPUT_PRIMITIVES_ARB:
> + case GL_CLIPPING_OUTPUT_PRIMITIVES_ARB:
> + if (ctx->Extensions.ARB_pipeline_statistics_query)
> + return get_pipe_stats_binding_point(ctx, target);
> + else
> + return NULL;
> +
> + case GL_GEOMETRY_SHADER_INVOCATIONS:
> + /* GL_GEOMETRY_SHADER_INVOCATIONS is defined in a non-sequential order */
> + target = GL_VERTICES_SUBMITTED_ARB + MAX_PIPELINE_STATISTICS - 1;
> + /* fallthrough */
> + case GL_GEOMETRY_SHADER_PRIMITIVES_EMITTED_ARB:
> + if (ctx->Version >= 32 || ctx->Extensions.ARB_geometry_shader4)
if (_mesa_has_geometry_shaders(ctx))
> + return get_pipe_stats_binding_point(ctx, target);
> + else
> + return NULL;
> +
> +#if 0 /* Not yet testable */
> + case GL_TESS_CONTROL_SHADER_PATCHES_ARB:
> + case GL_TESS_EVALUATION_SHADER_INVOCATIONS_ARB:
> + if (ctx->Version >= 40 || ctx->Extensions.ARB_tessellation_shader)
> + return get_pipe_stats_binding_point(ctx, target, index);
> + else
> + return NULL;
> +
> + case GL_COMPUTE_SHADER_INVOCATIONS_ARB:
> + if (ctx->Version >= 43 || ctx->Extensions.ARB_compute_shader)
> + return get_pipe_stats_binding_point(ctx, target, index);
> + return NULL;
> +#endif
> +
> default:
> return NULL;
> }
> @@ -553,6 +597,39 @@ _mesa_GetQueryIndexediv(GLenum target, GLuint index, GLenum pname,
> case GL_TRANSFORM_FEEDBACK_PRIMITIVES_WRITTEN:
> *params = ctx->Const.QueryCounterBits.PrimitivesWritten;
> break;
> + case GL_VERTICES_SUBMITTED_ARB:
> + *params = ctx->Const.QueryCounterBits.VerticesSubmitted;
> + break;
> + case GL_PRIMITIVES_SUBMITTED_ARB:
> + *params = ctx->Const.QueryCounterBits.PrimitivesSubmitted;
> + break;
> + case GL_VERTEX_SHADER_INVOCATIONS_ARB:
> + *params = ctx->Const.QueryCounterBits.VsInvocations;
> + break;
> + case GL_TESS_CONTROL_SHADER_PATCHES_ARB:
> + *params = ctx->Const.QueryCounterBits.TessPatches;
> + break;
> + case GL_TESS_EVALUATION_SHADER_INVOCATIONS_ARB:
> + *params = ctx->Const.QueryCounterBits.TessInvocations;
> + break;
> + case GL_GEOMETRY_SHADER_INVOCATIONS:
> + *params = ctx->Const.QueryCounterBits.GsInvocations;
> + break;
> + case GL_GEOMETRY_SHADER_PRIMITIVES_EMITTED_ARB:
> + *params = ctx->Const.QueryCounterBits.GsPrimitives;
> + break;
> + case GL_FRAGMENT_SHADER_INVOCATIONS_ARB:
> + *params = ctx->Const.QueryCounterBits.FsInvocations;
> + break;
> + case GL_COMPUTE_SHADER_INVOCATIONS_ARB:
> + *params = ctx->Const.QueryCounterBits.ComputeInvocations;
> + break;
> + case GL_CLIPPING_INPUT_PRIMITIVES_ARB:
> + *params = ctx->Const.QueryCounterBits.ClInPrimitives;
> + break;
> + case GL_CLIPPING_OUTPUT_PRIMITIVES_ARB:
> + *params = ctx->Const.QueryCounterBits.ClOutPrimitives;
> + break;
> default:
> _mesa_problem(ctx,
> "Unknown target in glGetQueryIndexediv(target = %s)",
>
More information about the mesa-dev
mailing list