From 2164f6b1eb7216c1045d547c94f26fe3ceaa9403 Mon Sep 17 00:00:00 2001 From: Aapo Alasuutari Date: Thu, 11 Aug 2022 16:56:56 +0300 Subject: perf(ops): Monomorphic sync op calls (#15337) Welcome to better optimised op calls! Currently opSync is called with parameters of every type and count. This most definitely makes the call megamorphic. Additionally, it seems that spread params leads to V8 not being able to optimise the calls quite as well (apparently Fast Calls cannot be used with spread params). Monomorphising op calls should lead to some improved performance. Now that unwrapping of sync ops results is done on Rust side, this is pretty simple: ``` opSync("op_foo", param1, param2); // -> turns to ops.op_foo(param1, param2); ``` This means sync op calls are now just directly calling the native binding function. When V8 Fast API Calls are enabled, this will enable those to be called on the optimised path. Monomorphising async ops likely requires using callbacks and is left as an exercise to the reader. --- ext/webgpu/src/01_webgpu.js | 232 ++++++++++++++++---------------------------- 1 file changed, 84 insertions(+), 148 deletions(-) (limited to 'ext/webgpu/src/01_webgpu.js') diff --git a/ext/webgpu/src/01_webgpu.js b/ext/webgpu/src/01_webgpu.js index 79d1f497c..9b8c7eb50 100644 --- a/ext/webgpu/src/01_webgpu.js +++ b/ext/webgpu/src/01_webgpu.js @@ -10,6 +10,7 @@ ((window) => { const core = window.Deno.core; + const ops = core.ops; const webidl = window.__bootstrap.webidl; const eventTarget = window.__bootstrap.eventTarget; const { DOMException } = window.__bootstrap.domException; @@ -940,8 +941,7 @@ context: "Argument 1", }); const device = assertDevice(this, { prefix, context: "this" }); - const { rid, err } = core.opSync( - "op_webgpu_create_buffer", + const { rid, err } = ops.op_webgpu_create_buffer( device.rid, descriptor.label, descriptor.size, @@ -991,7 +991,7 @@ context: "Argument 1", }); const device = assertDevice(this, { prefix, context: "this" }); - const { rid, err } = core.opSync("op_webgpu_create_texture", { + const { rid, err } = ops.op_webgpu_create_texture({ deviceRid: device.rid, ...descriptor, size: normalizeGPUExtent3D(descriptor.size), @@ -1019,7 +1019,7 @@ context: "Argument 1", }); const device = assertDevice(this, { prefix, context: "this" }); - const { rid, err } = core.opSync("op_webgpu_create_sampler", { + const { rid, err } = ops.op_webgpu_create_sampler({ deviceRid: device.rid, ...descriptor, }); @@ -1059,8 +1059,7 @@ } } - const { rid, err } = core.opSync( - "op_webgpu_create_bind_group_layout", + const { rid, err } = ops.op_webgpu_create_bind_group_layout( device.rid, descriptor.label, descriptor.entries, @@ -1102,8 +1101,7 @@ return rid; }, ); - const { rid, err } = core.opSync( - "op_webgpu_create_pipeline_layout", + const { rid, err } = ops.op_webgpu_create_pipeline_layout( device.rid, descriptor.label, bindGroupLayouts, @@ -1197,8 +1195,7 @@ } }); - const { rid, err } = core.opSync( - "op_webgpu_create_bind_group", + const { rid, err } = ops.op_webgpu_create_bind_group( device.rid, descriptor.label, layout, @@ -1227,8 +1224,7 @@ context: "Argument 1", }); const device = assertDevice(this, { prefix, context: "this" }); - const { rid, err } = core.opSync( - "op_webgpu_create_shader_module", + const { rid, err } = ops.op_webgpu_create_shader_module( device.rid, descriptor.label, descriptor.code, @@ -1278,8 +1274,7 @@ selfContext: "this", }); - const { rid, err } = core.opSync( - "op_webgpu_create_compute_pipeline", + const { rid, err } = ops.op_webgpu_create_compute_pipeline( device.rid, descriptor.label, layout, @@ -1350,7 +1345,7 @@ }; } - const { rid, err } = core.opSync("op_webgpu_create_render_pipeline", { + const { rid, err } = ops.op_webgpu_create_render_pipeline({ deviceRid: device.rid, label: descriptor.label, layout, @@ -1397,8 +1392,7 @@ context: "Argument 1", }); const device = assertDevice(this, { prefix, context: "this" }); - const { rid, err } = core.opSync( - "op_webgpu_create_command_encoder", + const { rid, err } = ops.op_webgpu_create_command_encoder( device.rid, descriptor.label, ); @@ -1430,8 +1424,7 @@ }, ); const device = assertDevice(this, { prefix, context: "this" }); - const { rid, err } = core.opSync( - "op_webgpu_create_render_bundle_encoder", + const { rid, err } = ops.op_webgpu_create_render_bundle_encoder( { deviceRid: device.rid, ...descriptor, @@ -1464,7 +1457,7 @@ }, ); const device = assertDevice(this, { prefix, context: "this" }); - const { rid, err } = core.opSync("op_webgpu_create_query_set", { + const { rid, err } = ops.op_webgpu_create_query_set({ deviceRid: device.rid, ...descriptor, }); @@ -1595,8 +1588,7 @@ return rid; }, ); - const { err } = core.opSync( - "op_webgpu_queue_submit", + const { err } = ops.op_webgpu_queue_submit( device.rid, commandBufferRids, ); @@ -1654,8 +1646,7 @@ selfContext: "this", resourceContext: "Argument 1", }); - const { err } = core.opSync( - "op_webgpu_write_buffer", + const { err } = ops.op_webgpu_write_buffer( device.rid, bufferRid, bufferOffset, @@ -1702,8 +1693,7 @@ selfContext: "this", resourceContext: "texture", }); - const { err } = core.opSync( - "op_webgpu_write_texture", + const { err } = ops.op_webgpu_write_texture( device.rid, { texture: textureRid, @@ -1960,8 +1950,7 @@ } const buffer = new ArrayBuffer(rangeSize); - const { rid } = core.opSync( - "op_webgpu_buffer_get_mapped_range", + const { rid } = ops.op_webgpu_buffer_get_mapped_range( bufferRid, offset, size, @@ -2015,8 +2004,7 @@ throw new DOMException(`${prefix}: invalid state.`, "OperationError"); } for (const [buffer, mappedRid] of mappedRanges) { - const { err } = core.opSync( - "op_webgpu_buffer_unmap", + const { err } = ops.op_webgpu_buffer_unmap( bufferRid, mappedRid, ...new SafeArrayIterator(write ? [new Uint8Array(buffer)] : []), @@ -2153,7 +2141,7 @@ }); const device = assertDevice(this, { prefix, context: "this" }); const textureRid = assertResource(this, { prefix, context: "this" }); - const { rid, err } = core.opSync("op_webgpu_create_texture_view", { + const { rid, err } = ops.op_webgpu_create_texture_view({ textureRid, ...descriptor, }); @@ -2536,11 +2524,11 @@ prefix, context: "this", }); - const { rid, label, err } = core.opSync( - "op_webgpu_compute_pipeline_get_bind_group_layout", - computePipelineRid, - index, - ); + const { rid, label, err } = ops + .op_webgpu_compute_pipeline_get_bind_group_layout( + computePipelineRid, + index, + ); device.pushError(err); const bindGroupLayout = createGPUBindGroupLayout( @@ -2613,11 +2601,11 @@ prefix, context: "this", }); - const { rid, label, err } = core.opSync( - "op_webgpu_render_pipeline_get_bind_group_layout", - renderPipelineRid, - index, - ); + const { rid, label, err } = ops + .op_webgpu_render_pipeline_get_bind_group_layout( + renderPipelineRid, + index, + ); device.pushError(err); const bindGroupLayout = createGPUBindGroupLayout( @@ -2807,8 +2795,7 @@ }, ); - const { rid } = core.opSync( - "op_webgpu_command_encoder_begin_render_pass", + const { rid } = ops.op_webgpu_command_encoder_begin_render_pass( commandEncoderRid, descriptor.label, colorAttachments, @@ -2842,8 +2829,7 @@ context: "this", }); - const { rid } = core.opSync( - "op_webgpu_command_encoder_begin_compute_pass", + const { rid } = ops.op_webgpu_command_encoder_begin_compute_pass( commandEncoderRid, descriptor.label, ); @@ -2919,8 +2905,7 @@ selfContext: "this", }); - const { err } = core.opSync( - "op_webgpu_command_encoder_copy_buffer_to_buffer", + const { err } = ops.op_webgpu_command_encoder_copy_buffer_to_buffer( commandEncoderRid, sourceRid, sourceOffset, @@ -2977,8 +2962,7 @@ selfContext: "this", }); - const { err } = core.opSync( - "op_webgpu_command_encoder_copy_buffer_to_texture", + const { err } = ops.op_webgpu_command_encoder_copy_buffer_to_texture( commandEncoderRid, { ...source, @@ -3042,8 +3026,7 @@ resourceContext: "buffer in Argument 2", selfContext: "this", }); - const { err } = core.opSync( - "op_webgpu_command_encoder_copy_texture_to_buffer", + const { err } = ops.op_webgpu_command_encoder_copy_texture_to_buffer( commandEncoderRid, { texture: sourceTextureRid, @@ -3107,8 +3090,7 @@ resourceContext: "texture in Argument 2", selfContext: "this", }); - const { err } = core.opSync( - "op_webgpu_command_encoder_copy_texture_to_texture", + const { err } = ops.op_webgpu_command_encoder_copy_texture_to_texture( commandEncoderRid, { texture: sourceTextureRid, @@ -3161,8 +3143,7 @@ prefix, context: "Argument 1", }); - const { err } = core.opSync( - "op_webgpu_command_encoder_clear_buffer", + const { err } = ops.op_webgpu_command_encoder_clear_buffer( commandEncoderRid, bufferRid, offset, @@ -3188,8 +3169,7 @@ prefix, context: "this", }); - const { err } = core.opSync( - "op_webgpu_command_encoder_push_debug_group", + const { err } = ops.op_webgpu_command_encoder_push_debug_group( commandEncoderRid, groupLabel, ); @@ -3204,8 +3184,7 @@ prefix, context: "this", }); - const { err } = core.opSync( - "op_webgpu_command_encoder_pop_debug_group", + const { err } = ops.op_webgpu_command_encoder_pop_debug_group( commandEncoderRid, ); device.pushError(err); @@ -3228,8 +3207,7 @@ prefix, context: "this", }); - const { err } = core.opSync( - "op_webgpu_command_encoder_insert_debug_marker", + const { err } = ops.op_webgpu_command_encoder_insert_debug_marker( commandEncoderRid, markerLabel, ); @@ -3267,8 +3245,7 @@ resourceContext: "Argument 1", selfContext: "this", }); - const { err } = core.opSync( - "op_webgpu_command_encoder_write_timestamp", + const { err } = ops.op_webgpu_command_encoder_write_timestamp( commandEncoderRid, querySetRid, queryIndex, @@ -3337,8 +3314,7 @@ resourceContext: "Argument 3", selfContext: "this", }); - const { err } = core.opSync( - "op_webgpu_command_encoder_resolve_query_set", + const { err } = ops.op_webgpu_command_encoder_resolve_query_set( commandEncoderRid, querySetRid, firstQuery, @@ -3365,8 +3341,7 @@ prefix, context: "this", }); - const { rid, err } = core.opSync( - "op_webgpu_command_encoder_finish", + const { rid, err } = ops.op_webgpu_command_encoder_finish( commandEncoderRid, descriptor.label, ); @@ -3465,7 +3440,7 @@ context: "encoder referenced by this", }); const renderPassRid = assertResource(this, { prefix, context: "this" }); - core.opSync("op_webgpu_render_pass_set_viewport", { + ops.op_webgpu_render_pass_set_viewport({ renderPassRid, x, y, @@ -3512,8 +3487,7 @@ context: "encoder referenced by this", }); const renderPassRid = assertResource(this, { prefix, context: "this" }); - core.opSync( - "op_webgpu_render_pass_set_scissor_rect", + ops.op_webgpu_render_pass_set_scissor_rect( renderPassRid, x, y, @@ -3543,8 +3517,7 @@ context: "encoder referenced by this", }); const renderPassRid = assertResource(this, { prefix, context: "this" }); - core.opSync( - "op_webgpu_render_pass_set_blend_constant", + ops.op_webgpu_render_pass_set_blend_constant( renderPassRid, normalizeGPUColor(color), ); @@ -3571,8 +3544,7 @@ context: "encoder referenced by this", }); const renderPassRid = assertResource(this, { prefix, context: "this" }); - core.opSync( - "op_webgpu_render_pass_set_stencil_reference", + ops.op_webgpu_render_pass_set_stencil_reference( renderPassRid, reference, ); @@ -3621,8 +3593,7 @@ resourceContext: "Argument 1", selfContext: "this", }); - core.opSync( - "op_webgpu_render_pass_begin_pipeline_statistics_query", + ops.op_webgpu_render_pass_begin_pipeline_statistics_query( renderPassRid, querySetRid, queryIndex, @@ -3642,8 +3613,7 @@ context: "encoder referenced by this", }); const renderPassRid = assertResource(this, { prefix, context: "this" }); - core.opSync( - "op_webgpu_render_pass_end_pipeline_statistics_query", + ops.op_webgpu_render_pass_end_pipeline_statistics_query( renderPassRid, ); } @@ -3683,8 +3653,7 @@ resourceContext: "Argument 1", selfContext: "this", }); - core.opSync( - "op_webgpu_render_pass_write_timestamp", + ops.op_webgpu_render_pass_write_timestamp( renderPassRid, querySetRid, queryIndex, @@ -3722,8 +3691,7 @@ }); return rid; }); - core.opSync( - "op_webgpu_render_pass_execute_bundles", + ops.op_webgpu_render_pass_execute_bundles( renderPassRid, bundleRids, ); @@ -3741,8 +3709,7 @@ context: "encoder referenced by this", }); const renderPassRid = assertResource(this, { prefix, context: "this" }); - const { err } = core.opSync( - "op_webgpu_render_pass_end", + const { err } = ops.op_webgpu_render_pass_end( commandEncoderRid, renderPassRid, ); @@ -3789,8 +3756,7 @@ dynamicOffsetsDataStart = 0; dynamicOffsetsDataLength = dynamicOffsetsData.length; } - core.opSync( - "op_webgpu_render_pass_set_bind_group", + ops.op_webgpu_render_pass_set_bind_group( renderPassRid, index, bindGroupRid, @@ -3821,8 +3787,7 @@ context: "encoder referenced by this", }); const renderPassRid = assertResource(this, { prefix, context: "this" }); - core.opSync( - "op_webgpu_render_pass_push_debug_group", + ops.op_webgpu_render_pass_push_debug_group( renderPassRid, groupLabel, ); @@ -3841,7 +3806,7 @@ context: "encoder referenced by this", }); const renderPassRid = assertResource(this, { prefix, context: "this" }); - core.opSync("op_webgpu_render_pass_pop_debug_group", renderPassRid); + ops.op_webgpu_render_pass_pop_debug_group(renderPassRid); } /** @@ -3865,8 +3830,7 @@ context: "encoder referenced by this", }); const renderPassRid = assertResource(this, { prefix, context: "this" }); - core.opSync( - "op_webgpu_render_pass_insert_debug_marker", + ops.op_webgpu_render_pass_insert_debug_marker( renderPassRid, markerLabel, ); @@ -3902,8 +3866,7 @@ resourceContext: "Argument 1", selfContext: "this", }); - core.opSync( - "op_webgpu_render_pass_set_pipeline", + ops.op_webgpu_render_pass_set_pipeline( renderPassRid, pipelineRid, ); @@ -3956,8 +3919,7 @@ resourceContext: "Argument 1", selfContext: "this", }); - core.opSync( - "op_webgpu_render_pass_set_index_buffer", + ops.op_webgpu_render_pass_set_index_buffer( renderPassRid, bufferRid, indexFormat, @@ -4013,8 +3975,7 @@ resourceContext: "Argument 2", selfContext: "this", }); - core.opSync( - "op_webgpu_render_pass_set_vertex_buffer", + ops.op_webgpu_render_pass_set_vertex_buffer( renderPassRid, slot, bufferRid, @@ -4058,8 +4019,7 @@ context: "encoder referenced by this", }); const renderPassRid = assertResource(this, { prefix, context: "this" }); - core.opSync( - "op_webgpu_render_pass_draw", + ops.op_webgpu_render_pass_draw( renderPassRid, vertexCount, instanceCount, @@ -4115,8 +4075,7 @@ context: "encoder referenced by this", }); const renderPassRid = assertResource(this, { prefix, context: "this" }); - core.opSync( - "op_webgpu_render_pass_draw_indexed", + ops.op_webgpu_render_pass_draw_indexed( renderPassRid, indexCount, instanceCount, @@ -4161,8 +4120,7 @@ resourceContext: "Argument 1", selfContext: "this", }); - core.opSync( - "op_webgpu_render_pass_draw_indirect", + ops.op_webgpu_render_pass_draw_indirect( renderPassRid, indirectBufferRid, indirectOffset, @@ -4204,8 +4162,7 @@ resourceContext: "Argument 1", selfContext: "this", }); - core.opSync( - "op_webgpu_render_pass_draw_indexed_indirect", + ops.op_webgpu_render_pass_draw_indexed_indirect( renderPassRid, indirectBufferRid, indirectOffset, @@ -4288,8 +4245,7 @@ resourceContext: "Argument 1", selfContext: "this", }); - core.opSync( - "op_webgpu_compute_pass_set_pipeline", + ops.op_webgpu_compute_pass_set_pipeline( computePassRid, pipelineRid, ); @@ -4330,8 +4286,7 @@ context: "encoder referenced by this", }); const computePassRid = assertResource(this, { prefix, context: "this" }); - core.opSync( - "op_webgpu_compute_pass_dispatch_workgroups", + ops.op_webgpu_compute_pass_dispatch_workgroups( computePassRid, workgroupCountX, workgroupCountY, @@ -4374,8 +4329,7 @@ resourceContext: "Argument 1", selfContext: "this", }); - core.opSync( - "op_webgpu_compute_pass_dispatch_workgroups_indirect", + ops.op_webgpu_compute_pass_dispatch_workgroups_indirect( computePassRid, indirectBufferRid, indirectOffset, @@ -4417,8 +4371,7 @@ resourceContext: "Argument 1", selfContext: "this", }); - core.opSync( - "op_webgpu_compute_pass_begin_pipeline_statistics_query", + ops.op_webgpu_compute_pass_begin_pipeline_statistics_query( computePassRid, querySetRid, queryIndex, @@ -4438,8 +4391,7 @@ context: "encoder referenced by this", }); const computePassRid = assertResource(this, { prefix, context: "this" }); - core.opSync( - "op_webgpu_compute_pass_end_pipeline_statistics_query", + ops.op_webgpu_compute_pass_end_pipeline_statistics_query( computePassRid, ); } @@ -4479,8 +4431,7 @@ resourceContext: "Argument 1", selfContext: "this", }); - core.opSync( - "op_webgpu_compute_pass_write_timestamp", + ops.op_webgpu_compute_pass_write_timestamp( computePassRid, querySetRid, queryIndex, @@ -4499,8 +4450,7 @@ context: "encoder referenced by this", }); const computePassRid = assertResource(this, { prefix, context: "this" }); - const { err } = core.opSync( - "op_webgpu_compute_pass_end", + const { err } = ops.op_webgpu_compute_pass_end( commandEncoderRid, computePassRid, ); @@ -4547,8 +4497,7 @@ dynamicOffsetsDataStart = 0; dynamicOffsetsDataLength = dynamicOffsetsData.length; } - core.opSync( - "op_webgpu_compute_pass_set_bind_group", + ops.op_webgpu_compute_pass_set_bind_group( computePassRid, index, bindGroupRid, @@ -4579,8 +4528,7 @@ context: "encoder referenced by this", }); const computePassRid = assertResource(this, { prefix, context: "this" }); - core.opSync( - "op_webgpu_compute_pass_push_debug_group", + ops.op_webgpu_compute_pass_push_debug_group( computePassRid, groupLabel, ); @@ -4599,7 +4547,7 @@ context: "encoder referenced by this", }); const computePassRid = assertResource(this, { prefix, context: "this" }); - core.opSync("op_webgpu_compute_pass_pop_debug_group", computePassRid); + ops.op_webgpu_compute_pass_pop_debug_group(computePassRid); } /** @@ -4623,8 +4571,7 @@ context: "encoder referenced by this", }); const computePassRid = assertResource(this, { prefix, context: "this" }); - core.opSync( - "op_webgpu_compute_pass_insert_debug_marker", + ops.op_webgpu_compute_pass_insert_debug_marker( computePassRid, markerLabel, ); @@ -4734,8 +4681,7 @@ prefix, context: "this", }); - const { rid, err } = core.opSync( - "op_webgpu_render_bundle_encoder_finish", + const { rid, err } = ops.op_webgpu_render_bundle_encoder_finish( renderBundleEncoderRid, descriptor.label, ); @@ -4786,8 +4732,7 @@ dynamicOffsetsDataStart = 0; dynamicOffsetsDataLength = dynamicOffsetsData.length; } - core.opSync( - "op_webgpu_render_bundle_encoder_set_bind_group", + ops.op_webgpu_render_bundle_encoder_set_bind_group( renderBundleEncoderRid, index, bindGroupRid, @@ -4814,8 +4759,7 @@ prefix, context: "this", }); - core.opSync( - "op_webgpu_render_bundle_encoder_push_debug_group", + ops.op_webgpu_render_bundle_encoder_push_debug_group( renderBundleEncoderRid, groupLabel, ); @@ -4830,8 +4774,7 @@ prefix, context: "this", }); - core.opSync( - "op_webgpu_render_bundle_encoder_pop_debug_group", + ops.op_webgpu_render_bundle_encoder_pop_debug_group( renderBundleEncoderRid, ); } @@ -4853,8 +4796,7 @@ prefix, context: "this", }); - core.opSync( - "op_webgpu_render_bundle_encoder_insert_debug_marker", + ops.op_webgpu_render_bundle_encoder_insert_debug_marker( renderBundleEncoderRid, markerLabel, ); @@ -4886,8 +4828,7 @@ resourceContext: "Argument 1", selfContext: "this", }); - core.opSync( - "op_webgpu_render_bundle_encoder_set_pipeline", + ops.op_webgpu_render_bundle_encoder_set_pipeline( renderBundleEncoderRid, pipelineRid, ); @@ -4934,8 +4875,7 @@ resourceContext: "Argument 1", selfContext: "this", }); - core.opSync( - "op_webgpu_render_bundle_encoder_set_index_buffer", + ops.op_webgpu_render_bundle_encoder_set_index_buffer( renderBundleEncoderRid, bufferRid, indexFormat, @@ -4985,8 +4925,7 @@ resourceContext: "Argument 2", selfContext: "this", }); - core.opSync( - "op_webgpu_render_bundle_encoder_set_vertex_buffer", + ops.op_webgpu_render_bundle_encoder_set_vertex_buffer( renderBundleEncoderRid, slot, bufferRid, @@ -5026,8 +4965,7 @@ prefix, context: "this", }); - core.opSync( - "op_webgpu_render_bundle_encoder_draw", + ops.op_webgpu_render_bundle_encoder_draw( renderBundleEncoderRid, vertexCount, instanceCount, @@ -5079,8 +5017,7 @@ prefix, context: "this", }); - core.opSync( - "op_webgpu_render_bundle_encoder_draw_indexed", + ops.op_webgpu_render_bundle_encoder_draw_indexed( renderBundleEncoderRid, indexCount, instanceCount, @@ -5121,8 +5058,7 @@ resourceContext: "Argument 1", selfContext: "this", }); - core.opSync( - "op_webgpu_render_bundle_encoder_draw_indirect", + ops.op_webgpu_render_bundle_encoder_draw_indirect( renderBundleEncoderRid, indirectBufferRid, indirectOffset, -- cgit v1.2.3