diff options
Diffstat (limited to 'src/intel/vulkan')
-rw-r--r-- | src/intel/vulkan/anv_allocator.c | 15 | ||||
-rw-r--r-- | src/intel/vulkan/anv_device.c | 22 | ||||
-rw-r--r-- | src/intel/vulkan/anv_private.h | 11 | ||||
-rw-r--r-- | src/intel/vulkan/gen7_pipeline.c | 11 | ||||
-rw-r--r-- | src/intel/vulkan/gen8_pipeline.c | 7 | ||||
-rw-r--r-- | src/intel/vulkan/genX_pipeline.c | 5 |
6 files changed, 28 insertions, 43 deletions
diff --git a/src/intel/vulkan/anv_allocator.c b/src/intel/vulkan/anv_allocator.c index a59ad3cd6f2..83a701e0b0d 100644 --- a/src/intel/vulkan/anv_allocator.c +++ b/src/intel/vulkan/anv_allocator.c @@ -924,8 +924,9 @@ anv_scratch_pool_alloc(struct anv_device *device, struct anv_scratch_pool *pool, if (size == 0) { /* We own the lock. Allocate a buffer */ - struct anv_physical_device *physical_device = + const struct anv_physical_device *physical_device = &device->instance->physicalDevice; + const struct gen_device_info *devinfo = &physical_device->info; /* WaCSScratchSize:hsw * @@ -944,14 +945,14 @@ anv_scratch_pool_alloc(struct anv_device *device, struct anv_scratch_pool *pool, */ const unsigned subslices = MAX2(physical_device->subslice_total, 1); const unsigned scratch_ids_per_subslice = - device->info.is_haswell ? 16 * 8 : physical_device->max_cs_threads; + device->info.is_haswell ? 16 * 8 : devinfo->max_cs_threads; uint32_t max_threads[] = { - [MESA_SHADER_VERTEX] = physical_device->max_vs_threads, - [MESA_SHADER_TESS_CTRL] = physical_device->max_hs_threads, - [MESA_SHADER_TESS_EVAL] = physical_device->max_ds_threads, - [MESA_SHADER_GEOMETRY] = physical_device->max_gs_threads, - [MESA_SHADER_FRAGMENT] = physical_device->max_wm_threads, + [MESA_SHADER_VERTEX] = devinfo->max_vs_threads, + [MESA_SHADER_TESS_CTRL] = devinfo->max_hs_threads, + [MESA_SHADER_TESS_EVAL] = devinfo->max_ds_threads, + [MESA_SHADER_GEOMETRY] = devinfo->max_gs_threads, + [MESA_SHADER_FRAGMENT] = devinfo->max_wm_threads, [MESA_SHADER_COMPUTE] = scratch_ids_per_subslice * subslices, }; diff --git a/src/intel/vulkan/anv_device.c b/src/intel/vulkan/anv_device.c index 6edbb1c8c1b..f786ebe027a 100644 --- a/src/intel/vulkan/anv_device.c +++ b/src/intel/vulkan/anv_device.c @@ -135,12 +135,6 @@ anv_physical_device_init(struct anv_physical_device *device, bool swizzled = anv_gem_get_bit6_swizzle(fd, I915_TILING_X); - device->max_vs_threads = device->info.max_vs_threads; - device->max_hs_threads = device->info.max_hs_threads; - device->max_ds_threads = device->info.max_ds_threads; - device->max_gs_threads = device->info.max_gs_threads; - device->max_wm_threads = device->info.max_wm_threads; - /* GENs prior to 8 do not support EU/Subslice info */ if (device->info.gen >= 8) { device->subslice_total = anv_gem_get_param(fd, I915_PARAM_SUBSLICE_TOTAL); @@ -161,13 +155,11 @@ anv_physical_device_init(struct anv_physical_device *device, if (device->info.is_cherryview && device->subslice_total > 0 && device->eu_total > 0) { /* Logical CS threads = EUs per subslice * 7 threads per EU */ - device->max_cs_threads = device->eu_total / device->subslice_total * 7; + uint32_t max_cs_threads = device->eu_total / device->subslice_total * 7; /* Fuse configurations may give more threads than expected, never less. */ - if (device->max_cs_threads < device->info.max_cs_threads) - device->max_cs_threads = device->info.max_cs_threads; - } else { - device->max_cs_threads = device->info.max_cs_threads; + if (max_cs_threads > device->info.max_cs_threads) + device->info.max_cs_threads = max_cs_threads; } close(fd); @@ -537,11 +529,11 @@ void anv_GetPhysicalDeviceProperties( .maxFragmentCombinedOutputResources = 8, .maxComputeSharedMemorySize = 32768, .maxComputeWorkGroupCount = { 65535, 65535, 65535 }, - .maxComputeWorkGroupInvocations = 16 * pdevice->max_cs_threads, + .maxComputeWorkGroupInvocations = 16 * devinfo->max_cs_threads, .maxComputeWorkGroupSize = { - 16 * pdevice->max_cs_threads, - 16 * pdevice->max_cs_threads, - 16 * pdevice->max_cs_threads, + 16 * devinfo->max_cs_threads, + 16 * devinfo->max_cs_threads, + 16 * devinfo->max_cs_threads, }, .subPixelPrecisionBits = 4 /* FIXME */, .subTexelPrecisionBits = 4 /* FIXME */, diff --git a/src/intel/vulkan/anv_private.h b/src/intel/vulkan/anv_private.h index 839f813d0d7..443c31fdbbe 100644 --- a/src/intel/vulkan/anv_private.h +++ b/src/intel/vulkan/anv_private.h @@ -573,17 +573,6 @@ struct anv_physical_device { uint32_t eu_total; uint32_t subslice_total; - /** - * Platform specific constants containing the maximum number of threads - * for each pipeline stage. - */ - uint32_t max_vs_threads; - uint32_t max_hs_threads; - uint32_t max_ds_threads; - uint32_t max_gs_threads; - uint32_t max_wm_threads; - uint32_t max_cs_threads; - struct anv_wsi_interface * wsi[VK_ICD_WSI_PLATFORM_MAX]; }; diff --git a/src/intel/vulkan/gen7_pipeline.c b/src/intel/vulkan/gen7_pipeline.c index 878308b7fbc..86d01e4fcca 100644 --- a/src/intel/vulkan/gen7_pipeline.c +++ b/src/intel/vulkan/gen7_pipeline.c @@ -45,8 +45,9 @@ genX(graphics_pipeline_create)( { ANV_FROM_HANDLE(anv_device, device, _device); ANV_FROM_HANDLE(anv_render_pass, pass, pCreateInfo->renderPass); - struct anv_physical_device *physical_device = + const struct anv_physical_device *physical_device = &device->instance->physicalDevice; + const struct gen_device_info *devinfo = &physical_device->info; struct anv_subpass *subpass = &pass->subpasses[pCreateInfo->subpass]; struct anv_pipeline *pipeline; VkResult result; @@ -125,7 +126,7 @@ genX(graphics_pipeline_create)( vs.VertexURBEntryReadLength = vs_prog_data->base.urb_read_length; vs.VertexURBEntryReadOffset = 0; - vs.MaximumNumberofThreads = physical_device->max_vs_threads - 1; + vs.MaximumNumberofThreads = devinfo->max_vs_threads - 1; vs.StatisticsEnable = true; vs.VSFunctionEnable = true; } @@ -154,7 +155,7 @@ genX(graphics_pipeline_create)( gs.DispatchGRFStartRegisterforURBData = gs_prog_data->base.base.dispatch_grf_start_reg; - gs.MaximumNumberofThreads = physical_device->max_gs_threads - 1; + gs.MaximumNumberofThreads = devinfo->max_gs_threads - 1; /* This in the next dword on HSW. */ gs.ControlDataFormat = gs_prog_data->control_data_format; gs.ControlDataHeaderSize = gs_prog_data->control_data_header_size_hwords; @@ -187,7 +188,7 @@ genX(graphics_pipeline_create)( * don't at least set the maximum number of threads. */ anv_batch_emit(&pipeline->batch, GENX(3DSTATE_PS), ps) { - ps.MaximumNumberofThreads = physical_device->max_wm_threads - 1; + ps.MaximumNumberofThreads = devinfo->max_wm_threads - 1; } } else { const struct brw_wm_prog_data *wm_prog_data = get_wm_prog_data(pipeline); @@ -209,7 +210,7 @@ genX(graphics_pipeline_create)( .offset = 0, }; ps.PerThreadScratchSpace = scratch_space(&wm_prog_data->base); - ps.MaximumNumberofThreads = physical_device->max_wm_threads - 1; + ps.MaximumNumberofThreads = devinfo->max_wm_threads - 1; ps.PushConstantEnable = wm_prog_data->base.nr_params > 0; ps.AttributeEnable = wm_prog_data->num_varying_inputs > 0; ps.oMaskPresenttoRenderTarget = wm_prog_data->uses_omask; diff --git a/src/intel/vulkan/gen8_pipeline.c b/src/intel/vulkan/gen8_pipeline.c index 9f16960ccc7..030f71ef147 100644 --- a/src/intel/vulkan/gen8_pipeline.c +++ b/src/intel/vulkan/gen8_pipeline.c @@ -55,8 +55,9 @@ genX(graphics_pipeline_create)( { ANV_FROM_HANDLE(anv_device, device, _device); ANV_FROM_HANDLE(anv_render_pass, pass, pCreateInfo->renderPass); - struct anv_physical_device *physical_device = + const struct anv_physical_device *physical_device = &device->instance->physicalDevice; + const struct gen_device_info *devinfo = &physical_device->info; struct anv_subpass *subpass = &pass->subpasses[pCreateInfo->subpass]; struct anv_pipeline *pipeline; VkResult result; @@ -144,7 +145,7 @@ genX(graphics_pipeline_create)( gs.DispatchGRFStartRegisterForURBData = gs_prog_data->base.base.dispatch_grf_start_reg; - gs.MaximumNumberofThreads = physical_device->max_gs_threads / 2 - 1; + gs.MaximumNumberofThreads = devinfo->max_gs_threads / 2 - 1; gs.ControlDataHeaderSize = gs_prog_data->control_data_header_size_hwords; gs.DispatchMode = gs_prog_data->base.dispatch_mode; gs.StatisticsEnable = true; @@ -215,7 +216,7 @@ genX(graphics_pipeline_create)( vs.VertexURBEntryReadLength = vs_prog_data->base.urb_read_length; vs.VertexURBEntryReadOffset = 0; - vs.MaximumNumberofThreads = physical_device->max_vs_threads - 1; + vs.MaximumNumberofThreads = devinfo->max_vs_threads - 1; vs.StatisticsEnable = false; vs.SIMD8DispatchEnable = pipeline->vs_simd8 != NO_KERNEL; vs.VertexCacheDisable = false; diff --git a/src/intel/vulkan/genX_pipeline.c b/src/intel/vulkan/genX_pipeline.c index e1929da25ff..e5ad6ba4af2 100644 --- a/src/intel/vulkan/genX_pipeline.c +++ b/src/intel/vulkan/genX_pipeline.c @@ -35,8 +35,9 @@ genX(compute_pipeline_create)( VkPipeline* pPipeline) { ANV_FROM_HANDLE(anv_device, device, _device); - struct anv_physical_device *physical_device = + const struct anv_physical_device *physical_device = &device->instance->physicalDevice; + const struct gen_device_info *devinfo = &physical_device->info; struct anv_pipeline *pipeline; VkResult result; @@ -120,7 +121,7 @@ genX(compute_pipeline_create)( vfe.GPGPUMode = true; #endif vfe.MaximumNumberofThreads = - physical_device->max_cs_threads * subslices - 1; + devinfo->max_cs_threads * subslices - 1; vfe.NumberofURBEntries = GEN_GEN <= 7 ? 0 : 2; vfe.ResetGatewayTimer = true; #if GEN_GEN <= 8 |