X-Git-Url: https://git.libre-soc.org/?a=blobdiff_plain;f=src%2Fgallium%2Fdrivers%2Fradeonsi%2Fsi_get.c;h=a5cb209b59ee00194563ceee206a59c9b64056d3;hb=ae2cb7280436ab3fe24afef510d30201e54b795c;hp=9d386364e21e52090416dbc7793f7503fcf647da;hpb=b86feec390f744f689a53d8c855b53ddee981d2c;p=mesa.git diff --git a/src/gallium/drivers/radeonsi/si_get.c b/src/gallium/drivers/radeonsi/si_get.c index 9d386364e21..a5cb209b59e 100644 --- a/src/gallium/drivers/radeonsi/si_get.c +++ b/src/gallium/drivers/radeonsi/si_get.c @@ -1,5 +1,6 @@ /* * Copyright 2017 Advanced Micro Devices, Inc. + * All Rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the "Software"), @@ -24,9 +25,11 @@ #include "si_pipe.h" #include "radeon/radeon_video.h" #include "radeon/radeon_vce.h" +#include "radeon/radeon_uvd_enc.h" #include "ac_llvm_util.h" #include "vl/vl_decoder.h" #include "vl/vl_video_buffer.h" +#include "util/u_screen.h" #include "util/u_video.h" #include "compiler/nir/nir.h" @@ -34,6 +37,9 @@ static const char *si_get_vendor(struct pipe_screen *pscreen) { + /* Don't change this. Games such as Alien Isolation are broken if this + * returns "Advanced Micro Devices, Inc." + */ return "X.Org"; } @@ -49,43 +55,6 @@ static const char *si_get_marketing_name(struct radeon_winsys *ws) return ws->get_chip_name(ws); } -const char *si_get_family_name(const struct si_screen *sscreen) -{ - switch (sscreen->b.info.family) { - case CHIP_TAHITI: return "AMD TAHITI"; - case CHIP_PITCAIRN: return "AMD PITCAIRN"; - case CHIP_VERDE: return "AMD CAPE VERDE"; - case CHIP_OLAND: return "AMD OLAND"; - case CHIP_HAINAN: return "AMD HAINAN"; - case CHIP_BONAIRE: return "AMD BONAIRE"; - case CHIP_KAVERI: return "AMD KAVERI"; - case CHIP_KABINI: return "AMD KABINI"; - case CHIP_HAWAII: return "AMD HAWAII"; - case CHIP_MULLINS: return "AMD MULLINS"; - case CHIP_TONGA: return "AMD TONGA"; - case CHIP_ICELAND: return "AMD ICELAND"; - case CHIP_CARRIZO: return "AMD CARRIZO"; - case CHIP_FIJI: return "AMD FIJI"; - case CHIP_POLARIS10: return "AMD POLARIS10"; - case CHIP_POLARIS11: return "AMD POLARIS11"; - case CHIP_POLARIS12: return "AMD POLARIS12"; - case CHIP_STONEY: return "AMD STONEY"; - case CHIP_VEGA10: return "AMD VEGA10"; - case CHIP_RAVEN: return "AMD RAVEN"; - default: return "AMD unknown"; - } -} - -static bool si_have_tgsi_compute(struct si_screen *sscreen) -{ - /* Old kernels disallowed some register writes for SI - * that are used for indirect dispatches. */ - return (sscreen->b.chip_class >= CIK || - sscreen->b.info.drm_major == 3 || - (sscreen->b.info.drm_major == 2 && - sscreen->b.info.drm_minor >= 45)); -} - static int si_get_param(struct pipe_screen *pscreen, enum pipe_cap param) { struct si_screen *sscreen = (struct si_screen *)pscreen; @@ -93,16 +62,16 @@ static int si_get_param(struct pipe_screen *pscreen, enum pipe_cap param) switch (param) { /* Supported features (boolean caps). */ case PIPE_CAP_ACCELERATED: - case PIPE_CAP_TWO_SIDED_STENCIL: case PIPE_CAP_MAX_DUAL_SOURCE_RENDER_TARGETS: case PIPE_CAP_ANISOTROPIC_FILTER: case PIPE_CAP_POINT_SPRITE: case PIPE_CAP_OCCLUSION_QUERY: - case PIPE_CAP_TEXTURE_SHADOW_MAP: case PIPE_CAP_TEXTURE_MIRROR_CLAMP: + case PIPE_CAP_TEXTURE_MIRROR_CLAMP_TO_EDGE: case PIPE_CAP_BLEND_EQUATION_SEPARATE: case PIPE_CAP_TEXTURE_SWIZZLE: case PIPE_CAP_DEPTH_CLIP_DISABLE: + case PIPE_CAP_DEPTH_CLIP_DISABLE_SEPARATE: case PIPE_CAP_SHADER_STENCIL_EXPORT: case PIPE_CAP_VERTEX_ELEMENT_INSTANCE_DIVISOR: case PIPE_CAP_MIXED_COLORBUFFER_FORMATS: @@ -118,7 +87,6 @@ static int si_get_param(struct pipe_screen *pscreen, enum pipe_cap param) case PIPE_CAP_INDEP_BLEND_FUNC: case PIPE_CAP_SEAMLESS_CUBE_MAP_PER_TEXTURE: case PIPE_CAP_VERTEX_COLOR_UNCLAMPED: - case PIPE_CAP_USER_CONSTANT_BUFFERS: case PIPE_CAP_START_INSTANCE: case PIPE_CAP_NPOT_TEXTURES: case PIPE_CAP_MIXED_FRAMEBUFFER_SIZES: @@ -157,6 +125,7 @@ static int si_get_param(struct pipe_screen *pscreen, enum pipe_cap param) case PIPE_CAP_TGSI_FS_FACE_IS_INTEGER_SYSVAL: case PIPE_CAP_INVALIDATE_BUFFER: case PIPE_CAP_SURFACE_REINTERPRET_BLOCKS: + case PIPE_CAP_QUERY_BUFFER_OBJECT: case PIPE_CAP_QUERY_MEMORY_INFO: case PIPE_CAP_TGSI_PACK_HALF_FLOAT: case PIPE_CAP_FRAMEBUFFER_NO_ATTACHMENT: @@ -188,31 +157,23 @@ static int si_get_param(struct pipe_screen *pscreen, enum pipe_cap param) case PIPE_CAP_ALLOW_MAPPED_BUFFERS_DURING_EXECUTION: case PIPE_CAP_TGSI_ANY_REG_AS_ADDRESS: case PIPE_CAP_SIGNED_VERTEX_BUFFER_OFFSET: - return 1; - - case PIPE_CAP_TGSI_VOTE: - return HAVE_LLVM >= 0x0400; - case PIPE_CAP_TGSI_BALLOT: - return HAVE_LLVM >= 0x0500; + case PIPE_CAP_TGSI_VOTE: + case PIPE_CAP_TGSI_FS_FBFETCH: + return 1; case PIPE_CAP_RESOURCE_FROM_USER_MEMORY: - return !SI_BIG_ENDIAN && sscreen->b.info.has_userptr; + return !SI_BIG_ENDIAN && sscreen->info.has_userptr; case PIPE_CAP_DEVICE_RESET_STATUS_QUERY: - return (sscreen->b.info.drm_major == 2 && - sscreen->b.info.drm_minor >= 43) || - sscreen->b.info.drm_major == 3; + return sscreen->info.has_gpu_reset_status_query || + sscreen->info.has_gpu_reset_counter_query; case PIPE_CAP_TEXTURE_MULTISAMPLE: - /* 2D tiling on CIK is supported since DRM 2.35.0 */ - return sscreen->b.chip_class < CIK || - (sscreen->b.info.drm_major == 2 && - sscreen->b.info.drm_minor >= 35) || - sscreen->b.info.drm_major == 3; + return sscreen->info.has_2d_tiling; case PIPE_CAP_MIN_MAP_BUFFER_ALIGNMENT: - return R600_MAP_BUFFER_ALIGNMENT; + return SI_MAP_BUFFER_ALIGNMENT; case PIPE_CAP_CONSTANT_BUFFER_OFFSET_ALIGNMENT: case PIPE_CAP_TEXTURE_BUFFER_OFFSET_ALIGNMENT: @@ -220,38 +181,35 @@ static int si_get_param(struct pipe_screen *pscreen, enum pipe_cap param) case PIPE_CAP_MAX_STREAM_OUTPUT_BUFFERS: case PIPE_CAP_MAX_VERTEX_STREAMS: case PIPE_CAP_SHADER_BUFFER_OFFSET_ALIGNMENT: + case PIPE_CAP_MAX_WINDOW_RECTANGLES: return 4; case PIPE_CAP_GLSL_FEATURE_LEVEL: - if (sscreen->b.debug_flags & DBG(NIR)) - return 140; /* no geometry and tessellation shaders yet */ - if (si_have_tgsi_compute(sscreen)) - return 450; + case PIPE_CAP_GLSL_FEATURE_LEVEL_COMPATIBILITY: + if (sscreen->info.has_indirect_compute_dispatch) + return 450; return 420; + case PIPE_CAP_MAX_TEXTURE_UPLOAD_MEMORY_BUDGET: + /* Optimal number for good TexSubImage performance on Polaris10. */ + return 64 * 1024 * 1024; + case PIPE_CAP_MAX_TEXTURE_BUFFER_SIZE: - return MIN2(sscreen->b.info.max_alloc_size, INT_MAX); + case PIPE_CAP_MAX_SHADER_BUFFER_SIZE: + return MIN2(sscreen->info.max_alloc_size, INT_MAX); case PIPE_CAP_VERTEX_BUFFER_OFFSET_4BYTE_ALIGNED_ONLY: case PIPE_CAP_VERTEX_BUFFER_STRIDE_4BYTE_ALIGNED_ONLY: case PIPE_CAP_VERTEX_ELEMENT_SRC_OFFSET_4BYTE_ALIGNED_ONLY: - /* SI doesn't support unaligned loads. - * CIK needs DRM 2.50.0 on radeon. */ - return sscreen->b.chip_class == SI || - (sscreen->b.info.drm_major == 2 && - sscreen->b.info.drm_minor < 50); + return !sscreen->info.has_unaligned_shader_loads; case PIPE_CAP_SPARSE_BUFFER_PAGE_SIZE: - /* TODO: GFX9 hangs. */ - if (sscreen->b.chip_class >= GFX9) - return 0; - /* Disable on SI due to VM faults in CP DMA. Enable once these - * faults are mitigated in software. - */ - if (sscreen->b.chip_class >= CIK && - sscreen->b.info.drm_major == 3 && - sscreen->b.info.drm_minor >= 13) - return RADEON_SPARSE_PAGE_SIZE; + return sscreen->info.has_sparse_vm_mappings ? + RADEON_SPARSE_PAGE_SIZE : 0; + + case PIPE_CAP_PACKED_UNIFORMS: + if (sscreen->debug_flags & DBG(NIR)) + return 1; return 0; /* Unsupported features. */ @@ -263,21 +221,30 @@ static int si_get_param(struct pipe_screen *pscreen, enum pipe_cap param) case PIPE_CAP_TEXTURE_GATHER_OFFSETS: case PIPE_CAP_VERTEXID_NOBASE: case PIPE_CAP_PRIMITIVE_RESTART_FOR_PATCHES: - case PIPE_CAP_MAX_WINDOW_RECTANGLES: - case PIPE_CAP_TGSI_FS_FBFETCH: case PIPE_CAP_TGSI_MUL_ZERO_WINS: case PIPE_CAP_UMA: case PIPE_CAP_POLYGON_MODE_FILL_RECTANGLE: case PIPE_CAP_POST_DEPTH_COVERAGE: case PIPE_CAP_TILE_RASTER_ORDER: case PIPE_CAP_MAX_COMBINED_SHADER_OUTPUT_RESOURCES: + case PIPE_CAP_CONTEXT_PRIORITY_MASK: + case PIPE_CAP_CONSERVATIVE_RASTER_POST_SNAP_TRIANGLES: + case PIPE_CAP_CONSERVATIVE_RASTER_POST_SNAP_POINTS_LINES: + case PIPE_CAP_CONSERVATIVE_RASTER_PRE_SNAP_TRIANGLES: + case PIPE_CAP_CONSERVATIVE_RASTER_PRE_SNAP_POINTS_LINES: + case PIPE_CAP_CONSERVATIVE_RASTER_POST_DEPTH_COVERAGE: + case PIPE_CAP_MAX_CONSERVATIVE_RASTER_SUBPIXEL_PRECISION_BIAS: + case PIPE_CAP_PROGRAMMABLE_SAMPLE_LOCATIONS: return 0; - case PIPE_CAP_NATIVE_FENCE_FD: - return sscreen->b.info.has_sync_file; + case PIPE_CAP_FENCE_SIGNAL: + return sscreen->info.has_syncobj; - case PIPE_CAP_QUERY_BUFFER_OBJECT: - return si_have_tgsi_compute(sscreen); + case PIPE_CAP_CONSTBUF0_FLAGS: + return SI_RESOURCE_FLAG_32BIT; + + case PIPE_CAP_NATIVE_FENCE_FD: + return sscreen->info.has_fence_to_handle; case PIPE_CAP_DRAW_PARAMETERS: case PIPE_CAP_MULTI_DRAW_INDIRECT: @@ -287,8 +254,11 @@ static int si_get_param(struct pipe_screen *pscreen, enum pipe_cap param) case PIPE_CAP_MAX_SHADER_PATCH_VARYINGS: return 30; + case PIPE_CAP_MAX_VARYINGS: + return 32; + case PIPE_CAP_TEXTURE_BORDER_COLOR_QUIRK: - return sscreen->b.chip_class <= VI ? + return sscreen->info.chip_class <= VI ? PIPE_QUIRK_TEXTURE_BORDER_COLOR_SWIZZLE_R600 : 0; /* Stream output. */ @@ -301,6 +271,10 @@ static int si_get_param(struct pipe_screen *pscreen, enum pipe_cap param) return 1024; case PIPE_CAP_MAX_GEOMETRY_TOTAL_OUTPUT_COMPONENTS: return 4095; + case PIPE_CAP_MAX_GS_INVOCATIONS: + /* The closed driver exposes 127, but 125 is the greatest + * number that works. */ + return 125; case PIPE_CAP_MAX_VERTEX_ATTRIB_STRIDE: return 2048; @@ -320,8 +294,11 @@ static int si_get_param(struct pipe_screen *pscreen, enum pipe_cap param) case PIPE_CAP_MAX_VIEWPORTS: return SI_MAX_VIEWPORTS; case PIPE_CAP_VIEWPORT_SUBPIXEL_BITS: + case PIPE_CAP_RASTERIZER_SUBPIXEL_BITS: case PIPE_CAP_MAX_RENDER_TARGETS: return 8; + case PIPE_CAP_FRAMEBUFFER_MSAA_CONSTRAINTS: + return sscreen->info.has_eqaa_surface_allocator ? 2 : 0; case PIPE_CAP_MIN_TEXTURE_GATHER_OFFSET: case PIPE_CAP_MIN_TEXEL_OFFSET: @@ -337,19 +314,21 @@ static int si_get_param(struct pipe_screen *pscreen, enum pipe_cap param) case PIPE_CAP_VENDOR_ID: return ATI_VENDOR_ID; case PIPE_CAP_DEVICE_ID: - return sscreen->b.info.pci_id; + return sscreen->info.pci_id; case PIPE_CAP_VIDEO_MEMORY: - return sscreen->b.info.vram_size >> 20; + return sscreen->info.vram_size >> 20; case PIPE_CAP_PCI_GROUP: - return sscreen->b.info.pci_domain; + return sscreen->info.pci_domain; case PIPE_CAP_PCI_BUS: - return sscreen->b.info.pci_bus; + return sscreen->info.pci_bus; case PIPE_CAP_PCI_DEVICE: - return sscreen->b.info.pci_dev; + return sscreen->info.pci_dev; case PIPE_CAP_PCI_FUNCTION: - return sscreen->b.info.pci_func; + return sscreen->info.pci_func; + + default: + return u_pipe_screen_get_param_defaults(pscreen, param); } - return 0; } static float si_get_paramf(struct pipe_screen* pscreen, enum pipe_capf param) @@ -357,17 +336,19 @@ static float si_get_paramf(struct pipe_screen* pscreen, enum pipe_capf param) switch (param) { case PIPE_CAPF_MAX_LINE_WIDTH: case PIPE_CAPF_MAX_LINE_WIDTH_AA: + /* This depends on the quant mode, though the precise interactions + * are unknown. */ + return 2048; case PIPE_CAPF_MAX_POINT_WIDTH: case PIPE_CAPF_MAX_POINT_WIDTH_AA: - return 8192.0f; + return SI_MAX_POINT_SIZE; case PIPE_CAPF_MAX_TEXTURE_ANISOTROPY: return 16.0f; case PIPE_CAPF_MAX_TEXTURE_LOD_BIAS: return 16.0f; - case PIPE_CAPF_GUARD_BAND_LEFT: - case PIPE_CAPF_GUARD_BAND_TOP: - case PIPE_CAPF_GUARD_BAND_RIGHT: - case PIPE_CAPF_GUARD_BAND_BOTTOM: + case PIPE_CAPF_MIN_CONSERVATIVE_RASTER_DILATE: + case PIPE_CAPF_MAX_CONSERVATIVE_RASTER_DILATE: + case PIPE_CAPF_CONSERVATIVE_RASTER_DILATE_GRANULARITY: return 0.0f; } return 0.0f; @@ -389,13 +370,10 @@ static int si_get_shader_param(struct pipe_screen* pscreen, break; case PIPE_SHADER_COMPUTE: switch (param) { - case PIPE_SHADER_CAP_PREFERRED_IR: - return PIPE_SHADER_IR_NATIVE; - case PIPE_SHADER_CAP_SUPPORTED_IRS: { int ir = 1 << PIPE_SHADER_IR_NATIVE; - if (si_have_tgsi_compute(sscreen)) + if (sscreen->info.has_indirect_compute_dispatch) ir |= 1 << PIPE_SHADER_IR_TGSI; return ir; @@ -434,7 +412,7 @@ static int si_get_shader_param(struct pipe_screen* pscreen, case PIPE_SHADER_CAP_MAX_TEMPS: return 256; /* Max native temporaries. */ case PIPE_SHADER_CAP_MAX_CONST_BUFFER_SIZE: - return 4096 * sizeof(float[4]); /* actually only memory limits this */ + return MIN2(sscreen->info.max_alloc_size, INT_MAX - 3); /* aligned to 4 */ case PIPE_SHADER_CAP_MAX_CONST_BUFFERS: return SI_NUM_CONST_BUFFERS; case PIPE_SHADER_CAP_MAX_TEXTURE_SAMPLERS: @@ -445,11 +423,11 @@ static int si_get_shader_param(struct pipe_screen* pscreen, case PIPE_SHADER_CAP_MAX_SHADER_IMAGES: return SI_NUM_IMAGES; case PIPE_SHADER_CAP_MAX_UNROLL_ITERATIONS_HINT: + if (sscreen->debug_flags & DBG(NIR)) + return 0; return 32; case PIPE_SHADER_CAP_PREFERRED_IR: - if (sscreen->b.debug_flags & DBG(NIR) && - (shader == PIPE_SHADER_VERTEX || - shader == PIPE_SHADER_FRAGMENT)) + if (sscreen->debug_flags & DBG(NIR)) return PIPE_SHADER_IR_NIR; return PIPE_SHADER_IR_TGSI; case PIPE_SHADER_CAP_LOWER_IF_THRESHOLD: @@ -473,12 +451,19 @@ static int si_get_shader_param(struct pipe_screen* pscreen, case PIPE_SHADER_CAP_INDIRECT_INPUT_ADDR: /* TODO: Indirect indexing of GS inputs is unimplemented. */ - return shader != PIPE_SHADER_GEOMETRY && - (sscreen->llvm_has_working_vgpr_indexing || - /* TCS and TES load inputs directly from LDS or - * offchip memory, so indirect indexing is trivial. */ - shader == PIPE_SHADER_TESS_CTRL || - shader == PIPE_SHADER_TESS_EVAL); + if (shader == PIPE_SHADER_GEOMETRY) + return 0; + + if (shader == PIPE_SHADER_VERTEX && + !sscreen->llvm_has_working_vgpr_indexing) + return 0; + + /* TCS and TES load inputs directly from LDS or offchip + * memory, so indirect indexing is always supported. + * PS has to support indirect indexing, because we can't + * lower that to TEMPs for INTERP instructions. + */ + return 1; case PIPE_SHADER_CAP_INDIRECT_OUTPUT_ADDR: return sscreen->llvm_has_working_vgpr_indexing || @@ -491,14 +476,16 @@ static int si_get_shader_param(struct pipe_screen* pscreen, case PIPE_SHADER_CAP_MAX_HW_ATOMIC_COUNTERS: case PIPE_SHADER_CAP_MAX_HW_ATOMIC_COUNTER_BUFFERS: return 0; + case PIPE_SHADER_CAP_SCALAR_ISA: + return 1; } return 0; } static const struct nir_shader_compiler_options nir_options = { - .vertex_id_zero_based = true, .lower_scmp = true, .lower_flrp32 = true, + .lower_flrp64 = true, .lower_fsat = true, .lower_fdiv = true, .lower_sub = true, @@ -535,14 +522,14 @@ static void si_get_device_uuid(struct pipe_screen *pscreen, char *uuid) { struct si_screen *sscreen = (struct si_screen *)pscreen; - ac_compute_device_uuid(&sscreen->b.info, uuid, PIPE_UUID_SIZE); + ac_compute_device_uuid(&sscreen->info, uuid, PIPE_UUID_SIZE); } static const char* si_get_name(struct pipe_screen *pscreen) { struct si_screen *sscreen = (struct si_screen*)pscreen; - return sscreen->b.renderer_string; + return sscreen->renderer_string; } static int si_get_video_param_no_decode(struct pipe_screen *screen, @@ -584,15 +571,20 @@ static int si_get_video_param(struct pipe_screen *screen, if (entrypoint == PIPE_VIDEO_ENTRYPOINT_ENCODE) { switch (param) { case PIPE_VIDEO_CAP_SUPPORTED: - return codec == PIPE_VIDEO_FORMAT_MPEG4_AVC && - (si_vce_is_fw_version_supported(&sscreen->b) || - sscreen->b.family == CHIP_RAVEN); + return (codec == PIPE_VIDEO_FORMAT_MPEG4_AVC && + (si_vce_is_fw_version_supported(sscreen) || + sscreen->info.family == CHIP_RAVEN || + sscreen->info.family == CHIP_RAVEN2)) || + (profile == PIPE_VIDEO_PROFILE_HEVC_MAIN && + (sscreen->info.family == CHIP_RAVEN || + sscreen->info.family == CHIP_RAVEN2 || + si_radeon_uvd_enc_supported(sscreen))); case PIPE_VIDEO_CAP_NPOT_TEXTURES: return 1; case PIPE_VIDEO_CAP_MAX_WIDTH: - return (sscreen->b.family < CHIP_TONGA) ? 2048 : 4096; + return (sscreen->info.family < CHIP_TONGA) ? 2048 : 4096; case PIPE_VIDEO_CAP_MAX_HEIGHT: - return (sscreen->b.family < CHIP_TONGA) ? 1152 : 2304; + return (sscreen->info.family < CHIP_TONGA) ? 1152 : 2304; case PIPE_VIDEO_CAP_PREFERED_FORMAT: return PIPE_FORMAT_NV12; case PIPE_VIDEO_CAP_PREFERS_INTERLACED: @@ -602,7 +594,7 @@ static int si_get_video_param(struct pipe_screen *screen, case PIPE_VIDEO_CAP_SUPPORTS_PROGRESSIVE: return true; case PIPE_VIDEO_CAP_STACKED_FRAMES: - return (sscreen->b.family < CHIP_TONGA) ? 1 : 2; + return (sscreen->info.family < CHIP_TONGA) ? 1 : 2; default: return 0; } @@ -616,9 +608,9 @@ static int si_get_video_param(struct pipe_screen *screen, case PIPE_VIDEO_FORMAT_MPEG4: return 1; case PIPE_VIDEO_FORMAT_MPEG4_AVC: - if ((sscreen->b.family == CHIP_POLARIS10 || - sscreen->b.family == CHIP_POLARIS11) && - sscreen->b.info.uvd_fw_version < UVD_FW_1_66_16 ) { + if ((sscreen->info.family == CHIP_POLARIS10 || + sscreen->info.family == CHIP_POLARIS11) && + sscreen->info.uvd_fw_version < UVD_FW_1_66_16 ) { RVID_ERR("POLARIS10/11 firmware version need to be updated.\n"); return false; } @@ -627,31 +619,39 @@ static int si_get_video_param(struct pipe_screen *screen, return true; case PIPE_VIDEO_FORMAT_HEVC: /* Carrizo only supports HEVC Main */ - if (sscreen->b.family >= CHIP_STONEY) + if (sscreen->info.family >= CHIP_STONEY) return (profile == PIPE_VIDEO_PROFILE_HEVC_MAIN || profile == PIPE_VIDEO_PROFILE_HEVC_MAIN_10); - else if (sscreen->b.family >= CHIP_CARRIZO) + else if (sscreen->info.family >= CHIP_CARRIZO) return profile == PIPE_VIDEO_PROFILE_HEVC_MAIN; return false; case PIPE_VIDEO_FORMAT_JPEG: - if (sscreen->b.family < CHIP_CARRIZO || sscreen->b.family >= CHIP_VEGA10) + if (sscreen->info.family == CHIP_RAVEN || + sscreen->info.family == CHIP_RAVEN2) + return true; + if (sscreen->info.family < CHIP_CARRIZO || sscreen->info.family >= CHIP_VEGA10) return false; - if (!(sscreen->b.info.drm_major == 3 && sscreen->b.info.drm_minor >= 19)) { + if (!(sscreen->info.drm_major == 3 && sscreen->info.drm_minor >= 19)) { RVID_ERR("No MJPEG support for the kernel version\n"); return false; } return true; + case PIPE_VIDEO_FORMAT_VP9: + if (sscreen->info.family < CHIP_RAVEN) + return false; + return true; default: return false; } case PIPE_VIDEO_CAP_NPOT_TEXTURES: return 1; case PIPE_VIDEO_CAP_MAX_WIDTH: - return (sscreen->b.family < CHIP_TONGA) ? 2048 : 4096; + return (sscreen->info.family < CHIP_TONGA) ? 2048 : 4096; case PIPE_VIDEO_CAP_MAX_HEIGHT: - return (sscreen->b.family < CHIP_TONGA) ? 1152 : 4096; + return (sscreen->info.family < CHIP_TONGA) ? 1152 : 4096; case PIPE_VIDEO_CAP_PREFERED_FORMAT: - if (profile == PIPE_VIDEO_PROFILE_HEVC_MAIN_10) + if (profile == PIPE_VIDEO_PROFILE_HEVC_MAIN_10 || + profile == PIPE_VIDEO_PROFILE_VP9_PROFILE2) return PIPE_FORMAT_P016; else return PIPE_FORMAT_NV12; @@ -664,6 +664,8 @@ static int si_get_video_param(struct pipe_screen *screen, return false; //The firmware doesn't support interlaced HEVC. else if (format == PIPE_VIDEO_FORMAT_JPEG) return false; + else if (format == PIPE_VIDEO_FORMAT_VP9) + return false; return true; } case PIPE_VIDEO_CAP_SUPPORTS_PROGRESSIVE: @@ -688,7 +690,7 @@ static int si_get_video_param(struct pipe_screen *screen, case PIPE_VIDEO_PROFILE_MPEG4_AVC_BASELINE: case PIPE_VIDEO_PROFILE_MPEG4_AVC_MAIN: case PIPE_VIDEO_PROFILE_MPEG4_AVC_HIGH: - return (sscreen->b.family < CHIP_TONGA) ? 41 : 52; + return (sscreen->info.family < CHIP_TONGA) ? 41 : 52; case PIPE_VIDEO_PROFILE_HEVC_MAIN: case PIPE_VIDEO_PROFILE_HEVC_MAIN_10: return 186; @@ -720,11 +722,11 @@ static boolean si_vid_is_format_supported(struct pipe_screen *screen, static unsigned get_max_threads_per_block(struct si_screen *screen, enum pipe_shader_ir ir_type) { - if (ir_type != PIPE_SHADER_IR_TGSI) + if (ir_type == PIPE_SHADER_IR_NATIVE) return 256; /* Only 16 waves per thread-group on gfx9. */ - if (screen->b.chip_class >= GFX9) + if (screen->info.chip_class >= GFX9) return 1024; /* Up to 40 waves per thread-group on GCN < gfx9. Expose a nice @@ -743,15 +745,10 @@ static int si_get_compute_param(struct pipe_screen *screen, //TODO: select these params by asic switch (param) { case PIPE_COMPUTE_CAP_IR_TARGET: { - const char *gpu; - const char *triple; - - if (HAVE_LLVM < 0x0400) - triple = "amdgcn--"; - else - triple = "amdgcn-mesa-mesa3d"; + const char *gpu, *triple; - gpu = ac_get_llvm_processor_name(sscreen->b.family); + triple = "amdgcn-mesa-mesa3d"; + gpu = ac_get_llvm_processor_name(sscreen->info.family); if (ret) { sprintf(ret, "%s-%s", gpu, triple); } @@ -813,8 +810,8 @@ static int si_get_compute_param(struct pipe_screen *screen, * 4 * MAX_MEM_ALLOC_SIZE. */ *max_global_size = MIN2(4 * max_mem_alloc_size, - MAX2(sscreen->b.info.gart_size, - sscreen->b.info.vram_size)); + MAX2(sscreen->info.gart_size, + sscreen->info.vram_size)); } return sizeof(uint64_t); @@ -838,21 +835,21 @@ static int si_get_compute_param(struct pipe_screen *screen, if (ret) { uint64_t *max_mem_alloc_size = ret; - *max_mem_alloc_size = sscreen->b.info.max_alloc_size; + *max_mem_alloc_size = sscreen->info.max_alloc_size; } return sizeof(uint64_t); case PIPE_COMPUTE_CAP_MAX_CLOCK_FREQUENCY: if (ret) { uint32_t *max_clock_frequency = ret; - *max_clock_frequency = sscreen->b.info.max_shader_clock; + *max_clock_frequency = sscreen->info.max_shader_clock; } return sizeof(uint32_t); case PIPE_COMPUTE_CAP_MAX_COMPUTE_UNITS: if (ret) { uint32_t *max_compute_units = ret; - *max_compute_units = sscreen->b.info.num_good_compute_units; + *max_compute_units = sscreen->info.num_good_compute_units; } return sizeof(uint32_t); @@ -873,10 +870,10 @@ static int si_get_compute_param(struct pipe_screen *screen, case PIPE_COMPUTE_CAP_MAX_VARIABLE_THREADS_PER_BLOCK: if (ret) { uint64_t *max_variable_threads_per_block = ret; - if (ir_type == PIPE_SHADER_IR_TGSI) - *max_variable_threads_per_block = SI_MAX_VARIABLE_THREADS_PER_BLOCK; - else + if (ir_type == PIPE_SHADER_IR_NATIVE) *max_variable_threads_per_block = 0; + else + *max_variable_threads_per_block = SI_MAX_VARIABLE_THREADS_PER_BLOCK; } return sizeof(uint64_t); } @@ -889,19 +886,19 @@ static uint64_t si_get_timestamp(struct pipe_screen *screen) { struct si_screen *sscreen = (struct si_screen*)screen; - return 1000000 * sscreen->b.ws->query_value(sscreen->b.ws, RADEON_TIMESTAMP) / - sscreen->b.info.clock_crystal_freq; + return 1000000 * sscreen->ws->query_value(sscreen->ws, RADEON_TIMESTAMP) / + sscreen->info.clock_crystal_freq; } static void si_query_memory_info(struct pipe_screen *screen, struct pipe_memory_info *info) { struct si_screen *sscreen = (struct si_screen*)screen; - struct radeon_winsys *ws = sscreen->b.ws; + struct radeon_winsys *ws = sscreen->ws; unsigned vram_usage, gtt_usage; - info->total_device_memory = sscreen->b.info.vram_size / 1024; - info->total_staging_memory = sscreen->b.info.gart_size / 1024; + info->total_device_memory = sscreen->info.vram_size / 1024; + info->total_staging_memory = sscreen->info.gart_size / 1024; /* The real TTM memory usage is somewhat random, because: * @@ -913,8 +910,8 @@ static void si_query_memory_info(struct pipe_screen *screen, * * Instead, return statistics of this process. */ - vram_usage = ws->query_value(ws, RADEON_REQUESTED_VRAM_MEMORY) / 1024; - gtt_usage = ws->query_value(ws, RADEON_REQUESTED_GTT_MEMORY) / 1024; + vram_usage = ws->query_value(ws, RADEON_VRAM_USAGE) / 1024; + gtt_usage = ws->query_value(ws, RADEON_GTT_USAGE) / 1024; info->avail_device_memory = vram_usage <= info->total_device_memory ? @@ -926,7 +923,7 @@ static void si_query_memory_info(struct pipe_screen *screen, info->device_memory_evicted = ws->query_value(ws, RADEON_NUM_BYTES_MOVED) / 1024; - if (sscreen->b.info.drm_major == 3 && sscreen->b.info.drm_minor >= 4) + if (sscreen->info.drm_major == 3 && sscreen->info.drm_minor >= 4) info->nr_device_memory_evictions = ws->query_value(ws, RADEON_NUM_EVICTIONS); else @@ -938,62 +935,62 @@ static struct disk_cache *si_get_disk_shader_cache(struct pipe_screen *pscreen) { struct si_screen *sscreen = (struct si_screen*)pscreen; - return sscreen->b.disk_shader_cache; + return sscreen->disk_shader_cache; } static void si_init_renderer_string(struct si_screen *sscreen) { - struct radeon_winsys *ws = sscreen->b.ws; - char family_name[32] = {}, llvm_string[32] = {}, kernel_version[128] = {}; + struct radeon_winsys *ws = sscreen->ws; + char first_name[256], second_name[32] = {}, kernel_version[128] = {}; struct utsname uname_data; - const char *chip_name = si_get_marketing_name(ws); + const char *marketing_name = si_get_marketing_name(ws); - if (chip_name) - snprintf(family_name, sizeof(family_name), "%s / ", - si_get_family_name(sscreen) + 4); - else - chip_name = si_get_family_name(sscreen); + if (marketing_name) { + snprintf(first_name, sizeof(first_name), "%s", marketing_name); + snprintf(second_name, sizeof(second_name), "%s, ", + sscreen->info.name); + } else { + snprintf(first_name, sizeof(first_name), "AMD %s", + sscreen->info.name); + } if (uname(&uname_data) == 0) snprintf(kernel_version, sizeof(kernel_version), - " / %s", uname_data.release); - - if (HAVE_LLVM > 0) { - snprintf(llvm_string, sizeof(llvm_string), - ", LLVM %i.%i.%i", (HAVE_LLVM >> 8) & 0xff, - HAVE_LLVM & 0xff, MESA_LLVM_VERSION_PATCH); - } - - snprintf(sscreen->b.renderer_string, sizeof(sscreen->b.renderer_string), - "%s (%sDRM %i.%i.%i%s%s)", - chip_name, family_name, sscreen->b.info.drm_major, - sscreen->b.info.drm_minor, sscreen->b.info.drm_patchlevel, - kernel_version, llvm_string); + ", %s", uname_data.release); + + snprintf(sscreen->renderer_string, sizeof(sscreen->renderer_string), + "%s (%sDRM %i.%i.%i%s, LLVM %i.%i.%i)", + first_name, second_name, sscreen->info.drm_major, + sscreen->info.drm_minor, sscreen->info.drm_patchlevel, + kernel_version, + (HAVE_LLVM >> 8) & 0xff, + HAVE_LLVM & 0xff, + MESA_LLVM_VERSION_PATCH); } void si_init_screen_get_functions(struct si_screen *sscreen) { - sscreen->b.b.get_name = si_get_name; - sscreen->b.b.get_vendor = si_get_vendor; - sscreen->b.b.get_device_vendor = si_get_device_vendor; - sscreen->b.b.get_param = si_get_param; - sscreen->b.b.get_paramf = si_get_paramf; - sscreen->b.b.get_compute_param = si_get_compute_param; - sscreen->b.b.get_timestamp = si_get_timestamp; - sscreen->b.b.get_shader_param = si_get_shader_param; - sscreen->b.b.get_compiler_options = si_get_compiler_options; - sscreen->b.b.get_device_uuid = si_get_device_uuid; - sscreen->b.b.get_driver_uuid = si_get_driver_uuid; - sscreen->b.b.query_memory_info = si_query_memory_info; - sscreen->b.b.get_disk_shader_cache = si_get_disk_shader_cache; - - if (sscreen->b.info.has_hw_decode) { - sscreen->b.b.get_video_param = si_get_video_param; - sscreen->b.b.is_video_format_supported = si_vid_is_format_supported; + sscreen->b.get_name = si_get_name; + sscreen->b.get_vendor = si_get_vendor; + sscreen->b.get_device_vendor = si_get_device_vendor; + sscreen->b.get_param = si_get_param; + sscreen->b.get_paramf = si_get_paramf; + sscreen->b.get_compute_param = si_get_compute_param; + sscreen->b.get_timestamp = si_get_timestamp; + sscreen->b.get_shader_param = si_get_shader_param; + sscreen->b.get_compiler_options = si_get_compiler_options; + sscreen->b.get_device_uuid = si_get_device_uuid; + sscreen->b.get_driver_uuid = si_get_driver_uuid; + sscreen->b.query_memory_info = si_query_memory_info; + sscreen->b.get_disk_shader_cache = si_get_disk_shader_cache; + + if (sscreen->info.has_hw_decode) { + sscreen->b.get_video_param = si_get_video_param; + sscreen->b.is_video_format_supported = si_vid_is_format_supported; } else { - sscreen->b.b.get_video_param = si_get_video_param_no_decode; - sscreen->b.b.is_video_format_supported = vl_video_buffer_is_format_supported; + sscreen->b.get_video_param = si_get_video_param_no_decode; + sscreen->b.is_video_format_supported = vl_video_buffer_is_format_supported; } si_init_renderer_string(sscreen);