X-Git-Url: https://git.libre-soc.org/?a=blobdiff_plain;f=src%2Fgallium%2Fdrivers%2Fetnaviv%2Fetnaviv_screen.c;h=8c80c1e95565284b5b1f22c94c7df9bd13add169;hb=a5e7c12cedb8a91236dd3caf99133f86349702a9;hp=6c0735e0fbf2ff4c67b7c2c0bb74b9c468f18222;hpb=5065549e2a75e2a56cd3bc8b0fbb6c9013e86cb4;p=mesa.git diff --git a/src/gallium/drivers/etnaviv/etnaviv_screen.c b/src/gallium/drivers/etnaviv/etnaviv_screen.c index 6c0735e0fbf..8c80c1e9556 100644 --- a/src/gallium/drivers/etnaviv/etnaviv_screen.c +++ b/src/gallium/drivers/etnaviv/etnaviv_screen.c @@ -38,15 +38,20 @@ #include "etnaviv_resource.h" #include "etnaviv_translate.h" -#include "os/os_time.h" +#include "util/hash_table.h" +#include "util/os_time.h" #include "util/u_math.h" #include "util/u_memory.h" +#include "util/u_screen.h" #include "util/u_string.h" #include "state_tracker/drm_driver.h" +#include "drm-uapi/drm_fourcc.h" + #define ETNA_DRM_VERSION(major, minor) ((major) << 16 | (minor)) #define ETNA_DRM_VERSION_FENCE_FD ETNA_DRM_VERSION(1, 1) +#define ETNA_DRM_VERSION_PERFMON ETNA_DRM_VERSION(1, 2) static const struct debug_named_value debug_options[] = { {"dbg_msgs", ETNA_DBG_MSGS, "Print debug messages"}, @@ -59,13 +64,15 @@ static const struct debug_named_value debug_options[] = { {"no_autodisable", ETNA_DBG_NO_AUTODISABLE, "Disable autodisable"}, {"no_supertile", ETNA_DBG_NO_SUPERTILE, "Disable supertiles"}, {"no_early_z", ETNA_DBG_NO_EARLY_Z, "Disable early z"}, - {"cflush_all", ETNA_DBG_CFLUSH_ALL, "Flush every cash before state update"}, + {"cflush_all", ETNA_DBG_CFLUSH_ALL, "Flush every cache before state update"}, {"msaa2x", ETNA_DBG_MSAA_2X, "Force 2x msaa"}, {"msaa4x", ETNA_DBG_MSAA_4X, "Force 4x msaa"}, {"flush_all", ETNA_DBG_FLUSH_ALL, "Flush after every rendered primitive"}, {"zero", ETNA_DBG_ZERO, "Zero all resources after allocation"}, {"draw_stall", ETNA_DBG_DRAW_STALL, "Stall FE/PE after each rendered primitive"}, {"shaderdb", ETNA_DBG_SHADERDB, "Enable shaderdb output"}, + {"no_singlebuffer",ETNA_DBG_NO_SINGLEBUF, "Disable single buffer feature"}, + {"nir", ETNA_DBG_NIR, "use new NIR compiler"}, DEBUG_NAMED_VALUE_END }; @@ -77,6 +84,12 @@ etna_screen_destroy(struct pipe_screen *pscreen) { struct etna_screen *screen = etna_screen(pscreen); + _mesa_set_destroy(screen->used_resources, NULL); + mtx_destroy(&screen->lock); + + if (screen->perfmon) + etna_perfmon_del(screen->perfmon); + if (screen->pipe) etna_pipe_del(screen->pipe); @@ -98,8 +111,8 @@ etna_screen_get_name(struct pipe_screen *pscreen) struct etna_screen *priv = etna_screen(pscreen); static char buffer[128]; - util_snprintf(buffer, sizeof(buffer), "Vivante GC%x rev %04x", priv->model, - priv->revision); + snprintf(buffer, sizeof(buffer), "Vivante GC%x rev %04x", priv->model, + priv->revision); return buffer; } @@ -123,223 +136,109 @@ etna_screen_get_param(struct pipe_screen *pscreen, enum pipe_cap param) switch (param) { /* Supported features (boolean caps). */ - case PIPE_CAP_TWO_SIDED_STENCIL: case PIPE_CAP_ANISOTROPIC_FILTER: case PIPE_CAP_POINT_SPRITE: - case PIPE_CAP_TEXTURE_SHADOW_MAP: case PIPE_CAP_BLEND_EQUATION_SEPARATE: case PIPE_CAP_TGSI_FS_COORD_ORIGIN_UPPER_LEFT: case PIPE_CAP_TGSI_FS_COORD_PIXEL_CENTER_HALF_INTEGER: - case PIPE_CAP_SM3: + case PIPE_CAP_FRAGMENT_SHADER_TEXTURE_LOD: + case PIPE_CAP_FRAGMENT_SHADER_DERIVATIVES: + case PIPE_CAP_VERTEX_SHADER_SATURATE: case PIPE_CAP_TEXTURE_BARRIER: case PIPE_CAP_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION: 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: - case PIPE_CAP_USER_CONSTANT_BUFFERS: case PIPE_CAP_TGSI_TEXCOORD: case PIPE_CAP_VERTEX_COLOR_UNCLAMPED: + case PIPE_CAP_MIXED_COLOR_DEPTH_BITS: return 1; case PIPE_CAP_NATIVE_FENCE_FD: return screen->drm_version >= ETNA_DRM_VERSION_FENCE_FD; + case PIPE_CAP_TGSI_FS_POSITION_IS_SYSVAL: + case PIPE_CAP_TGSI_FS_FACE_IS_INTEGER_SYSVAL: /* note: not integer */ + return DBG_ENABLED(ETNA_DBG_NIR); + case PIPE_CAP_TGSI_FS_POINT_IS_SYSVAL: + return 0; /* Memory */ case PIPE_CAP_CONSTANT_BUFFER_OFFSET_ALIGNMENT: return 256; case PIPE_CAP_MIN_MAP_BUFFER_ALIGNMENT: return 4; /* XXX could easily be supported */ - case PIPE_CAP_GLSL_FEATURE_LEVEL: - return 120; case PIPE_CAP_NPOT_TEXTURES: return true; /* VIV_FEATURE(priv->dev, chipMinorFeatures1, NON_POWER_OF_TWO); */ + case PIPE_CAP_TEXTURE_SWIZZLE: case PIPE_CAP_PRIMITIVE_RESTART: return VIV_FEATURE(screen, chipMinorFeatures1, HALTI0); - case PIPE_CAP_ENDIANNESS: - return PIPE_ENDIAN_LITTLE; /* on most Viv hw this is configurable (feature - ENDIANNESS_CONFIG) */ - /* Unsupported features. */ - case PIPE_CAP_SEAMLESS_CUBE_MAP: - case PIPE_CAP_TEXTURE_SWIZZLE: /* XXX supported on gc2000 */ - case PIPE_CAP_COMPUTE: /* XXX supported on gc2000 */ - case PIPE_CAP_MIXED_COLORBUFFER_FORMATS: /* only one colorbuffer supported, so mixing makes no sense */ - case PIPE_CAP_CONDITIONAL_RENDER: /* no occlusion queries */ - case PIPE_CAP_TGSI_INSTANCEID: /* no idea, really */ - case PIPE_CAP_START_INSTANCE: /* instancing not supported AFAIK */ - case PIPE_CAP_VERTEX_ELEMENT_INSTANCE_DIVISOR: /* instancing not supported AFAIK */ - case PIPE_CAP_SHADER_STENCIL_EXPORT: /* Fragment shader cannot export stencil value */ - case PIPE_CAP_MAX_DUAL_SOURCE_RENDER_TARGETS: /* no dual-source supported */ - case PIPE_CAP_TEXTURE_MULTISAMPLE: /* no texture multisample */ - case PIPE_CAP_TEXTURE_MIRROR_CLAMP: /* only mirrored repeat */ - case PIPE_CAP_INDEP_BLEND_ENABLE: - case PIPE_CAP_INDEP_BLEND_FUNC: - case PIPE_CAP_DEPTH_CLIP_DISABLE: - case PIPE_CAP_SEAMLESS_CUBE_MAP_PER_TEXTURE: - case PIPE_CAP_TGSI_FS_COORD_ORIGIN_LOWER_LEFT: - case PIPE_CAP_TGSI_FS_COORD_PIXEL_CENTER_INTEGER: - case PIPE_CAP_TGSI_CAN_COMPACT_CONSTANTS: /* Don't skip strict max uniform limit check */ - case PIPE_CAP_FRAGMENT_COLOR_CLAMPED: - case PIPE_CAP_VERTEX_COLOR_CLAMPED: - case PIPE_CAP_USER_VERTEX_BUFFERS: - case PIPE_CAP_TEXTURE_BUFFER_OBJECTS: case PIPE_CAP_TEXTURE_BUFFER_OFFSET_ALIGNMENT: - case PIPE_CAP_BUFFER_SAMPLER_VIEW_RGBA_ONLY: - case PIPE_CAP_MIXED_FRAMEBUFFER_SIZES: /* TODO: test me out with piglit */ - case PIPE_CAP_TGSI_VS_LAYER_VIEWPORT: - case PIPE_CAP_MAX_TEXTURE_GATHER_COMPONENTS: - case PIPE_CAP_TEXTURE_GATHER_SM5: - case PIPE_CAP_BUFFER_MAP_PERSISTENT_COHERENT: - case PIPE_CAP_FAKE_SW_MSAA: - case PIPE_CAP_TEXTURE_QUERY_LOD: - case PIPE_CAP_SAMPLE_SHADING: - case PIPE_CAP_TEXTURE_GATHER_OFFSETS: - case PIPE_CAP_TGSI_VS_WINDOW_SPACE_POSITION: - case PIPE_CAP_DRAW_INDIRECT: - case PIPE_CAP_TGSI_FS_FINE_DERIVATIVE: - case PIPE_CAP_CONDITIONAL_RENDER_INVERTED: - case PIPE_CAP_SAMPLER_VIEW_TARGET: - case PIPE_CAP_CLIP_HALFZ: - case PIPE_CAP_VERTEXID_NOBASE: - case PIPE_CAP_POLYGON_OFFSET_CLAMP: - case PIPE_CAP_MULTISAMPLE_Z_RESOLVE: - case PIPE_CAP_RESOURCE_FROM_USER_MEMORY: - case PIPE_CAP_DEVICE_RESET_STATUS_QUERY: - case PIPE_CAP_MAX_SHADER_PATCH_VARYINGS: - case PIPE_CAP_TEXTURE_FLOAT_LINEAR: - case PIPE_CAP_TEXTURE_HALF_FLOAT_LINEAR: - case PIPE_CAP_DEPTH_BOUNDS_TEST: - case PIPE_CAP_TGSI_TXQS: - case PIPE_CAP_FORCE_PERSAMPLE_INTERP: - case PIPE_CAP_SHAREABLE_SHADERS: - case PIPE_CAP_COPY_BETWEEN_COMPRESSED_AND_PLAIN_FORMATS: - case PIPE_CAP_CLEAR_TEXTURE: - case PIPE_CAP_DRAW_PARAMETERS: - case PIPE_CAP_TGSI_PACK_HALF_FLOAT: - case PIPE_CAP_MULTI_DRAW_INDIRECT: - case PIPE_CAP_MULTI_DRAW_INDIRECT_PARAMS: - case PIPE_CAP_TGSI_FS_POSITION_IS_SYSVAL: - case PIPE_CAP_TGSI_FS_FACE_IS_INTEGER_SYSVAL: - case PIPE_CAP_SHADER_BUFFER_OFFSET_ALIGNMENT: - case PIPE_CAP_INVALIDATE_BUFFER: - case PIPE_CAP_GENERATE_MIPMAP: - case PIPE_CAP_STRING_MARKER: - case PIPE_CAP_SURFACE_REINTERPRET_BLOCKS: - case PIPE_CAP_QUERY_BUFFER_OBJECT: - case PIPE_CAP_QUERY_MEMORY_INFO: - case PIPE_CAP_FRAMEBUFFER_NO_ATTACHMENT: - case PIPE_CAP_ROBUST_BUFFER_ACCESS_BEHAVIOR: - case PIPE_CAP_CULL_DISTANCE: - case PIPE_CAP_PRIMITIVE_RESTART_FOR_PATCHES: - case PIPE_CAP_TGSI_VOTE: - case PIPE_CAP_MAX_WINDOW_RECTANGLES: - case PIPE_CAP_POLYGON_OFFSET_UNITS_UNSCALED: - case PIPE_CAP_VIEWPORT_SUBPIXEL_BITS: - case PIPE_CAP_MIXED_COLOR_DEPTH_BITS: - case PIPE_CAP_TGSI_ARRAY_COMPONENTS: - case PIPE_CAP_STREAM_OUTPUT_INTERLEAVE_BUFFERS: - case PIPE_CAP_TGSI_CAN_READ_OUTPUTS: case PIPE_CAP_GLSL_OPTIMIZE_CONSERVATIVELY: - case PIPE_CAP_TGSI_FS_FBFETCH: - case PIPE_CAP_TGSI_MUL_ZERO_WINS: - case PIPE_CAP_DOUBLES: - case PIPE_CAP_INT64: - case PIPE_CAP_INT64_DIVMOD: - case PIPE_CAP_TGSI_TEX_TXF_LZ: - case PIPE_CAP_TGSI_CLOCK: - case PIPE_CAP_POLYGON_MODE_FILL_RECTANGLE: - case PIPE_CAP_SPARSE_BUFFER_PAGE_SIZE: - case PIPE_CAP_TGSI_BALLOT: - case PIPE_CAP_TGSI_TES_LAYER_VIEWPORT: - case PIPE_CAP_CAN_BIND_CONST_BUFFER_AS_VERTEX: case PIPE_CAP_ALLOW_MAPPED_BUFFERS_DURING_EXECUTION: - case PIPE_CAP_POST_DEPTH_COVERAGE: - case PIPE_CAP_BINDLESS_TEXTURE: return 0; /* Stream output. */ - case PIPE_CAP_MAX_STREAM_OUTPUT_BUFFERS: - case PIPE_CAP_STREAM_OUTPUT_PAUSE_RESUME: case PIPE_CAP_MAX_STREAM_OUTPUT_SEPARATE_COMPONENTS: case PIPE_CAP_MAX_STREAM_OUTPUT_INTERLEAVED_COMPONENTS: return 0; - /* Geometry shader output, unsupported. */ - case PIPE_CAP_MAX_GEOMETRY_OUTPUT_VERTICES: - case PIPE_CAP_MAX_GEOMETRY_TOTAL_OUTPUT_COMPONENTS: - case PIPE_CAP_MAX_VERTEX_STREAMS: - return 0; - case PIPE_CAP_MAX_VERTEX_ATTRIB_STRIDE: return 128; + case PIPE_CAP_MAX_VERTEX_ELEMENT_SRC_OFFSET: + return 255; /* Texturing. */ - case PIPE_CAP_MAX_TEXTURE_2D_LEVELS: + case PIPE_CAP_MAX_TEXTURE_2D_SIZE: + case PIPE_CAP_MAX_TEXTURE_ARRAY_LAYERS: /* TODO: verify */ + return screen->specs.max_texture_size; case PIPE_CAP_MAX_TEXTURE_CUBE_LEVELS: + case PIPE_CAP_MAX_TEXTURE_3D_LEVELS: { int log2_max_tex_size = util_last_bit(screen->specs.max_texture_size); assert(log2_max_tex_size > 0); return log2_max_tex_size; } - case PIPE_CAP_MAX_TEXTURE_3D_LEVELS: /* 3D textures not supported - fake it */ - return 5; - case PIPE_CAP_MAX_TEXTURE_ARRAY_LAYERS: - return 0; - case PIPE_CAP_CUBE_MAP_ARRAY: - return 0; + case PIPE_CAP_MIN_TEXTURE_GATHER_OFFSET: case PIPE_CAP_MIN_TEXEL_OFFSET: return -8; case PIPE_CAP_MAX_TEXTURE_GATHER_OFFSET: case PIPE_CAP_MAX_TEXEL_OFFSET: return 7; - case PIPE_CAP_TEXTURE_BORDER_COLOR_QUIRK: - return 0; - case PIPE_CAP_MAX_TEXTURE_BUFFER_SIZE: - return 65536; - - /* Render targets. */ - case PIPE_CAP_MAX_RENDER_TARGETS: - return 1; - - /* Viewports and scissors. */ - case PIPE_CAP_MAX_VIEWPORTS: - return 1; + case PIPE_CAP_SEAMLESS_CUBE_MAP_PER_TEXTURE: + return VIV_FEATURE(screen, chipMinorFeatures2, SEAMLESS_CUBE_MAP); /* Timer queries. */ - case PIPE_CAP_QUERY_TIME_ELAPSED: case PIPE_CAP_OCCLUSION_QUERY: - return 0; + return VIV_FEATURE(screen, chipMinorFeatures1, HALTI0); case PIPE_CAP_QUERY_TIMESTAMP: return 1; - case PIPE_CAP_QUERY_PIPELINE_STATISTICS: - return 0; /* Preferences */ case PIPE_CAP_PREFER_BLIT_BASED_TEXTURE_TRANSFER: return 0; + case PIPE_CAP_MAX_VARYINGS: + return screen->specs.max_varyings; + case PIPE_CAP_PCI_GROUP: case PIPE_CAP_PCI_BUS: case PIPE_CAP_PCI_DEVICE: case PIPE_CAP_PCI_FUNCTION: return 0; - case PIPE_CAP_VENDOR_ID: - case PIPE_CAP_DEVICE_ID: - return 0xFFFFFFFF; case PIPE_CAP_ACCELERATED: return 1; case PIPE_CAP_VIDEO_MEMORY: return 0; case PIPE_CAP_UMA: return 1; + default: + return u_pipe_screen_get_param_defaults(pscreen, param); } - - debug_printf("unknown param %d", param); - return 0; } static float @@ -357,10 +256,9 @@ etna_screen_get_paramf(struct pipe_screen *pscreen, enum pipe_capf param) return 16.0f; case PIPE_CAPF_MAX_TEXTURE_LOD_BIAS: return util_last_bit(screen->specs.max_texture_size); - 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; } @@ -422,6 +320,8 @@ etna_screen_get_shader_param(struct pipe_screen *pscreen, case PIPE_SHADER_CAP_TGSI_SQRT_SUPPORTED: return VIV_FEATURE(screen, chipMinorFeatures0, HAS_SQRT_TRIG); case PIPE_SHADER_CAP_INTEGERS: + case PIPE_SHADER_CAP_INT64_ATOMICS: + case PIPE_SHADER_CAP_FP16: return 0; case PIPE_SHADER_CAP_MAX_TEXTURE_SAMPLERS: case PIPE_SHADER_CAP_MAX_SAMPLER_VIEWS: @@ -429,11 +329,12 @@ etna_screen_get_shader_param(struct pipe_screen *pscreen, ? screen->specs.fragment_sampler_count : screen->specs.vertex_sampler_count; case PIPE_SHADER_CAP_PREFERRED_IR: - return PIPE_SHADER_IR_TGSI; + return DBG_ENABLED(ETNA_DBG_NIR) ? PIPE_SHADER_IR_NIR : PIPE_SHADER_IR_TGSI; case PIPE_SHADER_CAP_MAX_CONST_BUFFER_SIZE: return 4096; case PIPE_SHADER_CAP_TGSI_DROUND_SUPPORTED: case PIPE_SHADER_CAP_TGSI_DFRACEXP_DLDEXP_SUPPORTED: + case PIPE_SHADER_CAP_TGSI_LDEXP_SUPPORTED: case PIPE_SHADER_CAP_TGSI_FMA_SUPPORTED: case PIPE_SHADER_CAP_TGSI_ANY_INOUT_DECL_RANGE: return false; @@ -445,6 +346,9 @@ etna_screen_get_shader_param(struct pipe_screen *pscreen, case PIPE_SHADER_CAP_MAX_SHADER_IMAGES: case PIPE_SHADER_CAP_LOWER_IF_THRESHOLD: case PIPE_SHADER_CAP_TGSI_SKIP_MERGE_REGISTERS: + case PIPE_SHADER_CAP_MAX_HW_ATOMIC_COUNTERS: + case PIPE_SHADER_CAP_MAX_HW_ATOMIC_COUNTER_BUFFERS: + case PIPE_SHADER_CAP_SCALAR_ISA: return 0; } @@ -459,33 +363,69 @@ etna_screen_get_timestamp(struct pipe_screen *pscreen) } static bool -gpu_supports_texure_format(struct etna_screen *screen, uint32_t fmt) +gpu_supports_texture_target(struct etna_screen *screen, + enum pipe_texture_target target) +{ + if (target == PIPE_TEXTURE_CUBE_ARRAY) + return false; + + /* pre-halti has no array/3D */ + if (screen->specs.halti < 0 && + (target == PIPE_TEXTURE_1D_ARRAY || + target == PIPE_TEXTURE_2D_ARRAY || + target == PIPE_TEXTURE_3D)) + return false; + + return true; +} + +static bool +gpu_supports_texture_format(struct etna_screen *screen, uint32_t fmt, + enum pipe_format format) { + bool supported = true; + if (fmt == TEXTURE_FORMAT_ETC1) - return VIV_FEATURE(screen, chipFeatures, ETC1_TEXTURE_COMPRESSION); + supported = VIV_FEATURE(screen, chipFeatures, ETC1_TEXTURE_COMPRESSION); if (fmt >= TEXTURE_FORMAT_DXT1 && fmt <= TEXTURE_FORMAT_DXT4_DXT5) - return VIV_FEATURE(screen, chipFeatures, DXT_TEXTURE_COMPRESSION); + supported = VIV_FEATURE(screen, chipFeatures, DXT_TEXTURE_COMPRESSION); + + if (util_format_is_srgb(format)) + supported = VIV_FEATURE(screen, chipMinorFeatures1, HALTI0); + + if (fmt & EXT_FORMAT) + supported = VIV_FEATURE(screen, chipMinorFeatures1, HALTI0); + + if (fmt & ASTC_FORMAT) { + supported = screen->specs.tex_astc; + } + + if (!supported) + return false; + + if (texture_format_needs_swiz(format)) + return VIV_FEATURE(screen, chipMinorFeatures1, HALTI0); return true; } -static boolean +static bool etna_screen_is_format_supported(struct pipe_screen *pscreen, enum pipe_format format, enum pipe_texture_target target, - unsigned sample_count, unsigned usage) + unsigned sample_count, + unsigned storage_sample_count, + unsigned usage) { struct etna_screen *screen = etna_screen(pscreen); unsigned allowed = 0; - if (target != PIPE_BUFFER && - target != PIPE_TEXTURE_1D && - target != PIPE_TEXTURE_2D && - target != PIPE_TEXTURE_3D && - target != PIPE_TEXTURE_CUBE && - target != PIPE_TEXTURE_RECT) - return FALSE; + if (!gpu_supports_texture_target(screen, target)) + return false; + + if (MAX2(1, sample_count) != MAX2(1, storage_sample_count)) + return false; if (usage & PIPE_BIND_RENDER_TARGET) { /* if render target, must be RS-supported format */ @@ -494,7 +434,7 @@ etna_screen_is_format_supported(struct pipe_screen *pscreen, * must have MSAA'able format. */ if (sample_count > 1) { if (translate_samples_to_xyscale(sample_count, NULL, NULL, NULL) && - translate_msaa_format(format) != ETNA_NO_MATCH) { + translate_ts_format(format) != ETNA_NO_MATCH) { allowed |= PIPE_BIND_RENDER_TARGET; } } else { @@ -511,7 +451,7 @@ etna_screen_is_format_supported(struct pipe_screen *pscreen, if (usage & PIPE_BIND_SAMPLER_VIEW) { uint32_t fmt = translate_texture_format(format); - if (!gpu_supports_texure_format(screen, fmt)) + if (!gpu_supports_texture_format(screen, fmt, format)) fmt = ETNA_NO_MATCH; if (sample_count < 2 && fmt != ETNA_NO_MATCH) @@ -545,7 +485,48 @@ etna_screen_is_format_supported(struct pipe_screen *pscreen, return usage == allowed; } -static boolean +const uint64_t supported_modifiers[] = { + DRM_FORMAT_MOD_LINEAR, + DRM_FORMAT_MOD_VIVANTE_TILED, + DRM_FORMAT_MOD_VIVANTE_SUPER_TILED, + DRM_FORMAT_MOD_VIVANTE_SPLIT_TILED, + DRM_FORMAT_MOD_VIVANTE_SPLIT_SUPER_TILED, +}; + +static void +etna_screen_query_dmabuf_modifiers(struct pipe_screen *pscreen, + enum pipe_format format, int max, + uint64_t *modifiers, + unsigned int *external_only, int *count) +{ + struct etna_screen *screen = etna_screen(pscreen); + int i, num_modifiers = 0; + + if (max > ARRAY_SIZE(supported_modifiers)) + max = ARRAY_SIZE(supported_modifiers); + + if (!max) { + modifiers = NULL; + max = ARRAY_SIZE(supported_modifiers); + } + + for (i = 0; num_modifiers < max; i++) { + /* don't advertise split tiled formats on single pipe/buffer GPUs */ + if ((screen->specs.pixel_pipes == 1 || screen->specs.single_buffer) && + i >= 3) + break; + + if (modifiers) + modifiers[num_modifiers] = supported_modifiers[i]; + if (external_only) + external_only[num_modifiers] = util_format_is_yuv(format) ? 1 : 0; + num_modifiers++; + } + + *count = num_modifiers; +} + +static bool etna_get_specs(struct etna_screen *screen) { uint64_t val; @@ -604,13 +585,36 @@ etna_get_specs(struct etna_screen *screen) } screen->specs.num_constants = val; + /* Figure out gross GPU architecture. See rnndb/common.xml for a specific + * description of the differences. */ + if (VIV_FEATURE(screen, chipMinorFeatures5, HALTI5)) + screen->specs.halti = 5; /* New GC7000/GC8x00 */ + else if (VIV_FEATURE(screen, chipMinorFeatures5, HALTI4)) + screen->specs.halti = 4; /* Old GC7000/GC7400 */ + else if (VIV_FEATURE(screen, chipMinorFeatures5, HALTI3)) + screen->specs.halti = 3; /* None? */ + else if (VIV_FEATURE(screen, chipMinorFeatures4, HALTI2)) + screen->specs.halti = 2; /* GC2500/GC3000/GC5000/GC6400 */ + else if (VIV_FEATURE(screen, chipMinorFeatures2, HALTI1)) + screen->specs.halti = 1; /* GC900/GC4000/GC7000UL */ + else if (VIV_FEATURE(screen, chipMinorFeatures1, HALTI0)) + screen->specs.halti = 0; /* GC880/GC2000/GC7000TM */ + else + screen->specs.halti = -1; /* GC7000nanolite / pre-GC2000 except GC880 */ + if (screen->specs.halti >= 0) + DBG("etnaviv: GPU arch: HALTI%d", screen->specs.halti); + else + DBG("etnaviv: GPU arch: pre-HALTI"); + screen->specs.can_supertile = VIV_FEATURE(screen, chipMinorFeatures0, SUPER_TILED); screen->specs.bits_per_tile = VIV_FEATURE(screen, chipMinorFeatures0, 2BITPERTILE) ? 2 : 4; screen->specs.ts_clear_value = - VIV_FEATURE(screen, chipMinorFeatures0, 2BITPERTILE) ? 0x55555555 - : 0x11111111; + VIV_FEATURE(screen, chipMinorFeatures5, BLT_ENGINE) ? 0xffffffff : + VIV_FEATURE(screen, chipMinorFeatures0, 2BITPERTILE) ? 0x55555555 : + 0x11111111; + /* vertex and fragment samplers live in one address space */ screen->specs.vertex_sampler_offset = 8; @@ -626,10 +630,20 @@ etna_get_specs(struct etna_screen *screen) screen->model >= 0x1000 || screen->model == 0x880; screen->specs.npot_tex_any_wrap = VIV_FEATURE(screen, chipMinorFeatures1, NON_POWER_OF_TWO); - screen->specs.has_new_sin_cos = + screen->specs.has_new_transcendentals = VIV_FEATURE(screen, chipMinorFeatures3, HAS_FAST_TRANSCENDENTALS); - - if (VIV_FEATURE(screen, chipMinorFeatures3, INSTRUCTION_CACHE)) { + screen->specs.has_halti2_instructions = + VIV_FEATURE(screen, chipMinorFeatures4, HALTI2); + screen->specs.v4_compression = + VIV_FEATURE(screen, chipMinorFeatures6, V4_COMPRESSION); + + if (screen->specs.halti >= 5) { + /* GC7000 - this core must load shaders from memory. */ + screen->specs.vs_offset = 0; + screen->specs.ps_offset = 0; + screen->specs.max_instructions = 0; /* Do not program shaders manually */ + screen->specs.has_icache = true; + } else if (VIV_FEATURE(screen, chipMinorFeatures3, INSTRUCTION_CACHE)) { /* GC3000 - this core is capable of loading shaders from * memory. It can also run shaders from registers, as a fallback, but * "max_instructions" does not have the correct value. It has place for @@ -642,7 +656,8 @@ etna_get_specs(struct etna_screen *screen) * same. */ screen->specs.ps_offset = 0x8000 + 0x1000; - screen->specs.max_instructions = 256; + screen->specs.max_instructions = 256; /* maximum number instructions for non-icache use */ + screen->specs.has_icache = true; } else { if (instruction_count > 256) { /* unified instruction memory? */ screen->specs.vs_offset = 0xC000; @@ -653,6 +668,7 @@ etna_get_specs(struct etna_screen *screen) screen->specs.ps_offset = 0x6000; screen->specs.max_instructions = instruction_count / 2; } + screen->specs.has_icache = false; } if (VIV_FEATURE(screen, chipMinorFeatures1, HALTI0)) { @@ -681,6 +697,26 @@ etna_get_specs(struct etna_screen *screen) screen->specs.max_ps_uniforms = 256; } + if (screen->specs.halti >= 5) { + screen->specs.has_unified_uniforms = true; + screen->specs.vs_uniforms_offset = VIVS_SH_HALTI5_UNIFORMS_MIRROR(0); + screen->specs.ps_uniforms_offset = VIVS_SH_HALTI5_UNIFORMS(screen->specs.max_vs_uniforms*4); + } else if (screen->specs.halti >= 1) { + /* unified uniform memory on GC3000 - HALTI1 feature bit is just a guess + */ + screen->specs.has_unified_uniforms = true; + screen->specs.vs_uniforms_offset = VIVS_SH_UNIFORMS(0); + /* hardcode PS uniforms to start after end of VS uniforms - + * for more flexibility this offset could be variable based on the + * shader. + */ + screen->specs.ps_uniforms_offset = VIVS_SH_UNIFORMS(screen->specs.max_vs_uniforms*4); + } else { + screen->specs.has_unified_uniforms = false; + screen->specs.vs_uniforms_offset = VIVS_VS_UNIFORMS(0); + screen->specs.ps_uniforms_offset = VIVS_PS_UNIFORMS(0); + } + screen->specs.max_texture_size = VIV_FEATURE(screen, chipMinorFeatures0, TEXTURE_8K) ? 8192 : 2048; screen->specs.max_rendertarget_size = @@ -688,7 +724,11 @@ etna_get_specs(struct etna_screen *screen) screen->specs.single_buffer = VIV_FEATURE(screen, chipMinorFeatures4, SINGLE_BUFFER); if (screen->specs.single_buffer) - DBG("etnaviv: Single buffer mode enabled with %d pixel pipes\n", screen->specs.pixel_pipes); + DBG("etnaviv: Single buffer mode enabled with %d pixel pipes", screen->specs.pixel_pipes); + + screen->specs.tex_astc = VIV_FEATURE(screen, chipMinorFeatures4, TEXTURE_ASTC); + + screen->specs.use_blt = VIV_FEATURE(screen, chipMinorFeatures5, BLT_ENGINE); return true; @@ -696,25 +736,6 @@ fail: return false; } -boolean -etna_screen_bo_get_handle(struct pipe_screen *pscreen, struct etna_bo *bo, - unsigned stride, struct winsys_handle *whandle) -{ - whandle->stride = stride; - - if (whandle->type == DRM_API_HANDLE_TYPE_SHARED) { - return etna_bo_get_name(bo, &whandle->handle) == 0; - } else if (whandle->type == DRM_API_HANDLE_TYPE_KMS) { - whandle->handle = etna_bo_handle(bo); - return TRUE; - } else if (whandle->type == DRM_API_HANDLE_TYPE_FD) { - whandle->handle = etna_bo_dmabuf(bo); - return TRUE; - } else { - return FALSE; - } -} - struct etna_bo * etna_screen_bo_from_handle(struct pipe_screen *pscreen, struct winsys_handle *whandle, unsigned *out_stride) @@ -722,9 +743,9 @@ etna_screen_bo_from_handle(struct pipe_screen *pscreen, struct etna_screen *screen = etna_screen(pscreen); struct etna_bo *bo; - if (whandle->type == DRM_API_HANDLE_TYPE_SHARED) { + if (whandle->type == WINSYS_HANDLE_TYPE_SHARED) { bo = etna_bo_from_name(screen->dev, whandle->handle); - } else if (whandle->type == DRM_API_HANDLE_TYPE_FD) { + } else if (whandle->type == WINSYS_HANDLE_TYPE_FD) { bo = etna_bo_from_dmabuf(screen->dev, whandle->handle); } else { DBG("Attempt to import unsupported handle type %d", whandle->type); @@ -741,6 +762,13 @@ etna_screen_bo_from_handle(struct pipe_screen *pscreen, return bo; } +static const void * +etna_get_compiler_options(struct pipe_screen *pscreen, + enum pipe_shader_ir ir, unsigned shader) +{ + return &etna_screen(pscreen)->options; +} + struct pipe_screen * etna_screen_create(struct etna_device *dev, struct etna_gpu *gpu, struct renderonly *ro) @@ -757,6 +785,7 @@ etna_screen_create(struct etna_device *dev, struct etna_gpu *gpu, screen->dev = dev; screen->gpu = gpu; screen->ro = renderonly_dup(ro); + screen->refcnt = 1; if (!screen->ro) { DBG("could not create renderonly object"); @@ -833,9 +862,35 @@ etna_screen_create(struct etna_device *dev, struct etna_gpu *gpu, } screen->features[6] = val; + if (etna_gpu_get_param(screen->gpu, ETNA_GPU_FEATURES_7, &val)) { + DBG("could not get ETNA_GPU_FEATURES_7"); + goto fail; + } + screen->features[7] = val; + if (!etna_get_specs(screen)) goto fail; + screen->options = (nir_shader_compiler_options) { + .lower_fpow = true, + .lower_sub = true, + .lower_ftrunc = true, + .fuse_ffma = true, + .lower_bitops = true, + .lower_all_io_to_temps = true, + .vertex_id_zero_based = true, + .lower_flrp32 = true, + .lower_fmod = true, + .lower_vector_cmp = true, + .lower_fdph = true, + .lower_fdiv = true, /* !screen->specs.has_new_transcendentals */ + .lower_fsign = !screen->specs.has_sign_floor_ceil, + .lower_ffloor = !screen->specs.has_sign_floor_ceil, + .lower_fceil = !screen->specs.has_sign_floor_ceil, + .lower_fsqrt = !screen->specs.has_sin_cos_sqrt, + .lower_sincos = !screen->specs.has_sin_cos_sqrt, + }; + /* apply debug options that disable individual features */ if (DBG_ENABLED(ETNA_DBG_NO_EARLY_Z)) screen->features[viv_chipFeatures] |= chipFeatures_NO_EARLY_Z; @@ -845,11 +900,14 @@ etna_screen_create(struct etna_device *dev, struct etna_gpu *gpu, screen->features[viv_chipMinorFeatures1] &= ~chipMinorFeatures1_AUTO_DISABLE; if (DBG_ENABLED(ETNA_DBG_NO_SUPERTILE)) screen->specs.can_supertile = 0; + if (DBG_ENABLED(ETNA_DBG_NO_SINGLEBUF)) + screen->specs.single_buffer = 0; pscreen->destroy = etna_screen_destroy; pscreen->get_param = etna_screen_get_param; pscreen->get_paramf = etna_screen_get_paramf; pscreen->get_shader_param = etna_screen_get_shader_param; + pscreen->get_compiler_options = etna_get_compiler_options; pscreen->get_name = etna_screen_get_name; pscreen->get_vendor = etna_screen_get_vendor; @@ -858,15 +916,28 @@ etna_screen_create(struct etna_device *dev, struct etna_gpu *gpu, pscreen->get_timestamp = etna_screen_get_timestamp; pscreen->context_create = etna_context_create; pscreen->is_format_supported = etna_screen_is_format_supported; + pscreen->query_dmabuf_modifiers = etna_screen_query_dmabuf_modifiers; etna_fence_screen_init(pscreen); etna_query_screen_init(pscreen); etna_resource_screen_init(pscreen); + util_dynarray_init(&screen->supported_pm_queries, NULL); slab_create_parent(&screen->transfer_pool, sizeof(struct etna_transfer), 16); + if (screen->drm_version >= ETNA_DRM_VERSION_PERFMON) + etna_pm_query_setup(screen); + + mtx_init(&screen->lock, mtx_recursive); + screen->used_resources = _mesa_set_create(NULL, _mesa_hash_pointer, + _mesa_key_pointer_equal); + if (!screen->used_resources) + goto fail2; + return pscreen; +fail2: + mtx_destroy(&screen->lock); fail: etna_screen_destroy(pscreen); return NULL;