<div dir="ltr"><div dir="ltr"> </div> <div class="gmail_quote"><div dir="ltr" class="gmail_attr">On Wed, May 25, 2022 at 12:20 PM Alex Deucher <<a href="mailto:alexander.deucher@amd.com">alexander.deucher@amd.com</a>> wrote: </div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">From: Aurabindo Pillai <<a href="mailto:aurabindo.pillai@amd.com" target="_blank">aurabindo.pillai@amd.com</a>> GFX11 IP introduces new tiling mode. Various combinations of DCC settings are possible and the most preferred settings must be exposed for optimal use of the hardware. add_gfx11_modifiers() is based on recommendation from Marek for the preferred tiling modifier that are most efficient for the hardware. Signed-off-by: Aurabindo Pillai <<a href="mailto:aurabindo.pillai@amd.com" target="_blank">aurabindo.pillai@amd.com</a>> Acked-by: Alex Deucher <<a href="mailto:alexander.deucher@amd.com" target="_blank">alexander.deucher@amd.com</a>> Signed-off-by: Alex Deucher <<a href="mailto:alexander.deucher@amd.com" target="_blank">alexander.deucher@amd.com</a>> --- drivers/gpu/drm/amd/amdgpu/amdgpu_display.c | 40 ++++++++-- .../gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c | 74 ++++++++++++++++++- include/uapi/drm/drm_fourcc.h | 2 + 3 files changed, 108 insertions(+), 8 deletions(-) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c index ec395fe427f2..a54081a89282 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c @@ -30,6 +30,9 @@ #include "atom.h" #include "amdgpu_connectors.h" #include "amdgpu_display.h" +#include "soc15_common.h" +#include "gc/gc_11_0_0_offset.h" +#include "gc/gc_11_0_0_sh_mask.h" #include <asm/div64.h> #include <linux/pci.h> @@ -662,6 +665,11 @@ static int convert_tiling_flags_to_modifier(struct amdgpu_framebuffer *afb) { struct amdgpu_device *adev = drm_to_adev(afb-><a href="http://base.dev" rel="noreferrer" target="_blank">base.dev</a>); uint64_t modifier = 0; + int num_pipes = 0; + int num_pkrs = 0; + + num_pkrs = adev->gfx.config.gb_addr_config_fields.num_pkrs; + num_pipes = adev->gfx.config.gb_addr_config_fields.num_pipes; if (!afb->tiling_flags || !AMDGPU_TILING_GET(afb->tiling_flags, SWIZZLE_MODE)) { modifier = DRM_FORMAT_MOD_LINEAR; @@ -674,7 +682,7 @@ static int convert_tiling_flags_to_modifier(struct amdgpu_framebuffer *afb) int bank_xor_bits = 0; int packers = 0; int rb = 0; - int pipes = ilog2(adev->gfx.config.gb_addr_config_fields.num_pipes); + int pipes = ilog2(num_pipes); uint32_t dcc_offset = AMDGPU_TILING_GET(afb->tiling_flags, DCC_OFFSET_256B); switch (swizzle >> 2) { @@ -690,12 +698,17 @@ static int convert_tiling_flags_to_modifier(struct amdgpu_framebuffer *afb) case 6: /* 64 KiB _X */ block_size_bits = 16; break; + case 7: /* 256 KiB */ + block_size_bits = 18; + break; default: /* RESERVED or VAR */ return -EINVAL; } - if (adev->ip_versions[GC_HWIP][0] >= IP_VERSION(10, 3, 0)) + if (adev->ip_versions[GC_HWIP][0] >= IP_VERSION(11, 0, 0)) + version = AMD_FMT_MOD_TILE_VER_GFX11; + else if (adev->ip_versions[GC_HWIP][0] >= IP_VERSION(10, 3, 0)) version = AMD_FMT_MOD_TILE_VER_GFX10_RBPLUS; else if (adev->ip_versions[GC_HWIP][0] >= IP_VERSION(10, 0, 0)) version = AMD_FMT_MOD_TILE_VER_GFX10; @@ -718,7 +731,17 @@ static int convert_tiling_flags_to_modifier(struct amdgpu_framebuffer *afb) } </blockquote><div> </div><div>The switch statement right above this that is not in this patch and changes "version" should be skipped on >= gfx11. Under no circumstances should the version be changed on gfx11. </div><div> </div><div>Marek </div><div> </div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"> if (has_xor) { + if (num_pipes == num_pkrs && num_pkrs == 0) { + DRM_ERROR("invalid number of pipes and packers\n"); + return -EINVAL; + } + switch (version) { + case AMD_FMT_MOD_TILE_VER_GFX11: + pipe_xor_bits = min(block_size_bits - 8, pipes); + packers = min(block_size_bits - 8 - pipe_xor_bits, + ilog2(adev->gfx.config.gb_addr_config_fields.num_pkrs)); + break; case AMD_FMT_MOD_TILE_VER_GFX10_RBPLUS: pipe_xor_bits = min(block_size_bits - 8, pipes); packers = min(block_size_bits - 8 - pipe_xor_bits, @@ -752,9 +775,10 @@ static int convert_tiling_flags_to_modifier(struct amdgpu_framebuffer *afb) u64 render_dcc_offset; /* Enable constant encode on RAVEN2 and later. */ - bool dcc_constant_encode = adev->asic_type > CHIP_RAVEN || + bool dcc_constant_encode = (adev->asic_type > CHIP_RAVEN || (adev->asic_type == CHIP_RAVEN && - adev->external_rev_id >= 0x81); + adev->external_rev_id >= 0x81)) && + adev->ip_versions[GC_HWIP][0] < IP_VERSION(11, 0, 0); int max_cblock_size = dcc_i64b ? AMD_FMT_MOD_DCC_BLOCK_64B : dcc_i128b ? AMD_FMT_MOD_DCC_BLOCK_128B : @@ -869,10 +893,11 @@ static unsigned int get_dcc_block_size(uint64_t modifier, bool rb_aligned, return max(10 + (rb_aligned ? (int)AMD_FMT_MOD_GET(RB, modifier) : 0), 12); } case AMD_FMT_MOD_TILE_VER_GFX10: - case AMD_FMT_MOD_TILE_VER_GFX10_RBPLUS: { + case AMD_FMT_MOD_TILE_VER_GFX10_RBPLUS: + case AMD_FMT_MOD_TILE_VER_GFX11: { int pipes_log2 = AMD_FMT_MOD_GET(PIPE_XOR_BITS, modifier); - if (ver == AMD_FMT_MOD_TILE_VER_GFX10_RBPLUS && pipes_log2 > 1 && + if (ver >= AMD_FMT_MOD_TILE_VER_GFX10_RBPLUS && pipes_log2 > 1 && AMD_FMT_MOD_GET(PACKERS, modifier) == pipes_log2) ++pipes_log2; @@ -965,6 +990,9 @@ static int amdgpu_display_verify_sizes(struct amdgpu_framebuffer *rfb) case DC_SW_64KB_S_X: block_size_log2 = 16; break; + case DC_SW_VAR_S_X: + block_size_log2 = 18; + break; default: drm_dbg_kms(rfb-><a href="http://base.dev" rel="noreferrer" target="_blank">base.dev</a>, "Swizzle mode with unknown block size: %d\n", swizzle); diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c index a93affc37c53..badd136f5686 100644 --- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c +++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c @@ -88,10 +88,14 @@ #include "dcn/dcn_1_0_offset.h" #include "dcn/dcn_1_0_sh_mask.h" #include "soc15_hw_ip.h" +#include "soc15_common.h" #include "vega10_ip_offset.h" #include "soc15_common.h" +#include "gc/gc_11_0_0_offset.h" +#include "gc/gc_11_0_0_sh_mask.h" + #include "modules/inc/mod_freesync.h" #include "modules/power/power_helpers.h" #include "modules/inc/mod_info_packet.h" @@ -4885,7 +4889,9 @@ fill_gfx9_tiling_info_from_modifier(const struct amdgpu_device *adev, unsigned int mod_bank_xor_bits = AMD_FMT_MOD_GET(BANK_XOR_BITS, modifier); unsigned int mod_pipe_xor_bits = AMD_FMT_MOD_GET(PIPE_XOR_BITS, modifier); unsigned int pkrs_log2 = AMD_FMT_MOD_GET(PACKERS, modifier); - unsigned int pipes_log2 = min(4u, mod_pipe_xor_bits); + unsigned int pipes_log2; + + pipes_log2 = min(5u, mod_pipe_xor_bits); fill_gfx9_tiling_info_from_device(adev, tiling_info); @@ -5221,6 +5227,67 @@ add_gfx10_3_modifiers(const struct amdgpu_device *adev, AMD_FMT_MOD_SET(TILE_VERSION, AMD_FMT_MOD_TILE_VER_GFX9)); } +static void +add_gfx11_modifiers(const struct amdgpu_device *adev, + uint64_t **mods, uint64_t *size, uint64_t *capacity) +{ + int num_pipes = 0; + int pipe_xor_bits = 0; + int num_pkrs = 0; + int pkrs = 0; + u32 gb_addr_config; + unsigned swizzle_r_x; + uint64_t modifier_r_x; + uint64_t modifier_dcc_best; + uint64_t modifier_dcc_4k; + + /* TODO: GFX11 IP HW init hasnt finish and we get zero if we read from + * adev->gfx.config.gb_addr_config_fields.num_{pkrs,pipes} */ + gb_addr_config = RREG32_SOC15(GC, 0, regGB_ADDR_CONFIG); + ASSERT(gb_addr_config != 0); + + num_pkrs = 1 << REG_GET_FIELD(gb_addr_config, GB_ADDR_CONFIG, NUM_PKRS); + pkrs = ilog2(num_pkrs); + num_pipes = 1 << REG_GET_FIELD(gb_addr_config, GB_ADDR_CONFIG, NUM_PIPES); + pipe_xor_bits = ilog2(num_pipes); + + /* R_X swizzle modes are the best for rendering and DCC requires them. */ + swizzle_r_x = num_pipes > 16 ? AMD_FMT_MOD_TILE_GFX11_256K_R_X : + AMD_FMT_MOD_TILE_GFX9_64K_R_X; + + modifier_r_x = AMD_FMT_MOD | + AMD_FMT_MOD_SET(TILE_VERSION, AMD_FMT_MOD_TILE_VER_GFX11) | + AMD_FMT_MOD_SET(TILE, swizzle_r_x) | + AMD_FMT_MOD_SET(PIPE_XOR_BITS, pipe_xor_bits) | + AMD_FMT_MOD_SET(PACKERS, pkrs); + + /* DCC_CONSTANT_ENCODE is not set because it can't vary with gfx11 (it's implied to be 1). */ + modifier_dcc_best = modifier_r_x | + AMD_FMT_MOD_SET(DCC, 1) | + AMD_FMT_MOD_SET(DCC_INDEPENDENT_64B, 0) | + AMD_FMT_MOD_SET(DCC_INDEPENDENT_128B, 1) | + AMD_FMT_MOD_SET(DCC_MAX_COMPRESSED_BLOCK, AMD_FMT_MOD_DCC_BLOCK_128B); + + /* DCC settings for 4K and greater resolutions. (required by display hw) */ + modifier_dcc_4k = modifier_r_x | + AMD_FMT_MOD_SET(DCC, 1) | + AMD_FMT_MOD_SET(DCC_INDEPENDENT_64B, 1) | + AMD_FMT_MOD_SET(DCC_INDEPENDENT_128B, 1) | + AMD_FMT_MOD_SET(DCC_MAX_COMPRESSED_BLOCK, AMD_FMT_MOD_DCC_BLOCK_64B); + + add_modifier(mods, size, capacity, modifier_dcc_best); + add_modifier(mods, size, capacity, modifier_dcc_4k); + + add_modifier(mods, size, capacity, modifier_dcc_best | AMD_FMT_MOD_SET(DCC_RETILE, 1)); + add_modifier(mods, size, capacity, modifier_dcc_4k | AMD_FMT_MOD_SET(DCC_RETILE, 1)); + + add_modifier(mods, size, capacity, modifier_r_x); + + add_modifier(mods, size, capacity, AMD_FMT_MOD | + AMD_FMT_MOD_SET(TILE_VERSION, AMD_FMT_MOD_TILE_VER_GFX11) | + AMD_FMT_MOD_SET(TILE, AMD_FMT_MOD_TILE_GFX9_64K_D)); +} + static int get_plane_modifiers(const struct amdgpu_device *adev, unsigned int plane_type, uint64_t **mods) { @@ -5254,6 +5321,9 @@ get_plane_modifiers(const struct amdgpu_device *adev, unsigned int plane_type, u else add_gfx10_1_modifiers(adev, mods, &size, &capacity); break; + case AMDGPU_FAMILY_GC_11_0_0: + add_gfx11_modifiers(adev, mods, &size, &capacity); + break; } add_modifier(mods, &size, &capacity, DRM_FORMAT_MOD_LINEAR); @@ -5292,7 +5362,7 @@ fill_gfx9_plane_attributes_from_modifiers(struct amdgpu_device *adev, dcc->enable = 1; dcc->meta_pitch = afb->base.pitches[1]; dcc->independent_64b_blks = independent_64b_blks; - if (AMD_FMT_MOD_GET(TILE_VERSION, modifier) == AMD_FMT_MOD_TILE_VER_GFX10_RBPLUS) { + if (AMD_FMT_MOD_GET(TILE_VERSION, modifier) >= AMD_FMT_MOD_TILE_VER_GFX10_RBPLUS) { if (independent_64b_blks && independent_128b_blks) dcc->dcc_ind_blk = hubp_ind_block_64b_no_128bcl; else if (independent_128b_blks) diff --git a/include/uapi/drm/drm_fourcc.h b/include/uapi/drm/drm_fourcc.h index fc0c1454d275..14cb2dafb0fa 100644 --- a/include/uapi/drm/drm_fourcc.h +++ b/include/uapi/drm/drm_fourcc.h @@ -1294,6 +1294,7 @@ drm_fourcc_canonicalize_nvidia_format_mod(__u64 modifier) #define AMD_FMT_MOD_TILE_VER_GFX9 1 #define AMD_FMT_MOD_TILE_VER_GFX10 2 #define AMD_FMT_MOD_TILE_VER_GFX10_RBPLUS 3 +#define AMD_FMT_MOD_TILE_VER_GFX11 4 /* * 64K_S is the same for GFX9/GFX10/GFX10_RBPLUS and hence has GFX9 as canonical @@ -1309,6 +1310,7 @@ drm_fourcc_canonicalize_nvidia_format_mod(__u64 modifier) #define AMD_FMT_MOD_TILE_GFX9_64K_S_X 25 #define AMD_FMT_MOD_TILE_GFX9_64K_D_X 26 #define AMD_FMT_MOD_TILE_GFX9_64K_R_X 27 +#define AMD_FMT_MOD_TILE_GFX11_256K_R_X 31 #define AMD_FMT_MOD_DCC_BLOCK_64B 0 #define AMD_FMT_MOD_DCC_BLOCK_128B 1 -- 2.35.3 </blockquote></div></div>