#include <stdio.h>
#include <xf86drm.h>
-#include <amdgpu_drm.h>
+#include "drm-uapi/amdgpu_drm.h"
#include <amdgpu.h>
return value ? true : false;
}
+static bool has_timeline_syncobj(int fd)
+{
+ uint64_t value;
+ if (drmGetCap(fd, DRM_CAP_SYNCOBJ_TIMELINE, &value))
+ return false;
+ return value ? true : false;
+}
+
static uint64_t fix_vram_size(uint64_t size)
{
/* The VRAM size is underreported, so we need to fix it, because
return align64(size, 256*1024*1024);
}
+static uint32_t
+get_l2_cache_size(enum radeon_family family)
+{
+ switch (family) {
+ case CHIP_KABINI:
+ case CHIP_STONEY:
+ return 128 * 1024;
+ case CHIP_OLAND:
+ case CHIP_HAINAN:
+ case CHIP_ICELAND:
+ return 256 * 1024;
+ case CHIP_PITCAIRN:
+ case CHIP_VERDE:
+ case CHIP_BONAIRE:
+ case CHIP_KAVERI:
+ case CHIP_POLARIS12:
+ case CHIP_CARRIZO:
+ return 512 * 1024;
+ case CHIP_TAHITI:
+ case CHIP_TONGA:
+ return 768 * 1024;
+ break;
+ case CHIP_HAWAII:
+ case CHIP_POLARIS11:
+ return 1024 * 1024;
+ case CHIP_FIJI:
+ case CHIP_POLARIS10:
+ return 2048 * 1024;
+ break;
+ default:
+ return 4096 * 1024;
+ }
+}
+
bool ac_query_gpu_info(int fd, void *dev_p,
struct radeon_info *info,
struct amdgpu_gpu_info *amdinfo)
/* Set chip identification. */
info->pci_id = amdinfo->asic_id; /* TODO: is this correct? */
+ info->pci_rev_id = amdinfo->pci_rev_id;
info->vce_harvest_config = amdinfo->vce_harvest_config;
#define identify_chip2(asic, chipname) \
identify_chip(NAVI10);
identify_chip(NAVI12);
identify_chip(NAVI14);
+ identify_chip(SIENNA_CICHLID);
+ identify_chip(NAVY_FLOUNDER);
break;
}
return false;
}
- if (info->family >= CHIP_NAVI10)
+ if (info->family >= CHIP_SIENNA_CICHLID)
+ info->chip_class = GFX10_3;
+ else if (info->family >= CHIP_NAVI10)
info->chip_class = GFX10;
else if (info->family >= CHIP_VEGA10)
info->chip_class = GFX9;
else
info->max_alloc_size = info->gart_size * 0.7;
+ info->vram_type = amdinfo->vram_type;
+ info->vram_bit_width = amdinfo->vram_bit_width;
+ info->ce_ram_size = amdinfo->ce_ram_size;
+
+ info->l2_cache_size = get_l2_cache_size(info->family);
+ info->l1_cache_size = 16384;
+
/* Set which chips have uncached device memory. */
info->has_l2_uncached = info->chip_class >= GFX9;
/* Set hardware information. */
info->gds_size = gds.gds_total_size;
info->gds_gfx_partition_size = gds.gds_gfx_partition_size;
- /* convert the shader clock from KHz to MHz */
+ /* convert the shader/memory clocks from KHz to MHz */
info->max_shader_clock = amdinfo->max_engine_clk / 1000;
+ info->max_memory_clock = amdinfo->max_memory_clk / 1000;
info->num_tcc_blocks = device_info.num_tcc_blocks;
info->max_se = amdinfo->num_shader_engines;
info->max_sh_per_se = amdinfo->num_shader_arrays_per_engine;
uvd_enc.available_rings ? true : false;
info->has_userptr = true;
info->has_syncobj = has_syncobj(fd);
+ info->has_timeline_syncobj = has_timeline_syncobj(fd);
info->has_syncobj_wait_for_submit = info->has_syncobj && info->drm_minor >= 20;
info->has_fence_to_handle = info->has_syncobj && info->drm_minor >= 21;
info->has_ctx_priority = info->drm_minor >= 22;
info->has_unaligned_shader_loads = info->chip_class != GFX6;
/* Disable sparse mappings on GFX6 due to VM faults in CP DMA. Enable them once
* these faults are mitigated in software.
- * Disable sparse mappings on GFX9 due to hangs.
*/
- info->has_sparse_vm_mappings =
- info->chip_class >= GFX7 && info->chip_class <= GFX8 &&
- info->drm_minor >= 13;
+ info->has_sparse_vm_mappings = info->chip_class >= GFX7 && info->drm_minor >= 13;
info->has_2d_tiling = true;
info->has_read_registers_query = true;
info->has_scheduled_fence_dependency = info->drm_minor >= 28;
+ info->mid_command_buffer_preemption_enabled =
+ amdinfo->ids_flags & AMDGPU_IDS_FLAGS_PREEMPTION;
info->pa_sc_tile_steering_override = device_info.pa_sc_tile_steering_override;
info->num_render_backends = amdinfo->rb_pipes;
info->tcc_cache_line_size = 64;
}
info->gb_addr_config = amdinfo->gb_addr_cfg;
- if (info->chip_class == GFX9) {
+ if (info->chip_class >= GFX9) {
info->num_tile_pipes = 1 << G_0098F8_NUM_PIPES(amdinfo->gb_addr_cfg);
info->pipe_interleave_bytes =
256 << G_0098F8_PIPE_INTERLEAVE_SIZE_GFX9(amdinfo->gb_addr_cfg);
}
info->r600_has_virtual_memory = true;
+ /* LDS is 64KB per CU (4 SIMDs), which is 16KB per SIMD (usage above
+ * 16KB makes some SIMDs unoccupied).
+ *
+ * LDS is 128KB in WGP mode and 64KB in CU mode. Assume the WGP mode is used.
+ */
+ info->lds_size_per_workgroup = info->chip_class >= GFX10 ? 128 * 1024 : 64 * 1024;
+ info->lds_granularity = info->chip_class >= GFX7 ? 128 * 4 : 64 * 4;
+
assert(util_is_power_of_two_or_zero(dma.available_rings + 1));
assert(util_is_power_of_two_or_zero(compute.available_rings + 1));
info->num_rings[RING_VCN_ENC] = util_bitcount(vcn_enc.available_rings);
info->num_rings[RING_VCN_JPEG] = util_bitcount(vcn_jpeg.available_rings);
+ /* This is "align_mask" copied from the kernel, maximums of all IP versions. */
+ info->ib_pad_dw_mask[RING_GFX] = 0xff;
+ info->ib_pad_dw_mask[RING_COMPUTE] = 0xff;
+ info->ib_pad_dw_mask[RING_DMA] = 0xf;
+ info->ib_pad_dw_mask[RING_UVD] = 0xf;
+ info->ib_pad_dw_mask[RING_VCE] = 0x3f;
+ info->ib_pad_dw_mask[RING_UVD_ENC] = 0x3f;
+ info->ib_pad_dw_mask[RING_VCN_DEC] = 0xf;
+ info->ib_pad_dw_mask[RING_VCN_ENC] = 0x3f;
+ info->ib_pad_dw_mask[RING_VCN_JPEG] = 0xf;
+
/* The mere presence of CLEAR_STATE in the IB causes random GPU hangs
* on GFX6. Some CLEAR_STATE cause asic hang on radeon kernel, etc.
* SPI_VS_OUT_CONFIG. So only enable GFX7 CLEAR_STATE on amdgpu kernel.
info->family == CHIP_VEGA12 ||
info->family == CHIP_RAVEN ||
info->family == CHIP_RAVEN2 ||
- info->family == CHIP_RENOIR);
+ info->family == CHIP_RENOIR ||
+ info->chip_class >= GFX10_3);
info->has_out_of_order_rast = info->chip_class >= GFX8 &&
+ info->chip_class <= GFX9 &&
info->max_se >= 2;
+ /* Whether chips support double rate packed math instructions. */
+ info->has_packed_math_16bit = info->chip_class >= GFX9;
+
/* TODO: Figure out how to use LOAD_CONTEXT_REG on GFX6-GFX7. */
info->has_load_ctx_reg_pkt = info->chip_class >= GFX9 ||
(info->chip_class >= GFX8 &&
/* Get the number of good compute units. */
info->num_good_compute_units = 0;
- for (i = 0; i < info->max_se; i++)
- for (j = 0; j < info->max_sh_per_se; j++)
+ for (i = 0; i < info->max_se; i++) {
+ for (j = 0; j < info->max_sh_per_se; j++) {
+ /*
+ * The cu bitmap in amd gpu info structure is
+ * 4x4 size array, and it's usually suitable for Vega
+ * ASICs which has 4*2 SE/SH layout.
+ * But for Arcturus, SE/SH layout is changed to 8*1.
+ * To mostly reduce the impact, we make it compatible
+ * with current bitmap array as below:
+ * SE4,SH0 --> cu_bitmap[0][1]
+ * SE5,SH0 --> cu_bitmap[1][1]
+ * SE6,SH0 --> cu_bitmap[2][1]
+ * SE7,SH0 --> cu_bitmap[3][1]
+ */
+ info->cu_mask[i%4][j+i/4] = amdinfo->cu_bitmap[i%4][j+i/4];
info->num_good_compute_units +=
- util_bitcount(amdinfo->cu_bitmap[i][j]);
- info->num_good_cu_per_sh = info->num_good_compute_units /
- (info->max_se * info->max_sh_per_se);
+ util_bitcount(info->cu_mask[i][j]);
+ }
+ }
- /* Round down to the nearest multiple of 2, because the hw can't
- * disable CUs. It can only disable whole WGPs (dual-CUs).
+ /* On GFX10, only whole WGPs (in units of 2 CUs) can be disabled,
+ * and max - min <= 2.
*/
- if (info->chip_class >= GFX10)
- info->num_good_cu_per_sh -= info->num_good_cu_per_sh % 2;
+ unsigned cu_group = info->chip_class >= GFX10 ? 2 : 1;
+ info->max_good_cu_per_sa = DIV_ROUND_UP(info->num_good_compute_units,
+ (info->max_se * info->max_sh_per_se * cu_group)) * cu_group;
+ info->min_good_cu_per_sa = (info->num_good_compute_units /
+ (info->max_se * info->max_sh_per_se * cu_group)) * cu_group;
memcpy(info->si_tile_mode_array, amdinfo->gb_tile_mode,
sizeof(amdinfo->gb_tile_mode));
unsigned ib_align = 0;
ib_align = MAX2(ib_align, gfx.ib_start_alignment);
+ ib_align = MAX2(ib_align, gfx.ib_size_alignment);
ib_align = MAX2(ib_align, compute.ib_start_alignment);
+ ib_align = MAX2(ib_align, compute.ib_size_alignment);
ib_align = MAX2(ib_align, dma.ib_start_alignment);
+ ib_align = MAX2(ib_align, dma.ib_size_alignment);
ib_align = MAX2(ib_align, uvd.ib_start_alignment);
+ ib_align = MAX2(ib_align, uvd.ib_size_alignment);
ib_align = MAX2(ib_align, uvd_enc.ib_start_alignment);
+ ib_align = MAX2(ib_align, uvd_enc.ib_size_alignment);
ib_align = MAX2(ib_align, vce.ib_start_alignment);
+ ib_align = MAX2(ib_align, vce.ib_size_alignment);
ib_align = MAX2(ib_align, vcn_dec.ib_start_alignment);
+ ib_align = MAX2(ib_align, vcn_dec.ib_size_alignment);
ib_align = MAX2(ib_align, vcn_enc.ib_start_alignment);
+ ib_align = MAX2(ib_align, vcn_enc.ib_size_alignment);
ib_align = MAX2(ib_align, vcn_jpeg.ib_start_alignment);
- assert(ib_align);
- info->ib_start_alignment = ib_align;
-
- if (info->drm_minor >= 31 &&
- (info->family == CHIP_RAVEN ||
- info->family == CHIP_RAVEN2 ||
- info->family == CHIP_RENOIR)) {
+ ib_align = MAX2(ib_align, vcn_jpeg.ib_size_alignment);
+ /* GFX10 and maybe GFX9 need this alignment for cache coherency. */
+ if (info->chip_class >= GFX9)
+ ib_align = MAX2(ib_align, info->tcc_cache_line_size);
+ /* The kernel pads gfx and compute IBs to 256 dwords since:
+ * 66f3b2d527154bd258a57c8815004b5964aa1cf5
+ * Do the same.
+ */
+ ib_align = MAX2(ib_align, 1024);
+ info->ib_alignment = ib_align;
+
+ if ((info->drm_minor >= 31 &&
+ (info->family == CHIP_RAVEN ||
+ info->family == CHIP_RAVEN2 ||
+ info->family == CHIP_RENOIR)) ||
+ (info->drm_minor >= 34 &&
+ (info->family == CHIP_NAVI12 ||
+ info->family == CHIP_NAVI14)) ||
+ info->chip_class >= GFX10_3) {
if (info->num_render_backends == 1)
info->use_display_dcc_unaligned = true;
else
case CHIP_RENOIR:
case CHIP_NAVI10:
case CHIP_NAVI12:
+ case CHIP_SIENNA_CICHLID:
+ case CHIP_NAVY_FLOUNDER:
pc_lines = 1024;
break;
case CHIP_NAVI14:
if (info->chip_class >= GFX10)
info->num_sdp_interfaces = device_info.num_tcc_blocks;
- info->max_wave64_per_simd = info->family >= CHIP_POLARIS10 &&
- info->family <= CHIP_VEGAM ? 8 : 10;
+ if (info->chip_class >= GFX10_3)
+ info->max_wave64_per_simd = 16;
+ else if (info->chip_class == GFX10)
+ info->max_wave64_per_simd = 20;
+ else if (info->family >= CHIP_POLARIS10 && info->family <= CHIP_VEGAM)
+ info->max_wave64_per_simd = 8;
+ else
+ info->max_wave64_per_simd = 10;
- /* The number is per SIMD. There is enough SGPRs for the maximum number
- * of Wave32, which is double the number for Wave64.
- */
- if (info->chip_class >= GFX10)
- info->num_physical_sgprs_per_simd = 128 * info->max_wave64_per_simd * 2;
- else if (info->chip_class >= GFX8)
+ if (info->chip_class >= GFX10) {
+ info->num_physical_sgprs_per_simd = 128 * info->max_wave64_per_simd;
+ info->min_sgpr_alloc = 128;
+ info->sgpr_alloc_granularity = 128;
+ /* Don't use late alloc on small chips. */
+ info->use_late_alloc = info->num_render_backends > 4;
+ } else if (info->chip_class >= GFX8) {
info->num_physical_sgprs_per_simd = 800;
- else
+ info->min_sgpr_alloc = 16;
+ info->sgpr_alloc_granularity = 16;
+ info->use_late_alloc = true;
+ } else {
info->num_physical_sgprs_per_simd = 512;
+ info->min_sgpr_alloc = 8;
+ info->sgpr_alloc_granularity = 8;
+ /* Potential hang on Kabini: */
+ info->use_late_alloc = info->family != CHIP_KABINI;
+ }
+
+ info->max_sgpr_alloc = info->family == CHIP_TONGA ||
+ info->family == CHIP_ICELAND ? 96 : 104;
+
+ info->min_wave64_vgpr_alloc = 4;
+ info->max_vgpr_alloc = 256;
+ info->wave64_vgpr_alloc_granularity = 4;
info->num_physical_wave64_vgprs_per_simd = info->chip_class >= GFX10 ? 512 : 256;
+ info->num_simd_per_compute_unit = info->chip_class >= GFX10 ? 2 : 4;
+
return true;
}
printf(" marketing_name = %s\n", info->marketing_name);
printf(" is_pro_graphics = %u\n", info->is_pro_graphics);
printf(" pci_id = 0x%x\n", info->pci_id);
+ printf(" pci_rev_id = 0x%x\n", info->pci_rev_id);
printf(" family = %i\n", info->family);
printf(" chip_class = %i\n", info->chip_class);
printf(" family_id = %i\n", info->family_id);
printf(" gart_size = %i MB\n", (int)DIV_ROUND_UP(info->gart_size, 1024*1024));
printf(" vram_size = %i MB\n", (int)DIV_ROUND_UP(info->vram_size, 1024*1024));
printf(" vram_vis_size = %i MB\n", (int)DIV_ROUND_UP(info->vram_vis_size, 1024*1024));
+ printf(" vram_type = %i\n", info->vram_type);
+ printf(" vram_bit_width = %i\n", info->vram_bit_width);
printf(" gds_size = %u kB\n", info->gds_size / 1024);
printf(" gds_gfx_partition_size = %u kB\n", info->gds_gfx_partition_size / 1024);
printf(" max_alloc_size = %i MB\n",
printf(" tcc_cache_line_size = %u\n", info->tcc_cache_line_size);
printf(" tcc_harvested = %u\n", info->tcc_harvested);
printf(" pc_lines = %u\n", info->pc_lines);
+ printf(" lds_size_per_workgroup = %u\n", info->lds_size_per_workgroup);
+ printf(" lds_granularity = %i\n", info->lds_granularity);
+ printf(" max_memory_clock = %i\n", info->max_memory_clock);
+ printf(" ce_ram_size = %i\n", info->ce_ram_size);
+ printf(" l1_cache_size = %i\n", info->l1_cache_size);
+ printf(" l2_cache_size = %i\n", info->l2_cache_size);
printf("CP info:\n");
printf(" gfx_ib_pad_with_type2 = %i\n", info->gfx_ib_pad_with_type2);
- printf(" ib_start_alignment = %u\n", info->ib_start_alignment);
+ printf(" ib_alignment = %u\n", info->ib_alignment);
printf(" me_fw_version = %i\n", info->me_fw_version);
printf(" me_fw_feature = %i\n", info->me_fw_feature);
printf(" pfp_fw_version = %i\n", info->pfp_fw_version);
printf(" has_userptr = %i\n", info->has_userptr);
printf(" has_syncobj = %u\n", info->has_syncobj);
printf(" has_syncobj_wait_for_submit = %u\n", info->has_syncobj_wait_for_submit);
+ printf(" has_timeline_syncobj = %u\n", info->has_timeline_syncobj);
printf(" has_fence_to_handle = %u\n", info->has_fence_to_handle);
printf(" has_ctx_priority = %u\n", info->has_ctx_priority);
printf(" has_local_buffers = %u\n", info->has_local_buffers);
printf(" has_read_registers_query = %u\n", info->has_read_registers_query);
printf(" has_gds_ordered_append = %u\n", info->has_gds_ordered_append);
printf(" has_scheduled_fence_dependency = %u\n", info->has_scheduled_fence_dependency);
+ printf(" mid_command_buffer_preemption_enabled = %u\n", info->mid_command_buffer_preemption_enabled);
printf("Shader core info:\n");
printf(" max_shader_clock = %i\n", info->max_shader_clock);
printf(" num_good_compute_units = %i\n", info->num_good_compute_units);
- printf(" num_good_cu_per_sh = %i\n", info->num_good_cu_per_sh);
+ printf(" max_good_cu_per_sa = %i\n", info->max_good_cu_per_sa);
+ printf(" min_good_cu_per_sa = %i\n", info->min_good_cu_per_sa);
printf(" max_se = %i\n", info->max_se);
printf(" max_sh_per_se = %i\n", info->max_sh_per_se);
printf(" max_wave64_per_simd = %i\n", info->max_wave64_per_simd);
printf(" num_physical_sgprs_per_simd = %i\n", info->num_physical_sgprs_per_simd);
printf(" num_physical_wave64_vgprs_per_simd = %i\n", info->num_physical_wave64_vgprs_per_simd);
+ printf(" num_simd_per_compute_unit = %i\n", info->num_simd_per_compute_unit);
+ printf(" min_sgpr_alloc = %i\n", info->min_sgpr_alloc);
+ printf(" max_sgpr_alloc = %i\n", info->max_sgpr_alloc);
+ printf(" sgpr_alloc_granularity = %i\n", info->sgpr_alloc_granularity);
+ printf(" min_wave64_vgpr_alloc = %i\n", info->min_wave64_vgpr_alloc);
+ printf(" max_vgpr_alloc = %i\n", info->max_vgpr_alloc);
+ printf(" wave64_vgpr_alloc_granularity = %i\n", info->wave64_vgpr_alloc_granularity);
printf("Render backend info:\n");
printf(" pa_sc_tile_steering_override = 0x%x\n", info->pa_sc_tile_steering_override);
/* GFX6 */
if (max_waves_per_sh) {
unsigned limit_div16 = DIV_ROUND_UP(max_waves_per_sh, 16);
- compute_resource_limits |= S_00B854_WAVES_PER_SH_SI(limit_div16);
+ compute_resource_limits |= S_00B854_WAVES_PER_SH_GFX6(limit_div16);
}
}
return compute_resource_limits;