void si_init_config(struct radv_physical_device *physical_device,
struct radv_cmd_buffer *cmd_buffer)
{
unsigned num_rb = MIN2(physical_device->rad_info.num_render_backends, 16);
unsigned rb_mask = physical_device->rad_info.enabled_rb_mask;
unsigned raster_config, raster_config_1;
int i;
struct radeon_winsys_cs *cs = cmd_buffer->cs;
radeon_emit(cs, PKT3(PKT3_CONTEXT_CONTROL, 1, 0));
radeon_emit(cs, CONTEXT_CONTROL_LOAD_ENABLE(1));
radeon_emit(cs, CONTEXT_CONTROL_SHADOW_ENABLE(1));
radeon_set_context_reg(cs, R_028A18_VGT_HOS_MAX_TESS_LEVEL, fui(64));
radeon_set_context_reg(cs, R_028A1C_VGT_HOS_MIN_TESS_LEVEL, fui(0));
/* FIXME calculate these values somehow ??? */
radeon_set_context_reg(cs, R_028A54_VGT_GS_PER_ES, SI_GS_PER_ES);
radeon_set_context_reg(cs, R_028A58_VGT_ES_PER_GS, 0x40);
radeon_set_context_reg(cs, R_028A5C_VGT_GS_PER_VS, 0x2);
radeon_set_context_reg(cs, R_028A8C_VGT_PRIMITIVEID_RESET, 0x0);
radeon_set_context_reg(cs, R_028B28_VGT_STRMOUT_DRAW_OPAQUE_OFFSET, 0);
radeon_set_context_reg(cs, R_028B98_VGT_STRMOUT_BUFFER_CONFIG, 0x0);
radeon_set_context_reg(cs, R_028AB8_VGT_VTX_CNT_EN, 0x0);
if (physical_device->rad_info.chip_class < CIK)
radeon_set_config_reg(cs, R_008A14_PA_CL_ENHANCE, S_008A14_NUM_CLIP_SEQ(3) |
S_008A14_CLIP_VTX_REORDER_ENA(1));
radeon_set_context_reg(cs, R_028BD4_PA_SC_CENTROID_PRIORITY_0, 0x76543210);
radeon_set_context_reg(cs, R_028BD8_PA_SC_CENTROID_PRIORITY_1, 0xfedcba98);
radeon_set_context_reg(cs, R_02882C_PA_SU_PRIM_FILTER_CNTL, 0);
for (i = 0; i < 16; i++) {
radeon_set_context_reg(cs, R_0282D0_PA_SC_VPORT_ZMIN_0 + i*8, 0);
radeon_set_context_reg(cs, R_0282D4_PA_SC_VPORT_ZMAX_0 + i*8, fui(1.0));
}
switch (physical_device->rad_info.family) {
case CHIP_TAHITI:
case CHIP_PITCAIRN:
raster_config = 0x2a00126a;
raster_config_1 = 0x00000000;
break;
case CHIP_VERDE:
raster_config = 0x0000124a;
raster_config_1 = 0x00000000;
break;
case CHIP_OLAND:
raster_config = 0x00000082;
raster_config_1 = 0x00000000;
break;
case CHIP_HAINAN:
raster_config = 0x00000000;
raster_config_1 = 0x00000000;
break;
case CHIP_BONAIRE:
raster_config = 0x16000012;
raster_config_1 = 0x00000000;
break;
case CHIP_HAWAII:
raster_config = 0x3a00161a;
raster_config_1 = 0x0000002e;
break;
case CHIP_FIJI:
if (physical_device->rad_info.cik_macrotile_mode_array[0] == 0x000000e8) {
/* old kernels with old tiling config */
raster_config = 0x16000012;
raster_config_1 = 0x0000002a;
} else {
raster_config = 0x3a00161a;
raster_config_1 = 0x0000002e;
}
break;
case CHIP_POLARIS10:
raster_config = 0x16000012;
raster_config_1 = 0x0000002a;
break;
case CHIP_POLARIS11:
raster_config = 0x16000012;
raster_config_1 = 0x00000000;
break;
case CHIP_TONGA:
raster_config = 0x16000012;
raster_config_1 = 0x0000002a;
break;
case CHIP_ICELAND:
if (num_rb == 1)
raster_config = 0x00000000;
else
raster_config = 0x00000002;
raster_config_1 = 0x00000000;
break;
case CHIP_CARRIZO:
raster_config = 0x00000002;
raster_config_1 = 0x00000000;
break;
case CHIP_KAVERI:
/* KV should be 0x00000002, but that causes problems with radeon */
raster_config = 0x00000000; /* 0x00000002 */
raster_config_1 = 0x00000000;
break;
case CHIP_KABINI:
case CHIP_MULLINS:
case CHIP_STONEY:
raster_config = 0x00000000;
raster_config_1 = 0x00000000;
break;
default:
fprintf(stderr,
"radeonsi: Unknown GPU, using 0 for raster_config\n");
raster_config = 0x00000000;
raster_config_1 = 0x00000000;
break;
}
/* Always use the default config when all backends are enabled
* (or when we failed to determine the enabled backends).
*/
if (!rb_mask || util_bitcount(rb_mask) >= num_rb) {
radeon_set_context_reg(cs, R_028350_PA_SC_RASTER_CONFIG,
raster_config);
if (physical_device->rad_info.chip_class >= CIK)
radeon_set_context_reg(cs, R_028354_PA_SC_RASTER_CONFIG_1,
raster_config_1);
} else {
si_write_harvested_raster_configs(physical_device, cs, raster_config, raster_config_1);
}
radeon_set_context_reg(cs, R_028204_PA_SC_WINDOW_SCISSOR_TL, S_028204_WINDOW_OFFSET_DISABLE(1));
radeon_set_context_reg(cs, R_028240_PA_SC_GENERIC_SCISSOR_TL, S_028240_WINDOW_OFFSET_DISABLE(1));
radeon_set_context_reg(cs, R_028244_PA_SC_GENERIC_SCISSOR_BR,
S_028244_BR_X(16384) | S_028244_BR_Y(16384));
radeon_set_context_reg(cs, R_028030_PA_SC_SCREEN_SCISSOR_TL, 0);
radeon_set_context_reg(cs, R_028034_PA_SC_SCREEN_SCISSOR_BR,
S_028034_BR_X(16384) | S_028034_BR_Y(16384));
radeon_set_context_reg(cs, R_02820C_PA_SC_CLIPRECT_RULE, 0xFFFF);
radeon_set_context_reg(cs, R_028230_PA_SC_EDGERULE, 0xAAAAAAAA);
/* PA_SU_HARDWARE_SCREEN_OFFSET must be 0 due to hw bug on SI */
radeon_set_context_reg(cs, R_028234_PA_SU_HARDWARE_SCREEN_OFFSET, 0);
radeon_set_context_reg(cs, R_028820_PA_CL_NANINF_CNTL, 0);
radeon_set_context_reg(cs, R_028BE8_PA_CL_GB_VERT_CLIP_ADJ, fui(1.0));
radeon_set_context_reg(cs, R_028BEC_PA_CL_GB_VERT_DISC_ADJ, fui(1.0));
radeon_set_context_reg(cs, R_028BF0_PA_CL_GB_HORZ_CLIP_ADJ, fui(1.0));
radeon_set_context_reg(cs, R_028BF4_PA_CL_GB_HORZ_DISC_ADJ, fui(1.0));
radeon_set_context_reg(cs, R_028AC0_DB_SRESULTS_COMPARE_STATE0, 0x0);
radeon_set_context_reg(cs, R_028AC4_DB_SRESULTS_COMPARE_STATE1, 0x0);
radeon_set_context_reg(cs, R_028AC8_DB_PRELOAD_CONTROL, 0x0);
radeon_set_context_reg(cs, R_02800C_DB_RENDER_OVERRIDE,
S_02800C_FORCE_HIS_ENABLE0(V_02800C_FORCE_DISABLE) |
S_02800C_FORCE_HIS_ENABLE1(V_02800C_FORCE_DISABLE));
radeon_set_context_reg(cs, R_028400_VGT_MAX_VTX_INDX, ~0);
radeon_set_context_reg(cs, R_028404_VGT_MIN_VTX_INDX, 0);
radeon_set_context_reg(cs, R_028408_VGT_INDX_OFFSET, 0);
if (physical_device->rad_info.chip_class >= CIK) {
radeon_set_sh_reg(cs, R_00B41C_SPI_SHADER_PGM_RSRC3_HS, 0);
radeon_set_sh_reg(cs, R_00B31C_SPI_SHADER_PGM_RSRC3_ES, S_00B31C_CU_EN(0xffff));
radeon_set_sh_reg(cs, R_00B21C_SPI_SHADER_PGM_RSRC3_GS, S_00B21C_CU_EN(0xffff));
if (physical_device->rad_info.num_good_compute_units /
(physical_device->rad_info.max_se * physical_device->rad_info.max_sh_per_se) <= 4) {
/* Too few available compute units per SH. Disallowing
* VS to run on CU0 could hurt us more than late VS
* allocation would help.
*
* LATE_ALLOC_VS = 2 is the highest safe number.
*/
radeon_set_sh_reg(cs, R_00B51C_SPI_SHADER_PGM_RSRC3_LS, S_00B51C_CU_EN(0xffff));
radeon_set_sh_reg(cs, R_00B118_SPI_SHADER_PGM_RSRC3_VS, S_00B118_CU_EN(0xffff));
radeon_set_sh_reg(cs, R_00B11C_SPI_SHADER_LATE_ALLOC_VS, S_00B11C_LIMIT(2));
} else {
/* Set LATE_ALLOC_VS == 31. It should be less than
* the number of scratch waves. Limitations:
* - VS can't execute on CU0.
* - If HS writes outputs to LDS, LS can't execute on CU0.
*/
radeon_set_sh_reg(cs, R_00B51C_SPI_SHADER_PGM_RSRC3_LS, S_00B51C_CU_EN(0xfffe));
radeon_set_sh_reg(cs, R_00B118_SPI_SHADER_PGM_RSRC3_VS, S_00B118_CU_EN(0xfffe));
radeon_set_sh_reg(cs, R_00B11C_SPI_SHADER_LATE_ALLOC_VS, S_00B11C_LIMIT(31));
}
radeon_set_sh_reg(cs, R_00B01C_SPI_SHADER_PGM_RSRC3_PS, S_00B01C_CU_EN(0xffff));
}
if (physical_device->rad_info.chip_class >= VI) {
radeon_set_context_reg(cs, R_028424_CB_DCC_CONTROL,
S_028424_OVERWRITE_COMBINER_MRT_SHARING_DISABLE(1) |
S_028424_OVERWRITE_COMBINER_WATERMARK(4));
radeon_set_context_reg(cs, R_028C58_VGT_VERTEX_REUSE_BLOCK_CNTL, 30);
radeon_set_context_reg(cs, R_028C5C_VGT_OUT_DEALLOC_CNTL, 32);
radeon_set_context_reg(cs, R_028B50_VGT_TESS_DISTRIBUTION,
S_028B50_ACCUM_ISOLINE(32) |
S_028B50_ACCUM_TRI(11) |
S_028B50_ACCUM_QUAD(11) |
S_028B50_DONUT_SPLIT(16));
} else {
radeon_set_context_reg(cs, R_028C58_VGT_VERTEX_REUSE_BLOCK_CNTL, 14);
radeon_set_context_reg(cs, R_028C5C_VGT_OUT_DEALLOC_CNTL, 16);
}
if (physical_device->rad_info.family == CHIP_STONEY)
radeon_set_context_reg(cs, R_028C40_PA_SC_SHADER_CONTROL, 0);
si_init_compute(physical_device, cs);
}
static void si_pipe_shader_ps(struct pipe_context *ctx, struct si_pipe_shader *shader)
{
struct r600_context *rctx = (struct r600_context *)ctx;
struct si_pm4_state *pm4;
unsigned i, exports_ps, num_cout, spi_ps_in_control, db_shader_control;
unsigned num_sgprs, num_user_sgprs;
unsigned spi_baryc_cntl = 0, spi_ps_input_ena, spi_shader_z_format;
uint64_t va;
si_pm4_delete_state(rctx, ps, shader->pm4);
pm4 = shader->pm4 = si_pm4_alloc_state(rctx);
if (pm4 == NULL)
return;
db_shader_control = S_02880C_Z_ORDER(V_02880C_EARLY_Z_THEN_LATE_Z) |
S_02880C_ALPHA_TO_MASK_DISABLE(rctx->fb_cb0_is_integer);
for (i = 0; i < shader->shader.ninput; i++) {
switch (shader->shader.input[i].name) {
case TGSI_SEMANTIC_POSITION:
if (shader->shader.input[i].centroid) {
/* SPI_BARYC_CNTL.POS_FLOAT_LOCATION
* Possible vaules:
* 0 -> Position = pixel center (default)
* 1 -> Position = pixel centroid
* 2 -> Position = iterated sample number XXX:
* What does this mean?
*/
spi_baryc_cntl |= S_0286E0_POS_FLOAT_LOCATION(1);
}
/* Fall through */
case TGSI_SEMANTIC_FACE:
continue;
}
}
for (i = 0; i < shader->shader.noutput; i++) {
if (shader->shader.output[i].name == TGSI_SEMANTIC_POSITION)
db_shader_control |= S_02880C_Z_EXPORT_ENABLE(1);
if (shader->shader.output[i].name == TGSI_SEMANTIC_STENCIL)
db_shader_control |= S_02880C_STENCIL_TEST_VAL_EXPORT_ENABLE(1);
}
if (shader->shader.uses_kill || shader->key.ps.alpha_func != PIPE_FUNC_ALWAYS)
db_shader_control |= S_02880C_KILL_ENABLE(1);
exports_ps = 0;
num_cout = 0;
for (i = 0; i < shader->shader.noutput; i++) {
if (shader->shader.output[i].name == TGSI_SEMANTIC_POSITION ||
shader->shader.output[i].name == TGSI_SEMANTIC_STENCIL)
exports_ps |= 1;
else if (shader->shader.output[i].name == TGSI_SEMANTIC_COLOR) {
if (shader->shader.fs_write_all)
num_cout = shader->shader.nr_cbufs;
else
num_cout++;
}
}
if (!exports_ps) {
/* always at least export 1 component per pixel */
exports_ps = 2;
}
spi_ps_in_control = S_0286D8_NUM_INTERP(shader->shader.ninterp) |
S_0286D8_BC_OPTIMIZE_DISABLE(1);
si_pm4_set_reg(pm4, R_0286E0_SPI_BARYC_CNTL, spi_baryc_cntl);
spi_ps_input_ena = shader->spi_ps_input_ena;
/* we need to enable at least one of them, otherwise we hang the GPU */
assert(G_0286CC_PERSP_SAMPLE_ENA(spi_ps_input_ena) ||
G_0286CC_PERSP_CENTER_ENA(spi_ps_input_ena) ||
G_0286CC_PERSP_CENTROID_ENA(spi_ps_input_ena) ||
G_0286CC_PERSP_PULL_MODEL_ENA(spi_ps_input_ena) ||
G_0286CC_LINEAR_SAMPLE_ENA(spi_ps_input_ena) ||
G_0286CC_LINEAR_CENTER_ENA(spi_ps_input_ena) ||
G_0286CC_LINEAR_CENTROID_ENA(spi_ps_input_ena) ||
G_0286CC_LINE_STIPPLE_TEX_ENA(spi_ps_input_ena));
si_pm4_set_reg(pm4, R_0286CC_SPI_PS_INPUT_ENA, spi_ps_input_ena);
si_pm4_set_reg(pm4, R_0286D0_SPI_PS_INPUT_ADDR, spi_ps_input_ena);
si_pm4_set_reg(pm4, R_0286D8_SPI_PS_IN_CONTROL, spi_ps_in_control);
if (G_02880C_STENCIL_TEST_VAL_EXPORT_ENABLE(db_shader_control))
spi_shader_z_format = V_028710_SPI_SHADER_32_GR;
else if (G_02880C_Z_EXPORT_ENABLE(db_shader_control))
spi_shader_z_format = V_028710_SPI_SHADER_32_R;
else
spi_shader_z_format = 0;
si_pm4_set_reg(pm4, R_028710_SPI_SHADER_Z_FORMAT, spi_shader_z_format);
si_pm4_set_reg(pm4, R_028714_SPI_SHADER_COL_FORMAT,
shader->spi_shader_col_format);
si_pm4_set_reg(pm4, R_02823C_CB_SHADER_MASK, shader->cb_shader_mask);
va = r600_resource_va(ctx->screen, (void *)shader->bo);
si_pm4_add_bo(pm4, shader->bo, RADEON_USAGE_READ);
si_pm4_set_reg(pm4, R_00B020_SPI_SHADER_PGM_LO_PS, va >> 8);
si_pm4_set_reg(pm4, R_00B024_SPI_SHADER_PGM_HI_PS, va >> 40);
num_user_sgprs = SI_PS_NUM_USER_SGPR;
num_sgprs = shader->num_sgprs;
/* One SGPR after user SGPRs is pre-loaded with {prim_mask, lds_offset} */
if ((num_user_sgprs + 1) > num_sgprs) {
/* Last 2 reserved SGPRs are used for VCC */
num_sgprs = num_user_sgprs + 1 + 2;
}
assert(num_sgprs <= 104);
si_pm4_set_reg(pm4, R_00B028_SPI_SHADER_PGM_RSRC1_PS,
S_00B028_VGPRS((shader->num_vgprs - 1) / 4) |
S_00B028_SGPRS((num_sgprs - 1) / 8));
si_pm4_set_reg(pm4, R_00B02C_SPI_SHADER_PGM_RSRC2_PS,
S_00B02C_EXTRA_LDS_SIZE(shader->lds_size) |
S_00B02C_USER_SGPR(num_user_sgprs));
if (rctx->b.chip_class >= CIK) {
si_pm4_set_reg(pm4, R_00B01C_SPI_SHADER_PGM_RSRC3_PS,
S_00B01C_CU_EN(0xffff));
}
si_pm4_set_reg(pm4, R_02880C_DB_SHADER_CONTROL, db_shader_control);
shader->cb0_is_integer = rctx->fb_cb0_is_integer;
shader->sprite_coord_enable = rctx->sprite_coord_enable;
si_pm4_bind_state(rctx, ps, shader->pm4);
rctx->b.flags |= R600_CONTEXT_INV_SHADER_CACHE;
}
