* Christian König <christian.koenig@amd.com>
*/
-#include <byteswap.h>
-
#include "util/u_memory.h"
#include "util/u_framebuffer.h"
#include "util/u_blitter.h"
#include "util/u_upload_mgr.h"
#include "util/u_format_s3tc.h"
#include "tgsi/tgsi_parse.h"
-#include "radeonsi_pipe.h"
-#include "radeonsi_shader.h"
+#include "tgsi/tgsi_scan.h"
+#include "si_pipe.h"
+#include "si_shader.h"
#include "si_state.h"
+#include "../radeon/r600_cs.h"
#include "sid.h"
+static uint32_t cik_num_banks(struct si_screen *rscreen, unsigned bpe, unsigned tile_split)
+{
+ if (rscreen->b.info.cik_macrotile_mode_array_valid) {
+ unsigned index, tileb;
+
+ tileb = 8 * 8 * bpe;
+ tileb = MIN2(tile_split, tileb);
+
+ for (index = 0; tileb > 64; index++) {
+ tileb >>= 1;
+ }
+
+ assert(index < 16);
+
+ return (rscreen->b.info.cik_macrotile_mode_array[index] >> 6) & 0x3;
+ }
+
+ /* The old way. */
+ switch (rscreen->b.tiling_info.num_banks) {
+ case 2:
+ return V_02803C_ADDR_SURF_2_BANK;
+ case 4:
+ return V_02803C_ADDR_SURF_4_BANK;
+ case 8:
+ default:
+ return V_02803C_ADDR_SURF_8_BANK;
+ case 16:
+ return V_02803C_ADDR_SURF_16_BANK;
+ }
+}
+
+static unsigned cik_tile_split(unsigned tile_split)
+{
+ switch (tile_split) {
+ case 64:
+ tile_split = V_028040_ADDR_SURF_TILE_SPLIT_64B;
+ break;
+ case 128:
+ tile_split = V_028040_ADDR_SURF_TILE_SPLIT_128B;
+ break;
+ case 256:
+ tile_split = V_028040_ADDR_SURF_TILE_SPLIT_256B;
+ break;
+ case 512:
+ tile_split = V_028040_ADDR_SURF_TILE_SPLIT_512B;
+ break;
+ default:
+ case 1024:
+ tile_split = V_028040_ADDR_SURF_TILE_SPLIT_1KB;
+ break;
+ case 2048:
+ tile_split = V_028040_ADDR_SURF_TILE_SPLIT_2KB;
+ break;
+ case 4096:
+ tile_split = V_028040_ADDR_SURF_TILE_SPLIT_4KB;
+ break;
+ }
+ return tile_split;
+}
+
+static unsigned cik_macro_tile_aspect(unsigned macro_tile_aspect)
+{
+ switch (macro_tile_aspect) {
+ default:
+ case 1:
+ macro_tile_aspect = V_02803C_ADDR_SURF_MACRO_ASPECT_1;
+ break;
+ case 2:
+ macro_tile_aspect = V_02803C_ADDR_SURF_MACRO_ASPECT_2;
+ break;
+ case 4:
+ macro_tile_aspect = V_02803C_ADDR_SURF_MACRO_ASPECT_4;
+ break;
+ case 8:
+ macro_tile_aspect = V_02803C_ADDR_SURF_MACRO_ASPECT_8;
+ break;
+ }
+ return macro_tile_aspect;
+}
+
+static unsigned cik_bank_wh(unsigned bankwh)
+{
+ switch (bankwh) {
+ default:
+ case 1:
+ bankwh = V_02803C_ADDR_SURF_BANK_WIDTH_1;
+ break;
+ case 2:
+ bankwh = V_02803C_ADDR_SURF_BANK_WIDTH_2;
+ break;
+ case 4:
+ bankwh = V_02803C_ADDR_SURF_BANK_WIDTH_4;
+ break;
+ case 8:
+ bankwh = V_02803C_ADDR_SURF_BANK_WIDTH_8;
+ break;
+ }
+ return bankwh;
+}
+
+static unsigned cik_db_pipe_config(struct si_screen *rscreen, unsigned tile_mode)
+{
+ if (rscreen->b.info.si_tile_mode_array_valid) {
+ uint32_t gb_tile_mode = rscreen->b.info.si_tile_mode_array[tile_mode];
+
+ return G_009910_PIPE_CONFIG(gb_tile_mode);
+ }
+
+ /* This is probably broken for a lot of chips, but it's only used
+ * if the kernel cannot return the tile mode array for CIK. */
+ switch (rscreen->b.info.r600_num_tile_pipes) {
+ case 16:
+ return V_02803C_X_ADDR_SURF_P16_32X32_16X16;
+ case 8:
+ return V_02803C_X_ADDR_SURF_P8_32X32_16X16;
+ case 4:
+ default:
+ if (rscreen->b.info.r600_num_backends == 4)
+ return V_02803C_X_ADDR_SURF_P4_16X16;
+ else
+ return V_02803C_X_ADDR_SURF_P4_8X16;
+ case 2:
+ return V_02803C_ADDR_SURF_P2;
+ }
+}
+
/*
* inferred framebuffer and blender state
*/
-static void si_update_fb_blend_state(struct r600_context *rctx)
+static void si_update_fb_blend_state(struct si_context *rctx)
{
struct si_pm4_state *pm4;
struct si_state_blend *blend = rctx->queued.named.blend;
if (blend == NULL)
return;
- pm4 = CALLOC_STRUCT(si_pm4_state);
+ pm4 = si_pm4_alloc_state(rctx);
if (pm4 == NULL)
return;
return 0;
}
-static void *si_create_blend_state(struct pipe_context *ctx,
- const struct pipe_blend_state *state)
+static void *si_create_blend_state_mode(struct pipe_context *ctx,
+ const struct pipe_blend_state *state,
+ unsigned mode)
{
struct si_state_blend *blend = CALLOC_STRUCT(si_state_blend);
struct si_pm4_state *pm4 = &blend->pm4;
- uint32_t color_control;
+ uint32_t color_control = 0;
if (blend == NULL)
return NULL;
- color_control = S_028808_MODE(V_028808_CB_NORMAL);
+ blend->alpha_to_one = state->alpha_to_one;
+
if (state->logicop_enable) {
color_control |= S_028808_ROP3(state->logicop_func | (state->logicop_func << 4));
} else {
color_control |= S_028808_ROP3(0xcc);
}
- si_pm4_set_reg(pm4, R_028808_CB_COLOR_CONTROL, color_control);
- si_pm4_set_reg(pm4, R_028C38_PA_SC_AA_MASK_X0Y0_X1Y0, ~0);
- si_pm4_set_reg(pm4, R_028C3C_PA_SC_AA_MASK_X0Y1_X1Y1, ~0);
+ si_pm4_set_reg(pm4, R_028B70_DB_ALPHA_TO_MASK,
+ S_028B70_ALPHA_TO_MASK_ENABLE(state->alpha_to_coverage) |
+ S_028B70_ALPHA_TO_MASK_OFFSET0(2) |
+ S_028B70_ALPHA_TO_MASK_OFFSET1(2) |
+ S_028B70_ALPHA_TO_MASK_OFFSET2(2) |
+ S_028B70_ALPHA_TO_MASK_OFFSET3(2));
blend->cb_target_mask = 0;
for (int i = 0; i < 8; i++) {
si_pm4_set_reg(pm4, R_028780_CB_BLEND0_CONTROL + i * 4, blend_cntl);
}
+ if (blend->cb_target_mask) {
+ color_control |= S_028808_MODE(mode);
+ } else {
+ color_control |= S_028808_MODE(V_028808_CB_DISABLE);
+ }
+ si_pm4_set_reg(pm4, R_028808_CB_COLOR_CONTROL, color_control);
+
return blend;
}
+static void *si_create_blend_state(struct pipe_context *ctx,
+ const struct pipe_blend_state *state)
+{
+ return si_create_blend_state_mode(ctx, state, V_028808_CB_NORMAL);
+}
+
static void si_bind_blend_state(struct pipe_context *ctx, void *state)
{
- struct r600_context *rctx = (struct r600_context *)ctx;
+ struct si_context *rctx = (struct si_context *)ctx;
si_pm4_bind_state(rctx, blend, (struct si_state_blend *)state);
si_update_fb_blend_state(rctx);
}
static void si_delete_blend_state(struct pipe_context *ctx, void *state)
{
- struct r600_context *rctx = (struct r600_context *)ctx;
+ struct si_context *rctx = (struct si_context *)ctx;
si_pm4_delete_state(rctx, blend, (struct si_state_blend *)state);
}
static void si_set_blend_color(struct pipe_context *ctx,
const struct pipe_blend_color *state)
{
- struct r600_context *rctx = (struct r600_context *)ctx;
- struct si_pm4_state *pm4 = CALLOC_STRUCT(si_pm4_state);
+ struct si_context *rctx = (struct si_context *)ctx;
+ struct si_pm4_state *pm4 = si_pm4_alloc_state(rctx);
if (pm4 == NULL)
return;
static void si_set_clip_state(struct pipe_context *ctx,
const struct pipe_clip_state *state)
{
- struct r600_context *rctx = (struct r600_context *)ctx;
- struct si_pm4_state *pm4 = CALLOC_STRUCT(si_pm4_state);
+ struct si_context *rctx = (struct si_context *)ctx;
+ struct si_pm4_state *pm4 = si_pm4_alloc_state(rctx);
struct pipe_constant_buffer cb;
if (pm4 == NULL)
cb.user_buffer = state->ucp;
cb.buffer_offset = 0;
cb.buffer_size = 4*4*8;
- ctx->set_constant_buffer(ctx, PIPE_SHADER_VERTEX, 1, &cb);
+ ctx->set_constant_buffer(ctx, PIPE_SHADER_VERTEX, NUM_PIPE_CONST_BUFFERS, &cb);
pipe_resource_reference(&cb.buffer, NULL);
si_pm4_set_state(rctx, clip, pm4);
}
-static void si_set_scissor_state(struct pipe_context *ctx,
- const struct pipe_scissor_state *state)
+static void si_set_scissor_states(struct pipe_context *ctx,
+ unsigned start_slot,
+ unsigned num_scissors,
+ const struct pipe_scissor_state *state)
{
- struct r600_context *rctx = (struct r600_context *)ctx;
- struct si_pm4_state *pm4 = CALLOC_STRUCT(si_pm4_state);
+ struct si_context *rctx = (struct si_context *)ctx;
+ struct si_pm4_state *pm4 = si_pm4_alloc_state(rctx);
uint32_t tl, br;
if (pm4 == NULL)
si_pm4_set_state(rctx, scissor, pm4);
}
-static void si_set_viewport_state(struct pipe_context *ctx,
- const struct pipe_viewport_state *state)
+static void si_set_viewport_states(struct pipe_context *ctx,
+ unsigned start_slot,
+ unsigned num_viewports,
+ const struct pipe_viewport_state *state)
{
- struct r600_context *rctx = (struct r600_context *)ctx;
+ struct si_context *rctx = (struct si_context *)ctx;
struct si_state_viewport *viewport = CALLOC_STRUCT(si_state_viewport);
struct si_pm4_state *pm4 = &viewport->pm4;
return;
viewport->viewport = *state;
- si_pm4_set_reg(pm4, R_0282D0_PA_SC_VPORT_ZMIN_0, 0x00000000);
- si_pm4_set_reg(pm4, R_0282D4_PA_SC_VPORT_ZMAX_0, 0x3F800000);
si_pm4_set_reg(pm4, R_02843C_PA_CL_VPORT_XSCALE_0, fui(state->scale[0]));
si_pm4_set_reg(pm4, R_028440_PA_CL_VPORT_XOFFSET_0, fui(state->translate[0]));
si_pm4_set_reg(pm4, R_028444_PA_CL_VPORT_YSCALE_0, fui(state->scale[1]));
si_pm4_set_reg(pm4, R_028448_PA_CL_VPORT_YOFFSET_0, fui(state->translate[1]));
si_pm4_set_reg(pm4, R_02844C_PA_CL_VPORT_ZSCALE_0, fui(state->scale[2]));
si_pm4_set_reg(pm4, R_028450_PA_CL_VPORT_ZOFFSET_0, fui(state->translate[2]));
- si_pm4_set_reg(pm4, R_028818_PA_CL_VTE_CNTL, 0x0000043F);
si_pm4_set_state(rctx, viewport, viewport);
}
/*
* inferred state between framebuffer and rasterizer
*/
-static void si_update_fb_rs_state(struct r600_context *rctx)
+static void si_update_fb_rs_state(struct si_context *rctx)
{
struct si_state_rasterizer *rs = rctx->queued.named.rasterizer;
struct si_pm4_state *pm4;
return;
}
- pm4 = CALLOC_STRUCT(si_pm4_state);
+ pm4 = si_pm4_alloc_state(rctx);
+
+ if (pm4 == NULL)
+ return;
+
/* FIXME some of those reg can be computed with cso */
offset_db_fmt_cntl |= S_028B78_POLY_OFFSET_NEG_NUM_DB_BITS(depth);
si_pm4_set_reg(pm4, R_028B80_PA_SU_POLY_OFFSET_FRONT_SCALE,
}
rs->two_side = state->light_twoside;
+ rs->multisample_enable = state->multisample;
rs->clip_plane_enable = state->clip_plane_enable;
+ rs->line_stipple_enable = state->line_stipple_enable;
polygon_dual_mode = (state->fill_front != PIPE_POLYGON_MODE_FILL ||
state->fill_back != PIPE_POLYGON_MODE_FILL);
S_028810_PS_UCP_MODE(3) |
S_028810_ZCLIP_NEAR_DISABLE(!state->depth_clip) |
S_028810_ZCLIP_FAR_DISABLE(!state->depth_clip) |
+ S_028810_DX_RASTERIZATION_KILL(state->rasterizer_discard) |
S_028810_DX_LINEAR_ATTR_CLIP_ENA(1);
clip_rule = state->scissor ? 0xAAAA : 0xFFFF;
}
si_pm4_set_reg(pm4, R_0286D4_SPI_INTERP_CONTROL_0, tmp);
- si_pm4_set_reg(pm4, R_028820_PA_CL_NANINF_CNTL, 0x00000000);
/* point size 12.4 fixed point */
tmp = (unsigned)(state->point_size * 8.0);
si_pm4_set_reg(pm4, R_028A00_PA_SU_POINT_SIZE, S_028A00_HEIGHT(tmp) | S_028A00_WIDTH(tmp));
tmp = (unsigned)state->line_width * 8;
si_pm4_set_reg(pm4, R_028A08_PA_SU_LINE_CNTL, S_028A08_WIDTH(tmp));
si_pm4_set_reg(pm4, R_028A48_PA_SC_MODE_CNTL_0,
- S_028A48_LINE_STIPPLE_ENABLE(state->line_stipple_enable));
+ S_028A48_LINE_STIPPLE_ENABLE(state->line_stipple_enable) |
+ S_028A48_MSAA_ENABLE(state->multisample));
- si_pm4_set_reg(pm4, R_028BDC_PA_SC_LINE_CNTL, 0x00000400);
si_pm4_set_reg(pm4, R_028BE4_PA_SU_VTX_CNTL,
- S_028BE4_PIX_CENTER(state->half_pixel_center));
- si_pm4_set_reg(pm4, R_028BE8_PA_CL_GB_VERT_CLIP_ADJ, 0x3F800000);
- si_pm4_set_reg(pm4, R_028BEC_PA_CL_GB_VERT_DISC_ADJ, 0x3F800000);
- si_pm4_set_reg(pm4, R_028BF0_PA_CL_GB_HORZ_CLIP_ADJ, 0x3F800000);
- si_pm4_set_reg(pm4, R_028BF4_PA_CL_GB_HORZ_DISC_ADJ, 0x3F800000);
+ S_028BE4_PIX_CENTER(state->half_pixel_center) |
+ S_028BE4_QUANT_MODE(V_028BE4_X_16_8_FIXED_POINT_1_256TH));
si_pm4_set_reg(pm4, R_028B7C_PA_SU_POLY_OFFSET_CLAMP, fui(state->offset_clamp));
si_pm4_set_reg(pm4, R_02820C_PA_SC_CLIPRECT_RULE, clip_rule);
static void si_bind_rs_state(struct pipe_context *ctx, void *state)
{
- struct r600_context *rctx = (struct r600_context *)ctx;
+ struct si_context *rctx = (struct si_context *)ctx;
struct si_state_rasterizer *rs = (struct si_state_rasterizer *)state;
if (state == NULL)
static void si_delete_rs_state(struct pipe_context *ctx, void *state)
{
- struct r600_context *rctx = (struct r600_context *)ctx;
+ struct si_context *rctx = (struct si_context *)ctx;
si_pm4_delete_state(rctx, rasterizer, (struct si_state_rasterizer *)state);
}
/*
* infeered state between dsa and stencil ref
*/
-static void si_update_dsa_stencil_ref(struct r600_context *rctx)
+static void si_update_dsa_stencil_ref(struct si_context *rctx)
{
- struct si_pm4_state *pm4 = CALLOC_STRUCT(si_pm4_state);
+ struct si_pm4_state *pm4 = si_pm4_alloc_state(rctx);
struct pipe_stencil_ref *ref = &rctx->stencil_ref;
struct si_state_dsa *dsa = rctx->queued.named.dsa;
static void si_set_pipe_stencil_ref(struct pipe_context *ctx,
const struct pipe_stencil_ref *state)
{
- struct r600_context *rctx = (struct r600_context *)ctx;
+ struct si_context *rctx = (struct si_context *)ctx;
rctx->stencil_ref = *state;
si_update_dsa_stencil_ref(rctx);
}
struct si_state_dsa *dsa = CALLOC_STRUCT(si_state_dsa);
struct si_pm4_state *pm4 = &dsa->pm4;
unsigned db_depth_control;
- unsigned db_render_override, db_render_control;
+ unsigned db_render_control;
uint32_t db_stencil_control = 0;
if (dsa == NULL) {
if (state->alpha.enabled) {
dsa->alpha_func = state->alpha.func;
dsa->alpha_ref = state->alpha.ref_value;
+
+ si_pm4_set_reg(pm4, R_00B030_SPI_SHADER_USER_DATA_PS_0 +
+ SI_SGPR_ALPHA_REF * 4, fui(dsa->alpha_ref));
} else {
dsa->alpha_func = PIPE_FUNC_ALWAYS;
}
/* misc */
db_render_control = 0;
- db_render_override = S_02800C_FORCE_HIZ_ENABLE(V_02800C_FORCE_DISABLE) |
- S_02800C_FORCE_HIS_ENABLE0(V_02800C_FORCE_DISABLE) |
- S_02800C_FORCE_HIS_ENABLE1(V_02800C_FORCE_DISABLE);
- /* TODO db_render_override depends on query */
- si_pm4_set_reg(pm4, R_028020_DB_DEPTH_BOUNDS_MIN, 0x00000000);
- si_pm4_set_reg(pm4, R_028024_DB_DEPTH_BOUNDS_MAX, 0x00000000);
- si_pm4_set_reg(pm4, R_028028_DB_STENCIL_CLEAR, 0x00000000);
- si_pm4_set_reg(pm4, R_02802C_DB_DEPTH_CLEAR, 0x3F800000);
- //si_pm4_set_reg(pm4, R_028410_SX_ALPHA_TEST_CONTROL, alpha_test_control);
si_pm4_set_reg(pm4, R_028800_DB_DEPTH_CONTROL, db_depth_control);
si_pm4_set_reg(pm4, R_028000_DB_RENDER_CONTROL, db_render_control);
- si_pm4_set_reg(pm4, R_02800C_DB_RENDER_OVERRIDE, db_render_override);
si_pm4_set_reg(pm4, R_02842C_DB_STENCIL_CONTROL, db_stencil_control);
- si_pm4_set_reg(pm4, R_028AC0_DB_SRESULTS_COMPARE_STATE0, 0x0);
- si_pm4_set_reg(pm4, R_028AC4_DB_SRESULTS_COMPARE_STATE1, 0x0);
- si_pm4_set_reg(pm4, R_028AC8_DB_PRELOAD_CONTROL, 0x0);
- si_pm4_set_reg(pm4, R_028B70_DB_ALPHA_TO_MASK, 0x0000AA00);
- dsa->db_render_override = db_render_override;
return dsa;
}
static void si_bind_dsa_state(struct pipe_context *ctx, void *state)
{
- struct r600_context *rctx = (struct r600_context *)ctx;
+ struct si_context *rctx = (struct si_context *)ctx;
struct si_state_dsa *dsa = state;
if (state == NULL)
static void si_delete_dsa_state(struct pipe_context *ctx, void *state)
{
- struct r600_context *rctx = (struct r600_context *)ctx;
+ struct si_context *rctx = (struct si_context *)ctx;
si_pm4_delete_state(rctx, dsa, (struct si_state_dsa *)state);
}
-static void *si_create_db_flush_dsa(struct r600_context *rctx, bool copy_depth,
- bool copy_stencil)
+static void *si_create_db_flush_dsa(struct si_context *rctx, bool copy_depth,
+ bool copy_stencil, int sample)
{
struct pipe_depth_stencil_alpha_state dsa;
struct si_state_dsa *state;
memset(&dsa, 0, sizeof(dsa));
- state = rctx->context.create_depth_stencil_alpha_state(&rctx->context, &dsa);
+ state = rctx->b.b.create_depth_stencil_alpha_state(&rctx->b.b, &dsa);
if (copy_depth || copy_stencil) {
si_pm4_set_reg(&state->pm4, R_028000_DB_RENDER_CONTROL,
S_028000_DEPTH_COPY(copy_depth) |
S_028000_STENCIL_COPY(copy_stencil) |
- S_028000_COPY_CENTROID(1));
+ S_028000_COPY_CENTROID(1) |
+ S_028000_COPY_SAMPLE(sample));
} else {
si_pm4_set_reg(&state->pm4, R_028000_DB_RENDER_CONTROL,
S_028000_DEPTH_COMPRESS_DISABLE(1) |
S_028000_STENCIL_COMPRESS_DISABLE(1));
- si_pm4_set_reg(&state->pm4, R_02800C_DB_RENDER_OVERRIDE,
- S_02800C_FORCE_HIZ_ENABLE(V_02800C_FORCE_DISABLE) |
- S_02800C_FORCE_HIS_ENABLE0(V_02800C_FORCE_DISABLE) |
- S_02800C_FORCE_HIS_ENABLE1(V_02800C_FORCE_DISABLE) |
- S_02800C_DISABLE_TILE_RATE_TILES(1));
}
return state;
*/
static uint32_t si_translate_colorformat(enum pipe_format format)
{
- switch (format) {
- /* 8-bit buffers. */
- case PIPE_FORMAT_A8_UNORM:
- case PIPE_FORMAT_A8_SNORM:
- case PIPE_FORMAT_A8_UINT:
- case PIPE_FORMAT_A8_SINT:
- case PIPE_FORMAT_I8_UNORM:
- case PIPE_FORMAT_I8_SNORM:
- case PIPE_FORMAT_I8_UINT:
- case PIPE_FORMAT_I8_SINT:
- case PIPE_FORMAT_L8_UNORM:
- case PIPE_FORMAT_L8_SNORM:
- case PIPE_FORMAT_L8_UINT:
- case PIPE_FORMAT_L8_SINT:
- case PIPE_FORMAT_L8_SRGB:
- case PIPE_FORMAT_R8_UNORM:
- case PIPE_FORMAT_R8_SNORM:
- case PIPE_FORMAT_R8_UINT:
- case PIPE_FORMAT_R8_SINT:
- return V_028C70_COLOR_8;
-
- /* 16-bit buffers. */
- case PIPE_FORMAT_B5G6R5_UNORM:
- return V_028C70_COLOR_5_6_5;
-
- case PIPE_FORMAT_B5G5R5A1_UNORM:
- case PIPE_FORMAT_B5G5R5X1_UNORM:
- return V_028C70_COLOR_1_5_5_5;
-
- case PIPE_FORMAT_B4G4R4A4_UNORM:
- case PIPE_FORMAT_B4G4R4X4_UNORM:
- return V_028C70_COLOR_4_4_4_4;
-
- case PIPE_FORMAT_L8A8_UNORM:
- case PIPE_FORMAT_L8A8_SNORM:
- case PIPE_FORMAT_L8A8_UINT:
- case PIPE_FORMAT_L8A8_SINT:
- case PIPE_FORMAT_R8G8_SNORM:
- case PIPE_FORMAT_R8G8_UNORM:
- case PIPE_FORMAT_R8G8_UINT:
- case PIPE_FORMAT_R8G8_SINT:
- return V_028C70_COLOR_8_8;
+ const struct util_format_description *desc = util_format_description(format);
- case PIPE_FORMAT_Z16_UNORM:
- case PIPE_FORMAT_R16_UNORM:
- case PIPE_FORMAT_R16_SNORM:
- case PIPE_FORMAT_R16_UINT:
- case PIPE_FORMAT_R16_SINT:
- case PIPE_FORMAT_R16_FLOAT:
- case PIPE_FORMAT_L16_UNORM:
- case PIPE_FORMAT_L16_SNORM:
- case PIPE_FORMAT_L16_FLOAT:
- case PIPE_FORMAT_I16_UNORM:
- case PIPE_FORMAT_I16_SNORM:
- case PIPE_FORMAT_I16_FLOAT:
- case PIPE_FORMAT_A16_UNORM:
- case PIPE_FORMAT_A16_SNORM:
- case PIPE_FORMAT_A16_FLOAT:
- return V_028C70_COLOR_16;
-
- /* 32-bit buffers. */
- case PIPE_FORMAT_A8B8G8R8_SRGB:
- case PIPE_FORMAT_A8B8G8R8_UNORM:
- case PIPE_FORMAT_A8R8G8B8_UNORM:
- case PIPE_FORMAT_B8G8R8A8_SRGB:
- case PIPE_FORMAT_B8G8R8A8_UNORM:
- case PIPE_FORMAT_B8G8R8X8_UNORM:
- case PIPE_FORMAT_R8G8B8A8_SNORM:
- case PIPE_FORMAT_R8G8B8A8_UNORM:
- case PIPE_FORMAT_R8G8B8X8_UNORM:
- case PIPE_FORMAT_R8G8B8X8_SNORM:
- case PIPE_FORMAT_R8G8B8X8_SRGB:
- case PIPE_FORMAT_R8G8B8X8_UINT:
- case PIPE_FORMAT_R8G8B8X8_SINT:
- case PIPE_FORMAT_R8SG8SB8UX8U_NORM:
- case PIPE_FORMAT_X8B8G8R8_UNORM:
- case PIPE_FORMAT_X8R8G8B8_UNORM:
- case PIPE_FORMAT_R8G8B8A8_SSCALED:
- case PIPE_FORMAT_R8G8B8A8_USCALED:
- case PIPE_FORMAT_R8G8B8A8_SINT:
- case PIPE_FORMAT_R8G8B8A8_UINT:
- return V_028C70_COLOR_8_8_8_8;
-
- case PIPE_FORMAT_R10G10B10A2_UNORM:
- case PIPE_FORMAT_R10G10B10X2_SNORM:
- case PIPE_FORMAT_B10G10R10A2_UNORM:
- case PIPE_FORMAT_B10G10R10A2_UINT:
- case PIPE_FORMAT_B10G10R10X2_UNORM:
- case PIPE_FORMAT_R10SG10SB10SA2U_NORM:
- return V_028C70_COLOR_2_10_10_10;
-
- case PIPE_FORMAT_Z24X8_UNORM:
- case PIPE_FORMAT_Z24_UNORM_S8_UINT:
- return V_028C70_COLOR_8_24;
-
- case PIPE_FORMAT_S8X24_UINT:
- case PIPE_FORMAT_X8Z24_UNORM:
- case PIPE_FORMAT_S8_UINT_Z24_UNORM:
- return V_028C70_COLOR_24_8;
+#define HAS_SIZE(x,y,z,w) \
+ (desc->channel[0].size == (x) && desc->channel[1].size == (y) && \
+ desc->channel[2].size == (z) && desc->channel[3].size == (w))
- case PIPE_FORMAT_Z32_FLOAT_S8X24_UINT:
- return V_028C70_COLOR_X24_8_32_FLOAT;
-
- case PIPE_FORMAT_I32_FLOAT:
- case PIPE_FORMAT_L32_FLOAT:
- case PIPE_FORMAT_R32_FLOAT:
- case PIPE_FORMAT_A32_FLOAT:
- case PIPE_FORMAT_Z32_FLOAT:
- return V_028C70_COLOR_32;
-
- case PIPE_FORMAT_L16A16_UNORM:
- case PIPE_FORMAT_L16A16_SNORM:
- case PIPE_FORMAT_L16A16_FLOAT:
- case PIPE_FORMAT_R16G16_SSCALED:
- case PIPE_FORMAT_R16G16_UNORM:
- case PIPE_FORMAT_R16G16_SNORM:
- case PIPE_FORMAT_R16G16_UINT:
- case PIPE_FORMAT_R16G16_SINT:
- case PIPE_FORMAT_R16G16_FLOAT:
- return V_028C70_COLOR_16_16;
-
- case PIPE_FORMAT_R11G11B10_FLOAT:
+ if (format == PIPE_FORMAT_R11G11B10_FLOAT) /* isn't plain */
return V_028C70_COLOR_10_11_11;
- /* 64-bit buffers. */
- case PIPE_FORMAT_R16G16B16A16_UINT:
- case PIPE_FORMAT_R16G16B16A16_SINT:
- case PIPE_FORMAT_R16G16B16A16_USCALED:
- case PIPE_FORMAT_R16G16B16A16_SSCALED:
- case PIPE_FORMAT_R16G16B16A16_UNORM:
- case PIPE_FORMAT_R16G16B16A16_SNORM:
- case PIPE_FORMAT_R16G16B16A16_FLOAT:
- case PIPE_FORMAT_R16G16B16X16_UNORM:
- case PIPE_FORMAT_R16G16B16X16_SNORM:
- case PIPE_FORMAT_R16G16B16X16_FLOAT:
- case PIPE_FORMAT_R16G16B16X16_UINT:
- case PIPE_FORMAT_R16G16B16X16_SINT:
- return V_028C70_COLOR_16_16_16_16;
-
- case PIPE_FORMAT_L32A32_FLOAT:
- case PIPE_FORMAT_L32A32_UINT:
- case PIPE_FORMAT_L32A32_SINT:
- case PIPE_FORMAT_R32G32_FLOAT:
- case PIPE_FORMAT_R32G32_USCALED:
- case PIPE_FORMAT_R32G32_SSCALED:
- case PIPE_FORMAT_R32G32_SINT:
- case PIPE_FORMAT_R32G32_UINT:
- return V_028C70_COLOR_32_32;
-
- /* 128-bit buffers. */
- case PIPE_FORMAT_R32G32B32A32_SNORM:
- case PIPE_FORMAT_R32G32B32A32_UNORM:
- case PIPE_FORMAT_R32G32B32A32_SSCALED:
- case PIPE_FORMAT_R32G32B32A32_USCALED:
- case PIPE_FORMAT_R32G32B32A32_SINT:
- case PIPE_FORMAT_R32G32B32A32_UINT:
- case PIPE_FORMAT_R32G32B32A32_FLOAT:
- case PIPE_FORMAT_R32G32B32X32_FLOAT:
- case PIPE_FORMAT_R32G32B32X32_UINT:
- case PIPE_FORMAT_R32G32B32X32_SINT:
- return V_028C70_COLOR_32_32_32_32;
-
- /* YUV buffers. */
- case PIPE_FORMAT_UYVY:
- case PIPE_FORMAT_YUYV:
- /* 96-bit buffers. */
- case PIPE_FORMAT_R32G32B32_FLOAT:
- /* 8-bit buffers. */
- case PIPE_FORMAT_L4A4_UNORM:
- case PIPE_FORMAT_R4A4_UNORM:
- case PIPE_FORMAT_A4R4_UNORM:
- default:
- return V_028C70_COLOR_INVALID; /* Unsupported. */
+ if (desc->layout != UTIL_FORMAT_LAYOUT_PLAIN)
+ return V_028C70_COLOR_INVALID;
+
+ switch (desc->nr_channels) {
+ case 1:
+ switch (desc->channel[0].size) {
+ case 8:
+ return V_028C70_COLOR_8;
+ case 16:
+ return V_028C70_COLOR_16;
+ case 32:
+ return V_028C70_COLOR_32;
+ }
+ break;
+ case 2:
+ if (desc->channel[0].size == desc->channel[1].size) {
+ switch (desc->channel[0].size) {
+ case 8:
+ return V_028C70_COLOR_8_8;
+ case 16:
+ return V_028C70_COLOR_16_16;
+ case 32:
+ return V_028C70_COLOR_32_32;
+ }
+ } else if (HAS_SIZE(8,24,0,0)) {
+ return V_028C70_COLOR_24_8;
+ } else if (HAS_SIZE(24,8,0,0)) {
+ return V_028C70_COLOR_8_24;
+ }
+ break;
+ case 3:
+ if (HAS_SIZE(5,6,5,0)) {
+ return V_028C70_COLOR_5_6_5;
+ } else if (HAS_SIZE(32,8,24,0)) {
+ return V_028C70_COLOR_X24_8_32_FLOAT;
+ }
+ break;
+ case 4:
+ if (desc->channel[0].size == desc->channel[1].size &&
+ desc->channel[0].size == desc->channel[2].size &&
+ desc->channel[0].size == desc->channel[3].size) {
+ switch (desc->channel[0].size) {
+ case 4:
+ return V_028C70_COLOR_4_4_4_4;
+ case 8:
+ return V_028C70_COLOR_8_8_8_8;
+ case 16:
+ return V_028C70_COLOR_16_16_16_16;
+ case 32:
+ return V_028C70_COLOR_32_32_32_32;
+ }
+ } else if (HAS_SIZE(5,5,5,1)) {
+ return V_028C70_COLOR_1_5_5_5;
+ } else if (HAS_SIZE(10,10,10,2)) {
+ return V_028C70_COLOR_2_10_10_10;
+ }
+ break;
}
+ return V_028C70_COLOR_INVALID;
}
static uint32_t si_translate_colorswap(enum pipe_format format)
{
- switch (format) {
- /* 8-bit buffers. */
- case PIPE_FORMAT_L4A4_UNORM:
- case PIPE_FORMAT_A4R4_UNORM:
- return V_028C70_SWAP_ALT;
-
- case PIPE_FORMAT_A8_UNORM:
- case PIPE_FORMAT_A8_SNORM:
- case PIPE_FORMAT_A8_UINT:
- case PIPE_FORMAT_A8_SINT:
- case PIPE_FORMAT_R4A4_UNORM:
- return V_028C70_SWAP_ALT_REV;
- case PIPE_FORMAT_I8_UNORM:
- case PIPE_FORMAT_I8_SNORM:
- case PIPE_FORMAT_L8_UNORM:
- case PIPE_FORMAT_L8_SNORM:
- case PIPE_FORMAT_I8_UINT:
- case PIPE_FORMAT_I8_SINT:
- case PIPE_FORMAT_L8_UINT:
- case PIPE_FORMAT_L8_SINT:
- case PIPE_FORMAT_L8_SRGB:
- case PIPE_FORMAT_R8_UNORM:
- case PIPE_FORMAT_R8_SNORM:
- case PIPE_FORMAT_R8_UINT:
- case PIPE_FORMAT_R8_SINT:
- return V_028C70_SWAP_STD;
-
- /* 16-bit buffers. */
- case PIPE_FORMAT_B5G6R5_UNORM:
- return V_028C70_SWAP_STD_REV;
-
- case PIPE_FORMAT_B5G5R5A1_UNORM:
- case PIPE_FORMAT_B5G5R5X1_UNORM:
- return V_028C70_SWAP_ALT;
-
- case PIPE_FORMAT_B4G4R4A4_UNORM:
- case PIPE_FORMAT_B4G4R4X4_UNORM:
- return V_028C70_SWAP_ALT;
-
- case PIPE_FORMAT_Z16_UNORM:
- return V_028C70_SWAP_STD;
-
- case PIPE_FORMAT_L8A8_UNORM:
- case PIPE_FORMAT_L8A8_SNORM:
- case PIPE_FORMAT_L8A8_UINT:
- case PIPE_FORMAT_L8A8_SINT:
- return V_028C70_SWAP_ALT;
- case PIPE_FORMAT_R8G8_SNORM:
- case PIPE_FORMAT_R8G8_UNORM:
- case PIPE_FORMAT_R8G8_UINT:
- case PIPE_FORMAT_R8G8_SINT:
- return V_028C70_SWAP_STD;
-
- case PIPE_FORMAT_I16_UNORM:
- case PIPE_FORMAT_I16_SNORM:
- case PIPE_FORMAT_I16_FLOAT:
- case PIPE_FORMAT_L16_UNORM:
- case PIPE_FORMAT_L16_SNORM:
- case PIPE_FORMAT_L16_FLOAT:
- case PIPE_FORMAT_R16_UNORM:
- case PIPE_FORMAT_R16_SNORM:
- case PIPE_FORMAT_R16_UINT:
- case PIPE_FORMAT_R16_SINT:
- case PIPE_FORMAT_R16_FLOAT:
- return V_028C70_SWAP_STD;
-
- case PIPE_FORMAT_A16_UNORM:
- case PIPE_FORMAT_A16_SNORM:
- case PIPE_FORMAT_A16_FLOAT:
- return V_028C70_SWAP_ALT_REV;
-
- /* 32-bit buffers. */
- case PIPE_FORMAT_A8B8G8R8_SRGB:
- return V_028C70_SWAP_STD_REV;
- case PIPE_FORMAT_B8G8R8A8_SRGB:
- return V_028C70_SWAP_ALT;
-
- case PIPE_FORMAT_B8G8R8A8_UNORM:
- case PIPE_FORMAT_B8G8R8X8_UNORM:
- return V_028C70_SWAP_ALT;
-
- case PIPE_FORMAT_A8R8G8B8_UNORM:
- case PIPE_FORMAT_X8R8G8B8_UNORM:
- return V_028C70_SWAP_ALT_REV;
- case PIPE_FORMAT_R8G8B8A8_SNORM:
- case PIPE_FORMAT_R8G8B8A8_UNORM:
- case PIPE_FORMAT_R8G8B8A8_SSCALED:
- case PIPE_FORMAT_R8G8B8A8_USCALED:
- case PIPE_FORMAT_R8G8B8A8_SINT:
- case PIPE_FORMAT_R8G8B8A8_UINT:
- case PIPE_FORMAT_R8G8B8X8_UNORM:
- case PIPE_FORMAT_R8G8B8X8_SNORM:
- case PIPE_FORMAT_R8G8B8X8_SRGB:
- case PIPE_FORMAT_R8G8B8X8_UINT:
- case PIPE_FORMAT_R8G8B8X8_SINT:
- return V_028C70_SWAP_STD;
-
- case PIPE_FORMAT_A8B8G8R8_UNORM:
- case PIPE_FORMAT_X8B8G8R8_UNORM:
- /* case PIPE_FORMAT_R8SG8SB8UX8U_NORM: */
- return V_028C70_SWAP_STD_REV;
-
- case PIPE_FORMAT_Z24X8_UNORM:
- case PIPE_FORMAT_Z24_UNORM_S8_UINT:
- return V_028C70_SWAP_STD;
+ const struct util_format_description *desc = util_format_description(format);
- case PIPE_FORMAT_S8X24_UINT:
- case PIPE_FORMAT_X8Z24_UNORM:
- case PIPE_FORMAT_S8_UINT_Z24_UNORM:
- return V_028C70_SWAP_STD_REV;
+#define HAS_SWIZZLE(chan,swz) (desc->swizzle[chan] == UTIL_FORMAT_SWIZZLE_##swz)
- case PIPE_FORMAT_R10G10B10A2_UNORM:
- case PIPE_FORMAT_R10G10B10X2_SNORM:
- case PIPE_FORMAT_R10SG10SB10SA2U_NORM:
+ if (format == PIPE_FORMAT_R11G11B10_FLOAT) /* isn't plain */
return V_028C70_SWAP_STD;
- case PIPE_FORMAT_B10G10R10A2_UNORM:
- case PIPE_FORMAT_B10G10R10A2_UINT:
- case PIPE_FORMAT_B10G10R10X2_UNORM:
- return V_028C70_SWAP_ALT;
-
- case PIPE_FORMAT_R11G11B10_FLOAT:
- case PIPE_FORMAT_I32_FLOAT:
- case PIPE_FORMAT_L32_FLOAT:
- case PIPE_FORMAT_R32_FLOAT:
- case PIPE_FORMAT_R32_UINT:
- case PIPE_FORMAT_R32_SINT:
- case PIPE_FORMAT_Z32_FLOAT:
- case PIPE_FORMAT_R16G16_FLOAT:
- case PIPE_FORMAT_R16G16_UNORM:
- case PIPE_FORMAT_R16G16_SNORM:
- case PIPE_FORMAT_R16G16_UINT:
- case PIPE_FORMAT_R16G16_SINT:
- return V_028C70_SWAP_STD;
-
- case PIPE_FORMAT_L16A16_UNORM:
- case PIPE_FORMAT_L16A16_SNORM:
- case PIPE_FORMAT_L16A16_FLOAT:
- return V_028C70_SWAP_ALT;
-
- case PIPE_FORMAT_A32_FLOAT:
- return V_028C70_SWAP_ALT_REV;
-
- /* 64-bit buffers. */
- case PIPE_FORMAT_R32G32_FLOAT:
- case PIPE_FORMAT_R32G32_UINT:
- case PIPE_FORMAT_R32G32_SINT:
- case PIPE_FORMAT_R16G16B16A16_UNORM:
- case PIPE_FORMAT_R16G16B16A16_SNORM:
- case PIPE_FORMAT_R16G16B16A16_USCALED:
- case PIPE_FORMAT_R16G16B16A16_SSCALED:
- case PIPE_FORMAT_R16G16B16A16_UINT:
- case PIPE_FORMAT_R16G16B16A16_SINT:
- case PIPE_FORMAT_R16G16B16A16_FLOAT:
- case PIPE_FORMAT_R16G16B16X16_UNORM:
- case PIPE_FORMAT_R16G16B16X16_SNORM:
- case PIPE_FORMAT_R16G16B16X16_FLOAT:
- case PIPE_FORMAT_R16G16B16X16_UINT:
- case PIPE_FORMAT_R16G16B16X16_SINT:
- case PIPE_FORMAT_Z32_FLOAT_S8X24_UINT:
- return V_028C70_SWAP_STD;
+ if (desc->layout != UTIL_FORMAT_LAYOUT_PLAIN)
+ return ~0;
- case PIPE_FORMAT_L32A32_FLOAT:
- case PIPE_FORMAT_L32A32_UINT:
- case PIPE_FORMAT_L32A32_SINT:
- return V_028C70_SWAP_ALT;
-
- /* 128-bit buffers. */
- case PIPE_FORMAT_R32G32B32A32_FLOAT:
- case PIPE_FORMAT_R32G32B32A32_SNORM:
- case PIPE_FORMAT_R32G32B32A32_UNORM:
- case PIPE_FORMAT_R32G32B32A32_SSCALED:
- case PIPE_FORMAT_R32G32B32A32_USCALED:
- case PIPE_FORMAT_R32G32B32A32_SINT:
- case PIPE_FORMAT_R32G32B32A32_UINT:
- case PIPE_FORMAT_R32G32B32X32_FLOAT:
- case PIPE_FORMAT_R32G32B32X32_UINT:
- case PIPE_FORMAT_R32G32B32X32_SINT:
- return V_028C70_SWAP_STD;
- default:
- R600_ERR("unsupported colorswap format %d\n", format);
- return ~0U;
+ switch (desc->nr_channels) {
+ case 1:
+ if (HAS_SWIZZLE(0,X))
+ return V_028C70_SWAP_STD; /* X___ */
+ else if (HAS_SWIZZLE(3,X))
+ return V_028C70_SWAP_ALT_REV; /* ___X */
+ break;
+ case 2:
+ if ((HAS_SWIZZLE(0,X) && HAS_SWIZZLE(1,Y)) ||
+ (HAS_SWIZZLE(0,X) && HAS_SWIZZLE(1,NONE)) ||
+ (HAS_SWIZZLE(0,NONE) && HAS_SWIZZLE(1,Y)))
+ return V_028C70_SWAP_STD; /* XY__ */
+ else if ((HAS_SWIZZLE(0,Y) && HAS_SWIZZLE(1,X)) ||
+ (HAS_SWIZZLE(0,Y) && HAS_SWIZZLE(1,NONE)) ||
+ (HAS_SWIZZLE(0,NONE) && HAS_SWIZZLE(1,X)))
+ return V_028C70_SWAP_STD_REV; /* YX__ */
+ else if (HAS_SWIZZLE(0,X) && HAS_SWIZZLE(3,Y))
+ return V_028C70_SWAP_ALT; /* X__Y */
+ break;
+ case 3:
+ if (HAS_SWIZZLE(0,X))
+ return V_028C70_SWAP_STD; /* XYZ */
+ else if (HAS_SWIZZLE(0,Z))
+ return V_028C70_SWAP_STD_REV; /* ZYX */
+ break;
+ case 4:
+ /* check the middle channels, the 1st and 4th channel can be NONE */
+ if (HAS_SWIZZLE(1,Y) && HAS_SWIZZLE(2,Z))
+ return V_028C70_SWAP_STD; /* XYZW */
+ else if (HAS_SWIZZLE(1,Z) && HAS_SWIZZLE(2,Y))
+ return V_028C70_SWAP_STD_REV; /* WZYX */
+ else if (HAS_SWIZZLE(1,Y) && HAS_SWIZZLE(2,X))
+ return V_028C70_SWAP_ALT; /* ZYXW */
+ else if (HAS_SWIZZLE(1,X) && HAS_SWIZZLE(2,Y))
+ return V_028C70_SWAP_ALT_REV; /* WXYZ */
+ break;
}
return ~0U;
}
const struct util_format_description *desc,
int first_non_void)
{
- struct r600_screen *rscreen = (struct r600_screen*)screen;
- bool enable_s3tc = rscreen->info.drm_minor >= 31;
+ struct si_screen *rscreen = (struct si_screen*)screen;
+ bool enable_s3tc = rscreen->b.info.drm_minor >= 31;
boolean uniform = TRUE;
int i;
}
}
-static unsigned si_tex_dim(unsigned dim)
+static unsigned si_tex_dim(unsigned dim, unsigned nr_samples)
{
switch (dim) {
default:
return V_008F1C_SQ_RSRC_IMG_1D_ARRAY;
case PIPE_TEXTURE_2D:
case PIPE_TEXTURE_RECT:
- return V_008F1C_SQ_RSRC_IMG_2D;
+ return nr_samples > 1 ? V_008F1C_SQ_RSRC_IMG_2D_MSAA :
+ V_008F1C_SQ_RSRC_IMG_2D;
case PIPE_TEXTURE_2D_ARRAY:
- return V_008F1C_SQ_RSRC_IMG_2D_ARRAY;
+ return nr_samples > 1 ? V_008F1C_SQ_RSRC_IMG_2D_MSAA_ARRAY :
+ V_008F1C_SQ_RSRC_IMG_2D_ARRAY;
case PIPE_TEXTURE_3D:
return V_008F1C_SQ_RSRC_IMG_3D;
case PIPE_TEXTURE_CUBE:
util_format_get_first_non_void_channel(format)) != ~0U;
}
-static uint32_t si_translate_vertexformat(struct pipe_screen *screen,
- enum pipe_format format,
- const struct util_format_description *desc,
- int first_non_void)
+static uint32_t si_translate_buffer_dataformat(struct pipe_screen *screen,
+ const struct util_format_description *desc,
+ int first_non_void)
{
unsigned type = desc->channel[first_non_void].type;
int i;
if (type == UTIL_FORMAT_TYPE_FIXED)
return V_008F0C_BUF_DATA_FORMAT_INVALID;
+ if (desc->nr_channels == 4 &&
+ desc->channel[0].size == 10 &&
+ desc->channel[1].size == 10 &&
+ desc->channel[2].size == 10 &&
+ desc->channel[3].size == 2)
+ return V_008F0C_BUF_DATA_FORMAT_2_10_10_10;
+
/* See whether the components are of the same size. */
for (i = 0; i < desc->nr_channels; i++) {
if (desc->channel[first_non_void].size != desc->channel[i].size)
}
break;
case 32:
- if (type != UTIL_FORMAT_TYPE_FLOAT)
+ /* From the Southern Islands ISA documentation about MTBUF:
+ * 'Memory reads of data in memory that is 32 or 64 bits do not
+ * undergo any format conversion.'
+ */
+ if (type != UTIL_FORMAT_TYPE_FLOAT &&
+ !desc->channel[first_non_void].pure_integer)
return V_008F0C_BUF_DATA_FORMAT_INVALID;
switch (desc->nr_channels) {
return V_008F0C_BUF_DATA_FORMAT_INVALID;
}
+static uint32_t si_translate_buffer_numformat(struct pipe_screen *screen,
+ const struct util_format_description *desc,
+ int first_non_void)
+{
+ switch (desc->channel[first_non_void].type) {
+ case UTIL_FORMAT_TYPE_SIGNED:
+ if (desc->channel[first_non_void].normalized)
+ return V_008F0C_BUF_NUM_FORMAT_SNORM;
+ else if (desc->channel[first_non_void].pure_integer)
+ return V_008F0C_BUF_NUM_FORMAT_SINT;
+ else
+ return V_008F0C_BUF_NUM_FORMAT_SSCALED;
+ break;
+ case UTIL_FORMAT_TYPE_UNSIGNED:
+ if (desc->channel[first_non_void].normalized)
+ return V_008F0C_BUF_NUM_FORMAT_UNORM;
+ else if (desc->channel[first_non_void].pure_integer)
+ return V_008F0C_BUF_NUM_FORMAT_UINT;
+ else
+ return V_008F0C_BUF_NUM_FORMAT_USCALED;
+ break;
+ case UTIL_FORMAT_TYPE_FLOAT:
+ default:
+ return V_008F0C_BUF_NUM_FORMAT_FLOAT;
+ }
+}
+
static bool si_is_vertex_format_supported(struct pipe_screen *screen, enum pipe_format format)
{
const struct util_format_description *desc;
desc = util_format_description(format);
first_non_void = util_format_get_first_non_void_channel(format);
- data_format = si_translate_vertexformat(screen, format, desc, first_non_void);
+ data_format = si_translate_buffer_dataformat(screen, desc, first_non_void);
return data_format != V_008F0C_BUF_DATA_FORMAT_INVALID;
}
unsigned sample_count,
unsigned usage)
{
+ struct si_screen *rscreen = (struct si_screen *)screen;
unsigned retval = 0;
if (target >= PIPE_MAX_TEXTURE_TYPES) {
if (!util_format_is_supported(format, usage))
return FALSE;
- /* Multisample */
- if (sample_count > 1)
- return FALSE;
+ if (sample_count > 1) {
+ if (HAVE_LLVM < 0x0304)
+ return FALSE;
- if ((usage & PIPE_BIND_SAMPLER_VIEW) &&
- si_is_sampler_format_supported(screen, format)) {
- retval |= PIPE_BIND_SAMPLER_VIEW;
+ /* 2D tiling on CIK is supported since DRM 2.35.0 */
+ if (rscreen->b.chip_class >= CIK && rscreen->b.info.drm_minor < 35)
+ return FALSE;
+
+ switch (sample_count) {
+ case 2:
+ case 4:
+ case 8:
+ break;
+ default:
+ return FALSE;
+ }
+ }
+
+ if (usage & PIPE_BIND_SAMPLER_VIEW) {
+ if (target == PIPE_BUFFER) {
+ if (si_is_vertex_format_supported(screen, format))
+ retval |= PIPE_BIND_SAMPLER_VIEW;
+ } else {
+ if (si_is_sampler_format_supported(screen, format))
+ retval |= PIPE_BIND_SAMPLER_VIEW;
+ }
}
if ((usage & (PIPE_BIND_RENDER_TARGET |
return retval == usage;
}
-static unsigned si_tile_mode_index(struct r600_resource_texture *rtex, unsigned level, bool stencil)
+static unsigned si_tile_mode_index(struct r600_texture *rtex, unsigned level, bool stencil)
{
unsigned tile_mode_index = 0;
* framebuffer handling
*/
-static void si_cb(struct r600_context *rctx, struct si_pm4_state *pm4,
+static void si_cb(struct si_context *rctx, struct si_pm4_state *pm4,
const struct pipe_framebuffer_state *state, int cb)
{
- struct r600_resource_texture *rtex;
- struct r600_surface *surf;
+ struct r600_texture *rtex;
+ struct si_surface *surf;
unsigned level = state->cbufs[cb]->u.tex.level;
unsigned pitch, slice;
- unsigned color_info, color_attrib;
+ unsigned color_info, color_attrib, color_pitch, color_view;
unsigned tile_mode_index;
unsigned format, swap, ntype, endian;
uint64_t offset;
unsigned blend_clamp = 0, blend_bypass = 0;
unsigned max_comp_size;
- surf = (struct r600_surface *)state->cbufs[cb];
- rtex = (struct r600_resource_texture*)state->cbufs[cb]->texture;
+ surf = (struct si_surface *)state->cbufs[cb];
+ rtex = (struct r600_texture*)state->cbufs[cb]->texture;
offset = rtex->surface.level[level].offset;
- if (rtex->surface.level[level].mode < RADEON_SURF_MODE_1D) {
+
+ /* Layered rendering doesn't work with LINEAR_GENERAL.
+ * (LINEAR_ALIGNED and others work) */
+ if (rtex->surface.level[level].mode == RADEON_SURF_MODE_LINEAR) {
+ assert(state->cbufs[cb]->u.tex.first_layer == state->cbufs[cb]->u.tex.last_layer);
offset += rtex->surface.level[level].slice_size *
state->cbufs[cb]->u.tex.first_layer;
+ color_view = 0;
+ } else {
+ color_view = S_028C6C_SLICE_START(state->cbufs[cb]->u.tex.first_layer) |
+ S_028C6C_SLICE_MAX(state->cbufs[cb]->u.tex.last_layer);
}
+
pitch = (rtex->surface.level[level].nblk_x) / 8 - 1;
slice = (rtex->surface.level[level].nblk_x * rtex->surface.level[level].nblk_y) / 64;
if (slice) {
S_028C70_NUMBER_TYPE(ntype) |
S_028C70_ENDIAN(endian);
+ color_pitch = S_028C64_TILE_MAX(pitch);
+
color_attrib = S_028C74_TILE_MODE_INDEX(tile_mode_index) |
S_028C74_FORCE_DST_ALPHA_1(desc->swizzle[3] == UTIL_FORMAT_SWIZZLE_1);
- offset += r600_resource_va(rctx->context.screen, state->cbufs[cb]->texture);
+ if (rtex->resource.b.b.nr_samples > 1) {
+ unsigned log_samples = util_logbase2(rtex->resource.b.b.nr_samples);
+
+ color_attrib |= S_028C74_NUM_SAMPLES(log_samples) |
+ S_028C74_NUM_FRAGMENTS(log_samples);
+
+ if (rtex->fmask.size) {
+ color_info |= S_028C70_COMPRESSION(1);
+ unsigned fmask_bankh = util_logbase2(rtex->fmask.bank_height);
+
+ color_attrib |= S_028C74_FMASK_TILE_MODE_INDEX(rtex->fmask.tile_mode_index);
+
+ if (rctx->b.chip_class == SI) {
+ /* due to a hw bug, FMASK_BANK_HEIGHT must be set on SI too */
+ color_attrib |= S_028C74_FMASK_BANK_HEIGHT(fmask_bankh);
+ }
+ if (rctx->b.chip_class >= CIK) {
+ color_pitch |= S_028C64_FMASK_TILE_MAX(rtex->fmask.pitch / 8 - 1);
+ }
+ }
+ }
+
+ if (rtex->cmask.size) {
+ color_info |= S_028C70_FAST_CLEAR(1);
+ }
+
+ offset += r600_resource_va(rctx->b.b.screen, state->cbufs[cb]->texture);
offset >>= 8;
- /* FIXME handle enabling of CB beyond BASE8 which has different offset */
si_pm4_add_bo(pm4, &rtex->resource, RADEON_USAGE_READWRITE);
si_pm4_set_reg(pm4, R_028C60_CB_COLOR0_BASE + cb * 0x3C, offset);
- si_pm4_set_reg(pm4, R_028C64_CB_COLOR0_PITCH + cb * 0x3C, S_028C64_TILE_MAX(pitch));
+ si_pm4_set_reg(pm4, R_028C64_CB_COLOR0_PITCH + cb * 0x3C, color_pitch);
si_pm4_set_reg(pm4, R_028C68_CB_COLOR0_SLICE + cb * 0x3C, S_028C68_TILE_MAX(slice));
-
- if (rtex->surface.level[level].mode < RADEON_SURF_MODE_1D) {
- si_pm4_set_reg(pm4, R_028C6C_CB_COLOR0_VIEW + cb * 0x3C, 0x00000000);
- } else {
- si_pm4_set_reg(pm4, R_028C6C_CB_COLOR0_VIEW + cb * 0x3C,
- S_028C6C_SLICE_START(state->cbufs[cb]->u.tex.first_layer) |
- S_028C6C_SLICE_MAX(state->cbufs[cb]->u.tex.last_layer));
- }
+ si_pm4_set_reg(pm4, R_028C6C_CB_COLOR0_VIEW + cb * 0x3C, color_view);
si_pm4_set_reg(pm4, R_028C70_CB_COLOR0_INFO + cb * 0x3C, color_info);
si_pm4_set_reg(pm4, R_028C74_CB_COLOR0_ATTRIB + cb * 0x3C, color_attrib);
+ if (rtex->cmask.size) {
+ si_pm4_set_reg(pm4, R_028C7C_CB_COLOR0_CMASK + cb * 0x3C,
+ offset + (rtex->cmask.offset >> 8));
+ si_pm4_set_reg(pm4, R_028C80_CB_COLOR0_CMASK_SLICE + cb * 0x3C,
+ S_028C80_TILE_MAX(rtex->cmask.slice_tile_max));
+ }
+ if (rtex->fmask.size) {
+ si_pm4_set_reg(pm4, R_028C84_CB_COLOR0_FMASK + cb * 0x3C,
+ offset + (rtex->fmask.offset >> 8));
+ si_pm4_set_reg(pm4, R_028C88_CB_COLOR0_FMASK_SLICE + cb * 0x3C,
+ S_028C88_TILE_MAX(rtex->fmask.slice_tile_max));
+ }
+
+ /* set CB_COLOR1_INFO for possible dual-src blending */
+ if (state->nr_cbufs == 1) {
+ assert(cb == 0);
+ si_pm4_set_reg(pm4, R_028C70_CB_COLOR0_INFO + 1 * 0x3C, color_info);
+ }
+
/* Determine pixel shader export format */
max_comp_size = si_colorformat_max_comp_size(format);
if (ntype == V_028C70_NUMBER_SRGB ||
max_comp_size <= 10) ||
(ntype == V_028C70_NUMBER_FLOAT && max_comp_size <= 16)) {
rctx->export_16bpc |= 1 << cb;
+ /* set SPI_SHADER_COL_FORMAT for possible dual-src blending */
+ if (state->nr_cbufs == 1)
+ rctx->export_16bpc |= 1 << 1;
}
}
-static void si_db(struct r600_context *rctx, struct si_pm4_state *pm4,
+static void si_db(struct si_context *rctx, struct si_pm4_state *pm4,
const struct pipe_framebuffer_state *state)
{
- struct r600_resource_texture *rtex;
- struct r600_surface *surf;
- unsigned level, pitch, slice, format, tile_mode_index;
- uint32_t z_info, s_info;
+ struct si_screen *rscreen = rctx->screen;
+ struct r600_texture *rtex;
+ struct si_surface *surf;
+ unsigned level, pitch, slice, format, tile_mode_index, array_mode;
+ unsigned macro_aspect, tile_split, stile_split, bankh, bankw, nbanks, pipe_config;
+ uint32_t z_info, s_info, db_depth_info;
uint64_t z_offs, s_offs;
+ uint32_t db_htile_data_base, db_htile_surface;
if (state->zsbuf == NULL) {
si_pm4_set_reg(pm4, R_028040_DB_Z_INFO, S_028040_FORMAT(V_028040_Z_INVALID));
return;
}
- surf = (struct r600_surface *)state->zsbuf;
+ surf = (struct si_surface *)state->zsbuf;
level = surf->base.u.tex.level;
- rtex = (struct r600_resource_texture*)surf->base.texture;
+ rtex = (struct r600_texture*)surf->base.texture;
- format = si_translate_dbformat(rtex->real_format);
+ format = si_translate_dbformat(rtex->resource.b.b.format);
if (format == V_028040_Z_INVALID) {
- R600_ERR("Invalid DB format: %d, disabling DB.\n", rtex->real_format);
+ R600_ERR("Invalid DB format: %d, disabling DB.\n", rtex->resource.b.b.format);
}
assert(format != V_028040_Z_INVALID);
- s_offs = z_offs = r600_resource_va(rctx->context.screen, surf->base.texture);
+ s_offs = z_offs = r600_resource_va(rctx->b.b.screen, surf->base.texture);
z_offs += rtex->surface.level[level].offset;
s_offs += rtex->surface.stencil_level[level].offset;
slice = slice - 1;
}
+ db_depth_info = S_02803C_ADDR5_SWIZZLE_MASK(1);
+
z_info = S_028040_FORMAT(format);
+ if (rtex->resource.b.b.nr_samples > 1) {
+ z_info |= S_028040_NUM_SAMPLES(util_logbase2(rtex->resource.b.b.nr_samples));
+ }
+
if (rtex->surface.flags & RADEON_SURF_SBUFFER)
s_info = S_028044_FORMAT(V_028044_STENCIL_8);
else
s_info = S_028044_FORMAT(V_028044_STENCIL_INVALID);
- tile_mode_index = si_tile_mode_index(rtex, level, false);
- z_info |= S_028040_TILE_MODE_INDEX(tile_mode_index);
- tile_mode_index = si_tile_mode_index(rtex, level, true);
- s_info |= S_028044_TILE_MODE_INDEX(tile_mode_index);
+ if (rctx->b.chip_class >= CIK) {
+ switch (rtex->surface.level[level].mode) {
+ case RADEON_SURF_MODE_2D:
+ array_mode = V_02803C_ARRAY_2D_TILED_THIN1;
+ break;
+ case RADEON_SURF_MODE_1D:
+ case RADEON_SURF_MODE_LINEAR_ALIGNED:
+ case RADEON_SURF_MODE_LINEAR:
+ default:
+ array_mode = V_02803C_ARRAY_1D_TILED_THIN1;
+ break;
+ }
+ tile_split = rtex->surface.tile_split;
+ stile_split = rtex->surface.stencil_tile_split;
+ macro_aspect = rtex->surface.mtilea;
+ bankw = rtex->surface.bankw;
+ bankh = rtex->surface.bankh;
+ tile_split = cik_tile_split(tile_split);
+ stile_split = cik_tile_split(stile_split);
+ macro_aspect = cik_macro_tile_aspect(macro_aspect);
+ bankw = cik_bank_wh(bankw);
+ bankh = cik_bank_wh(bankh);
+ nbanks = cik_num_banks(rscreen, rtex->surface.bpe, rtex->surface.tile_split);
+ tile_mode_index = si_tile_mode_index(rtex, level, false);
+ pipe_config = cik_db_pipe_config(rscreen, tile_mode_index);
+
+ db_depth_info |= S_02803C_ARRAY_MODE(array_mode) |
+ S_02803C_PIPE_CONFIG(pipe_config) |
+ S_02803C_BANK_WIDTH(bankw) |
+ S_02803C_BANK_HEIGHT(bankh) |
+ S_02803C_MACRO_TILE_ASPECT(macro_aspect) |
+ S_02803C_NUM_BANKS(nbanks);
+ z_info |= S_028040_TILE_SPLIT(tile_split);
+ s_info |= S_028044_TILE_SPLIT(stile_split);
+ } else {
+ tile_mode_index = si_tile_mode_index(rtex, level, false);
+ z_info |= S_028040_TILE_MODE_INDEX(tile_mode_index);
+ tile_mode_index = si_tile_mode_index(rtex, level, true);
+ s_info |= S_028044_TILE_MODE_INDEX(tile_mode_index);
+ }
+
+ /* HiZ aka depth buffer htile */
+ /* use htile only for first level */
+ if (rtex->htile_buffer && !level) {
+ const struct util_format_description *fmt_desc;
+
+ z_info |= S_028040_TILE_SURFACE_ENABLE(1);
+
+ /* This is optimal for the clear value of 1.0 and using
+ * the LESS and LEQUAL test functions. Set this to 0
+ * for the opposite case. This can only be changed when
+ * clearing. */
+ z_info |= S_028040_ZRANGE_PRECISION(1);
+
+ fmt_desc = util_format_description(rtex->resource.b.b.format);
+ if (!util_format_has_stencil(fmt_desc)) {
+ /* Use all of the htile_buffer for depth */
+ s_info |= S_028044_TILE_STENCIL_DISABLE(1);
+ }
+
+ uint64_t va = r600_resource_va(&rctx->screen->b.b, &rtex->htile_buffer->b.b);
+ db_htile_data_base = va >> 8;
+ db_htile_surface = S_028ABC_FULL_CACHE(1);
+
+ si_pm4_add_bo(pm4, rtex->htile_buffer, RADEON_USAGE_READWRITE);
+ } else {
+ db_htile_data_base = 0;
+ db_htile_surface = 0;
+ }
si_pm4_set_reg(pm4, R_028008_DB_DEPTH_VIEW,
S_028008_SLICE_START(state->zsbuf->u.tex.first_layer) |
S_028008_SLICE_MAX(state->zsbuf->u.tex.last_layer));
+ si_pm4_set_reg(pm4, R_028014_DB_HTILE_DATA_BASE, db_htile_data_base);
- si_pm4_set_reg(pm4, R_02803C_DB_DEPTH_INFO, S_02803C_ADDR5_SWIZZLE_MASK(1));
+ si_pm4_set_reg(pm4, R_02803C_DB_DEPTH_INFO, db_depth_info);
si_pm4_set_reg(pm4, R_028040_DB_Z_INFO, z_info);
si_pm4_set_reg(pm4, R_028044_DB_STENCIL_INFO, s_info);
si_pm4_set_reg(pm4, R_028058_DB_DEPTH_SIZE, S_028058_PITCH_TILE_MAX(pitch));
si_pm4_set_reg(pm4, R_02805C_DB_DEPTH_SLICE, S_02805C_SLICE_TILE_MAX(slice));
+
+ si_pm4_set_reg(pm4, R_028ABC_DB_HTILE_SURFACE, db_htile_surface);
+}
+
+#define FILL_SREG(s0x, s0y, s1x, s1y, s2x, s2y, s3x, s3y) \
+ (((s0x) & 0xf) | (((s0y) & 0xf) << 4) | \
+ (((s1x) & 0xf) << 8) | (((s1y) & 0xf) << 12) | \
+ (((s2x) & 0xf) << 16) | (((s2y) & 0xf) << 20) | \
+ (((s3x) & 0xf) << 24) | (((s3y) & 0xf) << 28))
+
+/* 2xMSAA
+ * There are two locations (-4, 4), (4, -4). */
+static uint32_t sample_locs_2x[] = {
+ FILL_SREG(-4, 4, 4, -4, -4, 4, 4, -4),
+ FILL_SREG(-4, 4, 4, -4, -4, 4, 4, -4),
+ FILL_SREG(-4, 4, 4, -4, -4, 4, 4, -4),
+ FILL_SREG(-4, 4, 4, -4, -4, 4, 4, -4),
+};
+static unsigned max_dist_2x = 4;
+/* 4xMSAA
+ * There are 4 locations: (-2, -2), (2, 2), (-6, 6), (6, -6). */
+static uint32_t sample_locs_4x[] = {
+ FILL_SREG(-2, -2, 2, 2, -6, 6, 6, -6),
+ FILL_SREG(-2, -2, 2, 2, -6, 6, 6, -6),
+ FILL_SREG(-2, -2, 2, 2, -6, 6, 6, -6),
+ FILL_SREG(-2, -2, 2, 2, -6, 6, 6, -6),
+};
+static unsigned max_dist_4x = 6;
+/* Cayman/SI 8xMSAA */
+static uint32_t cm_sample_locs_8x[] = {
+ FILL_SREG(-2, -5, 3, -4, -1, 5, -6, -2),
+ FILL_SREG(-2, -5, 3, -4, -1, 5, -6, -2),
+ FILL_SREG(-2, -5, 3, -4, -1, 5, -6, -2),
+ FILL_SREG(-2, -5, 3, -4, -1, 5, -6, -2),
+ FILL_SREG( 6, 0, 0, 0, -5, 3, 4, 4),
+ FILL_SREG( 6, 0, 0, 0, -5, 3, 4, 4),
+ FILL_SREG( 6, 0, 0, 0, -5, 3, 4, 4),
+ FILL_SREG( 6, 0, 0, 0, -5, 3, 4, 4),
+};
+static unsigned cm_max_dist_8x = 8;
+/* Cayman/SI 16xMSAA */
+static uint32_t cm_sample_locs_16x[] = {
+ FILL_SREG(-7, -3, 7, 3, 1, -5, -5, 5),
+ FILL_SREG(-7, -3, 7, 3, 1, -5, -5, 5),
+ FILL_SREG(-7, -3, 7, 3, 1, -5, -5, 5),
+ FILL_SREG(-7, -3, 7, 3, 1, -5, -5, 5),
+ FILL_SREG(-3, -7, 3, 7, 5, -1, -1, 1),
+ FILL_SREG(-3, -7, 3, 7, 5, -1, -1, 1),
+ FILL_SREG(-3, -7, 3, 7, 5, -1, -1, 1),
+ FILL_SREG(-3, -7, 3, 7, 5, -1, -1, 1),
+ FILL_SREG(-8, -6, 4, 2, 2, -8, -2, 6),
+ FILL_SREG(-8, -6, 4, 2, 2, -8, -2, 6),
+ FILL_SREG(-8, -6, 4, 2, 2, -8, -2, 6),
+ FILL_SREG(-8, -6, 4, 2, 2, -8, -2, 6),
+ FILL_SREG(-4, -2, 0, 4, 6, -4, -6, 0),
+ FILL_SREG(-4, -2, 0, 4, 6, -4, -6, 0),
+ FILL_SREG(-4, -2, 0, 4, 6, -4, -6, 0),
+ FILL_SREG(-4, -2, 0, 4, 6, -4, -6, 0),
+};
+static unsigned cm_max_dist_16x = 8;
+
+static void si_get_sample_position(struct pipe_context *ctx,
+ unsigned sample_count,
+ unsigned sample_index,
+ float *out_value)
+{
+ int offset, index;
+ struct {
+ int idx:4;
+ } val;
+ switch (sample_count) {
+ case 1:
+ default:
+ out_value[0] = out_value[1] = 0.5;
+ break;
+ case 2:
+ offset = 4 * (sample_index * 2);
+ val.idx = (sample_locs_2x[0] >> offset) & 0xf;
+ out_value[0] = (float)(val.idx + 8) / 16.0f;
+ val.idx = (sample_locs_2x[0] >> (offset + 4)) & 0xf;
+ out_value[1] = (float)(val.idx + 8) / 16.0f;
+ break;
+ case 4:
+ offset = 4 * (sample_index * 2);
+ val.idx = (sample_locs_4x[0] >> offset) & 0xf;
+ out_value[0] = (float)(val.idx + 8) / 16.0f;
+ val.idx = (sample_locs_4x[0] >> (offset + 4)) & 0xf;
+ out_value[1] = (float)(val.idx + 8) / 16.0f;
+ break;
+ case 8:
+ offset = 4 * (sample_index % 4 * 2);
+ index = (sample_index / 4) * 4;
+ val.idx = (cm_sample_locs_8x[index] >> offset) & 0xf;
+ out_value[0] = (float)(val.idx + 8) / 16.0f;
+ val.idx = (cm_sample_locs_8x[index] >> (offset + 4)) & 0xf;
+ out_value[1] = (float)(val.idx + 8) / 16.0f;
+ break;
+ case 16:
+ offset = 4 * (sample_index % 4 * 2);
+ index = (sample_index / 4) * 4;
+ val.idx = (cm_sample_locs_16x[index] >> offset) & 0xf;
+ out_value[0] = (float)(val.idx + 8) / 16.0f;
+ val.idx = (cm_sample_locs_16x[index] >> (offset + 4)) & 0xf;
+ out_value[1] = (float)(val.idx + 8) / 16.0f;
+ break;
+ }
+}
+
+static void si_set_msaa_state(struct si_context *rctx, struct si_pm4_state *pm4, int nr_samples)
+{
+ unsigned max_dist = 0;
+
+ switch (nr_samples) {
+ default:
+ nr_samples = 0;
+ break;
+ case 2:
+ si_pm4_set_reg(pm4, R_028BF8_PA_SC_AA_SAMPLE_LOCS_PIXEL_X0Y0_0, sample_locs_2x[0]);
+ si_pm4_set_reg(pm4, R_028C08_PA_SC_AA_SAMPLE_LOCS_PIXEL_X1Y0_0, sample_locs_2x[1]);
+ si_pm4_set_reg(pm4, R_028C18_PA_SC_AA_SAMPLE_LOCS_PIXEL_X0Y1_0, sample_locs_2x[2]);
+ si_pm4_set_reg(pm4, R_028C28_PA_SC_AA_SAMPLE_LOCS_PIXEL_X1Y1_0, sample_locs_2x[3]);
+ max_dist = max_dist_2x;
+ break;
+ case 4:
+ si_pm4_set_reg(pm4, R_028BF8_PA_SC_AA_SAMPLE_LOCS_PIXEL_X0Y0_0, sample_locs_4x[0]);
+ si_pm4_set_reg(pm4, R_028C08_PA_SC_AA_SAMPLE_LOCS_PIXEL_X1Y0_0, sample_locs_4x[1]);
+ si_pm4_set_reg(pm4, R_028C18_PA_SC_AA_SAMPLE_LOCS_PIXEL_X0Y1_0, sample_locs_4x[2]);
+ si_pm4_set_reg(pm4, R_028C28_PA_SC_AA_SAMPLE_LOCS_PIXEL_X1Y1_0, sample_locs_4x[3]);
+ max_dist = max_dist_4x;
+ break;
+ case 8:
+ si_pm4_set_reg(pm4, R_028BF8_PA_SC_AA_SAMPLE_LOCS_PIXEL_X0Y0_0, cm_sample_locs_8x[0]);
+ si_pm4_set_reg(pm4, R_028BFC_PA_SC_AA_SAMPLE_LOCS_PIXEL_X0Y0_1, cm_sample_locs_8x[4]);
+ si_pm4_set_reg(pm4, R_028C00_PA_SC_AA_SAMPLE_LOCS_PIXEL_X0Y0_2, 0);
+ si_pm4_set_reg(pm4, R_028C04_PA_SC_AA_SAMPLE_LOCS_PIXEL_X0Y0_3, 0);
+ si_pm4_set_reg(pm4, R_028C08_PA_SC_AA_SAMPLE_LOCS_PIXEL_X1Y0_0, cm_sample_locs_8x[1]);
+ si_pm4_set_reg(pm4, R_028C0C_PA_SC_AA_SAMPLE_LOCS_PIXEL_X1Y0_1, cm_sample_locs_8x[5]);
+ si_pm4_set_reg(pm4, R_028C10_PA_SC_AA_SAMPLE_LOCS_PIXEL_X1Y0_2, 0);
+ si_pm4_set_reg(pm4, R_028C14_PA_SC_AA_SAMPLE_LOCS_PIXEL_X1Y0_3, 0);
+ si_pm4_set_reg(pm4, R_028C18_PA_SC_AA_SAMPLE_LOCS_PIXEL_X0Y1_0, cm_sample_locs_8x[2]);
+ si_pm4_set_reg(pm4, R_028C1C_PA_SC_AA_SAMPLE_LOCS_PIXEL_X0Y1_1, cm_sample_locs_8x[6]);
+ si_pm4_set_reg(pm4, R_028C20_PA_SC_AA_SAMPLE_LOCS_PIXEL_X0Y1_2, 0);
+ si_pm4_set_reg(pm4, R_028C24_PA_SC_AA_SAMPLE_LOCS_PIXEL_X0Y1_3, 0);
+ si_pm4_set_reg(pm4, R_028C28_PA_SC_AA_SAMPLE_LOCS_PIXEL_X1Y1_0, cm_sample_locs_8x[3]);
+ si_pm4_set_reg(pm4, R_028C2C_PA_SC_AA_SAMPLE_LOCS_PIXEL_X1Y1_1, cm_sample_locs_8x[7]);
+ max_dist = cm_max_dist_8x;
+ break;
+ case 16:
+ si_pm4_set_reg(pm4, R_028BF8_PA_SC_AA_SAMPLE_LOCS_PIXEL_X0Y0_0, cm_sample_locs_16x[0]);
+ si_pm4_set_reg(pm4, R_028BFC_PA_SC_AA_SAMPLE_LOCS_PIXEL_X0Y0_1, cm_sample_locs_16x[4]);
+ si_pm4_set_reg(pm4, R_028C00_PA_SC_AA_SAMPLE_LOCS_PIXEL_X0Y0_2, cm_sample_locs_16x[8]);
+ si_pm4_set_reg(pm4, R_028C04_PA_SC_AA_SAMPLE_LOCS_PIXEL_X0Y0_3, cm_sample_locs_16x[12]);
+ si_pm4_set_reg(pm4, R_028C08_PA_SC_AA_SAMPLE_LOCS_PIXEL_X1Y0_0, cm_sample_locs_16x[1]);
+ si_pm4_set_reg(pm4, R_028C0C_PA_SC_AA_SAMPLE_LOCS_PIXEL_X1Y0_1, cm_sample_locs_16x[5]);
+ si_pm4_set_reg(pm4, R_028C10_PA_SC_AA_SAMPLE_LOCS_PIXEL_X1Y0_2, cm_sample_locs_16x[9]);
+ si_pm4_set_reg(pm4, R_028C14_PA_SC_AA_SAMPLE_LOCS_PIXEL_X1Y0_3, cm_sample_locs_16x[13]);
+ si_pm4_set_reg(pm4, R_028C18_PA_SC_AA_SAMPLE_LOCS_PIXEL_X0Y1_0, cm_sample_locs_16x[2]);
+ si_pm4_set_reg(pm4, R_028C1C_PA_SC_AA_SAMPLE_LOCS_PIXEL_X0Y1_1, cm_sample_locs_16x[6]);
+ si_pm4_set_reg(pm4, R_028C20_PA_SC_AA_SAMPLE_LOCS_PIXEL_X0Y1_2, cm_sample_locs_16x[10]);
+ si_pm4_set_reg(pm4, R_028C24_PA_SC_AA_SAMPLE_LOCS_PIXEL_X0Y1_3, cm_sample_locs_16x[14]);
+ si_pm4_set_reg(pm4, R_028C28_PA_SC_AA_SAMPLE_LOCS_PIXEL_X1Y1_0, cm_sample_locs_16x[3]);
+ si_pm4_set_reg(pm4, R_028C2C_PA_SC_AA_SAMPLE_LOCS_PIXEL_X1Y1_1, cm_sample_locs_16x[7]);
+ si_pm4_set_reg(pm4, R_028C30_PA_SC_AA_SAMPLE_LOCS_PIXEL_X1Y1_2, cm_sample_locs_16x[11]);
+ si_pm4_set_reg(pm4, R_028C34_PA_SC_AA_SAMPLE_LOCS_PIXEL_X1Y1_3, cm_sample_locs_16x[15]);
+ max_dist = cm_max_dist_16x;
+ break;
+ }
+
+ if (nr_samples > 1) {
+ unsigned log_samples = util_logbase2(nr_samples);
+
+ si_pm4_set_reg(pm4, R_028BDC_PA_SC_LINE_CNTL,
+ S_028BDC_LAST_PIXEL(1) |
+ S_028BDC_EXPAND_LINE_WIDTH(1));
+ si_pm4_set_reg(pm4, R_028BE0_PA_SC_AA_CONFIG,
+ S_028BE0_MSAA_NUM_SAMPLES(log_samples) |
+ S_028BE0_MAX_SAMPLE_DIST(max_dist) |
+ S_028BE0_MSAA_EXPOSED_SAMPLES(log_samples));
+
+ si_pm4_set_reg(pm4, R_028804_DB_EQAA,
+ S_028804_MAX_ANCHOR_SAMPLES(log_samples) |
+ S_028804_PS_ITER_SAMPLES(log_samples) |
+ S_028804_MASK_EXPORT_NUM_SAMPLES(log_samples) |
+ S_028804_ALPHA_TO_MASK_NUM_SAMPLES(log_samples) |
+ S_028804_HIGH_QUALITY_INTERSECTIONS(1) |
+ S_028804_STATIC_ANCHOR_ASSOCIATIONS(1));
+ } else {
+ si_pm4_set_reg(pm4, R_028BDC_PA_SC_LINE_CNTL, S_028BDC_LAST_PIXEL(1));
+ si_pm4_set_reg(pm4, R_028BE0_PA_SC_AA_CONFIG, 0);
+
+ si_pm4_set_reg(pm4, R_028804_DB_EQAA,
+ S_028804_HIGH_QUALITY_INTERSECTIONS(1) |
+ S_028804_STATIC_ANCHOR_ASSOCIATIONS(1));
+ }
}
static void si_set_framebuffer_state(struct pipe_context *ctx,
const struct pipe_framebuffer_state *state)
{
- struct r600_context *rctx = (struct r600_context *)ctx;
- struct si_pm4_state *pm4 = CALLOC_STRUCT(si_pm4_state);
- uint32_t shader_mask, tl, br;
- int tl_x, tl_y, br_x, br_y;
+ struct si_context *rctx = (struct si_context *)ctx;
+ struct si_pm4_state *pm4 = si_pm4_alloc_state(rctx);
+ uint32_t tl, br;
+ int tl_x, tl_y, br_x, br_y, nr_samples, i;
if (pm4 == NULL)
return;
- si_pm4_inval_fb_cache(pm4, state->nr_cbufs);
-
- if (state->zsbuf)
- si_pm4_inval_zsbuf_cache(pm4);
+ if (rctx->framebuffer.nr_cbufs) {
+ rctx->b.flags |= R600_CONTEXT_FLUSH_AND_INV_CB |
+ R600_CONTEXT_FLUSH_AND_INV_CB_META;
+ }
+ if (rctx->framebuffer.zsbuf) {
+ rctx->b.flags |= R600_CONTEXT_FLUSH_AND_INV_DB |
+ R600_CONTEXT_FLUSH_AND_INV_DB_META;
+ }
util_copy_framebuffer_state(&rctx->framebuffer, state);
/* build states */
rctx->export_16bpc = 0;
- for (int i = 0; i < state->nr_cbufs; i++) {
+ rctx->fb_compressed_cb_mask = 0;
+ for (i = 0; i < state->nr_cbufs; i++) {
+ struct r600_texture *rtex;
+
+ if (!state->cbufs[i]) {
+ si_pm4_set_reg(pm4, R_028C70_CB_COLOR0_INFO + i * 0x3C,
+ S_028C70_FORMAT(V_028C70_COLOR_INVALID));
+ continue;
+ }
+
+ rtex = (struct r600_texture*)state->cbufs[i]->texture;
+
si_cb(rctx, pm4, state, i);
+
+ if (rtex->fmask.size || rtex->cmask.size) {
+ rctx->fb_compressed_cb_mask |= 1 << i;
+ }
}
+ for (; i < 8; i++) {
+ si_pm4_set_reg(pm4, R_028C70_CB_COLOR0_INFO + i * 0x3C,
+ S_028C70_FORMAT(V_028C70_COLOR_INVALID));
+ }
+
assert(!(rctx->export_16bpc & ~0xff));
si_db(rctx, pm4, state);
- shader_mask = 0;
- for (int i = 0; i < state->nr_cbufs; i++) {
- shader_mask |= 0xf << (i * 4);
- }
tl_x = 0;
tl_y = 0;
br_x = state->width;
si_pm4_set_reg(pm4, R_028034_PA_SC_SCREEN_SCISSOR_BR, br);
si_pm4_set_reg(pm4, R_028204_PA_SC_WINDOW_SCISSOR_TL, tl);
si_pm4_set_reg(pm4, R_028208_PA_SC_WINDOW_SCISSOR_BR, br);
- si_pm4_set_reg(pm4, R_028200_PA_SC_WINDOW_OFFSET, 0x00000000);
- si_pm4_set_reg(pm4, R_028230_PA_SC_EDGERULE, 0xAAAAAAAA);
- si_pm4_set_reg(pm4, R_02823C_CB_SHADER_MASK, shader_mask);
- si_pm4_set_reg(pm4, R_028BE0_PA_SC_AA_CONFIG, 0x00000000);
+
+ nr_samples = util_framebuffer_get_num_samples(state);
+
+ si_set_msaa_state(rctx, pm4, nr_samples);
+ rctx->fb_log_samples = util_logbase2(nr_samples);
+ rctx->fb_cb0_is_integer = state->nr_cbufs && state->cbufs[0] &&
+ util_format_is_pure_integer(state->cbufs[0]->format);
si_pm4_set_state(rctx, framebuffer, pm4);
si_update_fb_rs_state(rctx);
struct si_pipe_shader_selector *sel,
union si_shader_key *key)
{
- struct r600_context *rctx = (struct r600_context *)ctx;
+ struct si_context *rctx = (struct si_context *)ctx;
memset(key, 0, sizeof(*key));
if (sel->type == PIPE_SHADER_VERTEX) {
for (i = 0; i < rctx->vertex_elements->count; ++i)
key->vs.instance_divisors[i] = rctx->vertex_elements->elements[i].instance_divisor;
+ if (rctx->queued.named.rasterizer->clip_plane_enable & 0xf0)
+ key->vs.ucps_enabled |= 0x2;
+ if (rctx->queued.named.rasterizer->clip_plane_enable & 0xf)
+ key->vs.ucps_enabled |= 0x1;
} else if (sel->type == PIPE_SHADER_FRAGMENT) {
if (sel->fs_write_all)
key->ps.nr_cbufs = rctx->framebuffer.nr_cbufs;
key->ps.export_16bpc = rctx->export_16bpc;
+
if (rctx->queued.named.rasterizer) {
key->ps.color_two_side = rctx->queued.named.rasterizer->two_side;
key->ps.flatshade = rctx->queued.named.rasterizer->flatshade;
+
+ if (rctx->queued.named.blend) {
+ key->ps.alpha_to_one = rctx->queued.named.blend->alpha_to_one &&
+ rctx->queued.named.rasterizer->multisample_enable &&
+ !rctx->fb_cb0_is_integer;
+ }
}
if (rctx->queued.named.dsa) {
key->ps.alpha_func = rctx->queued.named.dsa->alpha_func;
- key->ps.alpha_ref = rctx->queued.named.dsa->alpha_ref;
+
+ /* Alpha-test should be disabled if colorbuffer 0 is integer. */
+ if (rctx->framebuffer.nr_cbufs &&
+ rctx->framebuffer.cbufs[0] &&
+ util_format_is_pure_integer(rctx->framebuffer.cbufs[0]->texture->format))
+ key->ps.alpha_func = PIPE_FUNC_ALWAYS;
} else {
key->ps.alpha_func = PIPE_FUNC_ALWAYS;
}
FREE(shader);
return r;
}
-
- /* We don't know the value of fs_write_all property until we built
- * at least one variant, so we may need to recompute the key (include
- * rctx->framebuffer.nr_cbufs) after building first variant. */
- if (sel->type == PIPE_SHADER_FRAGMENT &&
- sel->num_shaders == 0 &&
- shader->shader.fs_write_all) {
- sel->fs_write_all = 1;
- si_shader_selector_key(ctx, sel, &shader->key);
- }
-
sel->num_shaders++;
}
sel->tokens = tgsi_dup_tokens(state->tokens);
sel->so = state->stream_output;
+ if (pipe_shader_type == PIPE_SHADER_FRAGMENT) {
+ struct tgsi_shader_info info;
+
+ tgsi_scan_shader(state->tokens, &info);
+ sel->fs_write_all = info.color0_writes_all_cbufs;
+ }
+
r = si_shader_select(ctx, sel, NULL);
if (r) {
free(sel);
static void si_bind_vs_shader(struct pipe_context *ctx, void *state)
{
- struct r600_context *rctx = (struct r600_context *)ctx;
+ struct si_context *rctx = (struct si_context *)ctx;
struct si_pipe_shader_selector *sel = state;
if (rctx->vs_shader == sel)
return;
- rctx->vs_shader = sel;
+ if (!sel || !sel->current)
+ return;
- if (sel && sel->current)
- si_pm4_bind_state(rctx, vs, sel->current->pm4);
- else
- si_pm4_bind_state(rctx, vs, rctx->dummy_pixel_shader->pm4);
+ rctx->vs_shader = sel;
+ si_pm4_bind_state(rctx, vs, sel->current->pm4);
+ rctx->b.streamout.stride_in_dw = sel->so.stride;
+ rctx->b.flags |= R600_CONTEXT_INV_SHADER_CACHE;
}
static void si_bind_ps_shader(struct pipe_context *ctx, void *state)
{
- struct r600_context *rctx = (struct r600_context *)ctx;
+ struct si_context *rctx = (struct si_context *)ctx;
struct si_pipe_shader_selector *sel = state;
if (rctx->ps_shader == sel)
return;
- rctx->ps_shader = sel;
+ if (!sel || !sel->current)
+ sel = rctx->dummy_pixel_shader;
- if (sel && sel->current)
- si_pm4_bind_state(rctx, ps, sel->current->pm4);
- else
- si_pm4_bind_state(rctx, ps, rctx->dummy_pixel_shader->pm4);
+ rctx->ps_shader = sel;
+ si_pm4_bind_state(rctx, ps, sel->current->pm4);
+ rctx->b.flags |= R600_CONTEXT_INV_SHADER_CACHE;
}
static void si_delete_shader_selector(struct pipe_context *ctx,
struct si_pipe_shader_selector *sel)
{
- struct r600_context *rctx = (struct r600_context *)ctx;
+ struct si_context *rctx = (struct si_context *)ctx;
struct si_pipe_shader *p = sel->current, *c;
while (p) {
static void si_delete_vs_shader(struct pipe_context *ctx, void *state)
{
- struct r600_context *rctx = (struct r600_context *)ctx;
+ struct si_context *rctx = (struct si_context *)ctx;
struct si_pipe_shader_selector *sel = (struct si_pipe_shader_selector *)state;
if (rctx->vs_shader == sel) {
static void si_delete_ps_shader(struct pipe_context *ctx, void *state)
{
- struct r600_context *rctx = (struct r600_context *)ctx;
+ struct si_context *rctx = (struct si_context *)ctx;
struct si_pipe_shader_selector *sel = (struct si_pipe_shader_selector *)state;
if (rctx->ps_shader == sel) {
const struct pipe_sampler_view *state)
{
struct si_pipe_sampler_view *view = CALLOC_STRUCT(si_pipe_sampler_view);
- struct r600_resource_texture *tmp = (struct r600_resource_texture*)texture;
+ struct r600_texture *tmp = (struct r600_texture*)texture;
const struct util_format_description *desc;
unsigned format, num_format;
uint32_t pitch = 0;
/* initialize base object */
view->base = *state;
view->base.texture = NULL;
- pipe_reference(NULL, &texture->reference);
- view->base.texture = texture;
+ pipe_resource_reference(&view->base.texture, texture);
view->base.reference.count = 1;
view->base.context = ctx;
+ view->resource = &tmp->resource;
+
+ /* Buffer resource. */
+ if (texture->target == PIPE_BUFFER) {
+ unsigned stride;
+
+ desc = util_format_description(state->format);
+ first_non_void = util_format_get_first_non_void_channel(state->format);
+ stride = desc->block.bits / 8;
+ va = r600_resource_va(ctx->screen, texture) + state->u.buf.first_element*stride;
+ format = si_translate_buffer_dataformat(ctx->screen, desc, first_non_void);
+ num_format = si_translate_buffer_numformat(ctx->screen, desc, first_non_void);
+
+ view->state[0] = va;
+ view->state[1] = S_008F04_BASE_ADDRESS_HI(va >> 32) |
+ S_008F04_STRIDE(stride);
+ view->state[2] = state->u.buf.last_element + 1 - state->u.buf.first_element;
+ view->state[3] = S_008F0C_DST_SEL_X(si_map_swizzle(desc->swizzle[0])) |
+ S_008F0C_DST_SEL_Y(si_map_swizzle(desc->swizzle[1])) |
+ S_008F0C_DST_SEL_Z(si_map_swizzle(desc->swizzle[2])) |
+ S_008F0C_DST_SEL_W(si_map_swizzle(desc->swizzle[3])) |
+ S_008F0C_NUM_FORMAT(num_format) |
+ S_008F0C_DATA_FORMAT(format);
+ return &view->base;
+ }
state_swizzle[0] = state->swizzle_r;
state_swizzle[1] = state->swizzle_g;
format = 0;
}
- view->resource = &tmp->resource;
-
/* not supported any more */
//endian = si_colorformat_endian_swap(format);
va = r600_resource_va(ctx->screen, texture);
va += surflevel[0].offset;
+ va += tmp->mipmap_shift * surflevel[texture->last_level].slice_size;
view->state[0] = va >> 8;
view->state[1] = (S_008F14_BASE_ADDRESS_HI(va >> 40) |
S_008F14_DATA_FORMAT(format) |
S_008F1C_DST_SEL_Y(si_map_swizzle(swizzle[1])) |
S_008F1C_DST_SEL_Z(si_map_swizzle(swizzle[2])) |
S_008F1C_DST_SEL_W(si_map_swizzle(swizzle[3])) |
- S_008F1C_BASE_LEVEL(state->u.tex.first_level) |
- S_008F1C_LAST_LEVEL(state->u.tex.last_level) |
+ S_008F1C_BASE_LEVEL(texture->nr_samples > 1 ?
+ 0 : state->u.tex.first_level - tmp->mipmap_shift) |
+ S_008F1C_LAST_LEVEL(texture->nr_samples > 1 ?
+ util_logbase2(texture->nr_samples) :
+ state->u.tex.last_level - tmp->mipmap_shift) |
S_008F1C_TILING_INDEX(si_tile_mode_index(tmp, 0, false)) |
S_008F1C_POW2_PAD(texture->last_level > 0) |
- S_008F1C_TYPE(si_tex_dim(texture->target)));
+ S_008F1C_TYPE(si_tex_dim(texture->target, texture->nr_samples)));
view->state[4] = (S_008F20_DEPTH(depth - 1) | S_008F20_PITCH(pitch - 1));
view->state[5] = (S_008F24_BASE_ARRAY(state->u.tex.first_layer) |
S_008F24_LAST_ARRAY(state->u.tex.last_layer));
view->state[6] = 0;
view->state[7] = 0;
+ /* Initialize the sampler view for FMASK. */
+ if (tmp->fmask.size) {
+ uint64_t va = r600_resource_va(ctx->screen, texture) + tmp->fmask.offset;
+ uint32_t fmask_format;
+
+ switch (texture->nr_samples) {
+ case 2:
+ fmask_format = V_008F14_IMG_DATA_FORMAT_FMASK8_S2_F2;
+ break;
+ case 4:
+ fmask_format = V_008F14_IMG_DATA_FORMAT_FMASK8_S4_F4;
+ break;
+ case 8:
+ fmask_format = V_008F14_IMG_DATA_FORMAT_FMASK32_S8_F8;
+ break;
+ default:
+ assert(0);
+ fmask_format = V_008F14_IMG_DATA_FORMAT_INVALID;
+ }
+
+ view->fmask_state[0] = va >> 8;
+ view->fmask_state[1] = S_008F14_BASE_ADDRESS_HI(va >> 40) |
+ S_008F14_DATA_FORMAT(fmask_format) |
+ S_008F14_NUM_FORMAT(V_008F14_IMG_NUM_FORMAT_UINT);
+ view->fmask_state[2] = S_008F18_WIDTH(width - 1) |
+ S_008F18_HEIGHT(height - 1);
+ view->fmask_state[3] = S_008F1C_DST_SEL_X(V_008F1C_SQ_SEL_X) |
+ S_008F1C_DST_SEL_Y(V_008F1C_SQ_SEL_X) |
+ S_008F1C_DST_SEL_Z(V_008F1C_SQ_SEL_X) |
+ S_008F1C_DST_SEL_W(V_008F1C_SQ_SEL_X) |
+ S_008F1C_TILING_INDEX(tmp->fmask.tile_mode_index) |
+ S_008F1C_TYPE(si_tex_dim(texture->target, 0));
+ view->fmask_state[4] = S_008F20_DEPTH(depth - 1) |
+ S_008F20_PITCH(tmp->fmask.pitch - 1);
+ view->fmask_state[5] = S_008F24_BASE_ARRAY(state->u.tex.first_layer) |
+ S_008F24_LAST_ARRAY(state->u.tex.last_layer);
+ view->fmask_state[6] = 0;
+ view->fmask_state[7] = 0;
+ }
+
return &view->base;
}
FREE(resource);
}
+static bool wrap_mode_uses_border_color(unsigned wrap, bool linear_filter)
+{
+ return wrap == PIPE_TEX_WRAP_CLAMP_TO_BORDER ||
+ wrap == PIPE_TEX_WRAP_MIRROR_CLAMP_TO_BORDER ||
+ (linear_filter &&
+ (wrap == PIPE_TEX_WRAP_CLAMP ||
+ wrap == PIPE_TEX_WRAP_MIRROR_CLAMP));
+}
+
+static bool sampler_state_needs_border_color(const struct pipe_sampler_state *state)
+{
+ bool linear_filter = state->min_img_filter != PIPE_TEX_FILTER_NEAREST ||
+ state->mag_img_filter != PIPE_TEX_FILTER_NEAREST;
+
+ return (state->border_color.ui[0] || state->border_color.ui[1] ||
+ state->border_color.ui[2] || state->border_color.ui[3]) &&
+ (wrap_mode_uses_border_color(state->wrap_s, linear_filter) ||
+ wrap_mode_uses_border_color(state->wrap_t, linear_filter) ||
+ wrap_mode_uses_border_color(state->wrap_r, linear_filter));
+}
+
static void *si_create_sampler_state(struct pipe_context *ctx,
const struct pipe_sampler_state *state)
{
struct si_pipe_sampler_state *rstate = CALLOC_STRUCT(si_pipe_sampler_state);
- union util_color uc;
unsigned aniso_flag_offset = state->max_anisotropy > 1 ? 2 : 0;
unsigned border_color_type;
return NULL;
}
- util_pack_color(state->border_color.f, PIPE_FORMAT_A8R8G8B8_UNORM, &uc);
- switch (uc.ui) {
- case 0x000000FF:
- border_color_type = V_008F3C_SQ_TEX_BORDER_COLOR_OPAQUE_BLACK;
- break;
- case 0x00000000:
- border_color_type = V_008F3C_SQ_TEX_BORDER_COLOR_TRANS_BLACK;
- break;
- case 0xFFFFFFFF:
- border_color_type = V_008F3C_SQ_TEX_BORDER_COLOR_OPAQUE_WHITE;
- break;
- default: /* Use border color pointer */
+ if (sampler_state_needs_border_color(state))
border_color_type = V_008F3C_SQ_TEX_BORDER_COLOR_REGISTER;
- }
+ else
+ border_color_type = V_008F3C_SQ_TEX_BORDER_COLOR_TRANS_BLACK;
rstate->val[0] = (S_008F30_CLAMP_X(si_tex_wrap(state->wrap_s)) |
S_008F30_CLAMP_Y(si_tex_wrap(state->wrap_t)) |
rstate->val[3] = S_008F3C_BORDER_COLOR_TYPE(border_color_type);
if (border_color_type == V_008F3C_SQ_TEX_BORDER_COLOR_REGISTER) {
- memcpy(rstate->border_color, state->border_color.f,
+ memcpy(rstate->border_color, state->border_color.ui,
sizeof(rstate->border_color));
}
return rstate;
}
-static struct si_pm4_state *si_set_sampler_view(struct r600_context *rctx,
- unsigned count,
- struct pipe_sampler_view **views,
- struct r600_textures_info *samplers,
- unsigned user_data_reg)
+/* XXX consider moving this function to si_descriptors.c for gcc to inline
+ * the si_set_sampler_view calls. LTO might help too. */
+static void si_set_sampler_views(struct pipe_context *ctx,
+ unsigned shader, unsigned start,
+ unsigned count,
+ struct pipe_sampler_view **views)
{
- struct si_pipe_sampler_view **resource = (struct si_pipe_sampler_view **)views;
- struct si_pm4_state *pm4 = CALLOC_STRUCT(si_pm4_state);
- int i, j;
+ struct si_context *rctx = (struct si_context *)ctx;
+ struct si_textures_info *samplers = &rctx->samplers[shader];
+ struct si_pipe_sampler_view **rviews = (struct si_pipe_sampler_view **)views;
+ int i;
- if (!count)
- goto out;
+ if (shader != PIPE_SHADER_VERTEX && shader != PIPE_SHADER_FRAGMENT)
+ return;
- si_pm4_inval_texture_cache(pm4);
+ assert(start == 0);
- si_pm4_sh_data_begin(pm4);
for (i = 0; i < count; i++) {
- pipe_sampler_view_reference(
- (struct pipe_sampler_view **)&samplers->views[i],
- views[i]);
+ if (!views[i]) {
+ samplers->depth_texture_mask &= ~(1 << i);
+ samplers->compressed_colortex_mask &= ~(1 << i);
+ si_set_sampler_view(rctx, shader, i, NULL, NULL);
+ si_set_sampler_view(rctx, shader, FMASK_TEX_OFFSET + i,
+ NULL, NULL);
+ continue;
+ }
- if (views[i]) {
- struct r600_resource_texture *rtex =
- (struct r600_resource_texture*)views[i]->texture;
+ si_set_sampler_view(rctx, shader, i, views[i], rviews[i]->state);
+
+ if (views[i]->texture->target != PIPE_BUFFER) {
+ struct r600_texture *rtex =
+ (struct r600_texture*)views[i]->texture;
if (rtex->is_depth && !rtex->is_flushing_texture) {
samplers->depth_texture_mask |= 1 << i;
} else {
samplers->depth_texture_mask &= ~(1 << i);
}
+ if (rtex->cmask.size || rtex->fmask.size) {
+ samplers->compressed_colortex_mask |= 1 << i;
+ } else {
+ samplers->compressed_colortex_mask &= ~(1 << i);
+ }
- si_pm4_add_bo(pm4, resource[i]->resource, RADEON_USAGE_READ);
- } else {
- samplers->depth_texture_mask &= ~(1 << i);
- }
-
- for (j = 0; j < Elements(resource[i]->state); ++j) {
- si_pm4_sh_data_add(pm4, resource[i] ? resource[i]->state[j] : 0);
+ if (rtex->fmask.size) {
+ si_set_sampler_view(rctx, shader, FMASK_TEX_OFFSET + i,
+ views[i], rviews[i]->fmask_state);
+ } else {
+ si_set_sampler_view(rctx, shader, FMASK_TEX_OFFSET + i,
+ NULL, NULL);
+ }
}
}
-
- for (i = count; i < NUM_TEX_UNITS; i++) {
- if (samplers->views[i])
- pipe_sampler_view_reference((struct pipe_sampler_view **)&samplers->views[i], NULL);
+ for (; i < samplers->n_views; i++) {
+ samplers->depth_texture_mask &= ~(1 << i);
+ samplers->compressed_colortex_mask &= ~(1 << i);
+ si_set_sampler_view(rctx, shader, i, NULL, NULL);
+ si_set_sampler_view(rctx, shader, FMASK_TEX_OFFSET + i,
+ NULL, NULL);
}
- si_pm4_sh_data_end(pm4, user_data_reg, SI_SGPR_RESOURCE);
-
-out:
- rctx->ps_samplers.n_views = count;
- return pm4;
-}
-
-static void si_set_vs_sampler_view(struct pipe_context *ctx, unsigned count,
- struct pipe_sampler_view **views)
-{
- struct r600_context *rctx = (struct r600_context *)ctx;
- struct si_pm4_state *pm4;
-
- pm4 = si_set_sampler_view(rctx, count, views, &rctx->vs_samplers,
- R_00B130_SPI_SHADER_USER_DATA_VS_0);
- si_pm4_set_state(rctx, vs_sampler_views, pm4);
-}
-
-static void si_set_ps_sampler_view(struct pipe_context *ctx, unsigned count,
- struct pipe_sampler_view **views)
-{
- struct r600_context *rctx = (struct r600_context *)ctx;
- struct si_pm4_state *pm4;
-
- pm4 = si_set_sampler_view(rctx, count, views, &rctx->ps_samplers,
- R_00B030_SPI_SHADER_USER_DATA_PS_0);
- si_pm4_set_state(rctx, ps_sampler_views, pm4);
+ samplers->n_views = count;
+ rctx->b.flags |= R600_CONTEXT_INV_TEX_CACHE;
}
-static struct si_pm4_state *si_bind_sampler(struct r600_context *rctx, unsigned count,
- void **states,
- struct r600_textures_info *samplers,
- unsigned user_data_reg)
+static struct si_pm4_state *si_set_sampler_states(struct si_context *rctx, unsigned count,
+ void **states,
+ struct si_textures_info *samplers,
+ unsigned user_data_reg)
{
struct si_pipe_sampler_state **rstates = (struct si_pipe_sampler_state **)states;
- struct si_pm4_state *pm4 = CALLOC_STRUCT(si_pm4_state);
+ struct si_pm4_state *pm4 = si_pm4_alloc_state(rctx);
uint32_t *border_color_table = NULL;
int i, j;
if (!count)
goto out;
- si_pm4_inval_texture_cache(pm4);
+ rctx->b.flags |= R600_CONTEXT_INV_TEX_CACHE;
si_pm4_sh_data_begin(pm4);
for (i = 0; i < count; i++) {
if (!rctx->border_color_table ||
((rctx->border_color_offset + count - i) &
C_008F3C_BORDER_COLOR_PTR)) {
- si_resource_reference(&rctx->border_color_table, NULL);
+ r600_resource_reference(&rctx->border_color_table, NULL);
rctx->border_color_offset = 0;
rctx->border_color_table =
- si_resource_create_custom(&rctx->screen->screen,
+ r600_resource_create_custom(&rctx->screen->b.b,
PIPE_USAGE_STAGING,
4096 * 4 * 4);
}
if (!border_color_table) {
border_color_table =
- rctx->ws->buffer_map(rctx->border_color_table->cs_buf,
- rctx->cs,
+ rctx->b.ws->buffer_map(rctx->border_color_table->cs_buf,
+ rctx->b.rings.gfx.cs,
PIPE_TRANSFER_WRITE |
PIPE_TRANSFER_UNSYNCHRONIZED);
}
for (j = 0; j < 4; j++) {
- union fi border_color;
-
- border_color.f = rstates[i]->border_color[j];
border_color_table[4 * rctx->border_color_offset + j] =
- util_le32_to_cpu(border_color.i);
+ util_le32_to_cpu(rstates[i]->border_color[j]);
}
rstates[i]->val[3] &= C_008F3C_BORDER_COLOR_PTR;
if (border_color_table) {
uint64_t va_offset =
- r600_resource_va(&rctx->screen->screen,
+ r600_resource_va(&rctx->screen->b.b,
(void*)rctx->border_color_table);
si_pm4_set_reg(pm4, R_028080_TA_BC_BASE_ADDR, va_offset >> 8);
- rctx->ws->buffer_unmap(rctx->border_color_table->cs_buf);
+ if (rctx->b.chip_class >= CIK)
+ si_pm4_set_reg(pm4, R_028084_TA_BC_BASE_ADDR_HI, va_offset >> 40);
+ rctx->b.ws->buffer_unmap(rctx->border_color_table->cs_buf);
si_pm4_add_bo(pm4, rctx->border_color_table, RADEON_USAGE_READ);
}
return pm4;
}
-static void si_bind_vs_sampler(struct pipe_context *ctx, unsigned count, void **states)
+static void si_bind_vs_sampler_states(struct pipe_context *ctx, unsigned count, void **states)
{
- struct r600_context *rctx = (struct r600_context *)ctx;
+ struct si_context *rctx = (struct si_context *)ctx;
struct si_pm4_state *pm4;
- pm4 = si_bind_sampler(rctx, count, states, &rctx->vs_samplers,
+ pm4 = si_set_sampler_states(rctx, count, states, &rctx->samplers[PIPE_SHADER_VERTEX],
R_00B130_SPI_SHADER_USER_DATA_VS_0);
si_pm4_set_state(rctx, vs_sampler, pm4);
}
-static void si_bind_ps_sampler(struct pipe_context *ctx, unsigned count, void **states)
+static void si_bind_ps_sampler_states(struct pipe_context *ctx, unsigned count, void **states)
{
- struct r600_context *rctx = (struct r600_context *)ctx;
+ struct si_context *rctx = (struct si_context *)ctx;
struct si_pm4_state *pm4;
- pm4 = si_bind_sampler(rctx, count, states, &rctx->ps_samplers,
+ pm4 = si_set_sampler_states(rctx, count, states, &rctx->samplers[PIPE_SHADER_FRAGMENT],
R_00B030_SPI_SHADER_USER_DATA_PS_0);
si_pm4_set_state(rctx, ps_sampler, pm4);
}
-static void si_set_sample_mask(struct pipe_context *pipe, unsigned sample_mask)
-{
-}
-static void si_delete_sampler_state(struct pipe_context *ctx, void *state)
+static void si_bind_sampler_states(struct pipe_context *ctx, unsigned shader,
+ unsigned start, unsigned count,
+ void **states)
{
- free(state);
-}
+ assert(start == 0);
-/*
- * Constants
- */
-static void si_set_constant_buffer(struct pipe_context *ctx, uint shader, uint index,
- struct pipe_constant_buffer *input)
-{
- struct r600_context *rctx = (struct r600_context *)ctx;
- struct r600_constbuf_state *state = &rctx->constbuf_state[shader];
- struct pipe_constant_buffer *cb;
- const uint8_t *ptr;
-
- /* Note that the state tracker can unbind constant buffers by
- * passing NULL here.
- */
- if (unlikely(!input || (!input->buffer && !input->user_buffer))) {
- state->enabled_mask &= ~(1 << index);
- state->dirty_mask &= ~(1 << index);
- pipe_resource_reference(&state->cb[index].buffer, NULL);
- return;
- }
+ switch (shader) {
+ case PIPE_SHADER_VERTEX:
+ si_bind_vs_sampler_states(ctx, count, states);
+ break;
+ case PIPE_SHADER_FRAGMENT:
+ si_bind_ps_sampler_states(ctx, count, states);
+ break;
+ default:
+ ;
+ }
+}
- cb = &state->cb[index];
- cb->buffer_size = input->buffer_size;
- ptr = input->user_buffer;
- if (ptr) {
- /* Upload the user buffer. */
- if (R600_BIG_ENDIAN) {
- uint32_t *tmpPtr;
- unsigned i, size = input->buffer_size;
+static void si_set_sample_mask(struct pipe_context *ctx, unsigned sample_mask)
+{
+ struct si_context *rctx = (struct si_context *)ctx;
+ struct si_pm4_state *pm4 = si_pm4_alloc_state(rctx);
+ uint16_t mask = sample_mask;
- if (!(tmpPtr = malloc(size))) {
- R600_ERR("Failed to allocate BE swap buffer.\n");
- return;
- }
+ if (pm4 == NULL)
+ return;
- for (i = 0; i < size / 4; ++i) {
- tmpPtr[i] = bswap_32(((uint32_t *)ptr)[i]);
- }
+ si_pm4_set_reg(pm4, R_028C38_PA_SC_AA_MASK_X0Y0_X1Y0, mask | (mask << 16));
+ si_pm4_set_reg(pm4, R_028C3C_PA_SC_AA_MASK_X0Y1_X1Y1, mask | (mask << 16));
- u_upload_data(rctx->uploader, 0, size, tmpPtr, &cb->buffer_offset, &cb->buffer);
- free(tmpPtr);
- } else {
- u_upload_data(rctx->uploader, 0, input->buffer_size, ptr, &cb->buffer_offset, &cb->buffer);
- }
- } else {
- /* Setup the hw buffer. */
- cb->buffer_offset = input->buffer_offset;
- pipe_resource_reference(&cb->buffer, input->buffer);
- }
+ si_pm4_set_state(rctx, sample_mask, pm4);
+}
- state->enabled_mask |= 1 << index;
- state->dirty_mask |= 1 << index;
+static void si_delete_sampler_state(struct pipe_context *ctx, void *state)
+{
+ free(state);
}
/*
desc = util_format_description(elements[i].src_format);
first_non_void = util_format_get_first_non_void_channel(elements[i].src_format);
- data_format = si_translate_vertexformat(ctx->screen, elements[i].src_format,
- desc, first_non_void);
-
- switch (desc->channel[first_non_void].type) {
- case UTIL_FORMAT_TYPE_FIXED:
- num_format = V_008F0C_BUF_NUM_FORMAT_USCALED; /* XXX */
- break;
- case UTIL_FORMAT_TYPE_SIGNED:
- if (desc->channel[first_non_void].normalized)
- num_format = V_008F0C_BUF_NUM_FORMAT_SNORM;
- else if (desc->channel[first_non_void].pure_integer)
- num_format = V_008F0C_BUF_NUM_FORMAT_SINT;
- else
- num_format = V_008F0C_BUF_NUM_FORMAT_SSCALED;
- break;
- case UTIL_FORMAT_TYPE_UNSIGNED:
- if (desc->channel[first_non_void].normalized)
- num_format = V_008F0C_BUF_NUM_FORMAT_UNORM;
- else if (desc->channel[first_non_void].pure_integer)
- num_format = V_008F0C_BUF_NUM_FORMAT_UINT;
- else
- num_format = V_008F0C_BUF_NUM_FORMAT_USCALED;
- break;
- case UTIL_FORMAT_TYPE_FLOAT:
- default:
- num_format = V_008F14_IMG_NUM_FORMAT_FLOAT;
- }
+ data_format = si_translate_buffer_dataformat(ctx->screen, desc, first_non_void);
+ num_format = si_translate_buffer_numformat(ctx->screen, desc, first_non_void);
v->rsrc_word3[i] = S_008F0C_DST_SEL_X(si_map_swizzle(desc->swizzle[0])) |
S_008F0C_DST_SEL_Y(si_map_swizzle(desc->swizzle[1])) |
static void si_bind_vertex_elements(struct pipe_context *ctx, void *state)
{
- struct r600_context *rctx = (struct r600_context *)ctx;
+ struct si_context *rctx = (struct si_context *)ctx;
struct si_vertex_element *v = (struct si_vertex_element*)state;
rctx->vertex_elements = v;
static void si_delete_vertex_element(struct pipe_context *ctx, void *state)
{
- struct r600_context *rctx = (struct r600_context *)ctx;
+ struct si_context *rctx = (struct si_context *)ctx;
if (rctx->vertex_elements == state)
rctx->vertex_elements = NULL;
static void si_set_vertex_buffers(struct pipe_context *ctx, unsigned start_slot, unsigned count,
const struct pipe_vertex_buffer *buffers)
{
- struct r600_context *rctx = (struct r600_context *)ctx;
+ struct si_context *rctx = (struct si_context *)ctx;
util_set_vertex_buffers_count(rctx->vertex_buffer, &rctx->nr_vertex_buffers, buffers, start_slot, count);
}
static void si_set_index_buffer(struct pipe_context *ctx,
const struct pipe_index_buffer *ib)
{
- struct r600_context *rctx = (struct r600_context *)ctx;
+ struct si_context *rctx = (struct si_context *)ctx;
if (ib) {
pipe_resource_reference(&rctx->index_buffer.buffer, ib->buffer);
static void si_texture_barrier(struct pipe_context *ctx)
{
- struct r600_context *rctx = (struct r600_context *)ctx;
- struct si_pm4_state *pm4 = CALLOC_STRUCT(si_pm4_state);
+ struct si_context *rctx = (struct si_context *)ctx;
- si_pm4_inval_texture_cache(pm4);
- si_pm4_inval_fb_cache(pm4, rctx->framebuffer.nr_cbufs);
- si_pm4_set_state(rctx, texture_barrier, pm4);
+ rctx->b.flags |= R600_CONTEXT_INV_TEX_CACHE |
+ R600_CONTEXT_FLUSH_AND_INV_CB;
}
-void si_init_state_functions(struct r600_context *rctx)
+static void *si_create_blend_custom(struct si_context *rctx, unsigned mode)
{
- rctx->context.create_blend_state = si_create_blend_state;
- rctx->context.bind_blend_state = si_bind_blend_state;
- rctx->context.delete_blend_state = si_delete_blend_state;
- rctx->context.set_blend_color = si_set_blend_color;
+ struct pipe_blend_state blend;
- rctx->context.create_rasterizer_state = si_create_rs_state;
- rctx->context.bind_rasterizer_state = si_bind_rs_state;
- rctx->context.delete_rasterizer_state = si_delete_rs_state;
+ memset(&blend, 0, sizeof(blend));
+ blend.independent_blend_enable = true;
+ blend.rt[0].colormask = 0xf;
+ return si_create_blend_state_mode(&rctx->b.b, &blend, mode);
+}
- rctx->context.create_depth_stencil_alpha_state = si_create_dsa_state;
- rctx->context.bind_depth_stencil_alpha_state = si_bind_dsa_state;
- rctx->context.delete_depth_stencil_alpha_state = si_delete_dsa_state;
- rctx->custom_dsa_flush_depth_stencil = si_create_db_flush_dsa(rctx, true, true);
- rctx->custom_dsa_flush_depth = si_create_db_flush_dsa(rctx, true, false);
- rctx->custom_dsa_flush_stencil = si_create_db_flush_dsa(rctx, false, true);
- rctx->custom_dsa_flush_inplace = si_create_db_flush_dsa(rctx, false, false);
+static struct pipe_surface *r600_create_surface(struct pipe_context *pipe,
+ struct pipe_resource *texture,
+ const struct pipe_surface *surf_tmpl)
+{
+ struct r600_texture *rtex = (struct r600_texture*)texture;
+ struct si_surface *surface = CALLOC_STRUCT(si_surface);
+ unsigned level = surf_tmpl->u.tex.level;
- rctx->context.set_clip_state = si_set_clip_state;
- rctx->context.set_scissor_state = si_set_scissor_state;
- rctx->context.set_viewport_state = si_set_viewport_state;
- rctx->context.set_stencil_ref = si_set_pipe_stencil_ref;
+ if (surface == NULL)
+ return NULL;
- rctx->context.set_framebuffer_state = si_set_framebuffer_state;
+ assert(surf_tmpl->u.tex.first_layer <= util_max_layer(texture, surf_tmpl->u.tex.level));
+ assert(surf_tmpl->u.tex.last_layer <= util_max_layer(texture, surf_tmpl->u.tex.level));
- rctx->context.create_vs_state = si_create_vs_state;
- rctx->context.create_fs_state = si_create_fs_state;
- rctx->context.bind_vs_state = si_bind_vs_shader;
- rctx->context.bind_fs_state = si_bind_ps_shader;
- rctx->context.delete_vs_state = si_delete_vs_shader;
- rctx->context.delete_fs_state = si_delete_ps_shader;
+ pipe_reference_init(&surface->base.reference, 1);
+ pipe_resource_reference(&surface->base.texture, texture);
+ surface->base.context = pipe;
+ surface->base.format = surf_tmpl->format;
+ surface->base.width = rtex->surface.level[level].npix_x;
+ surface->base.height = rtex->surface.level[level].npix_y;
+ surface->base.texture = texture;
+ surface->base.u.tex.first_layer = surf_tmpl->u.tex.first_layer;
+ surface->base.u.tex.last_layer = surf_tmpl->u.tex.last_layer;
+ surface->base.u.tex.level = level;
- rctx->context.create_sampler_state = si_create_sampler_state;
- rctx->context.bind_vertex_sampler_states = si_bind_vs_sampler;
- rctx->context.bind_fragment_sampler_states = si_bind_ps_sampler;
- rctx->context.delete_sampler_state = si_delete_sampler_state;
+ return &surface->base;
+}
- rctx->context.create_sampler_view = si_create_sampler_view;
- rctx->context.set_vertex_sampler_views = si_set_vs_sampler_view;
- rctx->context.set_fragment_sampler_views = si_set_ps_sampler_view;
- rctx->context.sampler_view_destroy = si_sampler_view_destroy;
+static void r600_surface_destroy(struct pipe_context *pipe,
+ struct pipe_surface *surface)
+{
+ pipe_resource_reference(&surface->texture, NULL);
+ FREE(surface);
+}
- rctx->context.set_sample_mask = si_set_sample_mask;
+static boolean si_dma_copy(struct pipe_context *ctx,
+ struct pipe_resource *dst,
+ unsigned dst_level,
+ unsigned dst_x, unsigned dst_y, unsigned dst_z,
+ struct pipe_resource *src,
+ unsigned src_level,
+ const struct pipe_box *src_box)
+{
+ /* XXX implement this or share evergreen_dma_blit with r600g */
+ return FALSE;
+}
- rctx->context.set_constant_buffer = si_set_constant_buffer;
+void si_init_state_functions(struct si_context *rctx)
+{
+ int i;
- rctx->context.create_vertex_elements_state = si_create_vertex_elements;
- rctx->context.bind_vertex_elements_state = si_bind_vertex_elements;
- rctx->context.delete_vertex_elements_state = si_delete_vertex_element;
- rctx->context.set_vertex_buffers = si_set_vertex_buffers;
- rctx->context.set_index_buffer = si_set_index_buffer;
+ rctx->b.b.create_blend_state = si_create_blend_state;
+ rctx->b.b.bind_blend_state = si_bind_blend_state;
+ rctx->b.b.delete_blend_state = si_delete_blend_state;
+ rctx->b.b.set_blend_color = si_set_blend_color;
- rctx->context.create_stream_output_target = si_create_so_target;
- rctx->context.stream_output_target_destroy = si_so_target_destroy;
- rctx->context.set_stream_output_targets = si_set_so_targets;
+ rctx->b.b.create_rasterizer_state = si_create_rs_state;
+ rctx->b.b.bind_rasterizer_state = si_bind_rs_state;
+ rctx->b.b.delete_rasterizer_state = si_delete_rs_state;
- rctx->context.texture_barrier = si_texture_barrier;
- rctx->context.set_polygon_stipple = si_set_polygon_stipple;
+ rctx->b.b.create_depth_stencil_alpha_state = si_create_dsa_state;
+ rctx->b.b.bind_depth_stencil_alpha_state = si_bind_dsa_state;
+ rctx->b.b.delete_depth_stencil_alpha_state = si_delete_dsa_state;
- rctx->context.draw_vbo = si_draw_vbo;
+ for (i = 0; i < 8; i++) {
+ rctx->custom_dsa_flush_depth_stencil[i] = si_create_db_flush_dsa(rctx, true, true, i);
+ rctx->custom_dsa_flush_depth[i] = si_create_db_flush_dsa(rctx, true, false, i);
+ rctx->custom_dsa_flush_stencil[i] = si_create_db_flush_dsa(rctx, false, true, i);
+ }
+ rctx->custom_dsa_flush_inplace = si_create_db_flush_dsa(rctx, false, false, 0);
+ rctx->custom_blend_resolve = si_create_blend_custom(rctx, V_028808_CB_RESOLVE);
+ rctx->custom_blend_decompress = si_create_blend_custom(rctx, V_028808_CB_FMASK_DECOMPRESS);
+
+ rctx->b.b.set_clip_state = si_set_clip_state;
+ rctx->b.b.set_scissor_states = si_set_scissor_states;
+ rctx->b.b.set_viewport_states = si_set_viewport_states;
+ rctx->b.b.set_stencil_ref = si_set_pipe_stencil_ref;
+
+ rctx->b.b.set_framebuffer_state = si_set_framebuffer_state;
+ rctx->b.b.get_sample_position = si_get_sample_position;
+
+ rctx->b.b.create_vs_state = si_create_vs_state;
+ rctx->b.b.create_fs_state = si_create_fs_state;
+ rctx->b.b.bind_vs_state = si_bind_vs_shader;
+ rctx->b.b.bind_fs_state = si_bind_ps_shader;
+ rctx->b.b.delete_vs_state = si_delete_vs_shader;
+ rctx->b.b.delete_fs_state = si_delete_ps_shader;
+
+ rctx->b.b.create_sampler_state = si_create_sampler_state;
+ rctx->b.b.bind_sampler_states = si_bind_sampler_states;
+ rctx->b.b.delete_sampler_state = si_delete_sampler_state;
+
+ rctx->b.b.create_sampler_view = si_create_sampler_view;
+ rctx->b.b.set_sampler_views = si_set_sampler_views;
+ rctx->b.b.sampler_view_destroy = si_sampler_view_destroy;
+
+ rctx->b.b.set_sample_mask = si_set_sample_mask;
+
+ rctx->b.b.create_vertex_elements_state = si_create_vertex_elements;
+ rctx->b.b.bind_vertex_elements_state = si_bind_vertex_elements;
+ rctx->b.b.delete_vertex_elements_state = si_delete_vertex_element;
+ rctx->b.b.set_vertex_buffers = si_set_vertex_buffers;
+ rctx->b.b.set_index_buffer = si_set_index_buffer;
+
+ rctx->b.b.texture_barrier = si_texture_barrier;
+ rctx->b.b.set_polygon_stipple = si_set_polygon_stipple;
+ rctx->b.b.create_surface = r600_create_surface;
+ rctx->b.b.surface_destroy = r600_surface_destroy;
+ rctx->b.dma_copy = si_dma_copy;
+
+ rctx->b.b.draw_vbo = si_draw_vbo;
}
-void si_init_config(struct r600_context *rctx)
+void si_init_config(struct si_context *rctx)
{
- struct si_pm4_state *pm4 = CALLOC_STRUCT(si_pm4_state);
+ struct si_pm4_state *pm4 = si_pm4_alloc_state(rctx);
+
+ if (pm4 == NULL)
+ return;
si_cmd_context_control(pm4);
si_pm4_set_reg(pm4, R_028A40_VGT_GS_MODE, 0x0);
si_pm4_set_reg(pm4, R_028A84_VGT_PRIMITIVEID_EN, 0x0);
si_pm4_set_reg(pm4, R_028A8C_VGT_PRIMITIVEID_RESET, 0x0);
+ si_pm4_set_reg(pm4, R_028B28_VGT_STRMOUT_DRAW_OPAQUE_OFFSET, 0);
si_pm4_set_reg(pm4, R_028B94_VGT_STRMOUT_CONFIG, 0x0);
si_pm4_set_reg(pm4, R_028B98_VGT_STRMOUT_BUFFER_CONFIG, 0x0);
- si_pm4_set_reg(pm4, R_028AA8_IA_MULTI_VGT_PARAM,
- S_028AA8_SWITCH_ON_EOP(1) |
- S_028AA8_PARTIAL_VS_WAVE_ON(1) |
- S_028AA8_PRIMGROUP_SIZE(63));
+ if (rctx->b.chip_class == SI) {
+ si_pm4_set_reg(pm4, R_028AA8_IA_MULTI_VGT_PARAM,
+ S_028AA8_SWITCH_ON_EOP(1) |
+ S_028AA8_PARTIAL_VS_WAVE_ON(1) |
+ S_028AA8_PRIMGROUP_SIZE(63));
+ }
si_pm4_set_reg(pm4, R_028AB4_VGT_REUSE_OFF, 0x00000000);
si_pm4_set_reg(pm4, R_028AB8_VGT_VTX_CNT_EN, 0x0);
- si_pm4_set_reg(pm4, R_008A14_PA_CL_ENHANCE, S_008A14_NUM_CLIP_SEQ(3) |
- S_008A14_CLIP_VTX_REORDER_ENA(1));
+ if (rctx->b.chip_class < CIK)
+ si_pm4_set_reg(pm4, R_008A14_PA_CL_ENHANCE, S_008A14_NUM_CLIP_SEQ(3) |
+ S_008A14_CLIP_VTX_REORDER_ENA(1));
si_pm4_set_reg(pm4, R_028B54_VGT_SHADER_STAGES_EN, 0);
si_pm4_set_reg(pm4, R_028BD4_PA_SC_CENTROID_PRIORITY_0, 0x76543210);
si_pm4_set_reg(pm4, R_028BD8_PA_SC_CENTROID_PRIORITY_1, 0xfedcba98);
- si_pm4_set_reg(pm4, R_028804_DB_EQAA, 0x110000);
-
si_pm4_set_reg(pm4, R_02882C_PA_SU_PRIM_FILTER_CNTL, 0);
- switch (rctx->screen->family) {
- case CHIP_TAHITI:
- case CHIP_PITCAIRN:
- si_pm4_set_reg(pm4, R_028350_PA_SC_RASTER_CONFIG, 0x2a00126a);
- break;
- case CHIP_VERDE:
- si_pm4_set_reg(pm4, R_028350_PA_SC_RASTER_CONFIG, 0x0000124a);
- break;
- case CHIP_OLAND:
- si_pm4_set_reg(pm4, R_028350_PA_SC_RASTER_CONFIG, 0x00000082);
- break;
- case CHIP_HAINAN:
- si_pm4_set_reg(pm4, R_028350_PA_SC_RASTER_CONFIG, 0x00000000);
- break;
- default:
- si_pm4_set_reg(pm4, R_028350_PA_SC_RASTER_CONFIG, 0x00000000);
- break;
+ if (rctx->b.chip_class >= CIK) {
+ switch (rctx->screen->b.family) {
+ case CHIP_BONAIRE:
+ si_pm4_set_reg(pm4, R_028350_PA_SC_RASTER_CONFIG, 0x16000012);
+ si_pm4_set_reg(pm4, R_028354_PA_SC_RASTER_CONFIG_1, 0x00000000);
+ break;
+ case CHIP_HAWAII:
+ si_pm4_set_reg(pm4, R_028350_PA_SC_RASTER_CONFIG, 0x3a00161a);
+ si_pm4_set_reg(pm4, R_028354_PA_SC_RASTER_CONFIG_1, 0x0000002e);
+ break;
+ case CHIP_KAVERI:
+ /* XXX todo */
+ case CHIP_KABINI:
+ /* XXX todo */
+ default:
+ si_pm4_set_reg(pm4, R_028350_PA_SC_RASTER_CONFIG, 0x00000000);
+ si_pm4_set_reg(pm4, R_028354_PA_SC_RASTER_CONFIG_1, 0x00000000);
+ break;
+ }
+ } else {
+ switch (rctx->screen->b.family) {
+ case CHIP_TAHITI:
+ case CHIP_PITCAIRN:
+ si_pm4_set_reg(pm4, R_028350_PA_SC_RASTER_CONFIG, 0x2a00126a);
+ break;
+ case CHIP_VERDE:
+ si_pm4_set_reg(pm4, R_028350_PA_SC_RASTER_CONFIG, 0x0000124a);
+ break;
+ case CHIP_OLAND:
+ si_pm4_set_reg(pm4, R_028350_PA_SC_RASTER_CONFIG, 0x00000082);
+ break;
+ case CHIP_HAINAN:
+ si_pm4_set_reg(pm4, R_028350_PA_SC_RASTER_CONFIG, 0x00000000);
+ break;
+ default:
+ si_pm4_set_reg(pm4, R_028350_PA_SC_RASTER_CONFIG, 0x00000000);
+ break;
+ }
+ }
+
+ si_pm4_set_reg(pm4, R_028200_PA_SC_WINDOW_OFFSET, 0x00000000);
+ si_pm4_set_reg(pm4, R_028230_PA_SC_EDGERULE, 0xAAAAAAAA);
+ si_pm4_set_reg(pm4, R_0282D0_PA_SC_VPORT_ZMIN_0, 0x00000000);
+ si_pm4_set_reg(pm4, R_0282D4_PA_SC_VPORT_ZMAX_0, 0x3F800000);
+ si_pm4_set_reg(pm4, R_028818_PA_CL_VTE_CNTL, 0x0000043F);
+ si_pm4_set_reg(pm4, R_028820_PA_CL_NANINF_CNTL, 0x00000000);
+ si_pm4_set_reg(pm4, R_028BE8_PA_CL_GB_VERT_CLIP_ADJ, 0x3F800000);
+ si_pm4_set_reg(pm4, R_028BEC_PA_CL_GB_VERT_DISC_ADJ, 0x3F800000);
+ si_pm4_set_reg(pm4, R_028BF0_PA_CL_GB_HORZ_CLIP_ADJ, 0x3F800000);
+ si_pm4_set_reg(pm4, R_028BF4_PA_CL_GB_HORZ_DISC_ADJ, 0x3F800000);
+ si_pm4_set_reg(pm4, R_028020_DB_DEPTH_BOUNDS_MIN, 0x00000000);
+ si_pm4_set_reg(pm4, R_028024_DB_DEPTH_BOUNDS_MAX, 0x00000000);
+ si_pm4_set_reg(pm4, R_028028_DB_STENCIL_CLEAR, 0x00000000);
+ si_pm4_set_reg(pm4, R_02802C_DB_DEPTH_CLEAR, 0x3F800000);
+ si_pm4_set_reg(pm4, R_028AC0_DB_SRESULTS_COMPARE_STATE0, 0x0);
+ si_pm4_set_reg(pm4, R_028AC4_DB_SRESULTS_COMPARE_STATE1, 0x0);
+ si_pm4_set_reg(pm4, R_028AC8_DB_PRELOAD_CONTROL, 0x0);
+ si_pm4_set_reg(pm4, R_02800C_DB_RENDER_OVERRIDE,
+ S_02800C_FORCE_HIS_ENABLE0(V_02800C_FORCE_DISABLE) |
+ S_02800C_FORCE_HIS_ENABLE1(V_02800C_FORCE_DISABLE));
+ si_pm4_set_reg(pm4, R_028400_VGT_MAX_VTX_INDX, ~0);
+ si_pm4_set_reg(pm4, R_028404_VGT_MIN_VTX_INDX, 0);
+
+ if (rctx->b.chip_class >= CIK) {
+ si_pm4_set_reg(pm4, R_00B118_SPI_SHADER_PGM_RSRC3_VS, S_00B118_CU_EN(0xffff));
+ si_pm4_set_reg(pm4, R_00B11C_SPI_SHADER_LATE_ALLOC_VS, S_00B11C_LIMIT(0));
+ si_pm4_set_reg(pm4, R_00B01C_SPI_SHADER_PGM_RSRC3_PS, S_00B01C_CU_EN(0xffff));
}
si_pm4_set_state(rctx, init, pm4);
projects / mesa.git / blobdiff