From: Marek Olšák <[email protected]>
---
src/gallium/drivers/radeonsi/si_hw_context.c | 1 -
src/gallium/drivers/radeonsi/si_pipe.h | 1 -
src/gallium/drivers/radeonsi/si_state_draw.c | 19 -----------
src/gallium/drivers/radeonsi/si_state_shaders.c | 43 +++++++++++++++++++++++++
4 files changed, 43 insertions(+), 21 deletions(-)
diff --git a/src/gallium/drivers/radeonsi/si_hw_context.c
b/src/gallium/drivers/radeonsi/si_hw_context.c
index d862e26..e5da730 100644
--- a/src/gallium/drivers/radeonsi/si_hw_context.c
+++ b/src/gallium/drivers/radeonsi/si_hw_context.c
@@ -244,19 +244,18 @@ void si_begin_new_cs(struct si_context *ctx)
* the first draw call. */
si_invalidate_draw_sh_constants(ctx);
ctx->last_index_size = -1;
ctx->last_primitive_restart_en = -1;
ctx->last_restart_index = SI_RESTART_INDEX_UNKNOWN;
ctx->last_gs_out_prim = -1;
ctx->last_prim = -1;
ctx->last_multi_vgt_param = -1;
ctx->last_rast_prim = -1;
ctx->last_sc_line_stipple = ~0;
- ctx->last_vtx_reuse_depth = -1;
ctx->emit_scratch_reloc = true;
ctx->last_ls = NULL;
ctx->last_tcs = NULL;
ctx->last_tes_sh_base = -1;
ctx->last_num_tcs_input_cp = -1;
ctx->cs_shader_state.initialized = false;
}
diff --git a/src/gallium/drivers/radeonsi/si_pipe.h
b/src/gallium/drivers/radeonsi/si_pipe.h
index 13b1e34..18cd25c 100644
--- a/src/gallium/drivers/radeonsi/si_pipe.h
+++ b/src/gallium/drivers/radeonsi/si_pipe.h
@@ -319,21 +319,20 @@ struct si_context {
int last_start_instance;
int last_drawid;
int last_sh_base_reg;
int last_primitive_restart_en;
int last_restart_index;
int last_gs_out_prim;
int last_prim;
int last_multi_vgt_param;
int last_rast_prim;
unsigned last_sc_line_stipple;
- int last_vtx_reuse_depth;
int current_rast_prim; /* primitive type after TES,
GS */
bool gs_tri_strip_adj_fix;
/* Scratch buffer */
struct r600_resource *scratch_buffer;
bool emit_scratch_reloc;
unsigned scratch_waves;
unsigned spi_tmpring_size;
struct r600_resource *compute_scratch_buffer;
diff --git a/src/gallium/drivers/radeonsi/si_state_draw.c
b/src/gallium/drivers/radeonsi/si_state_draw.c
index cce5f30..3ce19aa 100644
--- a/src/gallium/drivers/radeonsi/si_state_draw.c
+++ b/src/gallium/drivers/radeonsi/si_state_draw.c
@@ -461,39 +461,20 @@ static void si_emit_rasterizer_prim_state(struct
si_context *sctx)
}
static void si_emit_draw_registers(struct si_context *sctx,
const struct pipe_draw_info *info)
{
struct radeon_winsys_cs *cs = sctx->b.gfx.cs;
unsigned prim = si_conv_pipe_prim(info->mode);
unsigned gs_out_prim = si_conv_prim_to_gs_out(sctx->current_rast_prim);
unsigned ia_multi_vgt_param, num_patches = 0;
- /* Polaris needs different VTX_REUSE_DEPTH settings depending on
- * whether the "fractional odd" tessellation spacing is used.
- */
- if (sctx->b.family >= CHIP_POLARIS10) {
- struct si_shader_selector *tes = sctx->tes_shader.cso;
- unsigned vtx_reuse_depth = 30;
-
- if (tes &&
- tes->info.properties[TGSI_PROPERTY_TES_SPACING] ==
- PIPE_TESS_SPACING_FRACTIONAL_ODD)
- vtx_reuse_depth = 14;
-
- if (vtx_reuse_depth != sctx->last_vtx_reuse_depth) {
- radeon_set_context_reg(cs,
R_028C58_VGT_VERTEX_REUSE_BLOCK_CNTL,
- vtx_reuse_depth);
- sctx->last_vtx_reuse_depth = vtx_reuse_depth;
- }
- }
-
if (sctx->tes_shader.cso)
si_emit_derived_tess_state(sctx, info, &num_patches);
ia_multi_vgt_param = si_get_ia_multi_vgt_param(sctx, info, num_patches);
/* Draw state. */
if (ia_multi_vgt_param != sctx->last_multi_vgt_param) {
if (sctx->b.chip_class >= CIK)
radeon_set_context_reg_idx(cs,
R_028AA8_IA_MULTI_VGT_PARAM, 1, ia_multi_vgt_param);
else
diff --git a/src/gallium/drivers/radeonsi/si_state_shaders.c
b/src/gallium/drivers/radeonsi/si_state_shaders.c
index 02f8d6c..94fffdb 100644
--- a/src/gallium/drivers/radeonsi/si_state_shaders.c
+++ b/src/gallium/drivers/radeonsi/si_state_shaders.c
@@ -324,20 +324,59 @@ static void si_set_tesseval_regs(struct si_screen
*sscreen,
} else
distribution_mode = V_028B6C_DISTRIBUTION_MODE_NO_DIST;
si_pm4_set_reg(pm4, R_028B6C_VGT_TF_PARAM,
S_028B6C_TYPE(type) |
S_028B6C_PARTITIONING(partitioning) |
S_028B6C_TOPOLOGY(topology) |
S_028B6C_DISTRIBUTION_MODE(distribution_mode));
}
+/* Polaris needs different VTX_REUSE_DEPTH settings depending on
+ * whether the "fractional odd" tessellation spacing is used.
+ *
+ * Possible VGT configurations and which state should set the register:
+ *
+ * Reg set in | VGT shader configuration | Value
+ * ------------------------------------------------------
+ * VS as VS | VS | 30
+ * VS as ES | ES -> GS -> VS | 30
+ * TES as VS | LS -> HS -> VS | 14 or 30
+ * TES as ES | LS -> HS -> ES -> GS -> VS | 14 or 30
+ */
+static void polaris_set_vgt_vertex_reuse(struct si_screen *sscreen,
+ struct si_shader *shader,
+ struct si_pm4_state *pm4)
+{
+ unsigned type = shader->selector->type;
+
+ if (sscreen->b.family < CHIP_POLARIS10)
+ return;
+
+ /* VS as VS, or VS as ES: */
+ if ((type == PIPE_SHADER_VERTEX &&
+ !shader->key.as_ls &&
+ !shader->is_gs_copy_shader) ||
+ /* TES as VS, or TES as ES: */
+ type == PIPE_SHADER_TESS_EVAL) {
+ unsigned vtx_reuse_depth = 30;
+
+ if (type == PIPE_SHADER_TESS_EVAL &&
+
shader->selector->info.properties[TGSI_PROPERTY_TES_SPACING] ==
+ PIPE_TESS_SPACING_FRACTIONAL_ODD)
+ vtx_reuse_depth = 14;
+
+ si_pm4_set_reg(pm4, R_028C58_VGT_VERTEX_REUSE_BLOCK_CNTL,
+ vtx_reuse_depth);
+ }
+}
+
static struct si_pm4_state *si_get_shader_pm4_state(struct si_shader *shader)
{
if (shader->pm4)
si_pm4_clear_state(shader->pm4);
else
shader->pm4 = CALLOC_STRUCT(si_pm4_state);
return shader->pm4;
}
@@ -431,20 +470,22 @@ static void si_shader_es(struct si_screen *sscreen,
struct si_shader *shader)
S_00B328_VGPR_COMP_CNT(vgpr_comp_cnt) |
S_00B328_DX10_CLAMP(1) |
S_00B328_FLOAT_MODE(shader->config.float_mode));
si_pm4_set_reg(pm4, R_00B32C_SPI_SHADER_PGM_RSRC2_ES,
S_00B32C_USER_SGPR(num_user_sgprs) |
S_00B32C_OC_LDS_EN(oc_lds_en) |
S_00B32C_SCRATCH_EN(shader->config.scratch_bytes_per_wave > 0));
if (shader->selector->type == PIPE_SHADER_TESS_EVAL)
si_set_tesseval_regs(sscreen, shader, pm4);
+
+ polaris_set_vgt_vertex_reuse(sscreen, shader, pm4);
}
/**
* Calculate the appropriate setting of VGT_GS_MODE when \p shader is a
* geometry shader.
*/
static uint32_t si_vgt_gs_mode(struct si_shader_selector *sel)
{
unsigned gs_max_vert_out = sel->gs_max_out_vertices;
unsigned cut_mode;
@@ -618,20 +659,22 @@ static void si_shader_vs(struct si_screen *sscreen,
struct si_shader *shader,
S_028818_VTX_XY_FMT(1) | S_028818_VTX_Z_FMT(1));
else
si_pm4_set_reg(pm4, R_028818_PA_CL_VTE_CNTL,
S_028818_VTX_W0_FMT(1) |
S_028818_VPORT_X_SCALE_ENA(1) |
S_028818_VPORT_X_OFFSET_ENA(1) |
S_028818_VPORT_Y_SCALE_ENA(1) |
S_028818_VPORT_Y_OFFSET_ENA(1) |
S_028818_VPORT_Z_SCALE_ENA(1) |
S_028818_VPORT_Z_OFFSET_ENA(1));
if (shader->selector->type == PIPE_SHADER_TESS_EVAL)
si_set_tesseval_regs(sscreen, shader, pm4);
+
+ polaris_set_vgt_vertex_reuse(sscreen, shader, pm4);
}
static unsigned si_get_ps_num_interp(struct si_shader *ps)
{
struct tgsi_shader_info *info = &ps->selector->info;
unsigned num_colors = !!(info->colors_read & 0x0f) +
!!(info->colors_read & 0xf0);
unsigned num_interp = ps->selector->info.num_inputs +
(ps->key.part.ps.prolog.color_two_side ?
num_colors : 0);
--
2.7.4
_______________________________________________
mesa-dev mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/mesa-dev
