radeonsi: fast exit si_emit_derived_tess_state early

Tested-by: Edmondo Tommasina <edmondo.tommasina@gmail.com>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>

diff --git a/src/gallium/drivers/radeonsi/si_pipe.h b/src/gallium/drivers/radeonsi/si_pipe.h
index 8e6a94deaded46d65f8fa5435a2ca43f04e9ca41..df2f1303c2d3c449913f4f46c25fde5502b5f746 100644 (file)
--- a/src/gallium/drivers/radeonsi/si_pipe.h
+++ b/src/gallium/drivers/radeonsi/si_pipe.h
@@ -336,6 +336,7 @@ struct si_context {
        struct si_shader_selector *last_tcs;
        int                     last_num_tcs_input_cp;
        int                     last_tes_sh_base;
+       unsigned                last_num_patches;
 
        /* Debug state. */
        bool                    is_debug;

diff --git a/src/gallium/drivers/radeonsi/si_state_draw.c b/src/gallium/drivers/radeonsi/si_state_draw.c
index affc156a3d1e8a087b47a9274246fd4c528cbd01..e904164fafad7c675f51c0c583dda72c915d1724 100644 (file)
--- a/src/gallium/drivers/radeonsi/si_state_draw.c
+++ b/src/gallium/drivers/radeonsi/si_state_draw.c
@@ -108,6 +108,19 @@ static void si_emit_derived_tess_state(struct si_context *sctx,
        unsigned tcs_in_layout, tcs_out_layout, tcs_out_offsets;
        unsigned offchip_layout, hardware_lds_size, ls_hs_config;
 
+       if (sctx->last_ls == ls->current &&
+           sctx->last_tcs == tcs &&
+           sctx->last_tes_sh_base == tes_sh_base &&
+           sctx->last_num_tcs_input_cp == num_tcs_input_cp) {
+               *num_patches = sctx->last_num_patches;
+               return;
+       }
+
+       sctx->last_ls = ls->current;
+       sctx->last_tcs = tcs;
+       sctx->last_tes_sh_base = tes_sh_base;
+       sctx->last_num_tcs_input_cp = num_tcs_input_cp;
+
        /* This calculates how shader inputs and outputs among VS, TCS, and TES
         * are laid out in LDS. */
        num_tcs_inputs = util_last_bit64(ls->cso->outputs_written);
@@ -153,6 +166,7 @@ static void si_emit_derived_tess_state(struct si_context *sctx,
         * specific value is taken from the proprietary driver.
         */
        *num_patches = MIN2(*num_patches, 40);
+       sctx->last_num_patches = *num_patches;
 
        output_patch0_offset = input_patch_size * *num_patches;
        perpatch_output_offset = output_patch0_offset + pervertex_output_patch_size;
@@ -168,17 +182,6 @@ static void si_emit_derived_tess_state(struct si_context *sctx,
                ls_rsrc2 |= S_00B52C_LDS_SIZE(align(lds_size, 256) / 256);
        }
 
-       if (sctx->last_ls == ls->current &&
-           sctx->last_tcs == tcs &&
-           sctx->last_tes_sh_base == tes_sh_base &&
-           sctx->last_num_tcs_input_cp == num_tcs_input_cp)
-               return;
-
-       sctx->last_ls = ls->current;
-       sctx->last_tcs = tcs;
-       sctx->last_tes_sh_base = tes_sh_base;
-       sctx->last_num_tcs_input_cp = num_tcs_input_cp;
-
        /* Due to a hw bug, RSRC2_LS must be written twice with another
         * LS register written in between. */
        if (sctx->b.chip_class == CIK && sctx->b.family != CHIP_HAWAII)