From: Jason Ekstrand <jason.ekstrand@intel.com>
Date: Fri, 21 Oct 2016 20:10:52 +0000 (-0700)
Subject: intel/blorp: Add support for vertex shaders
X-Git-Url: https://git.libre-soc.org/?a=commitdiff_plain;h=f7f768d19506dc9e480531ce41d03dba9b4fc792;p=mesa.git

intel/blorp: Add support for vertex shaders

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>
---

diff --git a/src/intel/blorp/blorp.c b/src/intel/blorp/blorp.c
index 5952a117eb5..08ce97d9e5d 100644
--- a/src/intel/blorp/blorp.c
+++ b/src/intel/blorp/blorp.c
@@ -190,6 +190,37 @@ blorp_compile_fs(struct blorp_context *blorp, void *mem_ctx,
    return program;
 }
 
+const unsigned *
+blorp_compile_vs(struct blorp_context *blorp, void *mem_ctx,
+                 struct nir_shader *nir,
+                 struct brw_vs_prog_data *vs_prog_data,
+                 unsigned *program_size)
+{
+   const struct brw_compiler *compiler = blorp->compiler;
+
+   nir->options =
+      compiler->glsl_compiler_options[MESA_SHADER_VERTEX].NirOptions;
+
+   nir = brw_preprocess_nir(compiler, nir);
+   nir_shader_gather_info(nir, nir_shader_get_entrypoint(nir));
+
+   vs_prog_data->inputs_read = nir->info->inputs_read;
+
+   brw_compute_vue_map(compiler->devinfo,
+                       &vs_prog_data->base.vue_map,
+                       nir->info->outputs_written,
+                       nir->info->separate_shader);
+
+   struct brw_vs_prog_key vs_key = { 0, };
+
+   const unsigned *program =
+      brw_compile_vs(compiler, blorp->driver_ctx, mem_ctx,
+                     &vs_key, vs_prog_data, nir,
+                     NULL, false, -1, program_size, NULL);
+
+   return program;
+}
+
 void
 blorp_gen6_hiz_op(struct blorp_batch *batch,
                   struct blorp_surf *surf, unsigned level, unsigned layer,
diff --git a/src/intel/blorp/blorp_genX_exec.h b/src/intel/blorp/blorp_genX_exec.h
index 78fb3ffe4ca..d278c1f58b4 100644
--- a/src/intel/blorp/blorp_genX_exec.h
+++ b/src/intel/blorp/blorp_genX_exec.h
@@ -201,8 +201,9 @@ blorp_emit_input_varying_data(struct blorp_batch *batch,
    const uint32_t *const inputs_src = (const uint32_t *)&params->wm_inputs;
    uint32_t *inputs = blorp_alloc_vertex_buffer(batch, *size, addr);
 
-   /* Zero data for the VUE header */
-   memset(inputs, 0, 4 * sizeof(uint32_t));
+   /* Copy in the VS inputs */
+   assert(sizeof(params->vs_inputs) == 16);
+   memcpy(inputs, &params->vs_inputs, sizeof(params->vs_inputs));
    inputs += 4;
 
    if (params->wm_prog_data) {
@@ -333,7 +334,7 @@ blorp_emit_vertex_elements(struct blorp_batch *batch,
    ve[0].Valid = true;
    ve[0].SourceElementFormat = ISL_FORMAT_R32G32B32A32_FLOAT;
    ve[0].SourceElementOffset = 0;
-   ve[0].Component0Control = VFCOMP_STORE_0;
+   ve[0].Component0Control = VFCOMP_STORE_SRC;
 
    /* From Gen8 onwards hardware is no more instructed to overwrite components
     * using an element specifier. Instead one has separate 3DSTATE_VF_SGVS
@@ -344,8 +345,8 @@ blorp_emit_vertex_elements(struct blorp_batch *batch,
 #else
    ve[0].Component1Control = VFCOMP_STORE_IID;
 #endif
-   ve[0].Component2Control = VFCOMP_STORE_0;
-   ve[0].Component3Control = VFCOMP_STORE_0;
+   ve[0].Component2Control = VFCOMP_STORE_SRC;
+   ve[0].Component3Control = VFCOMP_STORE_SRC;
 
    ve[1].VertexBufferIndex = 0;
    ve[1].Valid = true;
@@ -399,6 +400,35 @@ blorp_emit_vertex_elements(struct blorp_batch *batch,
 #endif
 }
 
+static void
+blorp_emit_vs_config(struct blorp_batch *batch,
+                     const struct blorp_params *params)
+{
+   struct brw_vs_prog_data *vs_prog_data = params->vs_prog_data;
+
+   blorp_emit(batch, GENX(3DSTATE_VS), vs) {
+      if (vs_prog_data) {
+         vs.FunctionEnable = true;
+
+         vs.KernelStartPointer = params->vs_prog_kernel;
+
+         vs.DispatchGRFStartRegisterForURBData =
+            vs_prog_data->base.base.dispatch_grf_start_reg;
+         vs.VertexURBEntryReadLength =
+            vs_prog_data->base.urb_read_length;
+         vs.VertexURBEntryReadOffset = 0;
+
+         vs.MaximumNumberofThreads =
+            batch->blorp->isl_dev->info->max_vs_threads - 1;
+
+#if GEN_GEN >= 8
+         vs.SIMD8DispatchEnable =
+            vs_prog_data->base.dispatch_mode == DISPATCH_MODE_SIMD8;
+#endif
+      }
+   }
+}
+
 static void
 blorp_emit_sf_config(struct blorp_batch *batch,
                      const struct blorp_params *params)
@@ -1315,7 +1345,7 @@ blorp_exec(struct blorp_batch *batch, const struct blorp_params *params)
     *
     * We've already done one at the start of the BLORP operation.
     */
-   blorp_emit(batch, GENX(3DSTATE_VS), vs);
+   blorp_emit_vs_config(batch, params);
 #if GEN_GEN >= 7
    blorp_emit(batch, GENX(3DSTATE_HS), hs);
    blorp_emit(batch, GENX(3DSTATE_TE), te);
diff --git a/src/intel/blorp/blorp_priv.h b/src/intel/blorp/blorp_priv.h
index 04aa152193d..b69babc0bab 100644
--- a/src/intel/blorp/blorp_priv.h
+++ b/src/intel/blorp/blorp_priv.h
@@ -152,6 +152,12 @@ struct brw_blorp_wm_inputs
    input; \
 })
 
+struct blorp_vs_inputs {
+   uint32_t base_layer;
+   uint32_t _instance_id; /* Set in hardware by SGVS */
+   uint32_t pad[2];
+};
+
 static inline unsigned
 brw_blorp_get_urb_length(const struct brw_wm_prog_data *prog_data)
 {
@@ -183,9 +189,12 @@ struct blorp_params
    enum blorp_fast_clear_op fast_clear_op;
    bool color_write_disable[4];
    struct brw_blorp_wm_inputs wm_inputs;
+   struct blorp_vs_inputs vs_inputs;
    unsigned num_samples;
    unsigned num_draw_buffers;
    unsigned num_layers;
+   uint32_t vs_prog_kernel;
+   struct brw_vs_prog_data *vs_prog_data;
    uint32_t wm_prog_kernel;
    struct brw_wm_prog_data *wm_prog_data;
 };
@@ -314,6 +323,12 @@ blorp_compile_fs(struct blorp_context *blorp, void *mem_ctx,
                  struct brw_wm_prog_data *wm_prog_data,
                  unsigned *program_size);
 
+const unsigned *
+blorp_compile_vs(struct blorp_context *blorp, void *mem_ctx,
+                 struct nir_shader *nir,
+                 struct brw_vs_prog_data *vs_prog_data,
+                 unsigned *program_size);
+
 /** \} */
 
 #ifdef __cplusplus