gallium: add a cap for VIEWPORT_SUBPIXEL_BITS (v2)

[mesa.git] / src / gallium / drivers / nouveau / nvc0 / nvc0_shader_state.c

diff --git a/src/gallium/drivers/nouveau/nvc0/nvc0_shader_state.c b/src/gallium/drivers/nouveau/nvc0/nvc0_shader_state.c
index 8595800592c231b7b286c315bacc519918aaed2d..01fe7ce9bfc34303a798a1cc1e7f3120f5246072 100644 (file)
--- a/src/gallium/drivers/nouveau/nvc0/nvc0_shader_state.c
+++ b/src/gallium/drivers/nouveau/nvc0/nvc0_shader_state.c
@@ -28,6 +28,8 @@
 #include "nvc0/nvc0_context.h"
 #include "nvc0/nvc0_query_hw.h"
 
+#include "nvc0/nvc0_compute.xml.h"
+
 static inline void
 nvc0_program_update_context_state(struct nvc0_context *nvc0,
                                   struct nvc0_program *prog, int stage)
@@ -37,11 +39,11 @@ nvc0_program_update_context_state(struct nvc0_context *nvc0,
    if (prog && prog->need_tls) {
       const uint32_t flags = NV_VRAM_DOMAIN(&nvc0->screen->base) | NOUVEAU_BO_RDWR;
       if (!nvc0->state.tls_required)
-         BCTX_REFN_bo(nvc0->bufctx_3d, TLS, flags, nvc0->screen->tls);
+         BCTX_REFN_bo(nvc0->bufctx_3d, 3D_TLS, flags, nvc0->screen->tls);
       nvc0->state.tls_required |= 1 << stage;
    } else {
       if (nvc0->state.tls_required == (1 << stage))
-         nouveau_bufctx_reset(nvc0->bufctx_3d, NVC0_BIND_TLS);
+         nouveau_bufctx_reset(nvc0->bufctx_3d, NVC0_BIND_3D_TLS);
       nvc0->state.tls_required &= ~(1 << stage);
    }
 
@@ -72,7 +74,7 @@ nvc0_program_validate(struct nvc0_context *nvc0, struct nvc0_program *prog)
 
    if (!prog->translated) {
       prog->translated = nvc0_program_translate(
-         prog, nvc0->screen->base.device->chipset);
+         prog, nvc0->screen->base.device->chipset, &nvc0->base.debug);
       if (!prog->translated)
          return false;
    }
@@ -152,7 +154,7 @@ nvc0_fragprog_validate(struct nvc0_context *nvc0)
                                      NVC0_3D_SHADE_MODEL_SMOOTH);
    }
 
-   if (fp->mem && !(nvc0->dirty & NVC0_NEW_FRAGPROG)) {
+   if (fp->mem && !(nvc0->dirty_3d & NVC0_NEW_3D_FRAGPROG)) {
       return;
    }
 
@@ -236,11 +238,8 @@ nvc0_gmtyprog_validate(struct nvc0_context *nvc0)
    struct nouveau_pushbuf *push = nvc0->base.pushbuf;
    struct nvc0_program *gp = nvc0->gmtyprog;
 
-   if (gp)
-      nvc0_program_validate(nvc0, gp);
-
    /* we allow GPs with no code for specifying stream output state only */
-   if (gp && gp->code_size) {
+   if (gp && nvc0_program_validate(nvc0, gp) && gp->code_size) {
       const bool gp_selects_layer = !!(gp->hdr[13] & (1 << 9));
 
       BEGIN_NVC0(push, NVC0_3D(MACRO_GP_SELECT), 1);
@@ -259,6 +258,19 @@ nvc0_gmtyprog_validate(struct nvc0_context *nvc0)
    nvc0_program_update_context_state(nvc0, gp, 3);
 }
 
+void
+nvc0_compprog_validate(struct nvc0_context *nvc0)
+{
+   struct nouveau_pushbuf *push = nvc0->base.pushbuf;
+   struct nvc0_program *cp = nvc0->compprog;
+
+   if (cp && !nvc0_program_validate(nvc0, cp))
+      return;
+
+   BEGIN_NVC0(push, NVC0_CP(FLUSH), 1);
+   PUSH_DATA (push, NVC0_COMPUTE_FLUSH_CODE);
+}
+
 void
 nvc0_tfb_validate(struct nvc0_context *nvc0)
 {
@@ -295,9 +307,8 @@ nvc0_tfb_validate(struct nvc0_context *nvc0)
    }
    nvc0->state.tfb = tfb;
 
-   if (!(nvc0->dirty & NVC0_NEW_TFB_TARGETS))
+   if (!(nvc0->dirty_3d & NVC0_NEW_3D_TFB_TARGETS))
       return;
-   nouveau_bufctx_reset(nvc0->bufctx_3d, NVC0_BIND_TFB);
 
    for (b = 0; b < nvc0->num_tfbbufs; ++b) {
       struct nvc0_so_target *targ = nvc0_so_target(nvc0->tfbbuf[b]);
@@ -313,13 +324,14 @@ nvc0_tfb_validate(struct nvc0_context *nvc0)
 
       buf = nv04_resource(targ->pipe.buffer);
 
-      BCTX_REFN(nvc0->bufctx_3d, TFB, buf, WR);
+      BCTX_REFN(nvc0->bufctx_3d, 3D_TFB, buf, WR);
 
       if (!(nvc0->tfbbuf_dirty & (1 << b)))
          continue;
 
       if (!targ->clean)
-         nvc0_hw_query_fifo_wait(push, nvc0_query(targ->pq));
+         nvc0_hw_query_fifo_wait(nvc0, nvc0_query(targ->pq));
+      nouveau_pushbuf_space(push, 0, 0, 1);
       BEGIN_NVC0(push, NVC0_3D(TFB_BUFFER_ENABLE(b)), 5);
       PUSH_DATA (push, 1);
       PUSH_DATAh(push, buf->address + targ->pipe.buffer_offset);