X-Git-Url: https://git.libre-soc.org/?a=blobdiff_plain;ds=sidebyside;f=src%2Fgallium%2Fdrivers%2Fnouveau%2Fnv50%2Fnv50_state.c;h=186d126305a6b79127169be02dd869d70b8836a8;hb=a2a1a5805fd617e7f3cc8be44dd79b50da07ebb9;hp=b6a180ef22f73e50a8c8774e2ac5cb7372752e33;hpb=a3ed98f7aa85636579a5696bf036ec13e5c9104a;p=mesa.git

diff --git a/src/gallium/drivers/nouveau/nv50/nv50_state.c b/src/gallium/drivers/nouveau/nv50/nv50_state.c
index b6a180ef22f..186d126305a 100644
--- a/src/gallium/drivers/nouveau/nv50/nv50_state.c
+++ b/src/gallium/drivers/nouveau/nv50/nv50_state.c
@@ -24,7 +24,7 @@
 #include "util/u_helpers.h"
 #include "util/u_inlines.h"
 #include "util/u_transfer.h"
-#include "util/u_format_srgb.h"
+#include "util/format_srgb.h"
 
 #include "tgsi/tgsi_parse.h"
 
@@ -57,16 +57,12 @@
  *  ! pipe_rasterizer_state.flatshade_first also applies to QUADS
  *    (There's a GL query for that, forcing an exception is just ridiculous.)
  *
- *  ! pipe_rasterizer_state.half_pixel_center is ignored - pixel centers
- *     are always at half integer coordinates and the top-left rule applies
- *    (There does not seem to be a hardware switch for this.)
- *
  *  ! pipe_rasterizer_state.sprite_coord_enable is masked with 0xff on NVC0
  *    (The hardware only has 8 slots meant for TexCoord and we have to assign
  *     in advance to maintain elegant separate shader objects.)
  */
 
-static INLINE uint32_t
+static inline uint32_t
 nv50_colormask(unsigned mask)
 {
    uint32_t ret = 0;
@@ -84,9 +80,9 @@ nv50_colormask(unsigned mask)
 }
 
 #define NV50_BLEND_FACTOR_CASE(a, b) \
-   case PIPE_BLENDFACTOR_##a: return NV50_3D_BLEND_FACTOR_##b
+   case PIPE_BLENDFACTOR_##a: return NV50_BLEND_FACTOR_##b
 
-static INLINE uint32_t
+static inline uint32_t
 nv50_blend_fac(unsigned factor)
 {
    switch (factor) {
@@ -110,7 +106,7 @@ nv50_blend_fac(unsigned factor)
    NV50_BLEND_FACTOR_CASE(INV_SRC1_COLOR, ONE_MINUS_SRC1_COLOR);
    NV50_BLEND_FACTOR_CASE(INV_SRC1_ALPHA, ONE_MINUS_SRC1_ALPHA);
    default:
-      return NV50_3D_BLEND_FACTOR_ZERO;
+      return NV50_BLEND_FACTOR_ZERO;
    }
 }
 
@@ -120,7 +116,7 @@ nv50_blend_state_create(struct pipe_context *pipe,
 {
    struct nv50_blend_stateobj *so = CALLOC_STRUCT(nv50_blend_stateobj);
    int i;
-   boolean emit_common_func = cso->rt[0].blend_enable;
+   bool emit_common_func = cso->rt[0].blend_enable;
    uint32_t ms;
 
    if (nv50_context(pipe)->screen->tesla->oclass >= NVA3_3D_CLASS) {
@@ -141,11 +137,11 @@ nv50_blend_state_create(struct pipe_context *pipe,
       for (i = 0; i < 8; ++i) {
          SB_DATA(so, cso->rt[i].blend_enable);
          if (cso->rt[i].blend_enable)
-            emit_common_func = TRUE;
+            emit_common_func = true;
       }
 
       if (nv50_context(pipe)->screen->tesla->oclass >= NVA3_3D_CLASS) {
-         emit_common_func = FALSE;
+         emit_common_func = false;
 
          for (i = 0; i < 8; ++i) {
             if (!cso->rt[i].blend_enable)
@@ -221,7 +217,7 @@ nv50_blend_state_delete(struct pipe_context *pipe, void *hwcso)
    FREE(hwcso);
 }
 
-/* NOTE: ignoring line_last_pixel, using FALSE (set on screen init) */
+/* NOTE: ignoring line_last_pixel */
 static void *
 nv50_rasterizer_state_create(struct pipe_context *pipe,
                              const struct pipe_rasterizer_state *cso)
@@ -235,8 +231,10 @@ nv50_rasterizer_state_create(struct pipe_context *pipe,
    so->pipe = *cso;
 
 #ifndef NV50_SCISSORS_CLIPPING
-   SB_BEGIN_3D(so, SCISSOR_ENABLE(0), 1);
-   SB_DATA    (so, cso->scissor);
+   for (int i = 0; i < NV50_MAX_VIEWPORTS; i++) {
+      SB_BEGIN_3D(so, SCISSOR_ENABLE(i), 1);
+      SB_DATA    (so, cso->scissor);
+   }
 #endif
 
    SB_BEGIN_3D(so, SHADE_MODEL, 1);
@@ -331,6 +329,12 @@ nv50_rasterizer_state_create(struct pipe_context *pipe,
    SB_BEGIN_3D(so, VIEW_VOLUME_CLIP_CTRL, 1);
    SB_DATA    (so, reg);
 
+   SB_BEGIN_3D(so, DEPTH_CLIP_NEGATIVE_Z, 1);
+   SB_DATA    (so, cso->clip_halfz);
+
+   SB_BEGIN_3D(so, PIXEL_CENTER_INTEGER, 1);
+   SB_DATA    (so, !cso->half_pixel_center);
+
    assert(so->size <= (sizeof(so->state) / sizeof(so->state[0])));
    return (void *)so;
 }
@@ -435,7 +439,7 @@ nv50_zsa_state_delete(struct pipe_context *pipe, void *hwcso)
 #define NV50_TSC_WRAP_CASE(n) \
     case PIPE_TEX_WRAP_##n: return NV50_TSC_WRAP_##n
 
-static INLINE unsigned
+static inline unsigned
 nv50_tsc_wrap_mode(unsigned wrap)
 {
    switch (wrap) {
@@ -556,22 +560,25 @@ nv50_sampler_state_delete(struct pipe_context *pipe, void *hwcso)
 {
    unsigned s, i;
 
-   for (s = 0; s < 3; ++s)
+   for (s = 0; s < 3; ++s) {
+      assert(nv50_context(pipe)->num_samplers[s] <= PIPE_MAX_SAMPLERS);
       for (i = 0; i < nv50_context(pipe)->num_samplers[s]; ++i)
          if (nv50_context(pipe)->samplers[s][i] == hwcso)
             nv50_context(pipe)->samplers[s][i] = NULL;
+   }
 
    nv50_screen_tsc_free(nv50_context(pipe)->screen, nv50_tsc_entry(hwcso));
 
    FREE(hwcso);
 }
 
-static INLINE void
+static inline void
 nv50_stage_sampler_states_bind(struct nv50_context *nv50, int s,
                                unsigned nr, void **hwcso)
 {
    unsigned i;
 
+   assert(nr <= PIPE_MAX_SAMPLERS);
    for (i = 0; i < nr; ++i) {
       struct nv50_tsc_entry *old = nv50->samplers[s][i];
 
@@ -579,9 +586,13 @@ nv50_stage_sampler_states_bind(struct nv50_context *nv50, int s,
       if (old)
          nv50_screen_tsc_unlock(nv50->screen, old);
    }
-   for (; i < nv50->num_samplers[s]; ++i)
-      if (nv50->samplers[s][i])
+   assert(nv50->num_samplers[s] <= PIPE_MAX_SAMPLERS);
+   for (; i < nv50->num_samplers[s]; ++i) {
+      if (nv50->samplers[s][i]) {
          nv50_screen_tsc_unlock(nv50->screen, nv50->samplers[s][i]);
+         nv50->samplers[s][i] = NULL;
+      }
+   }
 
    nv50->num_samplers[s] = nr;
 
@@ -639,13 +650,14 @@ nv50_sampler_view_destroy(struct pipe_context *pipe,
    FREE(nv50_tic_entry(view));
 }
 
-static INLINE void
+static inline void
 nv50_stage_set_sampler_views(struct nv50_context *nv50, int s,
                              unsigned nr,
                              struct pipe_sampler_view **views)
 {
    unsigned i;
 
+   assert(nr <= PIPE_MAX_SAMPLERS);
    for (i = 0; i < nr; ++i) {
       struct nv50_tic_entry *old = nv50_tic_entry(nv50->textures[s][i]);
       if (old)
@@ -654,6 +666,7 @@ nv50_stage_set_sampler_views(struct nv50_context *nv50, int s,
       pipe_sampler_view_reference(&nv50->textures[s][i], views[i]);
    }
 
+   assert(nv50->num_textures[s] <= PIPE_MAX_SAMPLERS);
    for (i = nr; i < nv50->num_textures[s]; ++i) {
       struct nv50_tic_entry *old = nv50_tic_entry(nv50->textures[s][i]);
       if (!old)
@@ -786,6 +799,7 @@ nv50_set_constant_buffer(struct pipe_context *pipe, uint shader, uint index,
    if (shader == PIPE_SHADER_COMPUTE)
       return;
 
+   assert(i < NV50_MAX_PIPE_CONSTBUFS);
    if (nv50->constbuf[s][i].user)
       nv50->constbuf[s][i].u.buf = NULL;
    else
@@ -794,15 +808,15 @@ nv50_set_constant_buffer(struct pipe_context *pipe, uint shader, uint index,
 
    pipe_resource_reference(&nv50->constbuf[s][i].u.buf, res);
 
-   nv50->constbuf[s][i].user = (cb && cb->user_buffer) ? TRUE : FALSE;
+   nv50->constbuf[s][i].user = (cb && cb->user_buffer) ? true : false;
    if (nv50->constbuf[s][i].user) {
       nv50->constbuf[s][i].u.data = cb->user_buffer;
-      nv50->constbuf[s][i].size = cb->buffer_size;
+      nv50->constbuf[s][i].size = MIN2(cb->buffer_size, 0x10000);
       nv50->constbuf_valid[s] |= 1 << i;
    } else
    if (res) {
       nv50->constbuf[s][i].offset = cb->buffer_offset;
-      nv50->constbuf[s][i].size = align(cb->buffer_size, 0x100);
+      nv50->constbuf[s][i].size = MIN2(align(cb->buffer_size, 0x100), 0x10000);
       nv50->constbuf_valid[s] |= 1 << i;
    } else {
       nv50->constbuf_valid[s] &= ~(1 << i);
@@ -855,6 +869,16 @@ nv50_set_sample_mask(struct pipe_context *pipe, unsigned sample_mask)
    nv50->dirty |= NV50_NEW_SAMPLE_MASK;
 }
 
+static void
+nv50_set_min_samples(struct pipe_context *pipe, unsigned min_samples)
+{
+   struct nv50_context *nv50 = nv50_context(pipe);
+
+   if (nv50->min_samples != min_samples) {
+      nv50->min_samples = min_samples;
+      nv50->dirty |= NV50_NEW_MIN_SAMPLES;
+   }
+}
 
 static void
 nv50_set_framebuffer_state(struct pipe_context *pipe,
@@ -897,9 +921,16 @@ nv50_set_scissor_states(struct pipe_context *pipe,
                         const struct pipe_scissor_state *scissor)
 {
    struct nv50_context *nv50 = nv50_context(pipe);
+   int i;
 
-   nv50->scissor = *scissor;
-   nv50->dirty |= NV50_NEW_SCISSOR;
+   assert(start_slot + num_scissors <= NV50_MAX_VIEWPORTS);
+   for (i = 0; i < num_scissors; i++) {
+      if (!memcmp(&nv50->scissors[start_slot + i], &scissor[i], sizeof(*scissor)))
+         continue;
+      nv50->scissors[start_slot + i] = scissor[i];
+      nv50->scissors_dirty |= 1 << (start_slot + i);
+      nv50->dirty |= NV50_NEW_SCISSOR;
+   }
 }
 
 static void
@@ -909,9 +940,16 @@ nv50_set_viewport_states(struct pipe_context *pipe,
                          const struct pipe_viewport_state *vpt)
 {
    struct nv50_context *nv50 = nv50_context(pipe);
+   int i;
 
-   nv50->viewport = *vpt;
-   nv50->dirty |= NV50_NEW_VIEWPORT;
+   assert(start_slot + num_viewports <= NV50_MAX_VIEWPORTS);
+   for (i = 0; i < num_viewports; i++) {
+      if (!memcmp(&nv50->viewports[start_slot + i], &vpt[i], sizeof(*vpt)))
+         continue;
+      nv50->viewports[start_slot + i] = vpt[i];
+      nv50->viewports_dirty |= 1 << (start_slot + i);
+      nv50->dirty |= NV50_NEW_VIEWPORT;
+   }
 }
 
 static void
@@ -988,13 +1026,14 @@ nv50_so_target_create(struct pipe_context *pipe,
                       struct pipe_resource *res,
                       unsigned offset, unsigned size)
 {
+   struct nv04_resource *buf = (struct nv04_resource *)res;
    struct nv50_so_target *targ = MALLOC_STRUCT(nv50_so_target);
    if (!targ)
       return NULL;
 
    if (nouveau_context(pipe)->screen->class_3d >= NVA0_3D_CLASS) {
       targ->pq = pipe->create_query(pipe,
-                                    NVA0_QUERY_STREAM_OUTPUT_BUFFER_OFFSET);
+                                    NVA0_QUERY_STREAM_OUTPUT_BUFFER_OFFSET, 0);
       if (!targ->pq) {
          FREE(targ);
          return NULL;
@@ -1002,7 +1041,7 @@ nv50_so_target_create(struct pipe_context *pipe,
    } else {
       targ->pq = NULL;
    }
-   targ->clean = TRUE;
+   targ->clean = true;
 
    targ->pipe.buffer_size = size;
    targ->pipe.buffer_offset = offset;
@@ -1011,6 +1050,9 @@ nv50_so_target_create(struct pipe_context *pipe,
    pipe_resource_reference(&targ->pipe.buffer, res);
    pipe_reference_init(&targ->pipe.reference, 1);
 
+   assert(buf->base.target == PIPE_BUFFER);
+   util_range_add(&buf->valid_buffer_range, offset, offset + size);
+
    return &targ->pipe;
 }
 
@@ -1029,35 +1071,36 @@ static void
 nv50_set_stream_output_targets(struct pipe_context *pipe,
                                unsigned num_targets,
                                struct pipe_stream_output_target **targets,
-                               unsigned append_mask)
+                               const unsigned *offsets)
 {
    struct nv50_context *nv50 = nv50_context(pipe);
    unsigned i;
-   boolean serialize = TRUE;
-   const boolean can_resume = nv50->screen->base.class_3d >= NVA0_3D_CLASS;
+   bool serialize = true;
+   const bool can_resume = nv50->screen->base.class_3d >= NVA0_3D_CLASS;
 
    assert(num_targets <= 4);
 
    for (i = 0; i < num_targets; ++i) {
-      const boolean changed = nv50->so_target[i] != targets[i];
-      if (!changed && (append_mask & (1 << i)))
+      const bool changed = nv50->so_target[i] != targets[i];
+      const bool append = (offsets[i] == (unsigned)-1);
+      if (!changed && append)
          continue;
       nv50->so_targets_dirty |= 1 << i;
 
       if (can_resume && changed && nv50->so_target[i]) {
          nva0_so_target_save_offset(pipe, nv50->so_target[i], i, serialize);
-         serialize = FALSE;
+         serialize = false;
       }
 
-      if (targets[i] && !(append_mask & (1 << i)))
-         nv50_so_target(targets[i])->clean = TRUE;
+      if (targets[i] && !append)
+         nv50_so_target(targets[i])->clean = true;
 
       pipe_so_target_reference(&nv50->so_target[i], targets[i]);
    }
    for (; i < nv50->num_so_targets; ++i) {
       if (can_resume && nv50->so_target[i]) {
          nva0_so_target_save_offset(pipe, nv50->so_target[i], i, serialize);
-         serialize = FALSE;
+         serialize = false;
       }
       pipe_so_target_reference(&nv50->so_target[i], NULL);
       nv50->so_targets_dirty |= 1 << i;
@@ -1107,6 +1150,7 @@ nv50_init_state_functions(struct nv50_context *nv50)
    pipe->set_stencil_ref = nv50_set_stencil_ref;
    pipe->set_clip_state = nv50_set_clip_state;
    pipe->set_sample_mask = nv50_set_sample_mask;
+   pipe->set_min_samples = nv50_set_min_samples;
    pipe->set_constant_buffer = nv50_set_constant_buffer;
    pipe->set_framebuffer_state = nv50_set_framebuffer_state;
    pipe->set_polygon_stipple = nv50_set_polygon_stipple;
@@ -1123,4 +1167,7 @@ nv50_init_state_functions(struct nv50_context *nv50)
    pipe->create_stream_output_target = nv50_so_target_create;
    pipe->stream_output_target_destroy = nv50_so_target_destroy;
    pipe->set_stream_output_targets = nv50_set_stream_output_targets;
+
+   nv50->sample_mask = ~0;
+   nv50->min_samples = 1;
 }