#include "common/os.h"
#include "jit_api.h"
-#include "state_llvm.h"
+#include "gen_state_llvm.h"
+#include "core/multisample.h"
+#include "core/state_funcs.h"
#include "gallivm/lp_bld_tgsi.h"
#include "util/u_format.h"
#include "util/u_inlines.h"
#include "util/u_helpers.h"
#include "util/u_framebuffer.h"
+#include "util/u_viewport.h"
+#include "util/u_prim.h"
#include "swr_state.h"
#include "swr_context.h"
-#include "swr_context_llvm.h"
+#include "gen_swr_context_llvm.h"
#include "swr_screen.h"
#include "swr_resource.h"
#include "swr_tex_sample.h"
// soState.streamToRasterizer not used
for (uint32_t i = 0; i < stream_output->num_outputs; i++) {
+ unsigned attrib_slot = stream_output->output[i].register_index;
+ attrib_slot = swr_so_adjust_attrib(attrib_slot, swr_vs);
swr_vs->soState.streamMasks[stream_output->output[i].stream] |=
- 1 << (stream_output->output[i].register_index - 1);
+ (1 << attrib_slot);
}
for (uint32_t i = 0; i < MAX_SO_STREAMS; i++) {
swr_vs->soState.streamNumEntries[i] =
{
struct swr_vertex_shader *swr_vs = (swr_vertex_shader *)vs;
FREE((void *)swr_vs->pipe.tokens);
- delete swr_vs;
+ struct swr_screen *screen = swr_screen(pipe->screen);
+
+ /* Defer deletion of vs state */
+ swr_fence_work_delete_vs(screen->flush_fence, swr_vs);
}
static void *
{
struct swr_fragment_shader *swr_fs = (swr_fragment_shader *)fs;
FREE((void *)swr_fs->pipe.tokens);
- delete swr_fs;
+ struct swr_screen *screen = swr_screen(pipe->screen);
+
+ /* Defer deleton of fs state */
+ swr_fence_work_delete_fs(screen->flush_fence, swr_fs);
+}
+
+static void *
+swr_create_gs_state(struct pipe_context *pipe,
+ const struct pipe_shader_state *gs)
+{
+ struct swr_geometry_shader *swr_gs = new swr_geometry_shader;
+ if (!swr_gs)
+ return NULL;
+
+ swr_gs->pipe.tokens = tgsi_dup_tokens(gs->tokens);
+
+ lp_build_tgsi_info(gs->tokens, &swr_gs->info);
+
+ return swr_gs;
+}
+
+
+static void
+swr_bind_gs_state(struct pipe_context *pipe, void *gs)
+{
+ struct swr_context *ctx = swr_context(pipe);
+
+ if (ctx->gs == gs)
+ return;
+
+ ctx->gs = (swr_geometry_shader *)gs;
+ ctx->dirty |= SWR_NEW_GS;
}
+static void
+swr_delete_gs_state(struct pipe_context *pipe, void *gs)
+{
+ struct swr_geometry_shader *swr_gs = (swr_geometry_shader *)gs;
+ FREE((void *)swr_gs->pipe.tokens);
+ struct swr_screen *screen = swr_screen(pipe->screen);
+
+ /* Defer deleton of fs state */
+ swr_fence_work_delete_gs(screen->flush_fence, swr_gs);
+}
static void
swr_set_constant_buffer(struct pipe_context *pipe,
- uint shader,
+ enum pipe_shader_type shader,
uint index,
const struct pipe_constant_buffer *cb)
{
/* note: reference counting */
util_copy_constant_buffer(&ctx->constants[shader][index], cb);
- if (shader == PIPE_SHADER_VERTEX || shader == PIPE_SHADER_GEOMETRY) {
+ if (shader == PIPE_SHADER_VERTEX) {
ctx->dirty |= SWR_NEW_VSCONSTANTS;
} else if (shader == PIPE_SHADER_FRAGMENT) {
ctx->dirty |= SWR_NEW_FSCONSTANTS;
+ } else if (shader == PIPE_SHADER_GEOMETRY) {
+ ctx->dirty |= SWR_NEW_GSCONSTANTS;
}
if (cb && cb->user_buffer) {
{
struct swr_vertex_element_state *velems;
assert(num_elements <= PIPE_MAX_ATTRIBS);
- velems = CALLOC_STRUCT(swr_vertex_element_state);
+ velems = new swr_vertex_element_state;
if (velems) {
+ memset(&velems->fsState, 0, sizeof(velems->fsState));
+ velems->fsState.bVertexIDOffsetEnable = true;
velems->fsState.numAttribs = num_elements;
for (unsigned i = 0; i < num_elements; i++) {
// XXX: we should do this keyed on the VS usage info
? ComponentControl::StoreSrc
: ComponentControl::Store1Fp;
velems->fsState.layout[i].ComponentPacking = ComponentEnable::XYZW;
- velems->fsState.layout[i].InstanceDataStepRate =
+ velems->fsState.layout[i].InstanceAdvancementState =
attribs[i].instance_divisor;
/* Calculate the pitch of each stream */
const SWR_FORMAT_INFO &swr_desc = GetFormatInfo(
mesa_to_swr_format(attribs[i].src_format));
velems->stream_pitch[attribs[i].vertex_buffer_index] += swr_desc.Bpp;
+
+ if (attribs[i].instance_divisor != 0) {
+ velems->instanced_bufs |= 1U << attribs[i].vertex_buffer_index;
+ uint32_t *min_instance_div =
+ &velems->min_instance_div[attribs[i].vertex_buffer_index];
+ if (!*min_instance_div ||
+ attribs[i].instance_divisor < *min_instance_div)
+ *min_instance_div = attribs[i].instance_divisor;
+ }
}
}
static void
swr_delete_vertex_elements_state(struct pipe_context *pipe, void *velems)
{
+ struct swr_vertex_element_state *swr_velems =
+ (struct swr_vertex_element_state *) velems;
/* XXX Need to destroy fetch shader? */
- FREE(velems);
+ delete swr_velems;
}
}
-static void
-swr_set_index_buffer(struct pipe_context *pipe,
- const struct pipe_index_buffer *ib)
-{
- struct swr_context *ctx = swr_context(pipe);
-
- if (ib)
- memcpy(&ctx->index_buffer, ib, sizeof(ctx->index_buffer));
- else
- memset(&ctx->index_buffer, 0, sizeof(ctx->index_buffer));
-
- ctx->dirty |= SWR_NEW_VERTEX;
-}
-
static void
swr_set_polygon_stipple(struct pipe_context *pipe,
const struct pipe_poly_stipple *stipple)
{
struct swr_context *ctx = swr_context(pipe);
- ctx->poly_stipple = *stipple; /* struct copy */
+ ctx->poly_stipple.pipe = *stipple; /* struct copy */
ctx->dirty |= SWR_NEW_STIPPLE;
}
assert(fb->height <= KNOB_GUARDBAND_HEIGHT);
if (changed) {
- unsigned i;
- for (i = 0; i < fb->nr_cbufs; ++i)
- pipe_surface_reference(&ctx->framebuffer.cbufs[i], fb->cbufs[i]);
- for (; i < ctx->framebuffer.nr_cbufs; ++i)
- pipe_surface_reference(&ctx->framebuffer.cbufs[i], NULL);
-
- ctx->framebuffer.nr_cbufs = fb->nr_cbufs;
+ util_copy_framebuffer_state(&ctx->framebuffer, fb);
- ctx->framebuffer.width = fb->width;
- ctx->framebuffer.height = fb->height;
-
- pipe_surface_reference(&ctx->framebuffer.zsbuf, fb->zsbuf);
+ /* 0 and 1 both indicate no msaa. Core doesn't understand 0 samples */
+ ctx->framebuffer.samples = std::max((ubyte)1, ctx->framebuffer.samples);
ctx->dirty |= SWR_NEW_FRAMEBUFFER;
}
}
}
+/*
+ * MSAA fixed sample position table
+ * used by update_derived and get_sample_position
+ * (integer locations on a 16x16 grid)
+ */
+static const uint8_t swr_sample_positions[][2] =
+{ /* 1x*/ { 8, 8},
+ /* 2x*/ {12,12},{ 4, 4},
+ /* 4x*/ { 6, 2},{14, 6},{ 2,10},{10,14},
+ /* 8x*/ { 9, 5},{ 7,11},{13, 9},{ 5, 3},
+ { 3,13},{ 1, 7},{11,15},{15, 1},
+ /*16x*/ { 9, 9},{ 7, 5},{ 5,10},{12, 7},
+ { 3, 6},{10,13},{13,11},{11, 3},
+ { 6,14},{ 8, 1},{ 4, 2},{ 2,12},
+ { 0, 8},{15, 4},{14,15},{ 1, 0} };
+
+static void
+swr_get_sample_position(struct pipe_context *pipe,
+ unsigned sample_count, unsigned sample_index,
+ float *out_value)
+{
+ /* validate sample_count */
+ sample_count = GetNumSamples(GetSampleCount(sample_count));
+
+ const uint8_t *sample = swr_sample_positions[sample_count-1 + sample_index];
+ out_value[0] = sample[0] / 16.0f;
+ out_value[1] = sample[1] / 16.0f;
+}
+
+
/*
* Update resource in-use status
* All resources bound to color or depth targets marked as WRITE resources.
/* VBO vertex buffers */
for (uint32_t i = 0; i < ctx->num_vertex_buffers; i++) {
struct pipe_vertex_buffer *vb = &ctx->vertex_buffer[i];
- if (!vb->user_buffer)
- swr_resource_read(vb->buffer);
+ if (!vb->is_user_buffer)
+ swr_resource_read(vb->buffer.resource);
}
/* VBO index buffer */
- if (p_draw_info && p_draw_info->indexed) {
- struct pipe_index_buffer *ib = &ctx->index_buffer;
- if (!ib->user_buffer)
- swr_resource_read(ib->buffer);
+ if (p_draw_info && p_draw_info->index_size) {
+ if (!p_draw_info->has_user_indices)
+ swr_resource_read(p_draw_info->index.resource);
+ }
+
+ /* transform feedback buffers */
+ for (uint32_t i = 0; i < ctx->num_so_targets; i++) {
+ struct pipe_stream_output_target *target = ctx->so_targets[i];
+ if (target && target->buffer)
+ swr_resource_write(target->buffer);
}
/* texture sampler views */
- for (uint32_t i = 0; i < PIPE_MAX_SHADER_SAMPLER_VIEWS; i++) {
- struct pipe_sampler_view *view =
- ctx->sampler_views[PIPE_SHADER_FRAGMENT][i];
- if (view)
- swr_resource_read(view->texture);
+ for (uint32_t j : {PIPE_SHADER_VERTEX, PIPE_SHADER_FRAGMENT}) {
+ for (uint32_t i = 0; i < ctx->num_sampler_views[j]; i++) {
+ struct pipe_sampler_view *view = ctx->sampler_views[j][i];
+ if (view)
+ swr_resource_read(view->texture);
+ }
+ }
+
+ /* constant buffers */
+ for (uint32_t j : {PIPE_SHADER_VERTEX, PIPE_SHADER_FRAGMENT}) {
+ for (uint32_t i = 0; i < PIPE_MAX_CONSTANT_BUFFERS; i++) {
+ struct pipe_constant_buffer *cb = &ctx->constants[j][i];
+ if (cb->buffer)
+ swr_resource_read(cb->buffer);
+ }
}
}
static void
swr_update_texture_state(struct swr_context *ctx,
- unsigned shader_type,
+ enum pipe_shader_type shader_type,
unsigned num_sampler_views,
swr_jit_texture *textures)
{
for (unsigned i = 0; i < num_sampler_views; i++) {
struct pipe_sampler_view *view =
ctx->sampler_views[shader_type][i];
+ struct swr_jit_texture *jit_tex = &textures[i];
+ memset(jit_tex, 0, sizeof(*jit_tex));
if (view) {
struct pipe_resource *res = view->texture;
struct swr_resource *swr_res = swr_resource(res);
- struct swr_jit_texture *jit_tex = &textures[i];
- memset(jit_tex, 0, sizeof(*jit_tex));
+ SWR_SURFACE_STATE *swr = &swr_res->swr;
+ size_t *mip_offsets = swr_res->mip_offsets;
+ if (swr_res->has_depth && swr_res->has_stencil &&
+ !util_format_has_depth(util_format_description(view->format))) {
+ swr = &swr_res->secondary;
+ mip_offsets = swr_res->secondary_mip_offsets;
+ }
+
jit_tex->width = res->width0;
jit_tex->height = res->height0;
- jit_tex->depth = res->depth0;
- jit_tex->first_level = view->u.tex.first_level;
- jit_tex->last_level = view->u.tex.last_level;
- jit_tex->base_ptr = swr_res->swr.pBaseAddress;
+ jit_tex->base_ptr = (uint8_t*)swr->xpBaseAddress;
+ if (view->target != PIPE_BUFFER) {
+ jit_tex->first_level = view->u.tex.first_level;
+ jit_tex->last_level = view->u.tex.last_level;
+ if (view->target == PIPE_TEXTURE_3D)
+ jit_tex->depth = res->depth0;
+ else
+ jit_tex->depth =
+ view->u.tex.last_layer - view->u.tex.first_layer + 1;
+ jit_tex->base_ptr += view->u.tex.first_layer *
+ swr->qpitch * swr->pitch;
+ } else {
+ unsigned view_blocksize = util_format_get_blocksize(view->format);
+ jit_tex->base_ptr += view->u.buf.offset;
+ jit_tex->width = view->u.buf.size / view_blocksize;
+ jit_tex->depth = 1;
+ }
for (unsigned level = jit_tex->first_level;
level <= jit_tex->last_level;
level++) {
- jit_tex->row_stride[level] = swr_res->row_stride[level];
- jit_tex->img_stride[level] = swr_res->img_stride[level];
- jit_tex->mip_offsets[level] = swr_res->mip_offsets[level];
+ jit_tex->row_stride[level] = swr->pitch;
+ jit_tex->img_stride[level] = swr->qpitch * swr->pitch;
+ jit_tex->mip_offsets[level] = mip_offsets[level];
}
}
}
static void
swr_update_sampler_state(struct swr_context *ctx,
- unsigned shader_type,
+ enum pipe_shader_type shader_type,
unsigned num_samplers,
swr_jit_sampler *samplers)
{
num_constants = pDC->num_constantsFS;
scratch = &ctx->scratch->fs_constants;
break;
+ case PIPE_SHADER_GEOMETRY:
+ constant = pDC->constantGS;
+ num_constants = pDC->num_constantsGS;
+ scratch = &ctx->scratch->gs_constants;
+ break;
default:
debug_printf("Unsupported shader type constants\n");
return;
}
}
+static bool
+swr_change_rt(struct swr_context *ctx,
+ unsigned attachment,
+ const struct pipe_surface *sf)
+{
+ swr_draw_context *pDC = &ctx->swrDC;
+ struct SWR_SURFACE_STATE *rt = &pDC->renderTargets[attachment];
+
+ /* Do nothing if the render target hasn't changed */
+ if ((!sf || !sf->texture) && (void*)(rt->xpBaseAddress) == nullptr)
+ return false;
+
+ /* Deal with disabling RT up front */
+ if (!sf || !sf->texture) {
+ /* If detaching attachment, mark tiles as RESOLVED so core
+ * won't try to load from non-existent target. */
+ swr_store_render_target(&ctx->pipe, attachment, SWR_TILE_RESOLVED);
+ *rt = {0};
+ return true;
+ }
+
+ const struct swr_resource *swr = swr_resource(sf->texture);
+ const SWR_SURFACE_STATE *swr_surface = &swr->swr;
+ SWR_FORMAT fmt = mesa_to_swr_format(sf->format);
+
+ if (attachment == SWR_ATTACHMENT_STENCIL && swr->secondary.xpBaseAddress) {
+ swr_surface = &swr->secondary;
+ fmt = swr_surface->format;
+ }
+
+ if (rt->xpBaseAddress == swr_surface->xpBaseAddress &&
+ rt->format == fmt &&
+ rt->lod == sf->u.tex.level &&
+ rt->arrayIndex == sf->u.tex.first_layer)
+ return false;
+
+ bool need_fence = false;
+
+ /* StoreTile for changed target */
+ if (rt->xpBaseAddress) {
+ /* If changing attachment to a new target, mark tiles as
+ * INVALID so they are reloaded from surface. */
+ swr_store_render_target(&ctx->pipe, attachment, SWR_TILE_INVALID);
+ need_fence = true;
+ } else {
+ /* if no previous attachment, invalidate tiles that may be marked
+ * RESOLVED because of an old attachment */
+ swr_invalidate_render_target(&ctx->pipe, attachment, sf->width, sf->height);
+ /* no need to set fence here */
+ }
+
+ /* Make new attachment */
+ *rt = *swr_surface;
+ rt->format = fmt;
+ rt->lod = sf->u.tex.level;
+ rt->arrayIndex = sf->u.tex.first_layer;
+
+ return need_fence;
+}
+
+/*
+ * for cases where resources are shared between contexts, invalidate
+ * this ctx's resource. so it can be fetched fresh. Old ctx's resource
+ * is already stored during a flush
+ */
+static inline void
+swr_invalidate_buffers_after_ctx_change(struct pipe_context *pipe)
+{
+ struct swr_context *ctx = swr_context(pipe);
+
+ for (uint32_t i = 0; i < ctx->framebuffer.nr_cbufs; i++) {
+ struct pipe_surface *cb = ctx->framebuffer.cbufs[i];
+ if (cb) {
+ struct swr_resource *res = swr_resource(cb->texture);
+ if (res->curr_pipe != pipe) {
+ /* if curr_pipe is NULL (first use), status should not be WRITE */
+ assert(res->curr_pipe || !(res->status & SWR_RESOURCE_WRITE));
+ if (res->status & SWR_RESOURCE_WRITE) {
+ swr_invalidate_render_target(pipe, i, cb->width, cb->height);
+ }
+ }
+ res->curr_pipe = pipe;
+ }
+ }
+ if (ctx->framebuffer.zsbuf) {
+ struct pipe_surface *zb = ctx->framebuffer.zsbuf;
+ if (zb) {
+ struct swr_resource *res = swr_resource(zb->texture);
+ if (res->curr_pipe != pipe) {
+ /* if curr_pipe is NULL (first use), status should not be WRITE */
+ assert(res->curr_pipe || !(res->status & SWR_RESOURCE_WRITE));
+ if (res->status & SWR_RESOURCE_WRITE) {
+ swr_invalidate_render_target(pipe, SWR_ATTACHMENT_DEPTH, zb->width, zb->height);
+ swr_invalidate_render_target(pipe, SWR_ATTACHMENT_STENCIL, zb->width, zb->height);
+ }
+ }
+ res->curr_pipe = pipe;
+ }
+ }
+}
+
+static inline void
+swr_user_vbuf_range(const struct pipe_draw_info *info,
+ const struct swr_vertex_element_state *velems,
+ const struct pipe_vertex_buffer *vb,
+ uint32_t i,
+ uint32_t *totelems,
+ uint32_t *base,
+ uint32_t *size)
+{
+ /* FIXME: The size is too large - we don't access the full extra stride. */
+ unsigned elems;
+ if (velems->instanced_bufs & (1U << i)) {
+ elems = info->instance_count / velems->min_instance_div[i] + 1;
+ *totelems = info->start_instance + elems;
+ *base = info->start_instance * vb->stride;
+ *size = elems * vb->stride;
+ } else if (vb->stride) {
+ elems = info->max_index - info->min_index + 1;
+ *totelems = (info->max_index + info->index_bias) + 1;
+ *base = (info->min_index + info->index_bias) * vb->stride;
+ *size = elems * vb->stride;
+ } else {
+ *totelems = 1;
+ *base = 0;
+ *size = velems->stream_pitch[i];
+ }
+}
+
+static void
+swr_update_poly_stipple(struct swr_context *ctx)
+{
+ struct swr_draw_context *pDC = &ctx->swrDC;
+
+ assert(sizeof(ctx->poly_stipple.pipe.stipple) == sizeof(pDC->polyStipple));
+ memcpy(pDC->polyStipple,
+ ctx->poly_stipple.pipe.stipple,
+ sizeof(ctx->poly_stipple.pipe.stipple));
+}
+
void
swr_update_derived(struct pipe_context *pipe,
const struct pipe_draw_info *p_draw_info)
{
struct swr_context *ctx = swr_context(pipe);
- struct swr_screen *screen = swr_screen(ctx->pipe.screen);
+ struct swr_screen *screen = swr_screen(pipe->screen);
+
+ /* When called from swr_clear (p_draw_info = null), set any null
+ * state-objects to the dummy state objects to prevent nullptr dereference
+ * in validation below.
+ *
+ * Important that this remains static for zero initialization. These
+ * aren't meant to be proper state objects, just empty structs. They will
+ * not be written to.
+ *
+ * Shaders can't be part of the union since they contain std::unordered_map
+ */
+ static struct {
+ union {
+ struct pipe_rasterizer_state rasterizer;
+ struct pipe_depth_stencil_alpha_state depth_stencil;
+ struct swr_blend_state blend;
+ } state;
+ struct swr_vertex_shader vs;
+ struct swr_fragment_shader fs;
+ } swr_dummy;
+
+ if (!p_draw_info) {
+ if (!ctx->rasterizer)
+ ctx->rasterizer = &swr_dummy.state.rasterizer;
+ if (!ctx->depth_stencil)
+ ctx->depth_stencil = &swr_dummy.state.depth_stencil;
+ if (!ctx->blend)
+ ctx->blend = &swr_dummy.state.blend;
+ if (!ctx->vs)
+ ctx->vs = &swr_dummy.vs;
+ if (!ctx->fs)
+ ctx->fs = &swr_dummy.fs;
+ }
/* Update screen->pipe to current pipe context. */
- if (screen->pipe != pipe)
- screen->pipe = pipe;
+ screen->pipe = pipe;
/* Any state that requires dirty flags to be re-triggered sets this mask */
/* For example, user_buffer vertex and index buffers. */
unsigned post_update_dirty_flags = 0;
+ /* bring resources that changed context up-to-date */
+ swr_invalidate_buffers_after_ctx_change(pipe);
+
/* Render Targets */
if (ctx->dirty & SWR_NEW_FRAMEBUFFER) {
struct pipe_framebuffer_state *fb = &ctx->framebuffer;
- SWR_SURFACE_STATE *new_attachment[SWR_NUM_ATTACHMENTS] = {0};
- UINT i;
+ const struct util_format_description *desc = NULL;
+ bool need_fence = false;
/* colorbuffer targets */
- if (fb->nr_cbufs)
- for (i = 0; i < fb->nr_cbufs; ++i)
- if (fb->cbufs[i]) {
- struct swr_resource *colorBuffer =
- swr_resource(fb->cbufs[i]->texture);
- new_attachment[SWR_ATTACHMENT_COLOR0 + i] = &colorBuffer->swr;
- }
-
- /* depth/stencil target */
- if (fb->zsbuf) {
- struct swr_resource *depthStencilBuffer =
- swr_resource(fb->zsbuf->texture);
- if (depthStencilBuffer->has_depth) {
- new_attachment[SWR_ATTACHMENT_DEPTH] = &depthStencilBuffer->swr;
-
- if (depthStencilBuffer->has_stencil)
- new_attachment[SWR_ATTACHMENT_STENCIL] =
- &depthStencilBuffer->secondary;
-
- } else if (depthStencilBuffer->has_stencil)
- new_attachment[SWR_ATTACHMENT_STENCIL] = &depthStencilBuffer->swr;
+ if (fb->nr_cbufs) {
+ for (unsigned i = 0; i < fb->nr_cbufs; ++i)
+ need_fence |= swr_change_rt(
+ ctx, SWR_ATTACHMENT_COLOR0 + i, fb->cbufs[i]);
}
+ for (unsigned i = fb->nr_cbufs; i < SWR_NUM_RENDERTARGETS; ++i)
+ need_fence |= swr_change_rt(ctx, SWR_ATTACHMENT_COLOR0 + i, NULL);
- /* Make the attachment updates */
- swr_draw_context *pDC = &ctx->swrDC;
- SWR_SURFACE_STATE *renderTargets = pDC->renderTargets;
- unsigned need_fence = FALSE;
- for (i = 0; i < SWR_NUM_ATTACHMENTS; i++) {
- void *new_base = nullptr;
- if (new_attachment[i])
- new_base = new_attachment[i]->pBaseAddress;
-
- /* StoreTile for changed target */
- if (renderTargets[i].pBaseAddress != new_base) {
- if (renderTargets[i].pBaseAddress) {
- /* If changing attachment to a new target, mark tiles as
- * INVALID so they are reloaded from surface.
- * If detaching attachment, mark tiles as RESOLVED so core
- * won't try to load from non-existent target. */
- enum SWR_TILE_STATE post_state = (new_attachment[i]
- ? SWR_TILE_INVALID : SWR_TILE_RESOLVED);
- swr_store_render_target(pipe, i, post_state);
-
- need_fence |= TRUE;
- }
+ /* depth/stencil target */
+ if (fb->zsbuf)
+ desc = util_format_description(fb->zsbuf->format);
+ if (fb->zsbuf && util_format_has_depth(desc))
+ need_fence |= swr_change_rt(ctx, SWR_ATTACHMENT_DEPTH, fb->zsbuf);
+ else
+ need_fence |= swr_change_rt(ctx, SWR_ATTACHMENT_DEPTH, NULL);
- /* Make new attachment */
- if (new_attachment[i])
- renderTargets[i] = *new_attachment[i];
- else
- if (renderTargets[i].pBaseAddress)
- renderTargets[i] = {0};
- }
- }
+ if (fb->zsbuf && util_format_has_stencil(desc))
+ need_fence |= swr_change_rt(ctx, SWR_ATTACHMENT_STENCIL, fb->zsbuf);
+ else
+ need_fence |= swr_change_rt(ctx, SWR_ATTACHMENT_STENCIL, NULL);
/* This fence ensures any attachment changes are resolved before the
* next draw */
rastState->pointSpriteTopOrigin =
rasterizer->sprite_coord_mode == PIPE_SPRITE_COORD_UPPER_LEFT;
- /* XXX TODO: Add multisample */
- rastState->msaaRastEnable = false;
- rastState->rastMode = SWR_MSAA_RASTMODE_OFF_PIXEL;
- rastState->sampleCount = SWR_MULTISAMPLE_1X;
+ /* If SWR_MSAA_FORCE_ENABLE is set, turn msaa on */
+ if (screen->msaa_force_enable && !rasterizer->multisample) {
+ /* Force enable and use the value the surface was created with */
+ rasterizer->multisample = true;
+ fb->samples = swr_resource(fb->cbufs[0]->texture)->swr.numSamples;
+ fprintf(stderr,"msaa force enable: %d samples\n", fb->samples);
+ }
+
+ rastState->sampleCount = GetSampleCount(fb->samples);
rastState->forcedSampleCount = false;
+ rastState->bIsCenterPattern = !rasterizer->multisample;
+ rastState->pixelLocation = SWR_PIXEL_LOCATION_CENTER;
+
+ /* Only initialize sample positions if msaa is enabled */
+ if (rasterizer->multisample) {
+ for (uint32_t i = 0; i < fb->samples; i++) {
+ const uint8_t *sample = swr_sample_positions[fb->samples-1 + i];
+ rastState->samplePositions.SetXi(i, sample[0] << 4);
+ rastState->samplePositions.SetYi(i, sample[1] << 4);
+ rastState->samplePositions.SetX (i, sample[0] / 16.0f);
+ rastState->samplePositions.SetY (i, sample[1] / 16.0f);
+ }
+ rastState->samplePositions.PrecalcSampleData(fb->samples);
+ }
bool do_offset = false;
switch (rasterizer->fill_front) {
rastState->slopeScaledDepthBias = 0;
rastState->depthBiasClamp = 0;
}
+
+ /* translate polygon mode, at least for the front==back case */
+ rastState->fillMode = swr_convert_fill_mode(rasterizer->fill_front);
+
struct pipe_surface *zb = fb->zsbuf;
if (zb && swr_resource(zb->texture)->has_depth)
rastState->depthFormat = swr_resource(zb->texture)->swr.format;
- rastState->depthClipEnable = rasterizer->depth_clip;
+ rastState->depthClipEnable = rasterizer->depth_clip_near;
+ rastState->clipHalfZ = rasterizer->clip_halfz;
- rastState->clipDistanceMask =
- ctx->vs->info.base.num_written_clipdistance ?
- ctx->vs->info.base.clipdist_writemask & rasterizer->clip_plane_enable :
- rasterizer->clip_plane_enable;
-
- rastState->cullDistanceMask =
- ctx->vs->info.base.culldist_writemask << ctx->vs->info.base.num_written_clipdistance;
-
- SwrSetRastState(ctx->swrContext, rastState);
- }
-
- /* Scissor */
- if (ctx->dirty & SWR_NEW_SCISSOR) {
- SwrSetScissorRects(ctx->swrContext, 1, &ctx->swr_scissor);
+ ctx->api.pfnSwrSetRastState(ctx->swrContext, rastState);
}
/* Viewport */
SWR_VIEWPORT_MATRICES *vpm = &ctx->derived.vpm;
vp->x = state->translate[0] - state->scale[0];
- vp->width = state->translate[0] + state->scale[0];
+ vp->width = 2 * state->scale[0];
vp->y = state->translate[1] - fabs(state->scale[1]);
- vp->height = state->translate[1] + fabs(state->scale[1]);
- if (rasterizer->clip_halfz == 0) {
- vp->minZ = state->translate[2] - state->scale[2];
- vp->maxZ = state->translate[2] + state->scale[2];
- } else {
- vp->minZ = state->translate[2];
- vp->maxZ = state->translate[2] + state->scale[2];
- }
+ vp->height = 2 * fabs(state->scale[1]);
+ util_viewport_zmin_zmax(state, rasterizer->clip_halfz,
+ &vp->minZ, &vp->maxZ);
vpm->m00[0] = state->scale[0];
vpm->m11[0] = state->scale[1];
/* Now that the matrix is calculated, clip the view coords to screen
* size. OpenGL allows for -ve x,y in the viewport. */
- vp->x = std::max(vp->x, 0.0f);
- vp->y = std::max(vp->y, 0.0f);
- vp->width = std::min(vp->width, (float)fb->width);
- vp->height = std::min(vp->height, (float)fb->height);
+ if (vp->x < 0.0f) {
+ vp->width += vp->x;
+ vp->x = 0.0f;
+ }
+ if (vp->y < 0.0f) {
+ vp->height += vp->y;
+ vp->y = 0.0f;
+ }
+ vp->width = std::min(vp->width, (float)fb->width - vp->x);
+ vp->height = std::min(vp->height, (float)fb->height - vp->y);
+
+ ctx->api.pfnSwrSetViewports(ctx->swrContext, 1, vp, vpm);
+ }
- SwrSetViewports(ctx->swrContext, 1, vp, vpm);
+ /* When called from swr_clear (p_draw_info = null), render targets,
+ * rasterState and viewports (dependent on render targets) are the only
+ * necessary validation. Defer remaining validation by setting
+ * post_update_dirty_flags and clear all dirty flags. BackendState is
+ * still unconditionally validated below */
+ if (!p_draw_info) {
+ post_update_dirty_flags = ctx->dirty & ~(SWR_NEW_FRAMEBUFFER |
+ SWR_NEW_RASTERIZER |
+ SWR_NEW_VIEWPORT);
+ ctx->dirty = 0;
+ }
+
+ /* Scissor */
+ if (ctx->dirty & SWR_NEW_SCISSOR) {
+ ctx->api.pfnSwrSetScissorRects(ctx->swrContext, 1, &ctx->swr_scissor);
}
/* Set vertex & index buffers */
- /* (using draw info if called by swr_draw_vbo) */
if (ctx->dirty & SWR_NEW_VERTEX) {
- uint32_t size, pitch, max_vertex, partial_inbounds;
- const uint8_t *p_data;
-
- /* If being called by swr_draw_vbo, copy draw details */
- struct pipe_draw_info info = {0};
- if (p_draw_info)
- info = *p_draw_info;
+ const struct pipe_draw_info &info = *p_draw_info;
/* vertex buffers */
SWR_VERTEX_BUFFER_STATE swrVertexBuffers[PIPE_MAX_ATTRIBS];
for (UINT i = 0; i < ctx->num_vertex_buffers; i++) {
+ uint32_t size, pitch, elems, partial_inbounds;
+ uint32_t min_vertex_index;
+ const uint8_t *p_data;
struct pipe_vertex_buffer *vb = &ctx->vertex_buffer[i];
pitch = vb->stride;
- if (!vb->user_buffer) {
- /* VBO
- * size is based on buffer->width0 rather than info.max_index
- * to prevent having to validate VBO on each draw */
- size = vb->buffer->width0;
- max_vertex = size / pitch;
- partial_inbounds = size % pitch;
+ if (!vb->is_user_buffer) {
+ /* VBO */
+ if (!pitch) {
+ /* If pitch=0 (ie vb->stride), buffer contains a single
+ * constant attribute. Use the stream_pitch which was
+ * calculated during creation of vertex_elements_state for the
+ * size of the attribute. */
+ size = ctx->velems->stream_pitch[i];
+ elems = 1;
+ partial_inbounds = 0;
+ min_vertex_index = 0;
+ } else {
+ /* size is based on buffer->width0 rather than info.max_index
+ * to prevent having to validate VBO on each draw. */
+ size = vb->buffer.resource->width0;
+ elems = size / pitch;
+ partial_inbounds = size % pitch;
+ min_vertex_index = 0;
+ }
- p_data = swr_resource_data(vb->buffer) + vb->buffer_offset;
+ p_data = swr_resource_data(vb->buffer.resource) + vb->buffer_offset;
} else {
/* Client buffer
* client memory is one-time use, re-trigger SWR_NEW_VERTEX to
* revalidate on each draw */
post_update_dirty_flags |= SWR_NEW_VERTEX;
- if (pitch) {
- size = (info.max_index - info.min_index + 1) * pitch;
- } else {
- /* pitch = 0, means constant value
- * set size to 1 vertex */
- size = ctx->velems->stream_pitch[i];
- }
-
- max_vertex = info.max_index + 1;
+ uint32_t base;
+ swr_user_vbuf_range(&info, ctx->velems, vb, i, &elems, &base, &size);
partial_inbounds = 0;
+ min_vertex_index = info.min_index + info.index_bias;
- /* Copy only needed vertices to scratch space */
size = AlignUp(size, 4);
- const void *ptr = (const uint8_t *) vb->user_buffer
- + info.min_index * pitch;
- ptr = swr_copy_to_scratch_space(
- ctx, &ctx->scratch->vertex_buffer, ptr, size);
- p_data = (const uint8_t *)ptr - info.min_index * pitch;
+ /* If size of client memory copy is too large, don't copy. The
+ * draw will access user-buffer directly and then block. This is
+ * faster than queuing many large client draws. */
+ if (size >= screen->client_copy_limit) {
+ post_update_dirty_flags |= SWR_LARGE_CLIENT_DRAW;
+ p_data = (const uint8_t *) vb->buffer.user;
+ } else {
+ /* Copy only needed vertices to scratch space */
+ const void *ptr = (const uint8_t *) vb->buffer.user + base;
+ ptr = (uint8_t *)swr_copy_to_scratch_space(
+ ctx, &ctx->scratch->vertex_buffer, ptr, size);
+ p_data = (const uint8_t *)ptr - base;
+ }
}
swrVertexBuffers[i] = {0};
swrVertexBuffers[i].index = i;
swrVertexBuffers[i].pitch = pitch;
- swrVertexBuffers[i].pData = p_data;
+ swrVertexBuffers[i].xpData = (gfxptr_t) p_data;
swrVertexBuffers[i].size = size;
- swrVertexBuffers[i].maxVertex = max_vertex;
+ swrVertexBuffers[i].minVertex = min_vertex_index;
+ swrVertexBuffers[i].maxVertex = elems;
swrVertexBuffers[i].partialInboundsSize = partial_inbounds;
}
- SwrSetVertexBuffers(
+ ctx->api.pfnSwrSetVertexBuffers(
ctx->swrContext, ctx->num_vertex_buffers, swrVertexBuffers);
/* index buffer, if required (info passed in by swr_draw_vbo) */
SWR_FORMAT index_type = R32_UINT; /* Default for non-indexed draws */
- if (info.indexed) {
- struct pipe_index_buffer *ib = &ctx->index_buffer;
+ if (info.index_size) {
+ const uint8_t *p_data;
+ uint32_t size, pitch;
- pitch = ib->index_size ? ib->index_size : sizeof(uint32_t);
+ pitch = info.index_size ? info.index_size : sizeof(uint32_t);
index_type = swr_convert_index_type(pitch);
- if (!ib->user_buffer) {
+ if (!info.has_user_indices) {
/* VBO
* size is based on buffer->width0 rather than info.count
* to prevent having to validate VBO on each draw */
- size = ib->buffer->width0;
- p_data = swr_resource_data(ib->buffer) + ib->offset;
+ size = info.index.resource->width0;
+ p_data = swr_resource_data(info.index.resource);
} else {
/* Client buffer
* client memory is one-time use, re-trigger SWR_NEW_VERTEX to
size = info.count * pitch;
size = AlignUp(size, 4);
-
- /* Copy indices to scratch space */
- const void *ptr = ib->user_buffer;
- ptr = swr_copy_to_scratch_space(
- ctx, &ctx->scratch->index_buffer, ptr, size);
- p_data = (const uint8_t *)ptr;
+ /* If size of client memory copy is too large, don't copy. The
+ * draw will access user-buffer directly and then block. This is
+ * faster than queuing many large client draws. */
+ if (size >= screen->client_copy_limit) {
+ post_update_dirty_flags |= SWR_LARGE_CLIENT_DRAW;
+ p_data = (const uint8_t *) info.index.user;
+ } else {
+ /* Copy indices to scratch space */
+ const void *ptr = info.index.user;
+ ptr = swr_copy_to_scratch_space(
+ ctx, &ctx->scratch->index_buffer, ptr, size);
+ p_data = (const uint8_t *)ptr;
+ }
}
SWR_INDEX_BUFFER_STATE swrIndexBuffer;
- swrIndexBuffer.format = swr_convert_index_type(ib->index_size);
- swrIndexBuffer.pIndices = p_data;
+ swrIndexBuffer.format = swr_convert_index_type(info.index_size);
+ swrIndexBuffer.xpIndices = (gfxptr_t) p_data;
swrIndexBuffer.size = size;
- SwrSetIndexBuffer(ctx->swrContext, &swrIndexBuffer);
+ ctx->api.pfnSwrSetIndexBuffer(ctx->swrContext, &swrIndexBuffer);
}
struct swr_vertex_element_state *velems = ctx->velems;
}
}
+ /* GeometryShader */
+ if (ctx->dirty & (SWR_NEW_GS |
+ SWR_NEW_VS |
+ SWR_NEW_SAMPLER |
+ SWR_NEW_SAMPLER_VIEW)) {
+ if (ctx->gs) {
+ swr_jit_gs_key key;
+ swr_generate_gs_key(key, ctx, ctx->gs);
+ auto search = ctx->gs->map.find(key);
+ PFN_GS_FUNC func;
+ if (search != ctx->gs->map.end()) {
+ func = search->second->shader;
+ } else {
+ func = swr_compile_gs(ctx, key);
+ }
+ ctx->api.pfnSwrSetGsFunc(ctx->swrContext, func);
+
+ /* JIT sampler state */
+ if (ctx->dirty & SWR_NEW_SAMPLER) {
+ swr_update_sampler_state(ctx,
+ PIPE_SHADER_GEOMETRY,
+ key.nr_samplers,
+ ctx->swrDC.samplersGS);
+ }
+
+ /* JIT sampler view state */
+ if (ctx->dirty & (SWR_NEW_SAMPLER_VIEW | SWR_NEW_FRAMEBUFFER)) {
+ swr_update_texture_state(ctx,
+ PIPE_SHADER_GEOMETRY,
+ key.nr_sampler_views,
+ ctx->swrDC.texturesGS);
+ }
+
+ ctx->api.pfnSwrSetGsState(ctx->swrContext, &ctx->gs->gsState);
+ } else {
+ SWR_GS_STATE state = { 0 };
+ ctx->api.pfnSwrSetGsState(ctx->swrContext, &state);
+ ctx->api.pfnSwrSetGsFunc(ctx->swrContext, NULL);
+ }
+ }
+
/* VertexShader */
if (ctx->dirty & (SWR_NEW_VS |
SWR_NEW_RASTERIZER | // for clip planes
} else {
func = swr_compile_vs(ctx, key);
}
- SwrSetVertexFunc(ctx->swrContext, func);
+ ctx->api.pfnSwrSetVertexFunc(ctx->swrContext, func);
/* JIT sampler state */
if (ctx->dirty & SWR_NEW_SAMPLER) {
}
}
+ /* work around the fact that poly stipple also affects lines */
+ /* and points, since we rasterize them as triangles, too */
+ /* Has to be before fragment shader, since it sets SWR_NEW_FS */
+ if (p_draw_info) {
+ bool new_prim_is_poly =
+ (u_reduced_prim(p_draw_info->mode) == PIPE_PRIM_TRIANGLES) &&
+ (ctx->derived.rastState.fillMode == SWR_FILLMODE_SOLID);
+ if (new_prim_is_poly != ctx->poly_stipple.prim_is_poly) {
+ ctx->dirty |= SWR_NEW_FS;
+ ctx->poly_stipple.prim_is_poly = new_prim_is_poly;
+ }
+ }
+
/* FragmentShader */
- if (ctx->dirty & (SWR_NEW_FS | SWR_NEW_SAMPLER | SWR_NEW_SAMPLER_VIEW
- | SWR_NEW_RASTERIZER | SWR_NEW_FRAMEBUFFER)) {
+ if (ctx->dirty & (SWR_NEW_FS |
+ SWR_NEW_VS |
+ SWR_NEW_GS |
+ SWR_NEW_RASTERIZER |
+ SWR_NEW_SAMPLER |
+ SWR_NEW_SAMPLER_VIEW |
+ SWR_NEW_FRAMEBUFFER)) {
swr_jit_fs_key key;
swr_generate_fs_key(key, ctx, ctx->fs);
auto search = ctx->fs->map.find(key);
psState.inputCoverage = SWR_INPUT_COVERAGE_NORMAL;
psState.writesODepth = ctx->fs->info.base.writes_z;
psState.usesSourceDepth = ctx->fs->info.base.reads_z;
- psState.shadingRate = SWR_SHADING_RATE_PIXEL; // XXX
- psState.numRenderTargets = ctx->framebuffer.nr_cbufs;
- psState.posOffset = SWR_PS_POSITION_SAMPLE_NONE; // XXX msaa
+ psState.shadingRate = SWR_SHADING_RATE_PIXEL;
+ psState.renderTargetMask = (1 << ctx->framebuffer.nr_cbufs) - 1;
+ psState.posOffset = SWR_PS_POSITION_SAMPLE_NONE;
uint32_t barycentricsMask = 0;
#if 0
// when we switch to mesa-master
psState.barycentricsMask = barycentricsMask;
psState.usesUAV = false; // XXX
psState.forceEarlyZ = false;
- SwrSetPixelShaderState(ctx->swrContext, &psState);
+ ctx->api.pfnSwrSetPixelShaderState(ctx->swrContext, &psState);
/* JIT sampler state */
- if (ctx->dirty & SWR_NEW_SAMPLER) {
+ if (ctx->dirty & (SWR_NEW_SAMPLER |
+ SWR_NEW_FS)) {
swr_update_sampler_state(ctx,
PIPE_SHADER_FRAGMENT,
key.nr_samplers,
}
/* JIT sampler view state */
- if (ctx->dirty & (SWR_NEW_SAMPLER_VIEW | SWR_NEW_FRAMEBUFFER)) {
+ if (ctx->dirty & (SWR_NEW_SAMPLER_VIEW |
+ SWR_NEW_FRAMEBUFFER |
+ SWR_NEW_FS)) {
swr_update_texture_state(ctx,
PIPE_SHADER_FRAGMENT,
key.nr_sampler_views,
swr_update_constants(ctx, PIPE_SHADER_FRAGMENT);
}
+ /* GeometryShader Constants */
+ if (ctx->dirty & SWR_NEW_GSCONSTANTS) {
+ swr_update_constants(ctx, PIPE_SHADER_GEOMETRY);
+ }
+
/* Depth/stencil state */
if (ctx->dirty & (SWR_NEW_DEPTH_STENCIL_ALPHA | SWR_NEW_FRAMEBUFFER)) {
struct pipe_depth_state *depth = &(ctx->depth_stencil->depth);
depthStencilState.depthTestEnable = depth->enabled;
depthStencilState.depthTestFunc = swr_convert_depth_func(depth->func);
depthStencilState.depthWriteEnable = depth->writemask;
- SwrSetDepthStencilState(ctx->swrContext, &depthStencilState);
+ ctx->api.pfnSwrSetDepthStencilState(ctx->swrContext, &depthStencilState);
depthBoundsState.depthBoundsTestEnable = depth->bounds_test;
depthBoundsState.depthBoundsTestMinValue = depth->bounds_min;
depthBoundsState.depthBoundsTestMaxValue = depth->bounds_max;
- SwrSetDepthBoundsState(ctx->swrContext, &depthBoundsState);
+ ctx->api.pfnSwrSetDepthBoundsState(ctx->swrContext, &depthBoundsState);
}
/* Blend State */
if (ctx->dirty & (SWR_NEW_BLEND |
+ SWR_NEW_RASTERIZER |
SWR_NEW_FRAMEBUFFER |
SWR_NEW_DEPTH_STENCIL_ALPHA)) {
struct pipe_framebuffer_state *fb = &ctx->framebuffer;
blendState.alphaTestReference =
*((uint32_t*)&ctx->depth_stencil->alpha.ref_value);
- // XXX MSAA
- blendState.sampleMask = 0;
- blendState.sampleCount = SWR_MULTISAMPLE_1X;
+ blendState.sampleMask = ctx->sample_mask;
+ blendState.sampleCount = GetSampleCount(fb->samples);
/* If there are no color buffers bound, disable writes on RT0
* and skip loop */
blendState.renderTarget[0].writeDisableGreen = 1;
blendState.renderTarget[0].writeDisableBlue = 1;
blendState.renderTarget[0].writeDisableAlpha = 1;
- SwrSetBlendFunc(ctx->swrContext, 0, NULL);
+ ctx->api.pfnSwrSetBlendFunc(ctx->swrContext, 0, NULL);
}
else
for (int target = 0;
compileState.blendState.logicOpEnable = false;
}
+ if (info.type[0] == SWR_TYPE_SINT || info.type[0] == SWR_TYPE_UINT)
+ compileState.blendState.blendEnable = false;
+
if (compileState.blendState.blendEnable == false &&
compileState.blendState.logicOpEnable == false &&
ctx->depth_stencil->alpha.enabled == 0) {
- SwrSetBlendFunc(ctx->swrContext, target, NULL);
+ ctx->api.pfnSwrSetBlendFunc(ctx->swrContext, target, NULL);
continue;
}
compileState.desc.alphaTestEnable =
ctx->depth_stencil->alpha.enabled;
compileState.desc.independentAlphaBlendEnable =
- ctx->blend->pipe.independent_blend_enable;
+ (compileState.blendState.sourceBlendFactor !=
+ compileState.blendState.sourceAlphaBlendFactor) ||
+ (compileState.blendState.destBlendFactor !=
+ compileState.blendState.destAlphaBlendFactor) ||
+ (compileState.blendState.colorBlendFunc !=
+ compileState.blendState.alphaBlendFunc);
compileState.desc.alphaToCoverageEnable =
ctx->blend->pipe.alpha_to_coverage;
- compileState.desc.sampleMaskEnable = 0; // XXX
- compileState.desc.numSamples = 1; // XXX
+ compileState.desc.sampleMaskEnable = (blendState.sampleMask != 0);
+ compileState.desc.numSamples = fb->samples;
compileState.alphaTestFunction =
swr_convert_depth_func(ctx->depth_stencil->alpha.func);
ctx->blendJIT->insert(std::make_pair(compileState, func));
}
- SwrSetBlendFunc(ctx->swrContext, target, func);
+ ctx->api.pfnSwrSetBlendFunc(ctx->swrContext, target, func);
}
- SwrSetBlendState(ctx->swrContext, &blendState);
+ ctx->api.pfnSwrSetBlendState(ctx->swrContext, &blendState);
}
if (ctx->dirty & SWR_NEW_STIPPLE) {
- /* XXX What to do with this one??? SWR doesn't stipple */
+ swr_update_poly_stipple(ctx);
}
if (ctx->dirty & (SWR_NEW_VS | SWR_NEW_SO | SWR_NEW_RASTERIZER)) {
ctx->vs->soState.rasterizerDisable =
ctx->rasterizer->rasterizer_discard;
- SwrSetSoState(ctx->swrContext, &ctx->vs->soState);
+ ctx->api.pfnSwrSetSoState(ctx->swrContext, &ctx->vs->soState);
pipe_stream_output_info *stream_output = &ctx->vs->pipe.stream_output;
continue;
buffer.enable = true;
buffer.pBuffer =
- (uint32_t *)swr_resource_data(ctx->so_targets[i]->buffer);
+ (uint32_t *)(swr_resource_data(ctx->so_targets[i]->buffer) +
+ ctx->so_targets[i]->buffer_offset);
buffer.bufferSize = ctx->so_targets[i]->buffer_size >> 2;
buffer.pitch = stream_output->stride[i];
- buffer.streamOffset = ctx->so_targets[i]->buffer_offset >> 2;
+ buffer.streamOffset = 0;
- SwrSetSoBuffers(ctx->swrContext, &buffer, i);
+ ctx->api.pfnSwrSetSoBuffers(ctx->swrContext, &buffer, i);
}
}
- if (ctx->dirty & SWR_NEW_CLIP) {
+ if (ctx->dirty & (SWR_NEW_CLIP | SWR_NEW_RASTERIZER | SWR_NEW_VS)) {
// shader exporting clip distances overrides all user clip planes
if (ctx->rasterizer->clip_plane_enable &&
!ctx->vs->info.base.num_written_clipdistance)
// set up backend state
SWR_BACKEND_STATE backendState = {0};
- backendState.numAttributes =
- ctx->vs->info.base.num_outputs - 1 +
- (ctx->rasterizer->sprite_coord_enable ? 1 : 0);
+ if (ctx->gs) {
+ backendState.numAttributes = ctx->gs->info.base.num_outputs - 1;
+ } else {
+ backendState.numAttributes = ctx->vs->info.base.num_outputs - 1;
+ if (ctx->fs->info.base.uses_primid) {
+ backendState.numAttributes++;
+ backendState.swizzleEnable = true;
+ for (unsigned i = 0; i < sizeof(backendState.numComponents); i++) {
+ backendState.swizzleMap[i].sourceAttrib = i;
+ }
+ backendState.swizzleMap[ctx->vs->info.base.num_outputs - 1].constantSource =
+ SWR_CONSTANT_SOURCE_PRIM_ID;
+ backendState.swizzleMap[ctx->vs->info.base.num_outputs - 1].componentOverrideMask = 1;
+ }
+ }
+ if (ctx->rasterizer->sprite_coord_enable)
+ backendState.numAttributes++;
+
+ backendState.numAttributes = std::min((size_t)backendState.numAttributes,
+ sizeof(backendState.numComponents));
for (unsigned i = 0; i < backendState.numAttributes; i++)
backendState.numComponents[i] = 4;
- backendState.constantInterpolationMask =
- ctx->rasterizer->flatshade ?
- ctx->fs->flatConstantMask :
- ctx->fs->constantMask;
+ backendState.constantInterpolationMask = ctx->fs->constantMask |
+ (ctx->rasterizer->flatshade ? ctx->fs->flatConstantMask : 0);
backendState.pointSpriteTexCoordMask = ctx->fs->pointSpriteMask;
- SwrSetBackendState(ctx->swrContext, &backendState);
+ struct tgsi_shader_info *pLastFE =
+ ctx->gs ?
+ &ctx->gs->info.base :
+ &ctx->vs->info.base;
+ backendState.readRenderTargetArrayIndex = pLastFE->writes_layer;
+ backendState.readViewportArrayIndex = pLastFE->writes_viewport_index;
+ backendState.vertexAttribOffset = VERTEX_ATTRIB_START_SLOT; // TODO: optimize
+
+ backendState.clipDistanceMask =
+ ctx->vs->info.base.num_written_clipdistance ?
+ ctx->vs->info.base.clipdist_writemask & ctx->rasterizer->clip_plane_enable :
+ ctx->rasterizer->clip_plane_enable;
+
+ backendState.cullDistanceMask =
+ ctx->vs->info.base.culldist_writemask << ctx->vs->info.base.num_written_clipdistance;
+
+ // Assume old layout of SGV, POSITION, CLIPCULL, ATTRIB
+ backendState.vertexClipCullOffset = backendState.vertexAttribOffset - 2;
+
+ ctx->api.pfnSwrSetBackendState(ctx->swrContext, &backendState);
/* Ensure that any in-progress attachment change StoreTiles finish */
if (swr_is_fence_pending(screen->flush_fence))
struct swr_context *swr = swr_context(pipe);
uint32_t i;
- assert(num_targets < MAX_SO_STREAMS);
+ assert(num_targets <= MAX_SO_STREAMS);
for (i = 0; i < num_targets; i++) {
pipe_so_target_reference(
swr->num_so_targets = num_targets;
- swr->dirty = SWR_NEW_SO;
+ swr->dirty |= SWR_NEW_SO;
}
pipe->bind_fs_state = swr_bind_fs_state;
pipe->delete_fs_state = swr_delete_fs_state;
+ pipe->create_gs_state = swr_create_gs_state;
+ pipe->bind_gs_state = swr_bind_gs_state;
+ pipe->delete_gs_state = swr_delete_gs_state;
+
pipe->set_constant_buffer = swr_set_constant_buffer;
pipe->create_vertex_elements_state = swr_create_vertex_elements_state;
pipe->delete_vertex_elements_state = swr_delete_vertex_elements_state;
pipe->set_vertex_buffers = swr_set_vertex_buffers;
- pipe->set_index_buffer = swr_set_index_buffer;
pipe->set_polygon_stipple = swr_set_polygon_stipple;
pipe->set_clip_state = swr_set_clip_state;
pipe->set_stencil_ref = swr_set_stencil_ref;
pipe->set_sample_mask = swr_set_sample_mask;
+ pipe->get_sample_position = swr_get_sample_position;
pipe->create_stream_output_target = swr_create_so_target;
pipe->stream_output_target_destroy = swr_destroy_so_target;
projects / mesa.git / blobdiff