/**************************************************************************
*
- * Copyright 2009 VMware, Inc.
+ * Copyright 2009-2010 VMware, Inc.
* All Rights Reserved.
*
* Permission is hereby granted, free of charge, to any person obtaining a
**************************************************************************/
+#include "pipe/p_screen.h"
#include "pipe/p_context.h"
#include "pipe/p_state.h"
#include "tgsi/tgsi_ureg.h"
#include "tgsi/tgsi_dump.h"
#include "tgsi/tgsi_sanity.h"
#include "util/u_debug.h"
+#include "util/u_inlines.h"
#include "util/u_memory.h"
#include "util/u_math.h"
+#include "util/u_bitmask.h"
union tgsi_any_token {
struct tgsi_header header;
struct tgsi_declaration decl;
struct tgsi_declaration_range decl_range;
struct tgsi_declaration_dimension decl_dim;
+ struct tgsi_declaration_interp decl_interp;
struct tgsi_declaration_semantic decl_semantic;
+ struct tgsi_declaration_sampler_view decl_sampler_view;
+ struct tgsi_declaration_array array;
struct tgsi_immediate imm;
union tgsi_immediate_data imm_data;
struct tgsi_instruction insn;
struct tgsi_instruction_predicate insn_predicate;
struct tgsi_instruction_label insn_label;
struct tgsi_instruction_texture insn_texture;
+ struct tgsi_texture_offset insn_texture_offset;
struct tgsi_src_register src;
+ struct tgsi_ind_register ind;
struct tgsi_dimension dim;
struct tgsi_dst_register dst;
unsigned value;
unsigned count;
};
-#define UREG_MAX_INPUT PIPE_MAX_ATTRIBS
+#define UREG_MAX_INPUT PIPE_MAX_SHADER_INPUTS
#define UREG_MAX_SYSTEM_VALUE PIPE_MAX_ATTRIBS
-#define UREG_MAX_OUTPUT PIPE_MAX_ATTRIBS
+#define UREG_MAX_OUTPUT PIPE_MAX_SHADER_OUTPUTS
#define UREG_MAX_CONSTANT_RANGE 32
-#define UREG_MAX_IMMEDIATE 32
-#define UREG_MAX_TEMP 256
-#define UREG_MAX_ADDR 2
+#define UREG_MAX_IMMEDIATE 4096
+#define UREG_MAX_ADDR 3
#define UREG_MAX_PRED 1
+#define UREG_MAX_ARRAY_TEMPS 256
struct const_decl {
struct {
struct ureg_program
{
unsigned processor;
- struct pipe_context *pipe;
+ bool supports_any_inout_decl_range;
struct {
unsigned semantic_name;
unsigned semantic_index;
unsigned interp;
unsigned char cylindrical_wrap;
- unsigned char centroid;
- } fs_input[UREG_MAX_INPUT];
- unsigned nr_fs_inputs;
+ unsigned interp_location;
+ unsigned first;
+ unsigned last;
+ unsigned array_id;
+ } input[UREG_MAX_INPUT];
+ unsigned nr_inputs, nr_input_regs;
- unsigned vs_inputs[UREG_MAX_INPUT/32];
+ unsigned vs_inputs[PIPE_MAX_ATTRIBS/32];
struct {
- unsigned index;
- unsigned semantic_name;
- unsigned semantic_index;
- } gs_input[UREG_MAX_INPUT];
- unsigned nr_gs_inputs;
-
- struct {
- unsigned index;
unsigned semantic_name;
unsigned semantic_index;
} system_value[UREG_MAX_SYSTEM_VALUE];
struct {
unsigned semantic_name;
unsigned semantic_index;
+ unsigned usage_mask; /* = TGSI_WRITEMASK_* */
+ unsigned first;
+ unsigned last;
+ unsigned array_id;
} output[UREG_MAX_OUTPUT];
- unsigned nr_outputs;
+ unsigned nr_outputs, nr_output_regs;
struct {
union {
struct ureg_src sampler[PIPE_MAX_SAMPLERS];
unsigned nr_samplers;
- unsigned temps_active[UREG_MAX_TEMP / 32];
+ struct {
+ unsigned index;
+ unsigned target;
+ unsigned return_type_x;
+ unsigned return_type_y;
+ unsigned return_type_z;
+ unsigned return_type_w;
+ } sampler_view[PIPE_MAX_SHADER_SAMPLER_VIEWS];
+ unsigned nr_sampler_views;
+
+ struct util_bitmask *free_temps;
+ struct util_bitmask *local_temps;
+ struct util_bitmask *decl_temps;
unsigned nr_temps;
+ unsigned array_temps[UREG_MAX_ARRAY_TEMPS];
+ unsigned nr_array_temps;
+
struct const_decl const_decls;
struct const_decl const_decls2D[PIPE_MAX_CONSTANT_BUFFERS];
- unsigned property_gs_input_prim;
- unsigned property_gs_output_prim;
- unsigned property_gs_max_vertices;
- unsigned char property_fs_coord_origin; /* = TGSI_FS_COORD_ORIGIN_* */
- unsigned char property_fs_coord_pixel_center; /* = TGSI_FS_COORD_PIXEL_CENTER_* */
+ unsigned properties[TGSI_PROPERTY_COUNT];
unsigned nr_addrs;
unsigned nr_preds;
return &ureg->domain[domain].tokens[nr];
}
-
-
-static INLINE struct ureg_dst
-ureg_dst_register( unsigned file,
- unsigned index )
-{
- struct ureg_dst dst;
-
- dst.File = file;
- dst.WriteMask = TGSI_WRITEMASK_XYZW;
- dst.Indirect = 0;
- dst.IndirectIndex = 0;
- dst.IndirectSwizzle = 0;
- dst.Saturate = 0;
- dst.Predicate = 0;
- dst.PredNegate = 0;
- dst.PredSwizzleX = TGSI_SWIZZLE_X;
- dst.PredSwizzleY = TGSI_SWIZZLE_Y;
- dst.PredSwizzleZ = TGSI_SWIZZLE_Z;
- dst.PredSwizzleW = TGSI_SWIZZLE_W;
- dst.Index = index;
-
- return dst;
-}
-
-
-void
-ureg_property_gs_input_prim(struct ureg_program *ureg,
- unsigned input_prim)
-{
- ureg->property_gs_input_prim = input_prim;
-}
-
-void
-ureg_property_gs_output_prim(struct ureg_program *ureg,
- unsigned output_prim)
-{
- ureg->property_gs_output_prim = output_prim;
-}
-
void
-ureg_property_gs_max_vertices(struct ureg_program *ureg,
- unsigned max_vertices)
+ureg_property(struct ureg_program *ureg, unsigned name, unsigned value)
{
- ureg->property_gs_max_vertices = max_vertices;
+ assert(name < Elements(ureg->properties));
+ ureg->properties[name] = value;
}
-void
-ureg_property_fs_coord_origin(struct ureg_program *ureg,
- unsigned fs_coord_origin)
-{
- ureg->property_fs_coord_origin = fs_coord_origin;
-}
-
-void
-ureg_property_fs_coord_pixel_center(struct ureg_program *ureg,
- unsigned fs_coord_pixel_center)
-{
- ureg->property_fs_coord_pixel_center = fs_coord_pixel_center;
-}
-
-
-
struct ureg_src
ureg_DECL_fs_input_cyl_centroid(struct ureg_program *ureg,
unsigned semantic_name,
unsigned semantic_index,
unsigned interp_mode,
unsigned cylindrical_wrap,
- unsigned centroid)
+ unsigned interp_location,
+ unsigned array_id,
+ unsigned array_size)
{
unsigned i;
- for (i = 0; i < ureg->nr_fs_inputs; i++) {
- if (ureg->fs_input[i].semantic_name == semantic_name &&
- ureg->fs_input[i].semantic_index == semantic_index) {
+ for (i = 0; i < ureg->nr_inputs; i++) {
+ if (ureg->input[i].semantic_name == semantic_name &&
+ ureg->input[i].semantic_index == semantic_index) {
+ assert(ureg->input[i].interp == interp_mode);
+ assert(ureg->input[i].cylindrical_wrap == cylindrical_wrap);
+ assert(ureg->input[i].interp_location == interp_location);
+ assert(ureg->input[i].array_id == array_id);
goto out;
}
}
- if (ureg->nr_fs_inputs < UREG_MAX_INPUT) {
- ureg->fs_input[i].semantic_name = semantic_name;
- ureg->fs_input[i].semantic_index = semantic_index;
- ureg->fs_input[i].interp = interp_mode;
- ureg->fs_input[i].cylindrical_wrap = cylindrical_wrap;
- ureg->fs_input[i].centroid = centroid;
- ureg->nr_fs_inputs++;
+ if (ureg->nr_inputs < UREG_MAX_INPUT) {
+ assert(array_size >= 1);
+ ureg->input[i].semantic_name = semantic_name;
+ ureg->input[i].semantic_index = semantic_index;
+ ureg->input[i].interp = interp_mode;
+ ureg->input[i].cylindrical_wrap = cylindrical_wrap;
+ ureg->input[i].interp_location = interp_location;
+ ureg->input[i].first = ureg->nr_input_regs;
+ ureg->input[i].last = ureg->nr_input_regs + array_size - 1;
+ ureg->input[i].array_id = array_id;
+ ureg->nr_input_regs += array_size;
+ ureg->nr_inputs++;
} else {
set_bad(ureg);
}
out:
- return ureg_src_register(TGSI_FILE_INPUT, i);
+ return ureg_src_array_register(TGSI_FILE_INPUT, ureg->input[i].first,
+ array_id);
}
unsigned index )
{
assert(ureg->processor == TGSI_PROCESSOR_VERTEX);
-
+ assert(index / 32 < ARRAY_SIZE(ureg->vs_inputs));
+
ureg->vs_inputs[index/32] |= 1 << (index % 32);
return ureg_src_register( TGSI_FILE_INPUT, index );
}
struct ureg_src
-ureg_DECL_gs_input(struct ureg_program *ureg,
- unsigned index,
- unsigned semantic_name,
- unsigned semantic_index)
+ureg_DECL_input(struct ureg_program *ureg,
+ unsigned semantic_name,
+ unsigned semantic_index,
+ unsigned array_id,
+ unsigned array_size)
{
- if (ureg->nr_gs_inputs < UREG_MAX_INPUT) {
- ureg->gs_input[ureg->nr_gs_inputs].index = index;
- ureg->gs_input[ureg->nr_gs_inputs].semantic_name = semantic_name;
- ureg->gs_input[ureg->nr_gs_inputs].semantic_index = semantic_index;
- ureg->nr_gs_inputs++;
- } else {
- set_bad(ureg);
- }
-
- /* XXX: Add suport for true 2D input registers. */
- return ureg_src_register(TGSI_FILE_INPUT, index);
+ return ureg_DECL_fs_input_cyl_centroid(ureg, semantic_name, semantic_index,
+ 0, 0, 0, array_id, array_size);
}
struct ureg_src
ureg_DECL_system_value(struct ureg_program *ureg,
- unsigned index,
unsigned semantic_name,
unsigned semantic_index)
{
+ unsigned i = 0;
+
if (ureg->nr_system_values < UREG_MAX_SYSTEM_VALUE) {
- ureg->system_value[ureg->nr_system_values].index = index;
ureg->system_value[ureg->nr_system_values].semantic_name = semantic_name;
ureg->system_value[ureg->nr_system_values].semantic_index = semantic_index;
+ i = ureg->nr_system_values;
ureg->nr_system_values++;
} else {
set_bad(ureg);
}
- return ureg_src_register(TGSI_FILE_SYSTEM_VALUE, index);
+ return ureg_src_register(TGSI_FILE_SYSTEM_VALUE, i);
}
struct ureg_dst
-ureg_DECL_output( struct ureg_program *ureg,
- unsigned name,
- unsigned index )
+ureg_DECL_output_masked(struct ureg_program *ureg,
+ unsigned name,
+ unsigned index,
+ unsigned usage_mask,
+ unsigned array_id,
+ unsigned array_size)
{
unsigned i;
+ assert(usage_mask != 0);
+
for (i = 0; i < ureg->nr_outputs; i++) {
if (ureg->output[i].semantic_name == name &&
- ureg->output[i].semantic_index == index)
+ ureg->output[i].semantic_index == index) {
+ assert(ureg->output[i].array_id == array_id);
+ ureg->output[i].usage_mask |= usage_mask;
goto out;
+ }
}
if (ureg->nr_outputs < UREG_MAX_OUTPUT) {
ureg->output[i].semantic_name = name;
ureg->output[i].semantic_index = index;
+ ureg->output[i].usage_mask = usage_mask;
+ ureg->output[i].first = ureg->nr_output_regs;
+ ureg->output[i].last = ureg->nr_output_regs + array_size - 1;
+ ureg->output[i].array_id = array_id;
+ ureg->nr_output_regs += array_size;
ureg->nr_outputs++;
}
else {
}
out:
- return ureg_dst_register( TGSI_FILE_OUTPUT, i );
+ return ureg_dst_array_register(TGSI_FILE_OUTPUT, ureg->output[i].first,
+ array_id);
+}
+
+
+struct ureg_dst
+ureg_DECL_output(struct ureg_program *ureg,
+ unsigned name,
+ unsigned index)
+{
+ return ureg_DECL_output_masked(ureg, name, index, TGSI_WRITEMASK_XYZW,
+ 0, 1);
+}
+
+struct ureg_dst
+ureg_DECL_output_array(struct ureg_program *ureg,
+ unsigned semantic_name,
+ unsigned semantic_index,
+ unsigned array_id,
+ unsigned array_size)
+{
+ return ureg_DECL_output_masked(ureg, semantic_name, semantic_index,
+ TGSI_WRITEMASK_XYZW,
+ array_id, array_size);
}
return ureg_src_register(TGSI_FILE_CONSTANT, index);
}
-
-/* Allocate a new temporary. Temporaries greater than UREG_MAX_TEMP
- * are legal, but will not be released.
- */
-struct ureg_dst ureg_DECL_temporary( struct ureg_program *ureg )
+static struct ureg_dst alloc_temporary( struct ureg_program *ureg,
+ boolean local )
{
unsigned i;
- for (i = 0; i < UREG_MAX_TEMP; i += 32) {
- int bit = ffs(~ureg->temps_active[i/32]);
- if (bit != 0) {
- i += bit - 1;
- goto out;
- }
+ /* Look for a released temporary.
+ */
+ for (i = util_bitmask_get_first_index(ureg->free_temps);
+ i != UTIL_BITMASK_INVALID_INDEX;
+ i = util_bitmask_get_next_index(ureg->free_temps, i + 1)) {
+ if (util_bitmask_get(ureg->local_temps, i) == local)
+ break;
}
- /* No reusable temps, so allocate a new one:
+ /* Or allocate a new one.
*/
- i = ureg->nr_temps++;
+ if (i == UTIL_BITMASK_INVALID_INDEX) {
+ i = ureg->nr_temps++;
-out:
- if (i < UREG_MAX_TEMP)
- ureg->temps_active[i/32] |= 1 << (i % 32);
+ if (local)
+ util_bitmask_set(ureg->local_temps, i);
+
+ /* Start a new declaration when the local flag changes */
+ if (!i || util_bitmask_get(ureg->local_temps, i - 1) != local)
+ util_bitmask_set(ureg->decl_temps, i);
+ }
- if (i >= ureg->nr_temps)
- ureg->nr_temps = i + 1;
+ util_bitmask_clear(ureg->free_temps, i);
return ureg_dst_register( TGSI_FILE_TEMPORARY, i );
}
+struct ureg_dst ureg_DECL_temporary( struct ureg_program *ureg )
+{
+ return alloc_temporary(ureg, FALSE);
+}
+
+struct ureg_dst ureg_DECL_local_temporary( struct ureg_program *ureg )
+{
+ return alloc_temporary(ureg, TRUE);
+}
+
+struct ureg_dst ureg_DECL_array_temporary( struct ureg_program *ureg,
+ unsigned size,
+ boolean local )
+{
+ unsigned i = ureg->nr_temps;
+ struct ureg_dst dst = ureg_dst_register( TGSI_FILE_TEMPORARY, i );
+
+ if (local)
+ util_bitmask_set(ureg->local_temps, i);
+
+ /* Always start a new declaration at the start */
+ util_bitmask_set(ureg->decl_temps, i);
+
+ ureg->nr_temps += size;
+
+ /* and also at the end of the array */
+ util_bitmask_set(ureg->decl_temps, ureg->nr_temps);
+
+ if (ureg->nr_array_temps < UREG_MAX_ARRAY_TEMPS) {
+ ureg->array_temps[ureg->nr_array_temps++] = i;
+ dst.ArrayID = ureg->nr_array_temps;
+ }
+
+ return dst;
+}
void ureg_release_temporary( struct ureg_program *ureg,
struct ureg_dst tmp )
{
if(tmp.File == TGSI_FILE_TEMPORARY)
- if (tmp.Index < UREG_MAX_TEMP)
- ureg->temps_active[tmp.Index/32] &= ~(1 << (tmp.Index % 32));
+ util_bitmask_set(ureg->free_temps, tmp.Index);
}
return ureg->sampler[0];
}
+/*
+ * Allocate a new shader sampler view.
+ */
+struct ureg_src
+ureg_DECL_sampler_view(struct ureg_program *ureg,
+ unsigned index,
+ unsigned target,
+ unsigned return_type_x,
+ unsigned return_type_y,
+ unsigned return_type_z,
+ unsigned return_type_w)
+{
+ struct ureg_src reg = ureg_src_register(TGSI_FILE_SAMPLER_VIEW, index);
+ uint i;
+
+ for (i = 0; i < ureg->nr_sampler_views; i++) {
+ if (ureg->sampler_view[i].index == index) {
+ return reg;
+ }
+ }
+
+ if (i < PIPE_MAX_SHADER_SAMPLER_VIEWS) {
+ ureg->sampler_view[i].index = index;
+ ureg->sampler_view[i].target = target;
+ ureg->sampler_view[i].return_type_x = return_type_x;
+ ureg->sampler_view[i].return_type_y = return_type_y;
+ ureg->sampler_view[i].return_type_z = return_type_z;
+ ureg->sampler_view[i].return_type_w = return_type_w;
+ ureg->nr_sampler_views++;
+ return reg;
+ }
+
+ assert(0);
+ return reg;
+}
+
+static int
+match_or_expand_immediate64( const unsigned *v,
+ int type,
+ unsigned nr,
+ unsigned *v2,
+ unsigned *pnr2,
+ unsigned *swizzle )
+{
+ unsigned nr2 = *pnr2;
+ unsigned i, j;
+ *swizzle = 0;
+
+ for (i = 0; i < nr; i += 2) {
+ boolean found = FALSE;
+
+ for (j = 0; j < nr2 && !found; j += 2) {
+ if (v[i] == v2[j] && v[i + 1] == v2[j + 1]) {
+ *swizzle |= (j << (i * 2)) | ((j + 1) << ((i + 1) * 2));
+ found = TRUE;
+ }
+ }
+ if (!found) {
+ if ((nr2) >= 4) {
+ return FALSE;
+ }
+
+ v2[nr2] = v[i];
+ v2[nr2 + 1] = v[i + 1];
+
+ *swizzle |= (nr2 << (i * 2)) | ((nr2 + 1) << ((i + 1) * 2));
+ nr2 += 2;
+ }
+ }
+
+ /* Actually expand immediate only when fully succeeded.
+ */
+ *pnr2 = nr2;
+ return TRUE;
+}
static int
match_or_expand_immediate( const unsigned *v,
+ int type,
unsigned nr,
unsigned *v2,
unsigned *pnr2,
unsigned nr2 = *pnr2;
unsigned i, j;
+ if (type == TGSI_IMM_FLOAT64)
+ return match_or_expand_immediate64(v, type, nr, v2, pnr2, swizzle);
+
*swizzle = 0;
for (i = 0; i < nr; i++) {
continue;
}
if (match_or_expand_immediate(v,
+ type,
nr,
ureg->immediate[i].value.u,
&ureg->immediate[i].nr,
i = ureg->nr_immediates++;
ureg->immediate[i].type = type;
if (match_or_expand_immediate(v,
+ type,
nr,
ureg->immediate[i].value.u,
&ureg->immediate[i].nr,
/* Make sure that all referenced elements are from this immediate.
* Has the effect of making size-one immediates into scalars.
*/
- for (j = nr; j < 4; j++) {
- swizzle |= (swizzle & 0x3) << (j * 2);
+ if (type == TGSI_IMM_FLOAT64) {
+ for (j = nr; j < 4; j+=2) {
+ swizzle |= (swizzle & 0xf) << (j * 2);
+ }
+ } else {
+ for (j = nr; j < 4; j++) {
+ swizzle |= (swizzle & 0x3) << (j * 2);
+ }
}
-
return ureg_swizzle(ureg_src_register(TGSI_FILE_IMMEDIATE, i),
(swizzle >> 0) & 0x3,
(swizzle >> 2) & 0x3,
return decl_immediate(ureg, fu.u, nr, TGSI_IMM_FLOAT32);
}
+struct ureg_src
+ureg_DECL_immediate_f64( struct ureg_program *ureg,
+ const double *v,
+ unsigned nr )
+{
+ union {
+ unsigned u[4];
+ double d[2];
+ } fu;
+ unsigned int i;
+
+ assert((nr / 2) < 3);
+ for (i = 0; i < nr / 2; i++) {
+ fu.d[i] = v[i];
+ }
+
+ return decl_immediate(ureg, fu.u, nr, TGSI_IMM_FLOAT64);
+}
struct ureg_src
ureg_DECL_immediate_uint( struct ureg_program *ureg,
unsigned n = 0;
assert(src.File != TGSI_FILE_NULL);
- assert(src.File != TGSI_FILE_OUTPUT);
assert(src.File < TGSI_FILE_COUNT);
out[n].value = 0;
if (src.Indirect) {
out[0].src.Indirect = 1;
out[n].value = 0;
- out[n].src.File = src.IndirectFile;
- out[n].src.SwizzleX = src.IndirectSwizzle;
- out[n].src.SwizzleY = src.IndirectSwizzle;
- out[n].src.SwizzleZ = src.IndirectSwizzle;
- out[n].src.SwizzleW = src.IndirectSwizzle;
- out[n].src.Index = src.IndirectIndex;
+ out[n].ind.File = src.IndirectFile;
+ out[n].ind.Swizzle = src.IndirectSwizzle;
+ out[n].ind.Index = src.IndirectIndex;
+ if (!ureg->supports_any_inout_decl_range &&
+ (src.File == TGSI_FILE_INPUT || src.File == TGSI_FILE_OUTPUT))
+ out[n].ind.ArrayID = 0;
+ else
+ out[n].ind.ArrayID = src.ArrayID;
n++;
}
if (src.Dimension) {
+ out[0].src.Dimension = 1;
+ out[n].dim.Dimension = 0;
+ out[n].dim.Padding = 0;
if (src.DimIndirect) {
- out[0].src.Dimension = 1;
out[n].dim.Indirect = 1;
- out[n].dim.Dimension = 0;
- out[n].dim.Padding = 0;
out[n].dim.Index = src.DimensionIndex;
n++;
out[n].value = 0;
- out[n].src.File = src.DimIndFile;
- out[n].src.SwizzleX = src.DimIndSwizzle;
- out[n].src.SwizzleY = src.DimIndSwizzle;
- out[n].src.SwizzleZ = src.DimIndSwizzle;
- out[n].src.SwizzleW = src.DimIndSwizzle;
- out[n].src.Index = src.DimIndIndex;
+ out[n].ind.File = src.DimIndFile;
+ out[n].ind.Swizzle = src.DimIndSwizzle;
+ out[n].ind.Index = src.DimIndIndex;
+ if (!ureg->supports_any_inout_decl_range &&
+ (src.File == TGSI_FILE_INPUT || src.File == TGSI_FILE_OUTPUT))
+ out[n].ind.ArrayID = 0;
+ else
+ out[n].ind.ArrayID = src.ArrayID;
} else {
- out[0].src.Dimension = 1;
out[n].dim.Indirect = 0;
- out[n].dim.Dimension = 0;
- out[n].dim.Padding = 0;
out[n].dim.Index = src.DimensionIndex;
}
n++;
ureg_emit_dst( struct ureg_program *ureg,
struct ureg_dst dst )
{
- unsigned size = (1 +
- (dst.Indirect ? 1 : 0));
+ unsigned size = 1 + (dst.Indirect ? 1 : 0) +
+ (dst.Dimension ? (dst.DimIndirect ? 2 : 1) : 0);
union tgsi_any_token *out = get_tokens( ureg, DOMAIN_INSN, size );
unsigned n = 0;
assert(dst.File != TGSI_FILE_CONSTANT);
assert(dst.File != TGSI_FILE_INPUT);
assert(dst.File != TGSI_FILE_SAMPLER);
+ assert(dst.File != TGSI_FILE_SAMPLER_VIEW);
assert(dst.File != TGSI_FILE_IMMEDIATE);
assert(dst.File < TGSI_FILE_COUNT);
-
+
out[n].value = 0;
out[n].dst.File = dst.File;
out[n].dst.WriteMask = dst.WriteMask;
if (dst.Indirect) {
out[n].value = 0;
- out[n].src.File = TGSI_FILE_ADDRESS;
- out[n].src.SwizzleX = dst.IndirectSwizzle;
- out[n].src.SwizzleY = dst.IndirectSwizzle;
- out[n].src.SwizzleZ = dst.IndirectSwizzle;
- out[n].src.SwizzleW = dst.IndirectSwizzle;
- out[n].src.Index = dst.IndirectIndex;
+ out[n].ind.File = dst.IndirectFile;
+ out[n].ind.Swizzle = dst.IndirectSwizzle;
+ out[n].ind.Index = dst.IndirectIndex;
+ if (!ureg->supports_any_inout_decl_range &&
+ (dst.File == TGSI_FILE_INPUT || dst.File == TGSI_FILE_OUTPUT))
+ out[n].ind.ArrayID = 0;
+ else
+ out[n].ind.ArrayID = dst.ArrayID;
+ n++;
+ }
+
+ if (dst.Dimension) {
+ out[0].dst.Dimension = 1;
+ out[n].dim.Dimension = 0;
+ out[n].dim.Padding = 0;
+ if (dst.DimIndirect) {
+ out[n].dim.Indirect = 1;
+ out[n].dim.Index = dst.DimensionIndex;
+ n++;
+ out[n].value = 0;
+ out[n].ind.File = dst.DimIndFile;
+ out[n].ind.Swizzle = dst.DimIndSwizzle;
+ out[n].ind.Index = dst.DimIndIndex;
+ if (!ureg->supports_any_inout_decl_range &&
+ (dst.File == TGSI_FILE_INPUT || dst.File == TGSI_FILE_OUTPUT))
+ out[n].ind.ArrayID = 0;
+ else
+ out[n].ind.ArrayID = dst.ArrayID;
+ } else {
+ out[n].dim.Indirect = 0;
+ out[n].dim.Index = dst.DimensionIndex;
+ }
n++;
}
#ifdef DEBUG
const struct tgsi_opcode_info *info = tgsi_get_opcode_info( opcode );
assert(info);
- if(info) {
+ if (info) {
assert(nr_dst == info->num_dst);
assert(nr_src == info->num_src);
}
}
+/**
+ * Emit a label token.
+ * \param label_token returns a token number indicating where the label
+ * needs to be patched later. Later, this value should be passed to the
+ * ureg_fixup_label() function.
+ */
void
ureg_emit_label(struct ureg_program *ureg,
unsigned extended_token,
{
union tgsi_any_token *out, *insn;
- if(!label_token)
+ if (!label_token)
return;
out = get_tokens( ureg, DOMAIN_INSN, 1 );
void
ureg_emit_texture(struct ureg_program *ureg,
unsigned extended_token,
- unsigned target )
+ unsigned target, unsigned num_offsets)
{
union tgsi_any_token *out, *insn;
out[0].value = 0;
out[0].insn_texture.Texture = target;
+ out[0].insn_texture.NumOffsets = num_offsets;
+}
+
+void
+ureg_emit_texture_offset(struct ureg_program *ureg,
+ const struct tgsi_texture_offset *offset)
+{
+ union tgsi_any_token *out;
+
+ out = get_tokens( ureg, DOMAIN_INSN, 1);
+
+ out[0].value = 0;
+ out[0].insn_texture_offset = *offset;
+
}
boolean negate = FALSE;
unsigned swizzle[4] = { 0 };
+ if (nr_dst && ureg_dst_is_empty(dst[0])) {
+ return;
+ }
+
saturate = nr_dst ? dst[0].Saturate : FALSE;
predicate = nr_dst ? dst[0].Predicate : FALSE;
if (predicate) {
const struct ureg_dst *dst,
unsigned nr_dst,
unsigned target,
+ const struct tgsi_texture_offset *texoffsets,
+ unsigned nr_offset,
const struct ureg_src *src,
unsigned nr_src )
{
boolean negate = FALSE;
unsigned swizzle[4] = { 0 };
+ if (nr_dst && ureg_dst_is_empty(dst[0])) {
+ return;
+ }
+
saturate = nr_dst ? dst[0].Saturate : FALSE;
predicate = nr_dst ? dst[0].Predicate : FALSE;
if (predicate) {
nr_dst,
nr_src);
- ureg_emit_texture( ureg, insn.extended_token, target );
+ ureg_emit_texture( ureg, insn.extended_token, target, nr_offset );
+
+ for (i = 0; i < nr_offset; i++)
+ ureg_emit_texture_offset( ureg, &texoffsets[i]);
for (i = 0; i < nr_dst; i++)
ureg_emit_dst( ureg, dst[i] );
static void
emit_decl_semantic(struct ureg_program *ureg,
unsigned file,
- unsigned index,
+ unsigned first,
+ unsigned last,
unsigned semantic_name,
- unsigned semantic_index)
+ unsigned semantic_index,
+ unsigned usage_mask,
+ unsigned array_id)
{
- union tgsi_any_token *out = get_tokens(ureg, DOMAIN_DECL, 3);
+ union tgsi_any_token *out = get_tokens(ureg, DOMAIN_DECL, array_id ? 4 : 3);
out[0].value = 0;
out[0].decl.Type = TGSI_TOKEN_TYPE_DECLARATION;
out[0].decl.NrTokens = 3;
out[0].decl.File = file;
- out[0].decl.UsageMask = TGSI_WRITEMASK_XYZW; /* FIXME! */
+ out[0].decl.UsageMask = usage_mask;
out[0].decl.Semantic = 1;
+ out[0].decl.Array = array_id != 0;
out[1].value = 0;
- out[1].decl_range.First = index;
- out[1].decl_range.Last = index;
+ out[1].decl_range.First = first;
+ out[1].decl_range.Last = last;
out[2].value = 0;
out[2].decl_semantic.Name = semantic_name;
out[2].decl_semantic.Index = semantic_index;
+
+ if (array_id) {
+ out[3].value = 0;
+ out[3].array.ArrayID = array_id;
+ }
}
static void
emit_decl_fs(struct ureg_program *ureg,
unsigned file,
- unsigned index,
+ unsigned first,
+ unsigned last,
unsigned semantic_name,
unsigned semantic_index,
unsigned interpolate,
unsigned cylindrical_wrap,
- unsigned centroid)
+ unsigned interpolate_location,
+ unsigned array_id)
{
- union tgsi_any_token *out = get_tokens(ureg, DOMAIN_DECL, 3);
+ union tgsi_any_token *out = get_tokens(ureg, DOMAIN_DECL,
+ array_id ? 5 : 4);
out[0].value = 0;
out[0].decl.Type = TGSI_TOKEN_TYPE_DECLARATION;
- out[0].decl.NrTokens = 3;
+ out[0].decl.NrTokens = 4;
out[0].decl.File = file;
out[0].decl.UsageMask = TGSI_WRITEMASK_XYZW; /* FIXME! */
- out[0].decl.Interpolate = interpolate;
+ out[0].decl.Interpolate = 1;
out[0].decl.Semantic = 1;
- out[0].decl.CylindricalWrap = cylindrical_wrap;
- out[0].decl.Centroid = centroid;
+ out[0].decl.Array = array_id != 0;
out[1].value = 0;
- out[1].decl_range.First = index;
- out[1].decl_range.Last = index;
+ out[1].decl_range.First = first;
+ out[1].decl_range.Last = last;
out[2].value = 0;
- out[2].decl_semantic.Name = semantic_name;
- out[2].decl_semantic.Index = semantic_index;
+ out[2].decl_interp.Interpolate = interpolate;
+ out[2].decl_interp.CylindricalWrap = cylindrical_wrap;
+ out[2].decl_interp.Location = interpolate_location;
+
+ out[3].value = 0;
+ out[3].decl_semantic.Name = semantic_name;
+ out[3].decl_semantic.Index = semantic_index;
+
+ if (array_id) {
+ out[4].value = 0;
+ out[4].array.ArrayID = array_id;
+ }
}
+static void
+emit_decl_temps( struct ureg_program *ureg,
+ unsigned first, unsigned last,
+ boolean local,
+ unsigned arrayid )
+{
+ union tgsi_any_token *out = get_tokens( ureg, DOMAIN_DECL,
+ arrayid ? 3 : 2 );
+
+ out[0].value = 0;
+ out[0].decl.Type = TGSI_TOKEN_TYPE_DECLARATION;
+ out[0].decl.NrTokens = 2;
+ out[0].decl.File = TGSI_FILE_TEMPORARY;
+ out[0].decl.UsageMask = TGSI_WRITEMASK_XYZW;
+ out[0].decl.Local = local;
+
+ out[1].value = 0;
+ out[1].decl_range.First = first;
+ out[1].decl_range.Last = last;
+
+ if (arrayid) {
+ out[0].decl.Array = 1;
+ out[2].value = 0;
+ out[2].array.ArrayID = arrayid;
+ }
+}
static void emit_decl_range( struct ureg_program *ureg,
unsigned file,
out[0].decl.NrTokens = 2;
out[0].decl.File = file;
out[0].decl.UsageMask = TGSI_WRITEMASK_XYZW;
- out[0].decl.Interpolate = TGSI_INTERPOLATE_CONSTANT;
out[0].decl.Semantic = 0;
out[1].value = 0;
out[0].decl.NrTokens = 3;
out[0].decl.File = file;
out[0].decl.UsageMask = TGSI_WRITEMASK_XYZW;
- out[0].decl.Interpolate = TGSI_INTERPOLATE_CONSTANT;
out[0].decl.Dimension = 1;
out[1].value = 0;
out[2].decl_dim.Index2D = index2D;
}
+static void
+emit_decl_sampler_view(struct ureg_program *ureg,
+ unsigned index,
+ unsigned target,
+ unsigned return_type_x,
+ unsigned return_type_y,
+ unsigned return_type_z,
+ unsigned return_type_w )
+{
+ union tgsi_any_token *out = get_tokens(ureg, DOMAIN_DECL, 3);
+
+ out[0].value = 0;
+ out[0].decl.Type = TGSI_TOKEN_TYPE_DECLARATION;
+ out[0].decl.NrTokens = 3;
+ out[0].decl.File = TGSI_FILE_SAMPLER_VIEW;
+ out[0].decl.UsageMask = 0xf;
+
+ out[1].value = 0;
+ out[1].decl_range.First = index;
+ out[1].decl_range.Last = index;
+
+ out[2].value = 0;
+ out[2].decl_sampler_view.Resource = target;
+ out[2].decl_sampler_view.ReturnTypeX = return_type_x;
+ out[2].decl_sampler_view.ReturnTypeY = return_type_y;
+ out[2].decl_sampler_view.ReturnTypeZ = return_type_z;
+ out[2].decl_sampler_view.ReturnTypeW = return_type_w;
+}
+
static void
emit_immediate( struct ureg_program *ureg,
const unsigned *v,
static void emit_decls( struct ureg_program *ureg )
{
- unsigned i;
-
- if (ureg->property_gs_input_prim != ~0) {
- assert(ureg->processor == TGSI_PROCESSOR_GEOMETRY);
-
- emit_property(ureg,
- TGSI_PROPERTY_GS_INPUT_PRIM,
- ureg->property_gs_input_prim);
- }
-
- if (ureg->property_gs_output_prim != ~0) {
- assert(ureg->processor == TGSI_PROCESSOR_GEOMETRY);
+ unsigned i,j;
- emit_property(ureg,
- TGSI_PROPERTY_GS_OUTPUT_PRIM,
- ureg->property_gs_output_prim);
- }
-
- if (ureg->property_gs_max_vertices != ~0) {
- assert(ureg->processor == TGSI_PROCESSOR_GEOMETRY);
-
- emit_property(ureg,
- TGSI_PROPERTY_GS_MAX_OUTPUT_VERTICES,
- ureg->property_gs_max_vertices);
- }
-
- if (ureg->property_fs_coord_origin) {
- assert(ureg->processor == TGSI_PROCESSOR_FRAGMENT);
-
- emit_property(ureg,
- TGSI_PROPERTY_FS_COORD_ORIGIN,
- ureg->property_fs_coord_origin);
- }
-
- if (ureg->property_fs_coord_pixel_center) {
- assert(ureg->processor == TGSI_PROCESSOR_FRAGMENT);
-
- emit_property(ureg,
- TGSI_PROPERTY_FS_COORD_PIXEL_CENTER,
- ureg->property_fs_coord_pixel_center);
- }
+ for (i = 0; i < Elements(ureg->properties); i++)
+ if (ureg->properties[i] != ~0)
+ emit_property(ureg, i, ureg->properties[i]);
if (ureg->processor == TGSI_PROCESSOR_VERTEX) {
- for (i = 0; i < UREG_MAX_INPUT; i++) {
+ for (i = 0; i < PIPE_MAX_ATTRIBS; i++) {
if (ureg->vs_inputs[i/32] & (1 << (i%32))) {
emit_decl_range( ureg, TGSI_FILE_INPUT, i, 1 );
}
}
} else if (ureg->processor == TGSI_PROCESSOR_FRAGMENT) {
- for (i = 0; i < ureg->nr_fs_inputs; i++) {
- emit_decl_fs(ureg,
- TGSI_FILE_INPUT,
- i,
- ureg->fs_input[i].semantic_name,
- ureg->fs_input[i].semantic_index,
- ureg->fs_input[i].interp,
- ureg->fs_input[i].cylindrical_wrap,
- ureg->fs_input[i].centroid);
+ if (ureg->supports_any_inout_decl_range) {
+ for (i = 0; i < ureg->nr_inputs; i++) {
+ emit_decl_fs(ureg,
+ TGSI_FILE_INPUT,
+ ureg->input[i].first,
+ ureg->input[i].last,
+ ureg->input[i].semantic_name,
+ ureg->input[i].semantic_index,
+ ureg->input[i].interp,
+ ureg->input[i].cylindrical_wrap,
+ ureg->input[i].interp_location,
+ ureg->input[i].array_id);
+ }
}
- } else {
- for (i = 0; i < ureg->nr_gs_inputs; i++) {
- emit_decl_semantic(ureg,
+ else {
+ for (i = 0; i < ureg->nr_inputs; i++) {
+ for (j = ureg->input[i].first; j <= ureg->input[i].last; j++) {
+ emit_decl_fs(ureg,
TGSI_FILE_INPUT,
- ureg->gs_input[i].index,
- ureg->gs_input[i].semantic_name,
- ureg->gs_input[i].semantic_index);
+ j, j,
+ ureg->input[i].semantic_name,
+ ureg->input[i].semantic_index +
+ (j - ureg->input[i].first),
+ ureg->input[i].interp,
+ ureg->input[i].cylindrical_wrap,
+ ureg->input[i].interp_location, 0);
+ }
+ }
+ }
+ } else {
+ if (ureg->supports_any_inout_decl_range) {
+ for (i = 0; i < ureg->nr_inputs; i++) {
+ emit_decl_semantic(ureg,
+ TGSI_FILE_INPUT,
+ ureg->input[i].first,
+ ureg->input[i].last,
+ ureg->input[i].semantic_name,
+ ureg->input[i].semantic_index,
+ TGSI_WRITEMASK_XYZW,
+ ureg->input[i].array_id);
+ }
+ }
+ else {
+ for (i = 0; i < ureg->nr_inputs; i++) {
+ for (j = ureg->input[i].first; j <= ureg->input[i].last; j++) {
+ emit_decl_semantic(ureg,
+ TGSI_FILE_INPUT,
+ j, j,
+ ureg->input[i].semantic_name,
+ ureg->input[i].semantic_index +
+ (j - ureg->input[i].first),
+ TGSI_WRITEMASK_XYZW, 0);
+ }
+ }
}
}
for (i = 0; i < ureg->nr_system_values; i++) {
emit_decl_semantic(ureg,
TGSI_FILE_SYSTEM_VALUE,
- ureg->system_value[i].index,
+ i,
+ i,
ureg->system_value[i].semantic_name,
- ureg->system_value[i].semantic_index);
+ ureg->system_value[i].semantic_index,
+ TGSI_WRITEMASK_XYZW, 0);
}
- for (i = 0; i < ureg->nr_outputs; i++) {
- emit_decl_semantic(ureg,
- TGSI_FILE_OUTPUT,
- i,
- ureg->output[i].semantic_name,
- ureg->output[i].semantic_index);
+ if (ureg->supports_any_inout_decl_range) {
+ for (i = 0; i < ureg->nr_outputs; i++) {
+ emit_decl_semantic(ureg,
+ TGSI_FILE_OUTPUT,
+ ureg->output[i].first,
+ ureg->output[i].last,
+ ureg->output[i].semantic_name,
+ ureg->output[i].semantic_index,
+ ureg->output[i].usage_mask,
+ ureg->output[i].array_id);
+ }
+ }
+ else {
+ for (i = 0; i < ureg->nr_outputs; i++) {
+ for (j = ureg->output[i].first; j <= ureg->output[i].last; j++) {
+ emit_decl_semantic(ureg,
+ TGSI_FILE_OUTPUT,
+ j, j,
+ ureg->output[i].semantic_name,
+ ureg->output[i].semantic_index +
+ (j - ureg->output[i].first),
+ ureg->output[i].usage_mask, 0);
+ }
+ }
}
for (i = 0; i < ureg->nr_samplers; i++) {
ureg->sampler[i].Index, 1 );
}
+ for (i = 0; i < ureg->nr_sampler_views; i++) {
+ emit_decl_sampler_view(ureg,
+ ureg->sampler_view[i].index,
+ ureg->sampler_view[i].target,
+ ureg->sampler_view[i].return_type_x,
+ ureg->sampler_view[i].return_type_y,
+ ureg->sampler_view[i].return_type_z,
+ ureg->sampler_view[i].return_type_w);
+ }
+
if (ureg->const_decls.nr_constant_ranges) {
for (i = 0; i < ureg->const_decls.nr_constant_ranges; i++) {
emit_decl_range(ureg,
}
if (ureg->nr_temps) {
- emit_decl_range( ureg,
- TGSI_FILE_TEMPORARY,
- 0, ureg->nr_temps );
+ unsigned array = 0;
+ for (i = 0; i < ureg->nr_temps;) {
+ boolean local = util_bitmask_get(ureg->local_temps, i);
+ unsigned first = i;
+ i = util_bitmask_get_next_index(ureg->decl_temps, i + 1);
+ if (i == UTIL_BITMASK_INVALID_INDEX)
+ i = ureg->nr_temps;
+
+ if (array < ureg->nr_array_temps && ureg->array_temps[array] == first)
+ emit_decl_temps( ureg, first, i - 1, local, ++array );
+ else
+ emit_decl_temps( ureg, first, i - 1, local, 0 );
+ }
}
if (ureg->nr_addrs) {
void *ureg_create_shader( struct ureg_program *ureg,
- struct pipe_context *pipe )
+ struct pipe_context *pipe,
+ const struct pipe_stream_output_info *so )
{
struct pipe_shader_state state;
if(!state.tokens)
return NULL;
- if (ureg->processor == TGSI_PROCESSOR_VERTEX)
- return pipe->create_vs_state( pipe, &state );
+ if (so)
+ state.stream_output = *so;
else
- return pipe->create_fs_state( pipe, &state );
+ memset(&state.stream_output, 0, sizeof(state.stream_output));
+
+ switch (ureg->processor) {
+ case TGSI_PROCESSOR_VERTEX:
+ return pipe->create_vs_state(pipe, &state);
+ case TGSI_PROCESSOR_TESS_CTRL:
+ return pipe->create_tcs_state(pipe, &state);
+ case TGSI_PROCESSOR_TESS_EVAL:
+ return pipe->create_tes_state(pipe, &state);
+ case TGSI_PROCESSOR_GEOMETRY:
+ return pipe->create_gs_state(pipe, &state);
+ case TGSI_PROCESSOR_FRAGMENT:
+ return pipe->create_fs_state(pipe, &state);
+ default:
+ return NULL;
+ }
}
}
-struct ureg_program *ureg_create( unsigned processor )
+struct ureg_program *
+ureg_create(unsigned processor)
{
+ return ureg_create_with_screen(processor, NULL);
+}
+
+
+struct ureg_program *
+ureg_create_with_screen(unsigned processor, struct pipe_screen *screen)
+{
+ int i;
struct ureg_program *ureg = CALLOC_STRUCT( ureg_program );
- if (ureg == NULL)
- return NULL;
+ if (!ureg)
+ goto no_ureg;
ureg->processor = processor;
- ureg->property_gs_input_prim = ~0;
- ureg->property_gs_output_prim = ~0;
- ureg->property_gs_max_vertices = ~0;
+ ureg->supports_any_inout_decl_range =
+ screen &&
+ screen->get_shader_param(screen,
+ util_pipe_shader_from_tgsi_processor(processor),
+ PIPE_SHADER_CAP_TGSI_ANY_INOUT_DECL_RANGE) != 0;
+
+ for (i = 0; i < Elements(ureg->properties); i++)
+ ureg->properties[i] = ~0;
+
+ ureg->free_temps = util_bitmask_create();
+ if (ureg->free_temps == NULL)
+ goto no_free_temps;
+
+ ureg->local_temps = util_bitmask_create();
+ if (ureg->local_temps == NULL)
+ goto no_local_temps;
+
+ ureg->decl_temps = util_bitmask_create();
+ if (ureg->decl_temps == NULL)
+ goto no_decl_temps;
+
return ureg;
+
+no_decl_temps:
+ util_bitmask_destroy(ureg->local_temps);
+no_local_temps:
+ util_bitmask_destroy(ureg->free_temps);
+no_free_temps:
+ FREE(ureg);
+no_ureg:
+ return NULL;
+}
+
+
+unsigned
+ureg_get_nr_outputs( const struct ureg_program *ureg )
+{
+ if (!ureg)
+ return 0;
+ return ureg->nr_outputs;
}
ureg->domain[i].tokens != error_tokens)
FREE(ureg->domain[i].tokens);
}
-
+
+ util_bitmask_destroy(ureg->free_temps);
+ util_bitmask_destroy(ureg->local_temps);
+ util_bitmask_destroy(ureg->decl_temps);
+
FREE(ureg);
}
projects / mesa.git / blobdiff