static void *
create_yuv_shader(struct pipe_context *pipe, struct ureg_program *ureg)
{
struct ureg_src y_sampler, u_sampler, v_sampler;
struct ureg_src pos;
struct ureg_src matrow0, matrow1, matrow2;
struct ureg_dst y, u, v, rgb;
struct ureg_dst out = ureg_DECL_output(ureg,
TGSI_SEMANTIC_COLOR,
0);
pos = ureg_DECL_fs_input(ureg,
TGSI_SEMANTIC_GENERIC, 0,
TGSI_INTERPOLATE_PERSPECTIVE);
rgb = ureg_DECL_temporary(ureg);
y = ureg_DECL_temporary(ureg);
u = ureg_DECL_temporary(ureg);
v = ureg_DECL_temporary(ureg);
y_sampler = ureg_DECL_sampler(ureg, 0);
u_sampler = ureg_DECL_sampler(ureg, 1);
v_sampler = ureg_DECL_sampler(ureg, 2);
matrow0 = ureg_DECL_constant(ureg, 0);
matrow1 = ureg_DECL_constant(ureg, 1);
matrow2 = ureg_DECL_constant(ureg, 2);
ureg_TEX(ureg, y, TGSI_TEXTURE_2D, pos, y_sampler);
ureg_TEX(ureg, u, TGSI_TEXTURE_2D, pos, u_sampler);
ureg_TEX(ureg, v, TGSI_TEXTURE_2D, pos, v_sampler);
ureg_SUB(ureg, u, ureg_src(u), ureg_scalar(matrow0, TGSI_SWIZZLE_W));
ureg_SUB(ureg, v, ureg_src(v), ureg_scalar(matrow0, TGSI_SWIZZLE_W));
ureg_MUL(ureg, rgb, ureg_scalar(ureg_src(y), TGSI_SWIZZLE_X), matrow0);
ureg_MAD(ureg, rgb,
ureg_scalar(ureg_src(u), TGSI_SWIZZLE_X), matrow1, ureg_src(rgb));
ureg_MAD(ureg, rgb,
ureg_scalar(ureg_src(v), TGSI_SWIZZLE_X), matrow2, ureg_src(rgb));
/* rgb.a = 1; */
ureg_MOV(ureg, ureg_writemask(rgb, TGSI_WRITEMASK_W),
ureg_scalar(matrow0, TGSI_SWIZZLE_X));
ureg_MOV(ureg, out, ureg_src(rgb));
ureg_release_temporary(ureg, rgb);
ureg_release_temporary(ureg, y);
ureg_release_temporary(ureg, u);
ureg_release_temporary(ureg, v);
ureg_END(ureg);
return ureg_create_shader_and_destroy(ureg, pipe);
}
static void *
create_vs(struct pipe_context *pipe, unsigned vs_traits)
{
struct ureg_program *ureg;
struct ureg_src src;
struct ureg_dst dst;
struct ureg_src const0, const1;
boolean is_fill = (vs_traits & VS_FILL) != 0;
boolean is_composite = (vs_traits & VS_COMPOSITE) != 0;
boolean has_mask = (vs_traits & VS_MASK) != 0;
boolean is_yuv = (vs_traits & VS_YUV) != 0;
unsigned input_slot = 0;
ureg = ureg_create(TGSI_PROCESSOR_VERTEX);
if (ureg == NULL)
return 0;
const0 = ureg_DECL_constant(ureg, 0);
const1 = ureg_DECL_constant(ureg, 1);
/* it has to be either a fill or a composite op */
debug_assert((is_fill ^ is_composite) ^ is_yuv);
src = ureg_DECL_vs_input(ureg, input_slot++);
dst = ureg_DECL_output(ureg, TGSI_SEMANTIC_POSITION, 0);
src = vs_normalize_coords(ureg, src, const0, const1);
ureg_MOV(ureg, dst, src);
if (is_yuv) {
src = ureg_DECL_vs_input(ureg, input_slot++);
dst = ureg_DECL_output(ureg, TGSI_SEMANTIC_GENERIC, 0);
ureg_MOV(ureg, dst, src);
}
if (is_composite) {
src = ureg_DECL_vs_input(ureg, input_slot++);
dst = ureg_DECL_output(ureg, TGSI_SEMANTIC_GENERIC, 0);
ureg_MOV(ureg, dst, src);
}
if (is_fill) {
src = ureg_DECL_vs_input(ureg, input_slot++);
dst = ureg_DECL_output(ureg, TGSI_SEMANTIC_COLOR, 0);
ureg_MOV(ureg, dst, src);
}
if (has_mask) {
src = ureg_DECL_vs_input(ureg, input_slot++);
dst = ureg_DECL_output(ureg, TGSI_SEMANTIC_GENERIC, 1);
ureg_MOV(ureg, dst, src);
}
ureg_END(ureg);
return ureg_create_shader_and_destroy(ureg, pipe);
}
/**
* Create a simple vertex shader that passes through position and the given
* attribute.
*/
static void *create_passthrough_vs(struct pipe_context *pipe, int semantic_name)
{
struct ureg_program *ureg;
struct ureg_src src[2], constants[3];
struct ureg_dst dst[2], tmp;
int i;
ureg = ureg_create(TGSI_PROCESSOR_VERTEX);
if (!ureg)
return NULL;
/* position is in user coordinates */
src[0] = ureg_DECL_vs_input(ureg, 0);
dst[0] = ureg_DECL_output(ureg, TGSI_SEMANTIC_POSITION, 0);
tmp = ureg_DECL_temporary(ureg);
for (i = 0; i < Elements(constants); i++)
constants[i] = ureg_DECL_constant(ureg, i);
/* transform to clipped coordinates */
ureg_DP4(ureg, ureg_writemask(tmp, TGSI_WRITEMASK_X), src[0], constants[0]);
ureg_DP4(ureg, ureg_writemask(tmp, TGSI_WRITEMASK_Y), src[0], constants[1]);
ureg_MOV(ureg, ureg_writemask(tmp, TGSI_WRITEMASK_Z), src[0]);
ureg_DP4(ureg, ureg_writemask(tmp, TGSI_WRITEMASK_W), src[0], constants[2]);
ureg_MOV(ureg, dst[0], ureg_src(tmp));
if (semantic_name >= 0) {
src[1] = ureg_DECL_vs_input(ureg, 1);
dst[1] = ureg_DECL_output(ureg, semantic_name, 0);
ureg_MOV(ureg, dst[1], src[1]);
}
ureg_END(ureg);
return ureg_create_shader_and_destroy(ureg, pipe);
}
/**
* Map a Mesa src register to a TGSI ureg_src register.
*/
static struct ureg_src
src_register( struct st_translate *t,
gl_register_file file,
GLint index )
{
switch( file ) {
case PROGRAM_UNDEFINED:
return ureg_src_undef();
case PROGRAM_TEMPORARY:
assert(index >= 0);
assert(index < Elements(t->temps));
if (ureg_dst_is_undef(t->temps[index]))
t->temps[index] = ureg_DECL_temporary( t->ureg );
return ureg_src(t->temps[index]);
case PROGRAM_ENV_PARAM:
case PROGRAM_LOCAL_PARAM:
case PROGRAM_UNIFORM:
assert(index >= 0);
return t->constants[index];
case PROGRAM_STATE_VAR:
case PROGRAM_CONSTANT: /* ie, immediate */
if (index < 0)
return ureg_DECL_constant( t->ureg, 0 );
else
return t->constants[index];
case PROGRAM_INPUT:
assert(t->inputMapping[index] < Elements(t->inputs));
return t->inputs[t->inputMapping[index]];
case PROGRAM_OUTPUT:
assert(t->outputMapping[index] < Elements(t->outputs));
return ureg_src(t->outputs[t->outputMapping[index]]); /* not needed? */
case PROGRAM_ADDRESS:
return ureg_src(t->address[index]);
case PROGRAM_SYSTEM_VALUE:
assert(index < Elements(t->systemValues));
return t->systemValues[index];
default:
debug_assert( 0 );
return ureg_src_undef();
}
}
static void *
create_frag_shader_palette(struct vl_compositor *c, bool include_cc)
{
struct ureg_program *shader;
struct ureg_src csc[3];
struct ureg_src tc;
struct ureg_src sampler;
struct ureg_src palette;
struct ureg_dst texel;
struct ureg_dst fragment;
unsigned i;
shader = ureg_create(TGSI_PROCESSOR_FRAGMENT);
if (!shader)
return false;
for (i = 0; include_cc && i < 3; ++i)
csc[i] = ureg_DECL_constant(shader, i);
tc = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_GENERIC, VS_O_VTEX, TGSI_INTERPOLATE_LINEAR);
sampler = ureg_DECL_sampler(shader, 0);
palette = ureg_DECL_sampler(shader, 1);
texel = ureg_DECL_temporary(shader);
fragment = ureg_DECL_output(shader, TGSI_SEMANTIC_COLOR, 0);
/*
* texel = tex(tc, sampler)
* fragment.xyz = tex(texel, palette) * csc
* fragment.a = texel.a
*/
ureg_TEX(shader, texel, TGSI_TEXTURE_2D, tc, sampler);
ureg_MOV(shader, ureg_writemask(fragment, TGSI_WRITEMASK_W), ureg_src(texel));
if (include_cc) {
ureg_TEX(shader, texel, TGSI_TEXTURE_1D, ureg_src(texel), palette);
for (i = 0; i < 3; ++i)
ureg_DP4(shader, ureg_writemask(fragment, TGSI_WRITEMASK_X << i), csc[i], ureg_src(texel));
} else {
ureg_TEX(shader, ureg_writemask(fragment, TGSI_WRITEMASK_XYZ),
TGSI_TEXTURE_1D, ureg_src(texel), palette);
}
ureg_release_temporary(shader, texel);
ureg_END(shader);
return ureg_create_shader_and_destroy(shader, c->pipe);
}
static void *
create_frag_shader_video_buffer(struct vl_compositor *c)
{
struct ureg_program *shader;
struct ureg_src tc;
struct ureg_src csc[3];
struct ureg_src sampler[3];
struct ureg_dst texel;
struct ureg_dst fragment;
unsigned i;
shader = ureg_create(TGSI_PROCESSOR_FRAGMENT);
if (!shader)
return false;
tc = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_GENERIC, VS_O_VTEX, TGSI_INTERPOLATE_LINEAR);
for (i = 0; i < 3; ++i) {
csc[i] = ureg_DECL_constant(shader, i);
sampler[i] = ureg_DECL_sampler(shader, i);
}
texel = ureg_DECL_temporary(shader);
fragment = ureg_DECL_output(shader, TGSI_SEMANTIC_COLOR, 0);
/*
* texel.xyz = tex(tc, sampler[i])
* fragment = csc * texel
*/
for (i = 0; i < 3; ++i)
ureg_TEX(shader, ureg_writemask(texel, TGSI_WRITEMASK_X << i), TGSI_TEXTURE_3D, tc, sampler[i]);
ureg_MOV(shader, ureg_writemask(texel, TGSI_WRITEMASK_W), ureg_imm1f(shader, 1.0f));
for (i = 0; i < 3; ++i)
ureg_DP4(shader, ureg_writemask(fragment, TGSI_WRITEMASK_X << i), csc[i], ureg_src(texel));
ureg_MOV(shader, ureg_writemask(fragment, TGSI_WRITEMASK_W), ureg_imm1f(shader, 1.0f));
ureg_release_temporary(shader, texel);
ureg_END(shader);
return ureg_create_shader_and_destroy(shader, c->pipe);
}
/**
* Emit the TGSI instructions for inverting and adjusting WPOS.
* This code is unavoidable because it also depends on whether
* a FBO is bound (STATE_FB_WPOS_Y_TRANSFORM).
*/
static void
emit_wpos_adjustment( struct st_translate *t,
const struct gl_program *program,
boolean invert,
GLfloat adjX, GLfloat adjY[2])
{
struct ureg_program *ureg = t->ureg;
/* Fragment program uses fragment position input.
* Need to replace instances of INPUT[WPOS] with temp T
* where T = INPUT[WPOS] by y is inverted.
*/
static const gl_state_index wposTransformState[STATE_LENGTH]
= { STATE_INTERNAL, STATE_FB_WPOS_Y_TRANSFORM, 0, 0, 0 };
/* XXX: note we are modifying the incoming shader here! Need to
* do this before emitting the constant decls below, or this
* will be missed:
*/
unsigned wposTransConst = _mesa_add_state_reference(program->Parameters,
wposTransformState);
struct ureg_src wpostrans = ureg_DECL_constant( ureg, wposTransConst );
struct ureg_dst wpos_temp = ureg_DECL_temporary( ureg );
struct ureg_src wpos_input = t->inputs[t->inputMapping[VARYING_SLOT_POS]];
/* First, apply the coordinate shift: */
if (adjX || adjY[0] || adjY[1]) {
if (adjY[0] != adjY[1]) {
/* Adjust the y coordinate by adjY[1] or adjY[0] respectively
* depending on whether inversion is actually going to be applied
* or not, which is determined by testing against the inversion
* state variable used below, which will be either +1 or -1.
*/
struct ureg_dst adj_temp = ureg_DECL_temporary(ureg);
ureg_CMP(ureg, adj_temp,
ureg_scalar(wpostrans, invert ? 2 : 0),
ureg_imm4f(ureg, adjX, adjY[0], 0.0f, 0.0f),
ureg_imm4f(ureg, adjX, adjY[1], 0.0f, 0.0f));
ureg_ADD(ureg, wpos_temp, wpos_input, ureg_src(adj_temp));
} else {
ureg_ADD(ureg, wpos_temp, wpos_input,
ureg_imm4f(ureg, adjX, adjY[0], 0.0f, 0.0f));
}
wpos_input = ureg_src(wpos_temp);
} else {
/* MOV wpos_temp, input[wpos]
*/
ureg_MOV( ureg, wpos_temp, wpos_input );
}
/* Now the conditional y flip: STATE_FB_WPOS_Y_TRANSFORM.xy/zw will be
* inversion/identity, or the other way around if we're drawing to an FBO.
*/
if (invert) {
/* MAD wpos_temp.y, wpos_input, wpostrans.xxxx, wpostrans.yyyy
*/
ureg_MAD( ureg,
ureg_writemask(wpos_temp, TGSI_WRITEMASK_Y ),
wpos_input,
ureg_scalar(wpostrans, 0),
ureg_scalar(wpostrans, 1));
} else {
/* MAD wpos_temp.y, wpos_input, wpostrans.zzzz, wpostrans.wwww
*/
ureg_MAD( ureg,
ureg_writemask(wpos_temp, TGSI_WRITEMASK_Y ),
wpos_input,
ureg_scalar(wpostrans, 2),
ureg_scalar(wpostrans, 3));
}
/* Use wpos_temp as position input from here on:
*/
t->inputs[t->inputMapping[VARYING_SLOT_POS]] = ureg_src(wpos_temp);
}
/**
* Translate Mesa program to TGSI format.
* \param program the program to translate
* \param numInputs number of input registers used
* \param inputMapping maps Mesa fragment program inputs to TGSI generic
* input indexes
* \param inputSemanticName the TGSI_SEMANTIC flag for each input
* \param inputSemanticIndex the semantic index (ex: which texcoord) for
* each input
* \param interpMode the TGSI_INTERPOLATE_LINEAR/PERSP mode for each input
* \param numOutputs number of output registers used
* \param outputMapping maps Mesa fragment program outputs to TGSI
* generic outputs
* \param outputSemanticName the TGSI_SEMANTIC flag for each output
* \param outputSemanticIndex the semantic index (ex: which texcoord) for
* each output
*
* \return PIPE_OK or PIPE_ERROR_OUT_OF_MEMORY
*/
enum pipe_error
st_translate_mesa_program(
struct gl_context *ctx,
uint procType,
struct ureg_program *ureg,
const struct gl_program *program,
GLuint numInputs,
const GLuint inputMapping[],
const ubyte inputSemanticName[],
const ubyte inputSemanticIndex[],
const GLuint interpMode[],
GLuint numOutputs,
const GLuint outputMapping[],
const ubyte outputSemanticName[],
const ubyte outputSemanticIndex[],
boolean passthrough_edgeflags,
boolean clamp_color)
{
struct st_translate translate, *t;
unsigned i;
enum pipe_error ret = PIPE_OK;
assert(numInputs <= ARRAY_SIZE(t->inputs));
assert(numOutputs <= ARRAY_SIZE(t->outputs));
t = &translate;
memset(t, 0, sizeof *t);
t->procType = procType;
t->inputMapping = inputMapping;
t->outputMapping = outputMapping;
t->ureg = ureg;
/*_mesa_print_program(program);*/
/*
* Declare input attributes.
*/
if (procType == TGSI_PROCESSOR_FRAGMENT) {
for (i = 0; i < numInputs; i++) {
t->inputs[i] = ureg_DECL_fs_input(ureg,
inputSemanticName[i],
inputSemanticIndex[i],
interpMode[i]);
}
if (program->InputsRead & VARYING_BIT_POS) {
/* Must do this after setting up t->inputs, and before
* emitting constant references, below:
*/
emit_wpos(st_context(ctx), t, program, ureg);
}
if (program->InputsRead & VARYING_BIT_FACE) {
emit_face_var( t, program );
}
/*
* Declare output attributes.
*/
for (i = 0; i < numOutputs; i++) {
switch (outputSemanticName[i]) {
case TGSI_SEMANTIC_POSITION:
t->outputs[i] = ureg_DECL_output( ureg,
TGSI_SEMANTIC_POSITION, /* Z / Depth */
outputSemanticIndex[i] );
t->outputs[i] = ureg_writemask( t->outputs[i],
TGSI_WRITEMASK_Z );
break;
case TGSI_SEMANTIC_STENCIL:
t->outputs[i] = ureg_DECL_output( ureg,
TGSI_SEMANTIC_STENCIL, /* Stencil */
outputSemanticIndex[i] );
t->outputs[i] = ureg_writemask( t->outputs[i],
TGSI_WRITEMASK_Y );
break;
case TGSI_SEMANTIC_COLOR:
t->outputs[i] = ureg_DECL_output( ureg,
TGSI_SEMANTIC_COLOR,
outputSemanticIndex[i] );
break;
default:
debug_assert(0);
return 0;
}
}
}
else if (procType == TGSI_PROCESSOR_GEOMETRY) {
for (i = 0; i < numInputs; i++) {
t->inputs[i] = ureg_DECL_input(ureg,
inputSemanticName[i],
inputSemanticIndex[i], 0, 1);
}
for (i = 0; i < numOutputs; i++) {
t->outputs[i] = ureg_DECL_output( ureg,
outputSemanticName[i],
outputSemanticIndex[i] );
}
}
else {
assert(procType == TGSI_PROCESSOR_VERTEX);
for (i = 0; i < numInputs; i++) {
t->inputs[i] = ureg_DECL_vs_input(ureg, i);
}
for (i = 0; i < numOutputs; i++) {
t->outputs[i] = ureg_DECL_output( ureg,
outputSemanticName[i],
outputSemanticIndex[i] );
if (outputSemanticName[i] == TGSI_SEMANTIC_FOG) {
/* force register to contain a fog coordinate in the form (F, 0, 0, 1). */
ureg_MOV(ureg,
ureg_writemask(t->outputs[i], TGSI_WRITEMASK_YZW),
ureg_imm4f(ureg, 0.0f, 0.0f, 0.0f, 1.0f));
t->outputs[i] = ureg_writemask(t->outputs[i], TGSI_WRITEMASK_X);
}
}
if (passthrough_edgeflags)
emit_edgeflags( t, program );
}
/* Declare address register.
*/
if (program->NumAddressRegs > 0) {
debug_assert( program->NumAddressRegs == 1 );
t->address[0] = ureg_DECL_address( ureg );
}
/* Declare misc input registers
*/
{
GLbitfield sysInputs = program->SystemValuesRead;
unsigned numSys = 0;
for (i = 0; sysInputs; i++) {
if (sysInputs & (1 << i)) {
unsigned semName = _mesa_sysval_to_semantic[i];
t->systemValues[i] = ureg_DECL_system_value(ureg, numSys, semName, 0);
if (semName == TGSI_SEMANTIC_INSTANCEID ||
semName == TGSI_SEMANTIC_VERTEXID) {
/* From Gallium perspective, these system values are always
* integer, and require native integer support. However, if
* native integer is supported on the vertex stage but not the
* pixel stage (e.g, i915g + draw), Mesa will generate IR that
* assumes these system values are floats. To resolve the
* inconsistency, we insert a U2F.
*/
struct st_context *st = st_context(ctx);
struct pipe_screen *pscreen = st->pipe->screen;
assert(procType == TGSI_PROCESSOR_VERTEX);
assert(pscreen->get_shader_param(pscreen, PIPE_SHADER_VERTEX, PIPE_SHADER_CAP_INTEGERS));
(void) pscreen; /* silence non-debug build warnings */
if (!ctx->Const.NativeIntegers) {
struct ureg_dst temp = ureg_DECL_local_temporary(t->ureg);
ureg_U2F( t->ureg, ureg_writemask(temp, TGSI_WRITEMASK_X), t->systemValues[i]);
t->systemValues[i] = ureg_scalar(ureg_src(temp), 0);
}
}
numSys++;
sysInputs &= ~(1 << i);
}
}
}
if (program->IndirectRegisterFiles & (1 << PROGRAM_TEMPORARY)) {
/* If temps are accessed with indirect addressing, declare temporaries
* in sequential order. Else, we declare them on demand elsewhere.
*/
for (i = 0; i < program->NumTemporaries; i++) {
/* XXX use TGSI_FILE_TEMPORARY_ARRAY when it's supported by ureg */
t->temps[i] = ureg_DECL_temporary( t->ureg );
}
}
/* Emit constants and immediates. Mesa uses a single index space
* for these, so we put all the translated regs in t->constants.
*/
if (program->Parameters) {
t->constants = calloc( program->Parameters->NumParameters,
sizeof t->constants[0] );
if (t->constants == NULL) {
ret = PIPE_ERROR_OUT_OF_MEMORY;
goto out;
}
for (i = 0; i < program->Parameters->NumParameters; i++) {
switch (program->Parameters->Parameters[i].Type) {
case PROGRAM_STATE_VAR:
case PROGRAM_UNIFORM:
t->constants[i] = ureg_DECL_constant( ureg, i );
break;
/* Emit immediates only when there's no indirect addressing of
* the const buffer.
* FIXME: Be smarter and recognize param arrays:
* indirect addressing is only valid within the referenced
* array.
*/
case PROGRAM_CONSTANT:
if (program->IndirectRegisterFiles & PROGRAM_ANY_CONST)
t->constants[i] = ureg_DECL_constant( ureg, i );
else
t->constants[i] =
ureg_DECL_immediate( ureg,
(const float*) program->Parameters->ParameterValues[i],
4 );
break;
default:
break;
}
}
}
/* texture samplers */
for (i = 0; i < ctx->Const.Program[MESA_SHADER_FRAGMENT].MaxTextureImageUnits; i++) {
if (program->SamplersUsed & (1 << i)) {
t->samplers[i] = ureg_DECL_sampler( ureg, i );
}
}
/* Emit each instruction in turn:
*/
for (i = 0; i < program->NumInstructions; i++) {
set_insn_start( t, ureg_get_instruction_number( ureg ));
compile_instruction( ctx, t, &program->Instructions[i], clamp_color );
}
/* Fix up all emitted labels:
*/
for (i = 0; i < t->labels_count; i++) {
ureg_fixup_label( ureg,
t->labels[i].token,
t->insn[t->labels[i].branch_target] );
}
out:
free(t->insn);
free(t->labels);
free(t->constants);
if (t->error) {
debug_printf("%s: translate error flag set\n", __func__);
}
return ret;
}
static void *
create_fs(struct pipe_context *pipe, unsigned fs_traits)
{
struct ureg_program *ureg;
struct ureg_src /*dst_sampler, */ src_sampler, mask_sampler;
struct ureg_src /*dst_pos, */ src_input, mask_pos;
struct ureg_dst src, mask;
struct ureg_dst out;
struct ureg_src imm0 = { 0 };
unsigned has_mask = (fs_traits & FS_MASK) != 0;
unsigned is_fill = (fs_traits & FS_FILL) != 0;
unsigned is_composite = (fs_traits & FS_COMPOSITE) != 0;
unsigned is_solid = (fs_traits & FS_SOLID_FILL) != 0;
unsigned is_lingrad = (fs_traits & FS_LINGRAD_FILL) != 0;
unsigned is_radgrad = (fs_traits & FS_RADGRAD_FILL) != 0;
unsigned comp_alpha_mask = fs_traits & FS_COMPONENT_ALPHA;
unsigned is_yuv = (fs_traits & FS_YUV) != 0;
unsigned src_repeat_none = (fs_traits & FS_SRC_REPEAT_NONE) != 0;
unsigned mask_repeat_none = (fs_traits & FS_MASK_REPEAT_NONE) != 0;
unsigned src_swizzle = (fs_traits & FS_SRC_SWIZZLE_RGB) != 0;
unsigned mask_swizzle = (fs_traits & FS_MASK_SWIZZLE_RGB) != 0;
unsigned src_set_alpha = (fs_traits & FS_SRC_SET_ALPHA) != 0;
unsigned mask_set_alpha = (fs_traits & FS_MASK_SET_ALPHA) != 0;
unsigned src_luminance = (fs_traits & FS_SRC_LUMINANCE) != 0;
unsigned mask_luminance = (fs_traits & FS_MASK_LUMINANCE) != 0;
unsigned dst_luminance = (fs_traits & FS_DST_LUMINANCE) != 0;
#if 0
print_fs_traits(fs_traits);
#else
(void)print_fs_traits;
#endif
ureg = ureg_create(TGSI_PROCESSOR_FRAGMENT);
if (ureg == NULL)
return 0;
/* it has to be either a fill, a composite op or a yuv conversion */
debug_assert((is_fill ^ is_composite) ^ is_yuv);
(void)is_yuv;
out = ureg_DECL_output(ureg, TGSI_SEMANTIC_COLOR, 0);
if (src_repeat_none || mask_repeat_none ||
src_set_alpha || mask_set_alpha || src_luminance) {
imm0 = ureg_imm4f(ureg, 0, 0, 0, 1);
}
if (is_composite) {
src_sampler = ureg_DECL_sampler(ureg, 0);
src_input = ureg_DECL_fs_input(ureg,
TGSI_SEMANTIC_GENERIC, 0,
TGSI_INTERPOLATE_PERSPECTIVE);
} else if (is_fill) {
if (is_solid)
src_input = ureg_DECL_fs_input(ureg,
TGSI_SEMANTIC_COLOR, 0,
TGSI_INTERPOLATE_PERSPECTIVE);
else
src_input = ureg_DECL_fs_input(ureg,
TGSI_SEMANTIC_POSITION, 0,
TGSI_INTERPOLATE_PERSPECTIVE);
} else {
debug_assert(is_yuv);
return create_yuv_shader(pipe, ureg);
}
if (has_mask) {
mask_sampler = ureg_DECL_sampler(ureg, 1);
mask_pos = ureg_DECL_fs_input(ureg,
TGSI_SEMANTIC_GENERIC, 1,
TGSI_INTERPOLATE_PERSPECTIVE);
}
#if 0 /* unused right now */
dst_sampler = ureg_DECL_sampler(ureg, 2);
dst_pos = ureg_DECL_fs_input(ureg,
TGSI_SEMANTIC_POSITION, 2,
TGSI_INTERPOLATE_PERSPECTIVE);
#endif
if (is_composite) {
if (has_mask || src_luminance || dst_luminance)
src = ureg_DECL_temporary(ureg);
else
src = out;
xrender_tex(ureg, src, src_input, src_sampler, imm0,
src_repeat_none, src_swizzle, src_set_alpha);
} else if (is_fill) {
if (is_solid) {
if (has_mask || src_luminance || dst_luminance)
src = ureg_dst(src_input);
else
ureg_MOV(ureg, out, src_input);
} else if (is_lingrad || is_radgrad) {
struct ureg_src coords, const0124, matrow0, matrow1, matrow2;
if (has_mask || src_luminance || dst_luminance)
src = ureg_DECL_temporary(ureg);
else
src = out;
coords = ureg_DECL_constant(ureg, 0);
const0124 = ureg_DECL_constant(ureg, 1);
matrow0 = ureg_DECL_constant(ureg, 2);
matrow1 = ureg_DECL_constant(ureg, 3);
matrow2 = ureg_DECL_constant(ureg, 4);
if (is_lingrad) {
linear_gradient(ureg, src,
src_input, src_sampler,
coords, const0124, matrow0, matrow1, matrow2);
} else if (is_radgrad) {
radial_gradient(ureg, src,
src_input, src_sampler,
coords, const0124, matrow0, matrow1, matrow2);
}
} else
debug_assert(!"Unknown fill type!");
}
if (src_luminance) {
ureg_MOV(ureg, src, ureg_scalar(ureg_src(src), TGSI_SWIZZLE_X));
ureg_MOV(ureg, ureg_writemask(src, TGSI_WRITEMASK_XYZ),
ureg_scalar(imm0, TGSI_SWIZZLE_X));
if (!has_mask && !dst_luminance)
ureg_MOV(ureg, out, ureg_src(src));
}
if (has_mask) {
mask = ureg_DECL_temporary(ureg);
xrender_tex(ureg, mask, mask_pos, mask_sampler, imm0,
mask_repeat_none, mask_swizzle, mask_set_alpha);
/* src IN mask */
src_in_mask(ureg, (dst_luminance) ? src : out, ureg_src(src),
ureg_src(mask),
comp_alpha_mask, mask_luminance);
ureg_release_temporary(ureg, mask);
}
if (dst_luminance) {
/*
* Make sure the alpha channel goes into the output L8 surface.
*/
ureg_MOV(ureg, out, ureg_scalar(ureg_src(src), TGSI_SWIZZLE_W));
}
ureg_END(ureg);
return ureg_create_shader_and_destroy(ureg, pipe);
}
static void *
create_frag_shader_weave(struct vl_compositor *c)
{
struct ureg_program *shader;
struct ureg_src i_tc[2];
struct ureg_src csc[3];
struct ureg_src sampler[3];
struct ureg_dst t_tc[2];
struct ureg_dst t_texel[2];
struct ureg_dst o_fragment;
unsigned i, j;
shader = ureg_create(TGSI_PROCESSOR_FRAGMENT);
if (!shader)
return false;
i_tc[0] = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_GENERIC, VS_O_VTOP, TGSI_INTERPOLATE_LINEAR);
i_tc[1] = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_GENERIC, VS_O_VBOTTOM, TGSI_INTERPOLATE_LINEAR);
for (i = 0; i < 3; ++i) {
csc[i] = ureg_DECL_constant(shader, i);
sampler[i] = ureg_DECL_sampler(shader, i);
}
for (i = 0; i < 2; ++i) {
t_tc[i] = ureg_DECL_temporary(shader);
t_texel[i] = ureg_DECL_temporary(shader);
}
o_fragment = ureg_DECL_output(shader, TGSI_SEMANTIC_COLOR, 0);
/* calculate the texture offsets
* t_tc.x = i_tc.x
* t_tc.y = (round(i_tc.y) + 0.5) / height * 2
*/
for (i = 0; i < 2; ++i) {
ureg_MOV(shader, ureg_writemask(t_tc[i], TGSI_WRITEMASK_X), i_tc[i]);
ureg_ROUND(shader, ureg_writemask(t_tc[i], TGSI_WRITEMASK_YZ), i_tc[i]);
ureg_MOV(shader, ureg_writemask(t_tc[i], TGSI_WRITEMASK_W),
ureg_imm1f(shader, i ? 0.75f : 0.25f));
ureg_ADD(shader, ureg_writemask(t_tc[i], TGSI_WRITEMASK_YZ),
ureg_src(t_tc[i]), ureg_imm1f(shader, 0.5f));
ureg_MUL(shader, ureg_writemask(t_tc[i], TGSI_WRITEMASK_Y),
ureg_src(t_tc[i]), ureg_scalar(i_tc[0], TGSI_SWIZZLE_W));
ureg_MUL(shader, ureg_writemask(t_tc[i], TGSI_WRITEMASK_Z),
ureg_src(t_tc[i]), ureg_scalar(i_tc[1], TGSI_SWIZZLE_W));
}
/* fetch the texels
* texel[0..1].x = tex(t_tc[0..1][0])
* texel[0..1].y = tex(t_tc[0..1][1])
* texel[0..1].z = tex(t_tc[0..1][2])
*/
for (i = 0; i < 2; ++i)
for (j = 0; j < 3; ++j) {
struct ureg_src src = ureg_swizzle(ureg_src(t_tc[i]),
TGSI_SWIZZLE_X, j ? TGSI_SWIZZLE_Z : TGSI_SWIZZLE_Y, TGSI_SWIZZLE_W, TGSI_SWIZZLE_W);
ureg_TEX(shader, ureg_writemask(t_texel[i], TGSI_WRITEMASK_X << j),
TGSI_TEXTURE_3D, src, sampler[j]);
}
/* calculate linear interpolation factor
* factor = |round(i_tc.y) - i_tc.y| * 2
*/
ureg_ROUND(shader, ureg_writemask(t_tc[0], TGSI_WRITEMASK_YZ), i_tc[0]);
ureg_ADD(shader, ureg_writemask(t_tc[0], TGSI_WRITEMASK_YZ),
ureg_src(t_tc[0]), ureg_negate(i_tc[0]));
ureg_MUL(shader, ureg_writemask(t_tc[0], TGSI_WRITEMASK_XY),
ureg_abs(ureg_src(t_tc[0])), ureg_imm1f(shader, 2.0f));
ureg_LRP(shader, t_texel[0], ureg_swizzle(ureg_src(t_tc[0]),
TGSI_SWIZZLE_Y, TGSI_SWIZZLE_Z, TGSI_SWIZZLE_Z, TGSI_SWIZZLE_Z),
ureg_src(t_texel[1]), ureg_src(t_texel[0]));
/* and finally do colour space transformation
* fragment = csc * texel
*/
ureg_MOV(shader, ureg_writemask(t_texel[0], TGSI_WRITEMASK_W), ureg_imm1f(shader, 1.0f));
for (i = 0; i < 3; ++i)
ureg_DP4(shader, ureg_writemask(o_fragment, TGSI_WRITEMASK_X << i), csc[i], ureg_src(t_texel[0]));
ureg_MOV(shader, ureg_writemask(o_fragment, TGSI_WRITEMASK_W), ureg_imm1f(shader, 1.0f));
for (i = 0; i < 2; ++i) {
ureg_release_temporary(shader, t_texel[i]);
ureg_release_temporary(shader, t_tc[i]);
}
ureg_END(shader);
return ureg_create_shader_and_destroy(shader, c->pipe);
}
/**
* Called when a new variant is needed, we need to translate
* the ATI fragment shader to TGSI
*/
enum pipe_error
st_translate_atifs_program(
struct ureg_program *ureg,
struct ati_fragment_shader *atifs,
struct gl_program *program,
GLuint numInputs,
const GLuint inputMapping[],
const ubyte inputSemanticName[],
const ubyte inputSemanticIndex[],
const GLuint interpMode[],
GLuint numOutputs,
const GLuint outputMapping[],
const ubyte outputSemanticName[],
const ubyte outputSemanticIndex[])
{
enum pipe_error ret = PIPE_OK;
unsigned pass, i, r;
struct st_translate translate, *t;
t = &translate;
memset(t, 0, sizeof *t);
t->inputMapping = inputMapping;
t->outputMapping = outputMapping;
t->ureg = ureg;
t->atifs = atifs;
/*
* Declare input attributes.
*/
for (i = 0; i < numInputs; i++) {
t->inputs[i] = ureg_DECL_fs_input(ureg,
inputSemanticName[i],
inputSemanticIndex[i],
interpMode[i]);
}
/*
* Declare output attributes:
* we always have numOutputs=1 and it's FRAG_RESULT_COLOR
*/
t->outputs[0] = ureg_DECL_output(ureg,
TGSI_SEMANTIC_COLOR,
outputSemanticIndex[0]);
/* Emit constants and immediates. Mesa uses a single index space
* for these, so we put all the translated regs in t->constants.
*/
if (program->Parameters) {
t->constants = calloc(program->Parameters->NumParameters,
sizeof t->constants[0]);
if (t->constants == NULL) {
ret = PIPE_ERROR_OUT_OF_MEMORY;
goto out;
}
for (i = 0; i < program->Parameters->NumParameters; i++) {
switch (program->Parameters->Parameters[i].Type) {
case PROGRAM_STATE_VAR:
case PROGRAM_UNIFORM:
t->constants[i] = ureg_DECL_constant(ureg, i);
break;
case PROGRAM_CONSTANT:
t->constants[i] =
ureg_DECL_immediate(ureg,
(const float*)program->Parameters->ParameterValues[i],
4);
break;
default:
break;
}
}
}
/* texture samplers */
for (i = 0; i < MAX_NUM_FRAGMENT_REGISTERS_ATI; i++) {
if (program->SamplersUsed & (1 << i)) {
t->samplers[i] = ureg_DECL_sampler(ureg, i);
/* the texture target is still unknown, it will be fixed in the draw call */
ureg_DECL_sampler_view(ureg, i, TGSI_TEXTURE_2D,
TGSI_RETURN_TYPE_FLOAT,
TGSI_RETURN_TYPE_FLOAT,
TGSI_RETURN_TYPE_FLOAT,
TGSI_RETURN_TYPE_FLOAT);
}
}
/* emit instructions */
for (pass = 0; pass < atifs->NumPasses; pass++) {
t->current_pass = pass;
for (r = 0; r < MAX_NUM_FRAGMENT_REGISTERS_ATI; r++) {
struct atifs_setupinst *texinst = &atifs->SetupInst[pass][r];
compile_setupinst(t, r, texinst);
}
for (i = 0; i < atifs->numArithInstr[pass]; i++) {
struct atifs_instruction *inst = &atifs->Instructions[pass][i];
compile_instruction(t, inst);
}
}
finalize_shader(t, atifs->NumPasses);
out:
free(t->constants);
if (t->error) {
debug_printf("%s: translate error flag set\n", __func__);
}
return ret;
}
static void *
combine_shaders(const struct shader_asm_info *shaders[SHADER_STAGES], int num_shaders,
struct pipe_context *pipe,
struct pipe_shader_state *shader)
{
VGboolean declare_input = VG_FALSE;
VGint start_const = -1, end_const = 0;
VGint start_temp = -1, end_temp = 0;
VGint start_sampler = -1, end_sampler = 0;
VGint i, current_shader = 0;
VGint num_consts, num_temps, num_samplers;
struct ureg_program *ureg;
struct ureg_src in[2];
struct ureg_src *sampler = NULL;
struct ureg_src *constant = NULL;
struct ureg_dst out, *temp = NULL;
void *p = NULL;
for (i = 0; i < num_shaders; ++i) {
if (shaders[i]->num_consts)
start_const = range_min(start_const, shaders[i]->start_const);
if (shaders[i]->num_temps)
start_temp = range_min(start_temp, shaders[i]->start_temp);
if (shaders[i]->num_samplers)
start_sampler = range_min(start_sampler, shaders[i]->start_sampler);
end_const = range_max(end_const, shaders[i]->start_const +
shaders[i]->num_consts);
end_temp = range_max(end_temp, shaders[i]->start_temp +
shaders[i]->num_temps);
end_sampler = range_max(end_sampler, shaders[i]->start_sampler +
shaders[i]->num_samplers);
if (shaders[i]->needs_position)
declare_input = VG_TRUE;
}
/* if they're still unitialized, initialize them */
if (start_const < 0)
start_const = 0;
if (start_temp < 0)
start_temp = 0;
if (start_sampler < 0)
start_sampler = 0;
num_consts = end_const - start_const;
num_temps = end_temp - start_temp;
num_samplers = end_sampler - start_sampler;
ureg = ureg_create(TGSI_PROCESSOR_FRAGMENT);
if (!ureg)
return NULL;
if (declare_input) {
in[0] = ureg_DECL_fs_input(ureg,
TGSI_SEMANTIC_POSITION,
0,
TGSI_INTERPOLATE_LINEAR);
in[1] = ureg_DECL_fs_input(ureg,
TGSI_SEMANTIC_GENERIC,
0,
TGSI_INTERPOLATE_PERSPECTIVE);
}
/* we always have a color output */
out = ureg_DECL_output(ureg, TGSI_SEMANTIC_COLOR, 0);
if (num_consts >= 1) {
constant = (struct ureg_src *) malloc(sizeof(struct ureg_src) * end_const);
for (i = start_const; i < end_const; i++) {
constant[i] = ureg_DECL_constant(ureg, i);
}
}
if (num_temps >= 1) {
temp = (struct ureg_dst *) malloc(sizeof(struct ureg_dst) * end_temp);
for (i = start_temp; i < end_temp; i++) {
temp[i] = ureg_DECL_temporary(ureg);
}
}
if (num_samplers >= 1) {
sampler = (struct ureg_src *) malloc(sizeof(struct ureg_src) * end_sampler);
for (i = start_sampler; i < end_sampler; i++) {
sampler[i] = ureg_DECL_sampler(ureg, i);
}
}
while (current_shader < num_shaders) {
if ((current_shader + 1) == num_shaders) {
shaders[current_shader]->func(ureg,
&out,
in,
sampler,
temp,
constant);
} else {
shaders[current_shader]->func(ureg,
&temp[0],
in,
sampler,
temp,
constant);
}
current_shader++;
}
ureg_END(ureg);
shader->tokens = ureg_finalize(ureg);
if(!shader->tokens)
return NULL;
p = pipe->create_fs_state(pipe, shader);
ureg_destroy(ureg);
if (num_temps >= 1) {
for (i = start_temp; i < end_temp; i++) {
ureg_release_temporary(ureg, temp[i]);
}
}
if (temp)
free(temp);
if (constant)
free(constant);
if (sampler)
free(sampler);
return p;
}
static void *
create_fs(struct st_context *st, bool download, enum pipe_texture_target target)
{
struct pipe_context *pipe = st->pipe;
struct pipe_screen *screen = pipe->screen;
struct ureg_program *ureg;
bool have_layer;
struct ureg_dst out;
struct ureg_src sampler;
struct ureg_src pos;
struct ureg_src layer;
struct ureg_src const0;
struct ureg_src const1;
struct ureg_dst temp0;
have_layer =
st->pbo.layers &&
(!download || target == PIPE_TEXTURE_1D_ARRAY
|| target == PIPE_TEXTURE_2D_ARRAY
|| target == PIPE_TEXTURE_3D
|| target == PIPE_TEXTURE_CUBE
|| target == PIPE_TEXTURE_CUBE_ARRAY);
ureg = ureg_create(PIPE_SHADER_FRAGMENT);
if (!ureg)
return NULL;
if (!download) {
out = ureg_DECL_output(ureg, TGSI_SEMANTIC_COLOR, 0);
} else {
struct ureg_src image;
/* writeonly images do not require an explicitly given format. */
image = ureg_DECL_image(ureg, 0, TGSI_TEXTURE_BUFFER, PIPE_FORMAT_NONE,
true, false);
out = ureg_dst(image);
}
sampler = ureg_DECL_sampler(ureg, 0);
if (screen->get_param(screen, PIPE_CAP_TGSI_FS_POSITION_IS_SYSVAL)) {
pos = ureg_DECL_system_value(ureg, TGSI_SEMANTIC_POSITION, 0);
} else {
pos = ureg_DECL_fs_input(ureg, TGSI_SEMANTIC_POSITION, 0,
TGSI_INTERPOLATE_LINEAR);
}
if (have_layer) {
layer = ureg_DECL_fs_input(ureg, TGSI_SEMANTIC_LAYER, 0,
TGSI_INTERPOLATE_CONSTANT);
}
const0 = ureg_DECL_constant(ureg, 0);
const1 = ureg_DECL_constant(ureg, 1);
temp0 = ureg_DECL_temporary(ureg);
/* Note: const0 = [ -xoffset + skip_pixels, -yoffset, stride, image_height ] */
/* temp0.xy = f2i(temp0.xy) */
ureg_F2I(ureg, ureg_writemask(temp0, TGSI_WRITEMASK_XY),
ureg_swizzle(pos,
TGSI_SWIZZLE_X, TGSI_SWIZZLE_Y,
TGSI_SWIZZLE_Y, TGSI_SWIZZLE_Y));
/* temp0.xy = temp0.xy + const0.xy */
ureg_UADD(ureg, ureg_writemask(temp0, TGSI_WRITEMASK_XY),
ureg_swizzle(ureg_src(temp0),
TGSI_SWIZZLE_X, TGSI_SWIZZLE_Y,
TGSI_SWIZZLE_Y, TGSI_SWIZZLE_Y),
ureg_swizzle(const0,
TGSI_SWIZZLE_X, TGSI_SWIZZLE_Y,
TGSI_SWIZZLE_Y, TGSI_SWIZZLE_Y));
/* temp0.x = const0.z * temp0.y + temp0.x */
ureg_UMAD(ureg, ureg_writemask(temp0, TGSI_WRITEMASK_X),
ureg_scalar(const0, TGSI_SWIZZLE_Z),
ureg_scalar(ureg_src(temp0), TGSI_SWIZZLE_Y),
ureg_scalar(ureg_src(temp0), TGSI_SWIZZLE_X));
if (have_layer) {
/* temp0.x = const0.w * layer + temp0.x */
ureg_UMAD(ureg, ureg_writemask(temp0, TGSI_WRITEMASK_X),
ureg_scalar(const0, TGSI_SWIZZLE_W),
ureg_scalar(layer, TGSI_SWIZZLE_X),
ureg_scalar(ureg_src(temp0), TGSI_SWIZZLE_X));
}
/* temp0.w = 0 */
ureg_MOV(ureg, ureg_writemask(temp0, TGSI_WRITEMASK_W), ureg_imm1u(ureg, 0));
if (download) {
struct ureg_dst temp1;
struct ureg_src op[2];
temp1 = ureg_DECL_temporary(ureg);
/* temp1.xy = pos.xy */
ureg_F2I(ureg, ureg_writemask(temp1, TGSI_WRITEMASK_XY), pos);
/* temp1.zw = 0 */
ureg_MOV(ureg, ureg_writemask(temp1, TGSI_WRITEMASK_ZW), ureg_imm1u(ureg, 0));
if (have_layer) {
struct ureg_dst temp1_layer =
ureg_writemask(temp1, target == PIPE_TEXTURE_1D_ARRAY ? TGSI_WRITEMASK_Y
: TGSI_WRITEMASK_Z);
/* temp1.y/z = layer */
ureg_MOV(ureg, temp1_layer, ureg_scalar(layer, TGSI_SWIZZLE_X));
if (target == PIPE_TEXTURE_3D) {
/* temp1.z += layer_offset */
ureg_UADD(ureg, temp1_layer,
ureg_scalar(ureg_src(temp1), TGSI_SWIZZLE_Z),
ureg_scalar(const1, TGSI_SWIZZLE_X));
}
}
/* temp1 = txf(sampler, temp1) */
ureg_TXF(ureg, temp1, util_pipe_tex_to_tgsi_tex(target, 1),
ureg_src(temp1), sampler);
/* store(out, temp0, temp1) */
op[0] = ureg_src(temp0);
op[1] = ureg_src(temp1);
ureg_memory_insn(ureg, TGSI_OPCODE_STORE, &out, 1, op, 2, 0,
TGSI_TEXTURE_BUFFER, PIPE_FORMAT_NONE);
ureg_release_temporary(ureg, temp1);
} else {
/* out = txf(sampler, temp0.x) */
ureg_TXF(ureg, out, TGSI_TEXTURE_BUFFER, ureg_src(temp0), sampler);
}
ureg_release_temporary(ureg, temp0);
ureg_END(ureg);
return ureg_create_shader_and_destroy(ureg, pipe);
}
