/**
* Allocate a renderbuffer for a an on-screen window (not a user-created
* renderbuffer). The window system code determines the format.
*/
struct gl_renderbuffer *
st_new_renderbuffer_fb(enum pipe_format format, int samples, boolean sw)
{
struct st_renderbuffer *strb;
strb = ST_CALLOC_STRUCT(st_renderbuffer);
if (!strb) {
_mesa_error(NULL, GL_OUT_OF_MEMORY, "creating renderbuffer");
return NULL;
}
_mesa_init_renderbuffer(&strb->Base, 0);
strb->Base.ClassID = 0x4242; /* just a unique value */
strb->Base.NumSamples = samples;
strb->Base.Format = st_pipe_format_to_mesa_format(format);
strb->Base._BaseFormat = _mesa_get_format_base_format(strb->Base.Format);
strb->software = sw;
switch (format) {
case PIPE_FORMAT_R8G8B8A8_UNORM:
case PIPE_FORMAT_B8G8R8A8_UNORM:
case PIPE_FORMAT_A8R8G8B8_UNORM:
strb->Base.InternalFormat = GL_RGBA8;
break;
case PIPE_FORMAT_R8G8B8X8_UNORM:
case PIPE_FORMAT_B8G8R8X8_UNORM:
case PIPE_FORMAT_X8R8G8B8_UNORM:
strb->Base.InternalFormat = GL_RGB8;
break;
case PIPE_FORMAT_B5G5R5A1_UNORM:
strb->Base.InternalFormat = GL_RGB5_A1;
break;
case PIPE_FORMAT_B4G4R4A4_UNORM:
strb->Base.InternalFormat = GL_RGBA4;
break;
case PIPE_FORMAT_B5G6R5_UNORM:
strb->Base.InternalFormat = GL_RGB565;
break;
case PIPE_FORMAT_Z16_UNORM:
strb->Base.InternalFormat = GL_DEPTH_COMPONENT16;
break;
case PIPE_FORMAT_Z32_UNORM:
strb->Base.InternalFormat = GL_DEPTH_COMPONENT32;
break;
case PIPE_FORMAT_Z24_UNORM_S8_UINT:
case PIPE_FORMAT_S8_UINT_Z24_UNORM:
strb->Base.InternalFormat = GL_DEPTH24_STENCIL8_EXT;
break;
case PIPE_FORMAT_Z24X8_UNORM:
case PIPE_FORMAT_X8Z24_UNORM:
strb->Base.InternalFormat = GL_DEPTH_COMPONENT24;
break;
case PIPE_FORMAT_S8_UINT:
strb->Base.InternalFormat = GL_STENCIL_INDEX8_EXT;
break;
case PIPE_FORMAT_R16G16B16A16_SNORM:
/* accum buffer */
strb->Base.InternalFormat = GL_RGBA16_SNORM;
break;
case PIPE_FORMAT_R16G16B16A16_UNORM:
strb->Base.InternalFormat = GL_RGBA16;
break;
case PIPE_FORMAT_R8_UNORM:
strb->Base.InternalFormat = GL_R8;
break;
case PIPE_FORMAT_R8G8_UNORM:
strb->Base.InternalFormat = GL_RG8;
break;
case PIPE_FORMAT_R16_UNORM:
strb->Base.InternalFormat = GL_R16;
break;
case PIPE_FORMAT_R16G16_UNORM:
strb->Base.InternalFormat = GL_RG16;
break;
case PIPE_FORMAT_R32G32B32A32_FLOAT:
strb->Base.InternalFormat = GL_RGBA32F;
break;
case PIPE_FORMAT_R16G16B16A16_FLOAT:
strb->Base.InternalFormat = GL_RGBA16F;
break;
default:
_mesa_problem(NULL,
"Unexpected format %s in st_new_renderbuffer_fb",
util_format_name(format));
free(strb);
return NULL;
}
/* st-specific methods */
strb->Base.Delete = st_renderbuffer_delete;
strb->Base.AllocStorage = st_renderbuffer_alloc_storage;
/* surface is allocated in st_renderbuffer_alloc_storage() */
strb->surface = NULL;
return &strb->Base;
}
/**
* Apply the given stencil operator to the array of stencil values.
* Don't touch stencil[i] if mask[i] is zero.
* Input: n - size of stencil array
* oper - the stencil buffer operator
* face - 0 or 1 for front or back face operation
* stencil - array of stencil values
* mask - array [n] of flag: 1=apply operator, 0=don't apply operator
* Output: stencil - modified values
*/
static void
apply_stencil_op( const GLcontext *ctx, GLenum oper, GLuint face,
GLuint n, GLstencil stencil[], const GLubyte mask[] )
{
const GLstencil ref = ctx->Stencil.Ref[face];
const GLstencil wrtmask = ctx->Stencil.WriteMask[face];
const GLstencil invmask = (GLstencil) (~wrtmask);
GLuint i;
switch (oper) {
case GL_KEEP:
/* do nothing */
break;
case GL_ZERO:
if (invmask==0) {
for (i=0;i<n;i++) {
if (mask[i]) {
stencil[i] = 0;
}
}
}
else {
for (i=0;i<n;i++) {
if (mask[i]) {
stencil[i] = (GLstencil) (stencil[i] & invmask);
}
}
}
break;
case GL_REPLACE:
if (invmask==0) {
for (i=0;i<n;i++) {
if (mask[i]) {
stencil[i] = ref;
}
}
}
else {
for (i=0;i<n;i++) {
if (mask[i]) {
GLstencil s = stencil[i];
stencil[i] = (GLstencil) ((invmask & s ) | (wrtmask & ref));
}
}
}
break;
case GL_INCR:
if (invmask==0) {
for (i=0;i<n;i++) {
if (mask[i]) {
GLstencil s = stencil[i];
if (s < STENCIL_MAX) {
stencil[i] = (GLstencil) (s+1);
}
}
}
}
else {
for (i=0;i<n;i++) {
if (mask[i]) {
/* VERIFY logic of adding 1 to a write-masked value */
GLstencil s = stencil[i];
if (s < STENCIL_MAX) {
stencil[i] = (GLstencil) ((invmask & s) | (wrtmask & (s+1)));
}
}
}
}
break;
case GL_DECR:
if (invmask==0) {
for (i=0;i<n;i++) {
if (mask[i]) {
GLstencil s = stencil[i];
if (s>0) {
stencil[i] = (GLstencil) (s-1);
}
}
}
}
else {
for (i=0;i<n;i++) {
if (mask[i]) {
/* VERIFY logic of subtracting 1 to a write-masked value */
GLstencil s = stencil[i];
if (s>0) {
stencil[i] = (GLstencil) ((invmask & s) | (wrtmask & (s-1)));
}
}
}
}
break;
case GL_INCR_WRAP_EXT:
if (invmask==0) {
for (i=0;i<n;i++) {
if (mask[i]) {
stencil[i]++;
}
}
}
else {
for (i=0;i<n;i++) {
if (mask[i]) {
GLstencil s = stencil[i];
stencil[i] = (GLstencil) ((invmask & s) | (wrtmask & (s+1)));
}
}
}
break;
case GL_DECR_WRAP_EXT:
if (invmask==0) {
for (i=0;i<n;i++) {
if (mask[i]) {
stencil[i]--;
}
}
}
else {
for (i=0;i<n;i++) {
if (mask[i]) {
GLstencil s = stencil[i];
stencil[i] = (GLstencil) ((invmask & s) | (wrtmask & (s-1)));
}
}
}
break;
case GL_INVERT:
if (invmask==0) {
for (i=0;i<n;i++) {
if (mask[i]) {
GLstencil s = stencil[i];
stencil[i] = (GLstencil) ~s;
}
}
}
else {
for (i=0;i<n;i++) {
if (mask[i]) {
GLstencil s = stencil[i];
stencil[i] = (GLstencil) ((invmask & s) | (wrtmask & ~s));
}
}
}
break;
default:
_mesa_problem(ctx, "Bad stencil op in apply_stencil_op");
}
}
/**
* Allocate a new driRenderbuffer object.
* Individual drivers are free to implement different versions of
* this function.
*
* At this time, this function can only be used for window-system
* renderbuffers, not user-created RBOs.
*
* \param format Either GL_RGBA, GL_DEPTH_COMPONENT16, GL_DEPTH_COMPONENT24,
* GL_DEPTH_COMPONENT32, or GL_STENCIL_INDEX8_EXT (for now).
* \param addr address in main memory of the buffer. Probably a memory
* mapped region.
* \param cpp chars or bytes per pixel
* \param offset start of renderbuffer with respect to start of framebuffer
* \param pitch pixels per row
*/
driRenderbuffer *
driNewRenderbuffer(gl_format format, GLvoid *addr,
GLint cpp, GLint offset, GLint pitch,
__DRIdrawablePrivate *dPriv)
{
driRenderbuffer *drb;
assert(format == GL_RGBA ||
format == GL_RGB5 ||
format == GL_RGBA8 ||
format == GL_DEPTH_COMPONENT16 ||
format == GL_DEPTH_COMPONENT24 ||
format == GL_DEPTH_COMPONENT32 ||
format == GL_STENCIL_INDEX8_EXT);
assert(cpp > 0);
assert(pitch > 0);
drb = _mesa_calloc(sizeof(driRenderbuffer));
if (drb) {
const GLuint name = 0;
_mesa_init_renderbuffer(&drb->Base, name);
/* Make sure we're using a null-valued GetPointer routine */
assert(drb->Base.GetPointer(NULL, &drb->Base, 0, 0) == NULL);
switch (format) {
case MESA_FORMAT_ARGB8888:
if (cpp == 2) {
/* override format */
format = MESA_FORMAT_RGB565;
}
drb->Base.DataType = GL_UNSIGNED_BYTE;
break;
case MESA_FORMAT_Z16:
/* Depth */
/* we always Get/Put 32-bit Z values */
drb->Base.DataType = GL_UNSIGNED_INT;
assert(cpp == 2);
break;
case MESA_FORMAT_Z32:
/* Depth */
/* we always Get/Put 32-bit Z values */
drb->Base.DataType = GL_UNSIGNED_INT;
assert(cpp == 4);
break;
case MESA_FORMAT_Z24_S8:
drb->Base.DataType = GL_UNSIGNED_INT_24_8_EXT;
assert(cpp == 4);
break;
case MESA_FORMAT_S8_Z24:
drb->Base.DataType = GL_UNSIGNED_INT_24_8_EXT;
assert(cpp == 4);
break;
case MESA_FORMAT_S8:
/* Stencil */
drb->Base.DataType = GL_UNSIGNED_BYTE;
break;
default:
_mesa_problem(NULL, "Bad format 0x%x in driNewRenderbuffer", format);
return NULL;
}
drb->Base.Format = format;
drb->Base.InternalFormat =
drb->Base._BaseFormat = _mesa_get_format_base_format(format);
drb->Base.AllocStorage = driRenderbufferStorage;
drb->Base.Delete = driDeleteRenderbuffer;
drb->Base.Data = addr;
/* DRI renderbuffer-specific fields: */
drb->dPriv = dPriv;
drb->offset = offset;
drb->pitch = pitch;
drb->cpp = cpp;
/* may be changed if page flipping is active: */
drb->flippedOffset = offset;
drb->flippedPitch = pitch;
drb->flippedData = addr;
}
return drb;
}
/**
* Append instructions to implement fog
*
* The \c fragment.fogcoord input is used to compute the fog blend factor.
*
* \param ctx The GL context
* \param fprog Fragment program that fog instructions will be appended to.
* \param fog_mode Fog mode. One of \c GL_EXP, \c GL_EXP2, or \c GL_LINEAR.
* \param saturate True if writes to color outputs should be clamped to [0, 1]
*
* \note
* This function sets \c VARYING_BIT_FOGC in \c fprog->Base.InputsRead.
*
* \todo With a little work, this function could be adapted to add fog code
* to vertex programs too.
*/
void
_mesa_append_fog_code(struct gl_context *ctx,
struct gl_fragment_program *fprog, GLenum fog_mode,
GLboolean saturate)
{
static const gl_state_index fogPStateOpt[STATE_LENGTH]
= { STATE_INTERNAL, STATE_FOG_PARAMS_OPTIMIZED, 0, 0, 0 };
static const gl_state_index fogColorState[STATE_LENGTH]
= { STATE_FOG_COLOR, 0, 0, 0, 0};
struct prog_instruction *newInst, *inst;
const GLuint origLen = fprog->Base.NumInstructions;
const GLuint newLen = origLen + 5;
GLuint i;
GLint fogPRefOpt, fogColorRef; /* state references */
GLuint colorTemp, fogFactorTemp; /* temporary registerss */
if (fog_mode == GL_NONE) {
_mesa_problem(ctx, "_mesa_append_fog_code() called for fragment program"
" with fog_mode == GL_NONE");
return;
}
if (!(fprog->Base.OutputsWritten & (1 << FRAG_RESULT_COLOR))) {
/* program doesn't output color, so nothing to do */
return;
}
/* Alloc storage for new instructions */
newInst = _mesa_alloc_instructions(newLen);
if (!newInst) {
_mesa_error(ctx, GL_OUT_OF_MEMORY,
"glProgramString(inserting fog_option code)");
return;
}
/* Copy orig instructions into new instruction buffer */
_mesa_copy_instructions(newInst, fprog->Base.Instructions, origLen);
/* PARAM fogParamsRefOpt = internal optimized fog params; */
fogPRefOpt
= _mesa_add_state_reference(fprog->Base.Parameters, fogPStateOpt);
/* PARAM fogColorRef = state.fog.color; */
fogColorRef
= _mesa_add_state_reference(fprog->Base.Parameters, fogColorState);
/* TEMP colorTemp; */
colorTemp = fprog->Base.NumTemporaries++;
/* TEMP fogFactorTemp; */
fogFactorTemp = fprog->Base.NumTemporaries++;
/* Scan program to find where result.color is written */
inst = newInst;
for (i = 0; i < fprog->Base.NumInstructions; i++) {
if (inst->Opcode == OPCODE_END)
break;
if (inst->DstReg.File == PROGRAM_OUTPUT &&
inst->DstReg.Index == FRAG_RESULT_COLOR) {
/* change the instruction to write to colorTemp w/ clamping */
inst->DstReg.File = PROGRAM_TEMPORARY;
inst->DstReg.Index = colorTemp;
inst->SaturateMode = saturate;
/* don't break (may be several writes to result.color) */
}
inst++;
}
assert(inst->Opcode == OPCODE_END); /* we'll overwrite this inst */
_mesa_init_instructions(inst, 5);
/* emit instructions to compute fog blending factor */
/* this is always clamped to [0, 1] regardless of fragment clamping */
if (fog_mode == GL_LINEAR) {
/* MAD fogFactorTemp.x, fragment.fogcoord.x, fogPRefOpt.x, fogPRefOpt.y; */
inst->Opcode = OPCODE_MAD;
inst->DstReg.File = PROGRAM_TEMPORARY;
inst->DstReg.Index = fogFactorTemp;
inst->DstReg.WriteMask = WRITEMASK_X;
inst->SrcReg[0].File = PROGRAM_INPUT;
inst->SrcReg[0].Index = VARYING_SLOT_FOGC;
inst->SrcReg[0].Swizzle = SWIZZLE_XXXX;
inst->SrcReg[1].File = PROGRAM_STATE_VAR;
inst->SrcReg[1].Index = fogPRefOpt;
inst->SrcReg[1].Swizzle = SWIZZLE_XXXX;
inst->SrcReg[2].File = PROGRAM_STATE_VAR;
inst->SrcReg[2].Index = fogPRefOpt;
inst->SrcReg[2].Swizzle = SWIZZLE_YYYY;
inst->SaturateMode = SATURATE_ZERO_ONE;
inst++;
}
else {
ASSERT(fog_mode == GL_EXP || fog_mode == GL_EXP2);
/* fogPRefOpt.z = d/ln(2), fogPRefOpt.w = d/sqrt(ln(2) */
/* EXP: MUL fogFactorTemp.x, fogPRefOpt.z, fragment.fogcoord.x; */
/* EXP2: MUL fogFactorTemp.x, fogPRefOpt.w, fragment.fogcoord.x; */
inst->Opcode = OPCODE_MUL;
inst->DstReg.File = PROGRAM_TEMPORARY;
inst->DstReg.Index = fogFactorTemp;
inst->DstReg.WriteMask = WRITEMASK_X;
inst->SrcReg[0].File = PROGRAM_STATE_VAR;
inst->SrcReg[0].Index = fogPRefOpt;
inst->SrcReg[0].Swizzle
= (fog_mode == GL_EXP) ? SWIZZLE_ZZZZ : SWIZZLE_WWWW;
inst->SrcReg[1].File = PROGRAM_INPUT;
inst->SrcReg[1].Index = VARYING_SLOT_FOGC;
inst->SrcReg[1].Swizzle = SWIZZLE_XXXX;
inst++;
if (fog_mode == GL_EXP2) {
/* MUL fogFactorTemp.x, fogFactorTemp.x, fogFactorTemp.x; */
inst->Opcode = OPCODE_MUL;
inst->DstReg.File = PROGRAM_TEMPORARY;
inst->DstReg.Index = fogFactorTemp;
inst->DstReg.WriteMask = WRITEMASK_X;
inst->SrcReg[0].File = PROGRAM_TEMPORARY;
inst->SrcReg[0].Index = fogFactorTemp;
inst->SrcReg[0].Swizzle = SWIZZLE_XXXX;
inst->SrcReg[1].File = PROGRAM_TEMPORARY;
inst->SrcReg[1].Index = fogFactorTemp;
inst->SrcReg[1].Swizzle = SWIZZLE_XXXX;
inst++;
}
/* EX2_SAT fogFactorTemp.x, -fogFactorTemp.x; */
inst->Opcode = OPCODE_EX2;
inst->DstReg.File = PROGRAM_TEMPORARY;
inst->DstReg.Index = fogFactorTemp;
inst->DstReg.WriteMask = WRITEMASK_X;
inst->SrcReg[0].File = PROGRAM_TEMPORARY;
inst->SrcReg[0].Index = fogFactorTemp;
inst->SrcReg[0].Negate = NEGATE_XYZW;
inst->SrcReg[0].Swizzle = SWIZZLE_XXXX;
inst->SaturateMode = SATURATE_ZERO_ONE;
inst++;
}
/* LRP result.color.xyz, fogFactorTemp.xxxx, colorTemp, fogColorRef; */
inst->Opcode = OPCODE_LRP;
inst->DstReg.File = PROGRAM_OUTPUT;
inst->DstReg.Index = FRAG_RESULT_COLOR;
inst->DstReg.WriteMask = WRITEMASK_XYZ;
inst->SrcReg[0].File = PROGRAM_TEMPORARY;
inst->SrcReg[0].Index = fogFactorTemp;
inst->SrcReg[0].Swizzle = SWIZZLE_XXXX;
inst->SrcReg[1].File = PROGRAM_TEMPORARY;
inst->SrcReg[1].Index = colorTemp;
inst->SrcReg[1].Swizzle = SWIZZLE_NOOP;
inst->SrcReg[2].File = PROGRAM_STATE_VAR;
inst->SrcReg[2].Index = fogColorRef;
inst->SrcReg[2].Swizzle = SWIZZLE_NOOP;
inst++;
/* MOV result.color.w, colorTemp.x; # copy alpha */
inst->Opcode = OPCODE_MOV;
inst->DstReg.File = PROGRAM_OUTPUT;
inst->DstReg.Index = FRAG_RESULT_COLOR;
inst->DstReg.WriteMask = WRITEMASK_W;
inst->SrcReg[0].File = PROGRAM_TEMPORARY;
inst->SrcReg[0].Index = colorTemp;
inst->SrcReg[0].Swizzle = SWIZZLE_NOOP;
inst++;
/* END; */
inst->Opcode = OPCODE_END;
inst++;
/* free old instructions */
_mesa_free_instructions(fprog->Base.Instructions, origLen);
/* install new instructions */
fprog->Base.Instructions = newInst;
fprog->Base.NumInstructions = inst - newInst;
fprog->Base.InputsRead |= VARYING_BIT_FOGC;
assert(fprog->Base.OutputsWritten & (1 << FRAG_RESULT_COLOR));
}
/**
* Apply the given stencil operator for each pixel in the array whose
* mask flag is set.
* \note This is for software stencil buffers only.
* Input: n - number of pixels in the span
* x, y - array of [n] pixels
* operator - the stencil buffer operator
* mask - array [n] of flag: 1=apply operator, 0=don't apply operator
*/
static void
apply_stencil_op_to_pixels( const GLcontext *ctx,
GLuint n, const GLint x[], const GLint y[],
GLenum oper, GLuint face, const GLubyte mask[] )
{
const GLstencil ref = ctx->Stencil.Ref[face];
const GLstencil wrtmask = ctx->Stencil.WriteMask[face];
const GLstencil invmask = (GLstencil) (~wrtmask);
GLuint i;
ASSERT(!SWRAST_CONTEXT(ctx)->Driver.WriteStencilSpan); /* software stencil buffer only! */
switch (oper) {
case GL_KEEP:
/* do nothing */
break;
case GL_ZERO:
if (invmask==0) {
for (i=0;i<n;i++) {
if (mask[i]) {
GLstencil *sptr = STENCIL_ADDRESS( x[i], y[i] );
*sptr = 0;
}
}
}
else {
for (i=0;i<n;i++) {
if (mask[i]) {
GLstencil *sptr = STENCIL_ADDRESS( x[i], y[i] );
*sptr = (GLstencil) (invmask & *sptr);
}
}
}
break;
case GL_REPLACE:
if (invmask==0) {
for (i=0;i<n;i++) {
if (mask[i]) {
GLstencil *sptr = STENCIL_ADDRESS( x[i], y[i] );
*sptr = ref;
}
}
}
else {
for (i=0;i<n;i++) {
if (mask[i]) {
GLstencil *sptr = STENCIL_ADDRESS( x[i], y[i] );
*sptr = (GLstencil) ((invmask & *sptr ) | (wrtmask & ref));
}
}
}
break;
case GL_INCR:
if (invmask==0) {
for (i=0;i<n;i++) {
if (mask[i]) {
GLstencil *sptr = STENCIL_ADDRESS( x[i], y[i] );
if (*sptr < STENCIL_MAX) {
*sptr = (GLstencil) (*sptr + 1);
}
}
}
}
else {
for (i=0;i<n;i++) {
if (mask[i]) {
GLstencil *sptr = STENCIL_ADDRESS( x[i], y[i] );
if (*sptr < STENCIL_MAX) {
*sptr = (GLstencil) ((invmask & *sptr) | (wrtmask & (*sptr+1)));
}
}
}
}
break;
case GL_DECR:
if (invmask==0) {
for (i=0;i<n;i++) {
if (mask[i]) {
GLstencil *sptr = STENCIL_ADDRESS( x[i], y[i] );
if (*sptr>0) {
*sptr = (GLstencil) (*sptr - 1);
}
}
}
}
else {
for (i=0;i<n;i++) {
if (mask[i]) {
GLstencil *sptr = STENCIL_ADDRESS( x[i], y[i] );
if (*sptr>0) {
*sptr = (GLstencil) ((invmask & *sptr) | (wrtmask & (*sptr-1)));
}
}
}
}
break;
case GL_INCR_WRAP_EXT:
if (invmask==0) {
for (i=0;i<n;i++) {
if (mask[i]) {
GLstencil *sptr = STENCIL_ADDRESS( x[i], y[i] );
*sptr = (GLstencil) (*sptr + 1);
}
}
}
else {
for (i=0;i<n;i++) {
if (mask[i]) {
GLstencil *sptr = STENCIL_ADDRESS( x[i], y[i] );
*sptr = (GLstencil) ((invmask & *sptr) | (wrtmask & (*sptr+1)));
}
}
}
break;
case GL_DECR_WRAP_EXT:
if (invmask==0) {
for (i=0;i<n;i++) {
if (mask[i]) {
GLstencil *sptr = STENCIL_ADDRESS( x[i], y[i] );
*sptr = (GLstencil) (*sptr - 1);
}
}
}
else {
for (i=0;i<n;i++) {
if (mask[i]) {
GLstencil *sptr = STENCIL_ADDRESS( x[i], y[i] );
*sptr = (GLstencil) ((invmask & *sptr) | (wrtmask & (*sptr-1)));
}
}
}
break;
case GL_INVERT:
if (invmask==0) {
for (i=0;i<n;i++) {
if (mask[i]) {
GLstencil *sptr = STENCIL_ADDRESS( x[i], y[i] );
*sptr = (GLstencil) (~*sptr);
}
}
}
else {
for (i=0;i<n;i++) {
if (mask[i]) {
GLstencil *sptr = STENCIL_ADDRESS( x[i], y[i] );
*sptr = (GLstencil) ((invmask & *sptr) | (wrtmask & ~*sptr));
}
}
}
break;
default:
_mesa_problem(ctx, "Bad stencilop in apply_stencil_op_to_pixels");
}
}
static const unsigned *
brw_cs_emit(struct brw_context *brw,
void *mem_ctx,
const struct brw_cs_prog_key *key,
struct brw_cs_prog_data *prog_data,
struct gl_compute_program *cp,
struct gl_shader_program *prog,
unsigned *final_assembly_size)
{
bool start_busy = false;
double start_time = 0;
if (unlikely(brw->perf_debug)) {
start_busy = (brw->batch.last_bo &&
drm_intel_bo_busy(brw->batch.last_bo));
start_time = get_time();
}
struct brw_shader *shader =
(struct brw_shader *) prog->_LinkedShaders[MESA_SHADER_COMPUTE];
if (unlikely(INTEL_DEBUG & DEBUG_CS))
brw_dump_ir("compute", prog, &shader->base, &cp->Base);
prog_data->local_size[0] = cp->LocalSize[0];
prog_data->local_size[1] = cp->LocalSize[1];
prog_data->local_size[2] = cp->LocalSize[2];
unsigned local_workgroup_size =
cp->LocalSize[0] * cp->LocalSize[1] * cp->LocalSize[2];
cfg_t *cfg = NULL;
const char *fail_msg = NULL;
int st_index = -1;
if (INTEL_DEBUG & DEBUG_SHADER_TIME)
st_index = brw_get_shader_time_index(brw, prog, &cp->Base, ST_CS);
/* Now the main event: Visit the shader IR and generate our CS IR for it.
*/
fs_visitor v8(brw->intelScreen->compiler, brw,
mem_ctx, MESA_SHADER_COMPUTE, key, &prog_data->base, prog,
&cp->Base, 8, st_index);
if (!v8.run_cs()) {
fail_msg = v8.fail_msg;
} else if (local_workgroup_size <= 8 * brw->max_cs_threads) {
cfg = v8.cfg;
prog_data->simd_size = 8;
}
fs_visitor v16(brw->intelScreen->compiler, brw,
mem_ctx, MESA_SHADER_COMPUTE, key, &prog_data->base, prog,
&cp->Base, 16, st_index);
if (likely(!(INTEL_DEBUG & DEBUG_NO16)) &&
!fail_msg && !v8.simd16_unsupported &&
local_workgroup_size <= 16 * brw->max_cs_threads) {
/* Try a SIMD16 compile */
v16.import_uniforms(&v8);
if (!v16.run_cs()) {
perf_debug("SIMD16 shader failed to compile: %s", v16.fail_msg);
if (!cfg) {
fail_msg =
"Couldn't generate SIMD16 program and not "
"enough threads for SIMD8";
}
} else {
cfg = v16.cfg;
prog_data->simd_size = 16;
}
}
if (unlikely(cfg == NULL)) {
assert(fail_msg);
prog->LinkStatus = false;
ralloc_strcat(&prog->InfoLog, fail_msg);
_mesa_problem(NULL, "Failed to compile compute shader: %s\n",
fail_msg);
return NULL;
}
fs_generator g(brw->intelScreen->compiler, brw,
mem_ctx, (void*) key, &prog_data->base, &cp->Base,
v8.promoted_constants, v8.runtime_check_aads_emit, "CS");
if (INTEL_DEBUG & DEBUG_CS) {
char *name = ralloc_asprintf(mem_ctx, "%s compute shader %d",
prog->Label ? prog->Label : "unnamed",
prog->Name);
g.enable_debug(name);
}
g.generate_code(cfg, prog_data->simd_size);
if (unlikely(brw->perf_debug) && shader) {
if (shader->compiled_once) {
_mesa_problem(&brw->ctx, "CS programs shouldn't need recompiles");
}
shader->compiled_once = true;
if (start_busy && !drm_intel_bo_busy(brw->batch.last_bo)) {
perf_debug("CS compile took %.03f ms and stalled the GPU\n",
(get_time() - start_time) * 1000);
}
}
return g.get_assembly(final_assembly_size);
}
/**
* Convert "classic" texture environment to ARB_texture_env_combine style
* environments.
*
* \param state texture_env_combine state vector to be filled-in.
* \param mode Classic texture environment mode (i.e., \c GL_REPLACE,
* \c GL_BLEND, \c GL_DECAL, etc.).
* \param texBaseFormat Base format of the texture associated with the
* texture unit.
*/
static void
calculate_derived_texenv( struct gl_tex_env_combine_state *state,
GLenum mode, GLenum texBaseFormat )
{
GLenum mode_rgb;
GLenum mode_a;
*state = default_combine_state;
switch (texBaseFormat) {
case GL_ALPHA:
state->SourceRGB[0] = GL_PREVIOUS;
break;
case GL_LUMINANCE_ALPHA:
case GL_INTENSITY:
case GL_RGBA:
break;
case GL_LUMINANCE:
case GL_RED:
case GL_RG:
case GL_RGB:
case GL_YCBCR_MESA:
state->SourceA[0] = GL_PREVIOUS;
break;
default:
_mesa_problem(NULL,
"Invalid texBaseFormat 0x%x in calculate_derived_texenv",
texBaseFormat);
return;
}
if (mode == GL_REPLACE_EXT)
mode = GL_REPLACE;
switch (mode) {
case GL_REPLACE:
case GL_MODULATE:
mode_rgb = (texBaseFormat == GL_ALPHA) ? GL_REPLACE : mode;
mode_a = mode;
break;
case GL_DECAL:
mode_rgb = GL_INTERPOLATE;
mode_a = GL_REPLACE;
state->SourceA[0] = GL_PREVIOUS;
/* Having alpha / luminance / intensity textures replace using the
* incoming fragment color matches the definition in NV_texture_shader.
* The 1.5 spec simply marks these as "undefined".
*/
switch (texBaseFormat) {
case GL_ALPHA:
case GL_LUMINANCE:
case GL_LUMINANCE_ALPHA:
case GL_INTENSITY:
state->SourceRGB[0] = GL_PREVIOUS;
break;
case GL_RED:
case GL_RG:
case GL_RGB:
case GL_YCBCR_MESA:
mode_rgb = GL_REPLACE;
break;
case GL_RGBA:
state->SourceRGB[2] = GL_TEXTURE;
break;
}
break;
case GL_BLEND:
mode_rgb = GL_INTERPOLATE;
mode_a = GL_MODULATE;
switch (texBaseFormat) {
case GL_ALPHA:
mode_rgb = GL_REPLACE;
break;
case GL_INTENSITY:
mode_a = GL_INTERPOLATE;
state->SourceA[0] = GL_CONSTANT;
state->OperandA[2] = GL_SRC_ALPHA;
/* FALLTHROUGH */
case GL_LUMINANCE:
case GL_RED:
case GL_RG:
case GL_RGB:
case GL_LUMINANCE_ALPHA:
case GL_RGBA:
case GL_YCBCR_MESA:
state->SourceRGB[2] = GL_TEXTURE;
state->SourceA[2] = GL_TEXTURE;
state->SourceRGB[0] = GL_CONSTANT;
state->OperandRGB[2] = GL_SRC_COLOR;
break;
}
break;
case GL_ADD:
mode_rgb = (texBaseFormat == GL_ALPHA) ? GL_REPLACE : GL_ADD;
mode_a = (texBaseFormat == GL_INTENSITY) ? GL_ADD : GL_MODULATE;
break;
default:
_mesa_problem(NULL,
"Invalid texture env mode 0x%x in calculate_derived_texenv",
mode);
return;
}
state->ModeRGB = (state->SourceRGB[0] != GL_PREVIOUS)
? mode_rgb : GL_REPLACE;
state->ModeA = (state->SourceA[0] != GL_PREVIOUS)
? mode_a : GL_REPLACE;
}
/**
* Clear the stencil buffer. If the buffer is a combined
* depth+stencil buffer, only the stencil bits will be touched.
*/
void
_swrast_clear_stencil_buffer(struct gl_context *ctx)
{
struct gl_renderbuffer *rb =
ctx->DrawBuffer->Attachment[BUFFER_STENCIL].Renderbuffer;
const GLubyte stencilBits = ctx->DrawBuffer->Visual.stencilBits;
const GLuint writeMask = ctx->Stencil.WriteMask[0];
const GLuint stencilMax = (1 << stencilBits) - 1;
GLint x, y, width, height;
GLubyte *map;
GLint rowStride, i, j;
GLbitfield mapMode;
if (!rb || writeMask == 0)
return;
/* compute region to clear */
x = ctx->DrawBuffer->_Xmin;
y = ctx->DrawBuffer->_Ymin;
width = ctx->DrawBuffer->_Xmax - ctx->DrawBuffer->_Xmin;
height = ctx->DrawBuffer->_Ymax - ctx->DrawBuffer->_Ymin;
mapMode = GL_MAP_WRITE_BIT;
if ((writeMask & stencilMax) != stencilMax) {
/* need to mask stencil values */
mapMode |= GL_MAP_READ_BIT;
}
else if (_mesa_get_format_bits(rb->Format, GL_DEPTH_BITS) > 0) {
/* combined depth+stencil, need to mask Z values */
mapMode |= GL_MAP_READ_BIT;
}
ctx->Driver.MapRenderbuffer(ctx, rb, x, y, width, height,
mapMode, &map, &rowStride);
if (!map) {
_mesa_error(ctx, GL_OUT_OF_MEMORY, "glClear(stencil)");
return;
}
switch (rb->Format) {
case MESA_FORMAT_S8:
{
GLubyte clear = ctx->Stencil.Clear & writeMask & 0xff;
GLubyte mask = (~writeMask) & 0xff;
if (mask != 0) {
/* masked clear */
for (i = 0; i < height; i++) {
GLubyte *row = map;
for (j = 0; j < width; j++) {
row[j] = (row[j] & mask) | clear;
}
map += rowStride;
}
}
else if (rowStride == width) {
/* clear whole buffer */
memset(map, clear, width * height);
}
else {
/* clear scissored */
for (i = 0; i < height; i++) {
memset(map, clear, width);
map += rowStride;
}
}
}
break;
case MESA_FORMAT_S8_Z24:
{
GLuint clear = (ctx->Stencil.Clear & writeMask & 0xff) << 24;
GLuint mask = (((~writeMask) & 0xff) << 24) | 0xffffff;
for (i = 0; i < height; i++) {
GLuint *row = (GLuint *) map;
for (j = 0; j < width; j++) {
row[j] = (row[j] & mask) | clear;
}
map += rowStride;
}
}
break;
case MESA_FORMAT_Z24_S8:
{
GLuint clear = ctx->Stencil.Clear & writeMask & 0xff;
GLuint mask = 0xffffff00 | ((~writeMask) & 0xff);
for (i = 0; i < height; i++) {
GLuint *row = (GLuint *) map;
for (j = 0; j < width; j++) {
row[j] = (row[j] & mask) | clear;
}
map += rowStride;
}
}
break;
default:
_mesa_problem(ctx, "Unexpected stencil buffer format %s"
" in _swrast_clear_stencil_buffer()",
_mesa_get_format_name(rb->Format));
}
ctx->Driver.UnmapRenderbuffer(ctx, rb);
}
/**
* glGetTexImage() helper: decompress a compressed texture by rendering
* a textured quad. Store the results in the user's buffer.
*/
static void
decompress_with_blit(struct gl_context * ctx,
GLenum format, GLenum type, GLvoid *pixels,
struct gl_texture_image *texImage)
{
struct st_context *st = st_context(ctx);
struct pipe_context *pipe = st->pipe;
struct st_texture_image *stImage = st_texture_image(texImage);
struct st_texture_object *stObj = st_texture_object(texImage->TexObject);
const GLuint width = texImage->Width;
const GLuint height = texImage->Height;
struct pipe_resource *dst_texture;
struct pipe_blit_info blit;
unsigned bind = (PIPE_BIND_RENDER_TARGET | PIPE_BIND_TRANSFER_READ);
struct pipe_transfer *tex_xfer;
ubyte *map;
/* create temp / dest surface */
if (!util_create_rgba_texture(pipe, width, height, bind,
&dst_texture)) {
_mesa_problem(ctx, "util_create_rgba_texture() failed "
"in decompress_with_blit()");
return;
}
blit.src.resource = stObj->pt;
blit.src.level = texImage->Level;
blit.src.format = util_format_linear(stObj->pt->format);
blit.dst.resource = dst_texture;
blit.dst.level = 0;
blit.dst.format = dst_texture->format;
blit.src.box.x = blit.dst.box.x = 0;
blit.src.box.y = blit.dst.box.y = 0;
blit.src.box.z = 0; /* XXX compressed array textures? */
blit.dst.box.z = 0;
blit.src.box.width = blit.dst.box.width = width;
blit.src.box.height = blit.dst.box.height = height;
blit.src.box.depth = blit.dst.box.depth = 1;
blit.mask = PIPE_MASK_RGBA;
blit.filter = PIPE_TEX_FILTER_NEAREST;
blit.scissor_enable = FALSE;
/* blit/render/decompress */
st->pipe->blit(st->pipe, &blit);
pixels = _mesa_map_pbo_dest(ctx, &ctx->Pack, pixels);
map = pipe_transfer_map(pipe, dst_texture, 0, 0,
PIPE_TRANSFER_READ,
0, 0, width, height, &tex_xfer);
if (!map) {
goto end;
}
/* copy/pack data into user buffer */
if (_mesa_format_matches_format_and_type(stImage->base.TexFormat,
format, type,
ctx->Pack.SwapBytes)) {
/* memcpy */
const uint bytesPerRow = width * util_format_get_blocksize(stImage->pt->format);
/* map the dst_surface so we can read from it */
GLuint row;
for (row = 0; row < height; row++) {
GLvoid *dest = _mesa_image_address2d(&ctx->Pack, pixels, width,
height, format, type, row, 0);
memcpy(dest, map, bytesPerRow);
map += tex_xfer->stride;
}
pipe_transfer_unmap(pipe, tex_xfer);
}
else {
/* format translation via floats */
GLuint row;
enum pipe_format pformat = util_format_linear(dst_texture->format);
GLfloat *rgba;
rgba = malloc(width * 4 * sizeof(GLfloat));
if (!rgba) {
_mesa_error(ctx, GL_OUT_OF_MEMORY, "glGetTexImage()");
goto end;
}
for (row = 0; row < height; row++) {
const GLbitfield transferOps = 0x0; /* bypassed for glGetTexImage() */
GLvoid *dest = _mesa_image_address2d(&ctx->Pack, pixels, width,
height, format, type, row, 0);
if (ST_DEBUG & DEBUG_FALLBACK)
debug_printf("%s: fallback format translation\n", __FUNCTION__);
/* get float[4] rgba row from surface */
pipe_get_tile_rgba_format(tex_xfer, map, 0, row, width, 1,
pformat, rgba);
_mesa_pack_rgba_span_float(ctx, width, (GLfloat (*)[4]) rgba, format,
type, dest, &ctx->Pack, transferOps);
}
free(rgba);
}
end:
if (map)
pipe_transfer_unmap(pipe, tex_xfer);
_mesa_unmap_pbo_dest(ctx, &ctx->Pack);
pipe_resource_reference(&dst_texture, NULL);
}
/**
* Check if the given register index is valid (doesn't exceed implementation-
* dependent limits).
* \return GL_TRUE if OK, GL_FALSE if bad index
*/
GLboolean
_mesa_valid_register_index(const struct gl_context *ctx,
gl_shader_type shaderType,
gl_register_file file, GLint index)
{
const struct gl_program_constants *c;
switch (shaderType) {
case MESA_SHADER_VERTEX:
c = &ctx->Const.VertexProgram;
break;
case MESA_SHADER_FRAGMENT:
c = &ctx->Const.FragmentProgram;
break;
case MESA_SHADER_GEOMETRY:
c = &ctx->Const.GeometryProgram;
break;
default:
_mesa_problem(ctx,
"unexpected shader type in _mesa_valid_register_index()");
return GL_FALSE;
}
switch (file) {
case PROGRAM_UNDEFINED:
return GL_TRUE; /* XXX or maybe false? */
case PROGRAM_TEMPORARY:
return index >= 0 && index < c->MaxTemps;
case PROGRAM_ENV_PARAM:
return index >= 0 && index < c->MaxEnvParams;
case PROGRAM_LOCAL_PARAM:
return index >= 0 && index < c->MaxLocalParams;
case PROGRAM_NAMED_PARAM:
return index >= 0 && index < c->MaxParameters;
case PROGRAM_UNIFORM:
case PROGRAM_STATE_VAR:
/* aka constant buffer */
return index >= 0 && index < c->MaxUniformComponents / 4;
case PROGRAM_CONSTANT:
/* constant buffer w/ possible relative negative addressing */
return (index > (int) c->MaxUniformComponents / -4 &&
index < c->MaxUniformComponents / 4);
case PROGRAM_INPUT:
if (index < 0)
return GL_FALSE;
switch (shaderType) {
case MESA_SHADER_VERTEX:
return index < VERT_ATTRIB_GENERIC0 + c->MaxAttribs;
case MESA_SHADER_FRAGMENT:
return index < FRAG_ATTRIB_VAR0 + ctx->Const.MaxVarying;
case MESA_SHADER_GEOMETRY:
return index < GEOM_ATTRIB_VAR0 + ctx->Const.MaxVarying;
default:
return GL_FALSE;
}
case PROGRAM_OUTPUT:
if (index < 0)
return GL_FALSE;
switch (shaderType) {
case MESA_SHADER_VERTEX:
return index < VERT_RESULT_VAR0 + ctx->Const.MaxVarying;
case MESA_SHADER_FRAGMENT:
return index < FRAG_RESULT_DATA0 + ctx->Const.MaxDrawBuffers;
case MESA_SHADER_GEOMETRY:
return index < GEOM_RESULT_VAR0 + ctx->Const.MaxVarying;
default:
return GL_FALSE;
}
case PROGRAM_ADDRESS:
return index >= 0 && index < c->MaxAddressRegs;
default:
_mesa_problem(ctx,
"unexpected register file in _mesa_valid_register_index()");
return GL_FALSE;
}
}
static int test_transform_function( transform_func func, int psize,
int mtype, unsigned long *cycles )
{
GLvector4f source[1], dest[1], ref[1];
GLmatrix mat[1];
GLfloat *m;
int i, j;
#ifdef RUN_DEBUG_BENCHMARK
int cycle_i; /* the counter for the benchmarks we run */
#endif
(void) cycles;
if ( psize > 4 ) {
_mesa_problem( NULL, "test_transform_function called with psize > 4\n" );
return 0;
}
mat->m = (GLfloat *) _mesa_align_malloc( 16 * sizeof(GLfloat), 16 );
mat->type = mtypes[mtype];
m = mat->m;
ASSERT( ((long)m & 15) == 0 );
init_matrix( m );
for ( i = 0 ; i < 4 ; i++ ) {
for ( j = 0 ; j < 4 ; j++ ) {
switch ( templates[mtype][i * 4 + j] ) {
case NIL:
m[j * 4 + i] = 0.0;
break;
case ONE:
m[j * 4 + i] = 1.0;
break;
case NEG:
m[j * 4 + i] = -1.0;
break;
case VAR:
break;
default:
ASSERT(0);
return 0;
}
}
}
for ( i = 0 ; i < TEST_COUNT ; i++) {
ASSIGN_4V( d[i], 0.0, 0.0, 0.0, 1.0 );
ASSIGN_4V( s[i], 0.0, 0.0, 0.0, 1.0 );
for ( j = 0 ; j < psize ; j++ )
s[i][j] = rnd();
}
source->data = (GLfloat(*)[4])s;
source->start = (GLfloat *)s;
source->count = TEST_COUNT;
source->stride = sizeof(s[0]);
source->size = 4;
source->flags = 0;
dest->data = (GLfloat(*)[4])d;
dest->start = (GLfloat *)d;
dest->count = TEST_COUNT;
dest->stride = sizeof(float[4]);
dest->size = 0;
dest->flags = 0;
ref->data = (GLfloat(*)[4])r;
ref->start = (GLfloat *)r;
ref->count = TEST_COUNT;
ref->stride = sizeof(float[4]);
ref->size = 0;
ref->flags = 0;
ref_transform( ref, mat, source );
if ( mesa_profile ) {
BEGIN_RACE( *cycles );
func( dest, mat->m, source );
END_RACE( *cycles );
}
else {
func( dest, mat->m, source );
}
for ( i = 0 ; i < TEST_COUNT ; i++ ) {
for ( j = 0 ; j < 4 ; j++ ) {
if ( significand_match( d[i][j], r[i][j] ) < REQUIRED_PRECISION ) {
printf("-----------------------------\n" );
printf("(i = %i, j = %i)\n", i, j );
printf("%f \t %f \t [diff = %e - %i bit missed]\n",
d[i][0], r[i][0], r[i][0]-d[i][0],
MAX_PRECISION - significand_match( d[i][0], r[i][0] ) );
printf("%f \t %f \t [diff = %e - %i bit missed]\n",
d[i][1], r[i][1], r[i][1]-d[i][1],
MAX_PRECISION - significand_match( d[i][1], r[i][1] ) );
printf("%f \t %f \t [diff = %e - %i bit missed]\n",
d[i][2], r[i][2], r[i][2]-d[i][2],
MAX_PRECISION - significand_match( d[i][2], r[i][2] ) );
printf("%f \t %f \t [diff = %e - %i bit missed]\n",
d[i][3], r[i][3], r[i][3]-d[i][3],
MAX_PRECISION - significand_match( d[i][3], r[i][3] ) );
return 0;
}
}
}
_mesa_align_free( mat->m );
return 1;
}
/**
* Combine two programs into one. Fix instructions so the outputs of
* the first program go to the inputs of the second program.
*/
struct gl_program *
_mesa_combine_programs(struct gl_context *ctx,
const struct gl_program *progA,
const struct gl_program *progB)
{
struct prog_instruction *newInst;
struct gl_program *newProg;
const GLuint lenA = progA->NumInstructions - 1; /* omit END instr */
const GLuint lenB = progB->NumInstructions;
const GLuint numParamsA = _mesa_num_parameters(progA->Parameters);
const GLuint newLength = lenA + lenB;
GLboolean usedTemps[MAX_PROGRAM_TEMPS];
GLuint firstTemp = 0;
GLbitfield inputsB;
GLuint i;
ASSERT(progA->Target == progB->Target);
newInst = _mesa_alloc_instructions(newLength);
if (!newInst)
return GL_FALSE;
_mesa_copy_instructions(newInst, progA->Instructions, lenA);
_mesa_copy_instructions(newInst + lenA, progB->Instructions, lenB);
/* adjust branch / instruction addresses for B's instructions */
for (i = 0; i < lenB; i++) {
newInst[lenA + i].BranchTarget += lenA;
}
newProg = ctx->Driver.NewProgram(ctx, progA->Target, 0);
newProg->Instructions = newInst;
newProg->NumInstructions = newLength;
/* find used temp regs (we may need new temps below) */
_mesa_find_used_registers(newProg, PROGRAM_TEMPORARY,
usedTemps, MAX_PROGRAM_TEMPS);
if (newProg->Target == GL_FRAGMENT_PROGRAM_ARB) {
struct gl_fragment_program *fprogA, *fprogB, *newFprog;
GLbitfield progB_inputsRead = progB->InputsRead;
GLint progB_colorFile, progB_colorIndex;
fprogA = (struct gl_fragment_program *) progA;
fprogB = (struct gl_fragment_program *) progB;
newFprog = (struct gl_fragment_program *) newProg;
newFprog->UsesKill = fprogA->UsesKill || fprogB->UsesKill;
/* We'll do a search and replace for instances
* of progB_colorFile/progB_colorIndex below...
*/
progB_colorFile = PROGRAM_INPUT;
progB_colorIndex = FRAG_ATTRIB_COL0;
/*
* The fragment program may get color from a state var rather than
* a fragment input (vertex output) if it's constant.
* See the texenvprogram.c code.
* So, search the program's parameter list now to see if the program
* gets color from a state var instead of a conventional fragment
* input register.
*/
for (i = 0; i < progB->Parameters->NumParameters; i++) {
struct gl_program_parameter *p = &progB->Parameters->Parameters[i];
if (p->Type == PROGRAM_STATE_VAR &&
p->StateIndexes[0] == STATE_INTERNAL &&
p->StateIndexes[1] == STATE_CURRENT_ATTRIB &&
(int) p->StateIndexes[2] == (int) VERT_ATTRIB_COLOR0) {
progB_inputsRead |= FRAG_BIT_COL0;
progB_colorFile = PROGRAM_STATE_VAR;
progB_colorIndex = i;
break;
}
}
/* Connect color outputs of fprogA to color inputs of fprogB, via a
* new temporary register.
*/
if ((progA->OutputsWritten & (1 << FRAG_RESULT_COLOR)) &&
(progB_inputsRead & FRAG_BIT_COL0)) {
GLint tempReg = _mesa_find_free_register(usedTemps, MAX_PROGRAM_TEMPS,
firstTemp);
if (tempReg < 0) {
_mesa_problem(ctx, "No free temp regs found in "
"_mesa_combine_programs(), using 31");
tempReg = 31;
}
firstTemp = tempReg + 1;
/* replace writes to result.color[0] with tempReg */
replace_registers(newInst, lenA,
PROGRAM_OUTPUT, FRAG_RESULT_COLOR,
PROGRAM_TEMPORARY, tempReg);
/* replace reads from the input color with tempReg */
replace_registers(newInst + lenA, lenB,
progB_colorFile, progB_colorIndex, /* search for */
PROGRAM_TEMPORARY, tempReg /* replace with */ );
}
/* compute combined program's InputsRead */
inputsB = progB_inputsRead;
if (progA->OutputsWritten & (1 << FRAG_RESULT_COLOR)) {
inputsB &= ~(1 << FRAG_ATTRIB_COL0);
}
newProg->InputsRead = progA->InputsRead | inputsB;
newProg->OutputsWritten = progB->OutputsWritten;
newProg->SamplersUsed = progA->SamplersUsed | progB->SamplersUsed;
}
else {
/* vertex program */
assert(0); /* XXX todo */
}
/*
* Merge parameters (uniforms, constants, etc)
*/
newProg->Parameters = _mesa_combine_parameter_lists(progA->Parameters,
progB->Parameters);
adjust_param_indexes(newInst + lenA, lenB, numParamsA);
return newProg;
}
/**
* Return a copy of a program.
* XXX Problem here if the program object is actually OO-derivation
* made by a device driver.
*/
struct gl_program *
_mesa_clone_program(struct gl_context *ctx, const struct gl_program *prog)
{
struct gl_program *clone;
clone = ctx->Driver.NewProgram(ctx, prog->Target, prog->Id);
if (!clone)
return NULL;
assert(clone->Target == prog->Target);
assert(clone->RefCount == 1);
clone->String = (GLubyte *) _mesa_strdup((char *) prog->String);
clone->Format = prog->Format;
clone->Instructions = _mesa_alloc_instructions(prog->NumInstructions);
if (!clone->Instructions) {
_mesa_reference_program(ctx, &clone, NULL);
return NULL;
}
_mesa_copy_instructions(clone->Instructions, prog->Instructions,
prog->NumInstructions);
clone->InputsRead = prog->InputsRead;
clone->OutputsWritten = prog->OutputsWritten;
clone->SamplersUsed = prog->SamplersUsed;
clone->ShadowSamplers = prog->ShadowSamplers;
memcpy(clone->TexturesUsed, prog->TexturesUsed, sizeof(prog->TexturesUsed));
if (prog->Parameters)
clone->Parameters = _mesa_clone_parameter_list(prog->Parameters);
memcpy(clone->LocalParams, prog->LocalParams, sizeof(clone->LocalParams));
if (prog->Varying)
clone->Varying = _mesa_clone_parameter_list(prog->Varying);
if (prog->Attributes)
clone->Attributes = _mesa_clone_parameter_list(prog->Attributes);
memcpy(clone->LocalParams, prog->LocalParams, sizeof(clone->LocalParams));
clone->IndirectRegisterFiles = prog->IndirectRegisterFiles;
clone->NumInstructions = prog->NumInstructions;
clone->NumTemporaries = prog->NumTemporaries;
clone->NumParameters = prog->NumParameters;
clone->NumAttributes = prog->NumAttributes;
clone->NumAddressRegs = prog->NumAddressRegs;
clone->NumNativeInstructions = prog->NumNativeInstructions;
clone->NumNativeTemporaries = prog->NumNativeTemporaries;
clone->NumNativeParameters = prog->NumNativeParameters;
clone->NumNativeAttributes = prog->NumNativeAttributes;
clone->NumNativeAddressRegs = prog->NumNativeAddressRegs;
clone->NumAluInstructions = prog->NumAluInstructions;
clone->NumTexInstructions = prog->NumTexInstructions;
clone->NumTexIndirections = prog->NumTexIndirections;
clone->NumNativeAluInstructions = prog->NumNativeAluInstructions;
clone->NumNativeTexInstructions = prog->NumNativeTexInstructions;
clone->NumNativeTexIndirections = prog->NumNativeTexIndirections;
switch (prog->Target) {
case GL_VERTEX_PROGRAM_ARB:
{
const struct gl_vertex_program *vp
= (const struct gl_vertex_program *) prog;
struct gl_vertex_program *vpc = (struct gl_vertex_program *) clone;
vpc->IsPositionInvariant = vp->IsPositionInvariant;
vpc->IsNVProgram = vp->IsNVProgram;
}
break;
case GL_FRAGMENT_PROGRAM_ARB:
{
const struct gl_fragment_program *fp
= (const struct gl_fragment_program *) prog;
struct gl_fragment_program *fpc = (struct gl_fragment_program *) clone;
fpc->FogOption = fp->FogOption;
fpc->UsesKill = fp->UsesKill;
fpc->OriginUpperLeft = fp->OriginUpperLeft;
fpc->PixelCenterInteger = fp->PixelCenterInteger;
}
break;
case MESA_GEOMETRY_PROGRAM:
{
const struct gl_geometry_program *gp
= (const struct gl_geometry_program *) prog;
struct gl_geometry_program *gpc = (struct gl_geometry_program *) clone;
gpc->VerticesOut = gp->VerticesOut;
gpc->InputType = gp->InputType;
gpc->OutputType = gp->OutputType;
}
break;
default:
_mesa_problem(NULL, "Unexpected target in _mesa_clone_program");
}
return clone;
}
/**
* glGetTexImage() helper: decompress a compressed texture by rendering
* a textured quad. Store the results in the user's buffer.
*/
static void
decompress_with_blit(struct gl_context * ctx,
GLenum format, GLenum type, GLvoid *pixels,
struct gl_texture_image *texImage)
{
struct st_context *st = st_context(ctx);
struct pipe_context *pipe = st->pipe;
struct st_texture_image *stImage = st_texture_image(texImage);
struct st_texture_object *stObj = st_texture_object(texImage->TexObject);
struct pipe_sampler_view *src_view;
const GLuint width = texImage->Width;
const GLuint height = texImage->Height;
struct pipe_surface *dst_surface;
struct pipe_resource *dst_texture;
struct pipe_transfer *tex_xfer;
unsigned bind = (PIPE_BIND_RENDER_TARGET | /* util_blit may choose to render */
PIPE_BIND_TRANSFER_READ);
/* create temp / dest surface */
if (!util_create_rgba_surface(pipe, width, height, bind,
&dst_texture, &dst_surface)) {
_mesa_problem(ctx, "util_create_rgba_surface() failed "
"in decompress_with_blit()");
return;
}
/* Disable conditional rendering. */
if (st->render_condition) {
pipe->render_condition(pipe, NULL, 0);
}
/* Create sampler view that limits fetches to the source mipmap level */
{
struct pipe_sampler_view sv_temp;
u_sampler_view_default_template(&sv_temp, stObj->pt, stObj->pt->format);
sv_temp.format = util_format_linear(sv_temp.format);
sv_temp.u.tex.first_level =
sv_temp.u.tex.last_level = texImage->Level;
src_view = pipe->create_sampler_view(pipe, stObj->pt, &sv_temp);
if (!src_view) {
_mesa_error(ctx, GL_OUT_OF_MEMORY, "glGetTexImage");
return;
}
}
/* blit/render/decompress */
util_blit_pixels_tex(st->blit,
src_view, /* pipe_resource (src) */
0, 0, /* src x0, y0 */
width, height, /* src x1, y1 */
dst_surface, /* pipe_surface (dst) */
0, 0, /* dst x0, y0 */
width, height, /* dst x1, y1 */
0.0, /* z */
PIPE_TEX_MIPFILTER_NEAREST);
/* Restore conditional rendering state. */
if (st->render_condition) {
pipe->render_condition(pipe, st->render_condition,
st->condition_mode);
}
/* map the dst_surface so we can read from it */
tex_xfer = pipe_get_transfer(pipe,
dst_texture, 0, 0,
PIPE_TRANSFER_READ,
0, 0, width, height);
pixels = _mesa_map_pbo_dest(ctx, &ctx->Pack, pixels);
/* copy/pack data into user buffer */
if (_mesa_format_matches_format_and_type(stImage->base.TexFormat,
format, type,
ctx->Pack.SwapBytes)) {
/* memcpy */
const uint bytesPerRow = width * util_format_get_blocksize(stImage->pt->format);
ubyte *map = pipe_transfer_map(pipe, tex_xfer);
GLuint row;
for (row = 0; row < height; row++) {
GLvoid *dest = _mesa_image_address2d(&ctx->Pack, pixels, width,
height, format, type, row, 0);
memcpy(dest, map, bytesPerRow);
map += tex_xfer->stride;
}
pipe_transfer_unmap(pipe, tex_xfer);
}
else {
/* format translation via floats */
GLuint row;
enum pipe_format pformat = util_format_linear(dst_texture->format);
GLfloat *rgba;
rgba = (GLfloat *) malloc(width * 4 * sizeof(GLfloat));
if (!rgba) {
_mesa_error(ctx, GL_OUT_OF_MEMORY, "glGetTexImage()");
goto end;
}
for (row = 0; row < height; row++) {
const GLbitfield transferOps = 0x0; /* bypassed for glGetTexImage() */
GLvoid *dest = _mesa_image_address2d(&ctx->Pack, pixels, width,
height, format, type, row, 0);
if (ST_DEBUG & DEBUG_FALLBACK)
debug_printf("%s: fallback format translation\n", __FUNCTION__);
/* get float[4] rgba row from surface */
pipe_get_tile_rgba_format(pipe, tex_xfer, 0, row, width, 1,
pformat, rgba);
_mesa_pack_rgba_span_float(ctx, width, (GLfloat (*)[4]) rgba, format,
type, dest, &ctx->Pack, transferOps);
}
free(rgba);
}
end:
_mesa_unmap_pbo_dest(ctx, &ctx->Pack);
pipe->transfer_destroy(pipe, tex_xfer);
/* destroy the temp / dest surface */
util_destroy_rgba_surface(dst_texture, dst_surface);
pipe_sampler_view_release(pipe, &src_view);
}
/**
* Helper for _mesa_meta_CopyTexSubImage1/2/3D() functions.
* Have to be careful with locking and meta state for pixel transfer.
*/
static void
copy_tex_sub_image(struct gl_context *ctx,
GLuint dims,
struct gl_texture_image *texImage,
GLint xoffset, GLint yoffset, GLint zoffset,
struct gl_renderbuffer *rb,
GLint x, GLint y,
GLsizei width, GLsizei height)
{
struct gl_texture_object *texObj = texImage->TexObject;
const GLenum target = texObj->Target;
GLenum format, type;
GLint bpp;
void *buf;
/* Choose format/type for temporary image buffer */
format = _mesa_get_format_base_format(texImage->TexFormat);
if (format == GL_LUMINANCE ||
format == GL_LUMINANCE_ALPHA ||
format == GL_INTENSITY) {
/* We don't want to use GL_LUMINANCE, GL_INTENSITY, etc. for the
* temp image buffer because glReadPixels will do L=R+G+B which is
* not what we want (should be L=R).
*/
format = GL_RGBA;
}
if (_mesa_is_format_integer_color(texImage->TexFormat)) {
_mesa_problem(ctx, "unsupported integer color copyteximage");
return;
}
type = get_temp_image_type(ctx, format);
bpp = _mesa_bytes_per_pixel(format, type);
if (bpp <= 0) {
_mesa_problem(ctx, "Bad bpp in meta copy_tex_sub_image()");
return;
}
/*
* Alloc image buffer (XXX could use a PBO)
*/
buf = malloc(width * height * bpp);
if (!buf) {
_mesa_error(ctx, GL_OUT_OF_MEMORY, "glCopyTexSubImage%uD", dims);
return;
}
_mesa_unlock_texture(ctx, texObj); /* need to unlock first */
/*
* Read image from framebuffer (disable pixel transfer ops)
*/
_mesa_meta_begin(ctx, MESA_META_PIXEL_STORE | MESA_META_PIXEL_TRANSFER);
ctx->Driver.ReadPixels(ctx, x, y, width, height,
format, type, &ctx->Pack, buf);
_mesa_meta_end(ctx);
_mesa_update_state(ctx); /* to update pixel transfer state */
/*
* Store texture data (with pixel transfer ops)
*/
_mesa_meta_begin(ctx, MESA_META_PIXEL_STORE);
if (target == GL_TEXTURE_1D) {
ctx->Driver.TexSubImage1D(ctx, texImage,
xoffset, width,
format, type, buf, &ctx->Unpack);
}
else {
ctx->Driver.TexSubImage2D(ctx, texImage,
xoffset, yoffset, width, height,
format, type, buf, &ctx->Unpack);
}
_mesa_meta_end(ctx);
_mesa_lock_texture(ctx, texObj); /* re-lock */
free(buf);
}
/**
* Vert/Geom/Frag programs have per-context variants. Free all the
* variants attached to the given program which match the given context.
*/
static void
destroy_program_variants(struct st_context *st, struct gl_program *program)
{
if (!program)
return;
switch (program->Target) {
case GL_VERTEX_PROGRAM_ARB:
{
struct st_vertex_program *stvp = (struct st_vertex_program *) program;
struct st_vp_variant *vpv, **prevPtr = &stvp->variants;
for (vpv = stvp->variants; vpv; ) {
struct st_vp_variant *next = vpv->next;
if (vpv->key.st == st) {
/* unlink from list */
*prevPtr = next;
/* destroy this variant */
delete_vp_variant(st, vpv);
}
else {
prevPtr = &vpv->next;
}
vpv = next;
}
}
break;
case GL_FRAGMENT_PROGRAM_ARB:
{
struct st_fragment_program *stfp =
(struct st_fragment_program *) program;
struct st_fp_variant *fpv, **prevPtr = &stfp->variants;
for (fpv = stfp->variants; fpv; ) {
struct st_fp_variant *next = fpv->next;
if (fpv->key.st == st) {
/* unlink from list */
*prevPtr = next;
/* destroy this variant */
delete_fp_variant(st, fpv);
}
else {
prevPtr = &fpv->next;
}
fpv = next;
}
}
break;
case MESA_GEOMETRY_PROGRAM:
{
struct st_geometry_program *stgp =
(struct st_geometry_program *) program;
struct st_gp_variant *gpv, **prevPtr = &stgp->variants;
for (gpv = stgp->variants; gpv; ) {
struct st_gp_variant *next = gpv->next;
if (gpv->key.st == st) {
/* unlink from list */
*prevPtr = next;
/* destroy this variant */
delete_gp_variant(st, gpv);
}
else {
prevPtr = &gpv->next;
}
gpv = next;
}
}
break;
default:
_mesa_problem(NULL, "Unexpected program target 0x%x in "
"destroy_program_variants_cb()", program->Target);
}
}
/**
* Generate assembly for a Vec4 IR instruction.
*
* \param instruction The Vec4 IR instruction to generate code for.
* \param dst The destination register.
* \param src An array of up to three source registers.
*/
void
vec4_generator::generate_vec4_instruction(vec4_instruction *instruction,
struct brw_reg dst,
struct brw_reg *src)
{
vec4_instruction *inst = (vec4_instruction *) instruction;
if (dst.width == BRW_WIDTH_4) {
/* This happens in attribute fixups for "dual instanced" geometry
* shaders, since they use attributes that are vec4's. Since the exec
* width is only 4, it's essential that the caller set
* force_writemask_all in order to make sure the instruction is executed
* regardless of which channels are enabled.
*/
assert(inst->force_writemask_all);
/* Fix up any <8;8,1> or <0;4,1> source registers to <4;4,1> to satisfy
* the following register region restrictions (from Graphics BSpec:
* 3D-Media-GPGPU Engine > EU Overview > Registers and Register Regions
* > Register Region Restrictions)
*
* 1. ExecSize must be greater than or equal to Width.
*
* 2. If ExecSize = Width and HorzStride != 0, VertStride must be set
* to Width * HorzStride."
*/
for (int i = 0; i < 3; i++) {
if (src[i].file == BRW_GENERAL_REGISTER_FILE)
src[i] = stride(src[i], 4, 4, 1);
}
}
switch (inst->opcode) {
case BRW_OPCODE_MOV:
brw_MOV(p, dst, src[0]);
break;
case BRW_OPCODE_ADD:
brw_ADD(p, dst, src[0], src[1]);
break;
case BRW_OPCODE_MUL:
brw_MUL(p, dst, src[0], src[1]);
break;
case BRW_OPCODE_MACH:
brw_set_acc_write_control(p, 1);
brw_MACH(p, dst, src[0], src[1]);
brw_set_acc_write_control(p, 0);
break;
case BRW_OPCODE_MAD:
assert(brw->gen >= 6);
brw_MAD(p, dst, src[0], src[1], src[2]);
break;
case BRW_OPCODE_FRC:
brw_FRC(p, dst, src[0]);
break;
case BRW_OPCODE_RNDD:
brw_RNDD(p, dst, src[0]);
break;
case BRW_OPCODE_RNDE:
brw_RNDE(p, dst, src[0]);
break;
case BRW_OPCODE_RNDZ:
brw_RNDZ(p, dst, src[0]);
break;
case BRW_OPCODE_AND:
brw_AND(p, dst, src[0], src[1]);
break;
case BRW_OPCODE_OR:
brw_OR(p, dst, src[0], src[1]);
break;
case BRW_OPCODE_XOR:
brw_XOR(p, dst, src[0], src[1]);
break;
case BRW_OPCODE_NOT:
brw_NOT(p, dst, src[0]);
break;
case BRW_OPCODE_ASR:
brw_ASR(p, dst, src[0], src[1]);
break;
case BRW_OPCODE_SHR:
brw_SHR(p, dst, src[0], src[1]);
break;
case BRW_OPCODE_SHL:
brw_SHL(p, dst, src[0], src[1]);
break;
case BRW_OPCODE_CMP:
brw_CMP(p, dst, inst->conditional_mod, src[0], src[1]);
break;
case BRW_OPCODE_SEL:
brw_SEL(p, dst, src[0], src[1]);
break;
case BRW_OPCODE_DPH:
brw_DPH(p, dst, src[0], src[1]);
break;
case BRW_OPCODE_DP4:
brw_DP4(p, dst, src[0], src[1]);
break;
case BRW_OPCODE_DP3:
brw_DP3(p, dst, src[0], src[1]);
break;
case BRW_OPCODE_DP2:
brw_DP2(p, dst, src[0], src[1]);
break;
case BRW_OPCODE_F32TO16:
assert(brw->gen >= 7);
brw_F32TO16(p, dst, src[0]);
break;
case BRW_OPCODE_F16TO32:
assert(brw->gen >= 7);
brw_F16TO32(p, dst, src[0]);
break;
case BRW_OPCODE_LRP:
assert(brw->gen >= 6);
brw_LRP(p, dst, src[0], src[1], src[2]);
break;
case BRW_OPCODE_BFREV:
assert(brw->gen >= 7);
/* BFREV only supports UD type for src and dst. */
brw_BFREV(p, retype(dst, BRW_REGISTER_TYPE_UD),
retype(src[0], BRW_REGISTER_TYPE_UD));
break;
case BRW_OPCODE_FBH:
assert(brw->gen >= 7);
/* FBH only supports UD type for dst. */
brw_FBH(p, retype(dst, BRW_REGISTER_TYPE_UD), src[0]);
break;
case BRW_OPCODE_FBL:
assert(brw->gen >= 7);
/* FBL only supports UD type for dst. */
brw_FBL(p, retype(dst, BRW_REGISTER_TYPE_UD), src[0]);
break;
case BRW_OPCODE_CBIT:
assert(brw->gen >= 7);
/* CBIT only supports UD type for dst. */
brw_CBIT(p, retype(dst, BRW_REGISTER_TYPE_UD), src[0]);
break;
case BRW_OPCODE_ADDC:
assert(brw->gen >= 7);
brw_set_acc_write_control(p, 1);
brw_ADDC(p, dst, src[0], src[1]);
brw_set_acc_write_control(p, 0);
break;
case BRW_OPCODE_SUBB:
assert(brw->gen >= 7);
brw_set_acc_write_control(p, 1);
brw_SUBB(p, dst, src[0], src[1]);
brw_set_acc_write_control(p, 0);
break;
case BRW_OPCODE_BFE:
assert(brw->gen >= 7);
brw_BFE(p, dst, src[0], src[1], src[2]);
break;
case BRW_OPCODE_BFI1:
assert(brw->gen >= 7);
brw_BFI1(p, dst, src[0], src[1]);
break;
case BRW_OPCODE_BFI2:
assert(brw->gen >= 7);
brw_BFI2(p, dst, src[0], src[1], src[2]);
break;
case BRW_OPCODE_IF:
if (inst->src[0].file != BAD_FILE) {
/* The instruction has an embedded compare (only allowed on gen6) */
assert(brw->gen == 6);
gen6_IF(p, inst->conditional_mod, src[0], src[1]);
} else {
struct brw_instruction *brw_inst = brw_IF(p, BRW_EXECUTE_8);
brw_inst->header.predicate_control = inst->predicate;
}
break;
case BRW_OPCODE_ELSE:
brw_ELSE(p);
break;
case BRW_OPCODE_ENDIF:
brw_ENDIF(p);
break;
case BRW_OPCODE_DO:
brw_DO(p, BRW_EXECUTE_8);
break;
case BRW_OPCODE_BREAK:
brw_BREAK(p);
brw_set_predicate_control(p, BRW_PREDICATE_NONE);
break;
case BRW_OPCODE_CONTINUE:
/* FINISHME: We need to write the loop instruction support still. */
if (brw->gen >= 6)
gen6_CONT(p);
else
brw_CONT(p);
brw_set_predicate_control(p, BRW_PREDICATE_NONE);
break;
case BRW_OPCODE_WHILE:
brw_WHILE(p);
break;
case SHADER_OPCODE_RCP:
case SHADER_OPCODE_RSQ:
case SHADER_OPCODE_SQRT:
case SHADER_OPCODE_EXP2:
case SHADER_OPCODE_LOG2:
case SHADER_OPCODE_SIN:
case SHADER_OPCODE_COS:
if (brw->gen == 6) {
generate_math1_gen6(inst, dst, src[0]);
} else {
/* Also works for Gen7. */
generate_math1_gen4(inst, dst, src[0]);
}
break;
case SHADER_OPCODE_POW:
case SHADER_OPCODE_INT_QUOTIENT:
case SHADER_OPCODE_INT_REMAINDER:
if (brw->gen >= 7) {
generate_math2_gen7(inst, dst, src[0], src[1]);
} else if (brw->gen == 6) {
generate_math2_gen6(inst, dst, src[0], src[1]);
} else {
generate_math2_gen4(inst, dst, src[0], src[1]);
}
break;
case SHADER_OPCODE_TEX:
case SHADER_OPCODE_TXD:
case SHADER_OPCODE_TXF:
case SHADER_OPCODE_TXF_CMS:
case SHADER_OPCODE_TXF_MCS:
case SHADER_OPCODE_TXL:
case SHADER_OPCODE_TXS:
case SHADER_OPCODE_TG4:
case SHADER_OPCODE_TG4_OFFSET:
generate_tex(inst, dst, src[0]);
break;
case VS_OPCODE_URB_WRITE:
generate_vs_urb_write(inst);
break;
case SHADER_OPCODE_GEN4_SCRATCH_READ:
generate_scratch_read(inst, dst, src[0]);
break;
case SHADER_OPCODE_GEN4_SCRATCH_WRITE:
generate_scratch_write(inst, dst, src[0], src[1]);
break;
case VS_OPCODE_PULL_CONSTANT_LOAD:
generate_pull_constant_load(inst, dst, src[0], src[1]);
break;
case VS_OPCODE_PULL_CONSTANT_LOAD_GEN7:
generate_pull_constant_load_gen7(inst, dst, src[0], src[1]);
break;
case GS_OPCODE_URB_WRITE:
generate_gs_urb_write(inst);
break;
case GS_OPCODE_THREAD_END:
generate_gs_thread_end(inst);
break;
case GS_OPCODE_SET_WRITE_OFFSET:
generate_gs_set_write_offset(dst, src[0], src[1]);
break;
case GS_OPCODE_SET_VERTEX_COUNT:
generate_gs_set_vertex_count(dst, src[0]);
break;
case GS_OPCODE_SET_DWORD_2_IMMED:
generate_gs_set_dword_2_immed(dst, src[0]);
break;
case GS_OPCODE_PREPARE_CHANNEL_MASKS:
generate_gs_prepare_channel_masks(dst);
break;
case GS_OPCODE_SET_CHANNEL_MASKS:
generate_gs_set_channel_masks(dst, src[0]);
break;
case GS_OPCODE_GET_INSTANCE_ID:
generate_gs_get_instance_id(dst);
break;
case SHADER_OPCODE_SHADER_TIME_ADD:
brw_shader_time_add(p, src[0],
prog_data->base.binding_table.shader_time_start);
brw_mark_surface_used(&prog_data->base,
prog_data->base.binding_table.shader_time_start);
break;
case SHADER_OPCODE_UNTYPED_ATOMIC:
generate_untyped_atomic(inst, dst, src[0], src[1]);
break;
case SHADER_OPCODE_UNTYPED_SURFACE_READ:
generate_untyped_surface_read(inst, dst, src[0]);
break;
case VS_OPCODE_UNPACK_FLAGS_SIMD4X2:
generate_unpack_flags(inst, dst);
break;
default:
if (inst->opcode < (int) ARRAY_SIZE(opcode_descs)) {
_mesa_problem(&brw->ctx, "Unsupported opcode in `%s' in vec4\n",
opcode_descs[inst->opcode].name);
} else {
_mesa_problem(&brw->ctx, "Unsupported opcode %d in vec4", inst->opcode);
}
abort();
}
}
/**
* Prepare the source or destination resource, including:
* - Error checking
* - Creating texture wrappers for renderbuffers
* \param name the texture or renderbuffer name
* \param target GL_TEXTURE target or GL_RENDERBUFFER. For the later, will
* be changed to a compatible GL_TEXTURE target.
* \param level mipmap level
* \param tex_obj returns a pointer to a texture object
* \param tex_image returns a pointer to a texture image
* \param tmp_tex returns temporary texture object name
* \return true if success, false if error
*/
static bool
prepare_target(struct gl_context *ctx, GLuint name, GLenum *target, int level,
struct gl_texture_object **tex_obj,
struct gl_texture_image **tex_image, GLuint *tmp_tex,
GLuint *width,
GLuint *height,
const char *dbg_prefix)
{
if (name == 0) {
_mesa_error(ctx, GL_INVALID_VALUE,
"glCopyImageSubData(%sName = %d)", dbg_prefix, name);
return false;
}
/*
* INVALID_ENUM is generated
* * if either <srcTarget> or <dstTarget>
* - is not RENDERBUFFER or a valid non-proxy texture target
* - is TEXTURE_BUFFER, or
* - is one of the cubemap face selectors described in table 3.17,
*/
switch (*target) {
case GL_RENDERBUFFER:
/* Not a texture target, but valid */
case GL_TEXTURE_1D:
case GL_TEXTURE_1D_ARRAY:
case GL_TEXTURE_2D:
case GL_TEXTURE_3D:
case GL_TEXTURE_CUBE_MAP:
case GL_TEXTURE_RECTANGLE:
case GL_TEXTURE_2D_ARRAY:
case GL_TEXTURE_CUBE_MAP_ARRAY:
case GL_TEXTURE_2D_MULTISAMPLE:
case GL_TEXTURE_2D_MULTISAMPLE_ARRAY:
/* These are all valid */
break;
case GL_TEXTURE_EXTERNAL_OES:
/* Only exists in ES */
case GL_TEXTURE_BUFFER:
default:
_mesa_error(ctx, GL_INVALID_ENUM,
"glCopyImageSubData(%sTarget = %s)", dbg_prefix,
_mesa_enum_to_string(*target));
return false;
}
if (*target == GL_RENDERBUFFER) {
struct gl_renderbuffer *rb = _mesa_lookup_renderbuffer(ctx, name);
if (!rb) {
_mesa_error(ctx, GL_INVALID_VALUE,
"glCopyImageSubData(%sName = %u)", dbg_prefix, name);
return false;
}
if (!rb->Name) {
_mesa_error(ctx, GL_INVALID_OPERATION,
"glCopyImageSubData(%sName incomplete)", dbg_prefix);
return false;
}
if (level != 0) {
_mesa_error(ctx, GL_INVALID_VALUE,
"glCopyImageSubData(%sLevel = %u)", dbg_prefix, level);
return false;
}
if (rb->NumSamples > 1)
*target = GL_TEXTURE_2D_MULTISAMPLE;
else
*target = GL_TEXTURE_2D;
*tmp_tex = 0;
_mesa_GenTextures(1, tmp_tex);
if (*tmp_tex == 0)
return false; /* Error already set by GenTextures */
_mesa_BindTexture(*target, *tmp_tex);
*tex_obj = _mesa_lookup_texture(ctx, *tmp_tex);
*tex_image = _mesa_get_tex_image(ctx, *tex_obj, *target, 0);
*width = rb->Width;
*height = rb->Height;
if (!ctx->Driver.BindRenderbufferTexImage(ctx, rb, *tex_image)) {
_mesa_problem(ctx, "Failed to create texture from renderbuffer");
return false;
}
if (ctx->Driver.FinishRenderTexture && !rb->NeedsFinishRenderTexture) {
rb->NeedsFinishRenderTexture = true;
ctx->Driver.FinishRenderTexture(ctx, rb);
}
} else {
*tex_obj = _mesa_lookup_texture(ctx, name);
if (!*tex_obj) {
_mesa_error(ctx, GL_INVALID_VALUE,
"glCopyImageSubData(%sName = %u)", dbg_prefix, name);
return false;
}
_mesa_test_texobj_completeness(ctx, *tex_obj);
if (!(*tex_obj)->_BaseComplete ||
(level != 0 && !(*tex_obj)->_MipmapComplete)) {
_mesa_error(ctx, GL_INVALID_OPERATION,
"glCopyImageSubData(%sName incomplete)", dbg_prefix);
return false;
}
if ((*tex_obj)->Target != *target) {
_mesa_error(ctx, GL_INVALID_ENUM,
"glCopyImageSubData(%sTarget = %s)", dbg_prefix,
_mesa_enum_to_string(*target));
return false;
}
if (level < 0 || level >= MAX_TEXTURE_LEVELS) {
_mesa_error(ctx, GL_INVALID_VALUE,
"glCopyImageSubData(%sLevel = %d)", dbg_prefix, level);
return false;
}
*tex_image = _mesa_select_tex_image(*tex_obj, *target, level);
if (!*tex_image) {
_mesa_error(ctx, GL_INVALID_VALUE,
"glCopyImageSubData(%sLevel = %u)", dbg_prefix, level);
return false;
}
*width = (*tex_image)->Width;
*height = (*tex_image)->Height;
}
return true;
}
static void ctx_emit_cs(GLcontext *ctx, struct radeon_state_atom *atom)
{
r100ContextPtr r100 = R100_CONTEXT(ctx);
BATCH_LOCALS(&r100->radeon);
struct radeon_renderbuffer *rrb, *drb;
uint32_t cbpitch = 0;
uint32_t zbpitch = 0;
uint32_t dwords = atom->check(ctx, atom);
uint32_t depth_fmt;
rrb = radeon_get_colorbuffer(&r100->radeon);
if (!rrb || !rrb->bo) {
fprintf(stderr, "no rrb\n");
return;
}
atom->cmd[CTX_RB3D_CNTL] &= ~(0xf << 10);
if (rrb->cpp == 4)
atom->cmd[CTX_RB3D_CNTL] |= RADEON_COLOR_FORMAT_ARGB8888;
else switch (rrb->base.Format) {
case MESA_FORMAT_RGB565:
atom->cmd[CTX_RB3D_CNTL] |= RADEON_COLOR_FORMAT_RGB565;
break;
case MESA_FORMAT_ARGB4444:
atom->cmd[CTX_RB3D_CNTL] |= RADEON_COLOR_FORMAT_ARGB4444;
break;
case MESA_FORMAT_ARGB1555:
atom->cmd[CTX_RB3D_CNTL] |= RADEON_COLOR_FORMAT_ARGB1555;
break;
default:
_mesa_problem(ctx, "unexpected format in ctx_emit_cs()");
}
cbpitch = (rrb->pitch / rrb->cpp);
if (rrb->bo->flags & RADEON_BO_FLAGS_MACRO_TILE)
cbpitch |= R200_COLOR_TILE_ENABLE;
drb = radeon_get_depthbuffer(&r100->radeon);
if (drb) {
zbpitch = (drb->pitch / drb->cpp);
if (drb->cpp == 4)
depth_fmt = RADEON_DEPTH_FORMAT_24BIT_INT_Z;
else
depth_fmt = RADEON_DEPTH_FORMAT_16BIT_INT_Z;
atom->cmd[CTX_RB3D_ZSTENCILCNTL] &= ~RADEON_DEPTH_FORMAT_MASK;
atom->cmd[CTX_RB3D_ZSTENCILCNTL] |= depth_fmt;
}
BEGIN_BATCH_NO_AUTOSTATE(dwords);
/* In the CS case we need to split this up */
OUT_BATCH(CP_PACKET0(packet[0].start, 3));
OUT_BATCH_TABLE((atom->cmd + 1), 4);
if (drb) {
OUT_BATCH(CP_PACKET0(RADEON_RB3D_DEPTHOFFSET, 0));
OUT_BATCH_RELOC(0, drb->bo, 0, 0, RADEON_GEM_DOMAIN_VRAM, 0);
OUT_BATCH(CP_PACKET0(RADEON_RB3D_DEPTHPITCH, 0));
OUT_BATCH(zbpitch);
}
OUT_BATCH(CP_PACKET0(RADEON_RB3D_ZSTENCILCNTL, 0));
OUT_BATCH(atom->cmd[CTX_RB3D_ZSTENCILCNTL]);
OUT_BATCH(CP_PACKET0(RADEON_PP_CNTL, 1));
OUT_BATCH(atom->cmd[CTX_PP_CNTL]);
OUT_BATCH(atom->cmd[CTX_RB3D_CNTL]);
if (rrb) {
OUT_BATCH(CP_PACKET0(RADEON_RB3D_COLOROFFSET, 0));
OUT_BATCH_RELOC(0, rrb->bo, 0, 0, RADEON_GEM_DOMAIN_VRAM, 0);
OUT_BATCH(CP_PACKET0(RADEON_RB3D_COLORPITCH, 0));
OUT_BATCH_RELOC(cbpitch, rrb->bo, cbpitch, 0, RADEON_GEM_DOMAIN_VRAM, 0);
}
// if (atom->cmd_size == CTX_STATE_SIZE_NEWDRM) {
// OUT_BATCH_TABLE((atom->cmd + 14), 4);
// }
END_BATCH();
BEGIN_BATCH_NO_AUTOSTATE(4);
OUT_BATCH(CP_PACKET0(RADEON_RE_TOP_LEFT, 0));
OUT_BATCH(0);
OUT_BATCH(CP_PACKET0(RADEON_RE_WIDTH_HEIGHT, 0));
if (rrb) {
OUT_BATCH(((rrb->base.Width - 1) << RADEON_RE_WIDTH_SHIFT) |
((rrb->base.Height - 1) << RADEON_RE_HEIGHT_SHIFT));
} else {
OUT_BATCH(0);
}
END_BATCH();
}
static const struct gl_texture_format *
mgaChooseTextureFormat( GLcontext *ctx, GLint internalFormat,
GLenum format, GLenum type )
{
mgaContextPtr mmesa = MGA_CONTEXT(ctx);
const GLboolean do32bpt = mmesa->default32BitTextures;
switch ( internalFormat ) {
case 4:
case GL_RGBA:
case GL_COMPRESSED_RGBA:
if ( format == GL_BGRA ) {
if ( type == GL_UNSIGNED_INT_8_8_8_8_REV ) {
return &_mesa_texformat_argb8888;
}
else if ( type == GL_UNSIGNED_SHORT_4_4_4_4_REV ) {
return &_mesa_texformat_argb4444;
}
else if ( type == GL_UNSIGNED_SHORT_1_5_5_5_REV ) {
return &_mesa_texformat_argb1555;
}
}
return do32bpt ? &_mesa_texformat_argb8888 : &_mesa_texformat_argb4444;
case 3:
case GL_RGB:
case GL_COMPRESSED_RGB:
if ( format == GL_RGB && type == GL_UNSIGNED_SHORT_5_6_5 ) {
return &_mesa_texformat_rgb565;
}
return do32bpt ? &_mesa_texformat_argb8888 : &_mesa_texformat_rgb565;
case GL_RGBA8:
case GL_RGB10_A2:
case GL_RGBA12:
case GL_RGBA16:
return do32bpt ? &_mesa_texformat_argb8888 : &_mesa_texformat_argb4444;
case GL_RGBA4:
case GL_RGBA2:
return &_mesa_texformat_argb4444;
case GL_RGB5_A1:
return &_mesa_texformat_argb1555;
case GL_RGB8:
case GL_RGB10:
case GL_RGB12:
case GL_RGB16:
return do32bpt ? &_mesa_texformat_argb8888 : &_mesa_texformat_rgb565;
case GL_RGB5:
case GL_RGB4:
case GL_R3_G3_B2:
return &_mesa_texformat_rgb565;
case GL_ALPHA:
case GL_ALPHA4:
case GL_ALPHA8:
case GL_ALPHA12:
case GL_ALPHA16:
case GL_COMPRESSED_ALPHA:
/* FIXME: This will report incorrect component sizes... */
return MGA_IS_G400(mmesa) ? &_mesa_texformat_al88 : &_mesa_texformat_argb4444;
case 1:
case GL_LUMINANCE:
case GL_LUMINANCE4:
case GL_LUMINANCE8:
case GL_LUMINANCE12:
case GL_LUMINANCE16:
case GL_COMPRESSED_LUMINANCE:
/* FIXME: This will report incorrect component sizes... */
return MGA_IS_G400(mmesa) ? &_mesa_texformat_al88 : &_mesa_texformat_rgb565;
case 2:
case GL_LUMINANCE_ALPHA:
case GL_LUMINANCE4_ALPHA4:
case GL_LUMINANCE6_ALPHA2:
case GL_LUMINANCE8_ALPHA8:
case GL_LUMINANCE12_ALPHA4:
case GL_LUMINANCE12_ALPHA12:
case GL_LUMINANCE16_ALPHA16:
case GL_COMPRESSED_LUMINANCE_ALPHA:
/* FIXME: This will report incorrect component sizes... */
return MGA_IS_G400(mmesa) ? &_mesa_texformat_al88 : &_mesa_texformat_argb4444;
case GL_INTENSITY:
case GL_INTENSITY4:
case GL_INTENSITY8:
case GL_INTENSITY12:
case GL_INTENSITY16:
case GL_COMPRESSED_INTENSITY:
/* FIXME: This will report incorrect component sizes... */
return MGA_IS_G400(mmesa) ? &_mesa_texformat_i8 : &_mesa_texformat_argb4444;
case GL_YCBCR_MESA:
if (MGA_IS_G400(mmesa) &&
(type == GL_UNSIGNED_SHORT_8_8_APPLE ||
type == GL_UNSIGNED_BYTE))
return &_mesa_texformat_ycbcr;
else
return &_mesa_texformat_ycbcr_rev;
case GL_COLOR_INDEX:
case GL_COLOR_INDEX1_EXT:
case GL_COLOR_INDEX2_EXT:
case GL_COLOR_INDEX4_EXT:
case GL_COLOR_INDEX8_EXT:
case GL_COLOR_INDEX12_EXT:
case GL_COLOR_INDEX16_EXT:
return &_mesa_texformat_ci8;
default:
_mesa_problem( ctx, "unexpected texture format in %s", __FUNCTION__ );
return NULL;
}
return NULL; /* never get here */
}
/**
* Examine texture unit's combine/env state to update derived state.
*/
static void
update_tex_combine(struct gl_context *ctx,
struct gl_texture_unit *texUnit,
struct gl_fixedfunc_texture_unit *fftexUnit)
{
struct gl_tex_env_combine_state *combine;
/* No combiners will apply to this. */
if (texUnit->_Current->Target == GL_TEXTURE_BUFFER)
return;
/* Set the texUnit->_CurrentCombine field to point to the user's combiner
* state, or the combiner state which is derived from traditional texenv
* mode.
*/
if (fftexUnit->EnvMode == GL_COMBINE ||
fftexUnit->EnvMode == GL_COMBINE4_NV) {
fftexUnit->_CurrentCombine = & fftexUnit->Combine;
}
else {
const struct gl_texture_object *texObj = texUnit->_Current;
GLenum format = texObj->Image[0][texObj->BaseLevel]->_BaseFormat;
if (format == GL_DEPTH_COMPONENT || format == GL_DEPTH_STENCIL_EXT) {
format = texObj->DepthMode;
}
calculate_derived_texenv(&fftexUnit->_EnvMode, fftexUnit->EnvMode, format);
fftexUnit->_CurrentCombine = & fftexUnit->_EnvMode;
}
combine = fftexUnit->_CurrentCombine;
/* Determine number of source RGB terms in the combiner function */
switch (combine->ModeRGB) {
case GL_REPLACE:
combine->_NumArgsRGB = 1;
break;
case GL_ADD:
case GL_ADD_SIGNED:
if (fftexUnit->EnvMode == GL_COMBINE4_NV)
combine->_NumArgsRGB = 4;
else
combine->_NumArgsRGB = 2;
break;
case GL_MODULATE:
case GL_SUBTRACT:
case GL_DOT3_RGB:
case GL_DOT3_RGBA:
case GL_DOT3_RGB_EXT:
case GL_DOT3_RGBA_EXT:
combine->_NumArgsRGB = 2;
break;
case GL_INTERPOLATE:
case GL_MODULATE_ADD_ATI:
case GL_MODULATE_SIGNED_ADD_ATI:
case GL_MODULATE_SUBTRACT_ATI:
combine->_NumArgsRGB = 3;
break;
default:
combine->_NumArgsRGB = 0;
_mesa_problem(ctx, "invalid RGB combine mode in update_texture_state");
return;
}
/* Determine number of source Alpha terms in the combiner function */
switch (combine->ModeA) {
case GL_REPLACE:
combine->_NumArgsA = 1;
break;
case GL_ADD:
case GL_ADD_SIGNED:
if (fftexUnit->EnvMode == GL_COMBINE4_NV)
combine->_NumArgsA = 4;
else
combine->_NumArgsA = 2;
break;
case GL_MODULATE:
case GL_SUBTRACT:
combine->_NumArgsA = 2;
break;
case GL_INTERPOLATE:
case GL_MODULATE_ADD_ATI:
case GL_MODULATE_SIGNED_ADD_ATI:
case GL_MODULATE_SUBTRACT_ATI:
combine->_NumArgsA = 3;
break;
default:
combine->_NumArgsA = 0;
_mesa_problem(ctx, "invalid Alpha combine mode in update_texture_state");
break;
}
pack_tex_combine(fftexUnit);
}
/**
* As above, but color index mode.
*/
void
_swrast_fog_ci_span( const GLcontext *ctx, struct sw_span *span )
{
const SWcontext *swrast = SWRAST_CONTEXT(ctx);
const GLuint haveW = (span->interpMask & SPAN_W);
const GLuint fogIndex = (GLuint) ctx->Fog.Index;
GLuint *index = span->array->index;
ASSERT(swrast->_FogEnabled);
ASSERT(span->arrayMask & SPAN_INDEX);
ASSERT((span->interpMask | span->arrayMask) & SPAN_FOG);
/* we need to compute fog blend factors */
if (swrast->_PreferPixelFog) {
/* The span's fog values are fog coordinates, now compute blend factors
* and blend the fragment colors with the fog color.
*/
switch (ctx->Fog.Mode) {
case GL_LINEAR:
{
const GLfloat fogEnd = ctx->Fog.End;
const GLfloat fogScale = (ctx->Fog.Start == ctx->Fog.End)
? 1.0F : 1.0F / (ctx->Fog.End - ctx->Fog.Start);
const GLfloat fogStep = span->fogStep;
GLfloat fogCoord = span->fog;
const GLfloat wStep = haveW ? span->dwdx : 0.0F;
GLfloat w = haveW ? span->w : 1.0F;
GLuint i;
for (i = 0; i < span->end; i++) {
GLfloat f = (fogEnd - fogCoord / w) * fogScale;
f = CLAMP(f, 0.0F, 1.0F);
index[i] = (GLuint) ((GLfloat) index[i] + (1.0F - f) * fogIndex);
fogCoord += fogStep;
w += wStep;
}
}
break;
case GL_EXP:
{
const GLfloat density = -ctx->Fog.Density;
const GLfloat fogStep = span->fogStep;
GLfloat fogCoord = span->fog;
const GLfloat wStep = haveW ? span->dwdx : 0.0F;
GLfloat w = haveW ? span->w : 1.0F;
GLuint i;
for (i = 0; i < span->end; i++) {
GLfloat f = (GLfloat) exp(density * fogCoord / w);
f = CLAMP(f, 0.0F, 1.0F);
index[i] = (GLuint) ((GLfloat) index[i] + (1.0F - f) * fogIndex);
fogCoord += fogStep;
w += wStep;
}
}
break;
case GL_EXP2:
{
const GLfloat negDensitySquared = -ctx->Fog.Density * ctx->Fog.Density;
const GLfloat fogStep = span->fogStep;
GLfloat fogCoord = span->fog;
const GLfloat wStep = haveW ? span->dwdx : 0.0F;
GLfloat w = haveW ? span->w : 1.0F;
GLuint i;
for (i = 0; i < span->end; i++) {
const GLfloat coord = fogCoord / w;
GLfloat tmp = negDensitySquared * coord * coord;
GLfloat f;
#if defined(__alpha__) || defined(__alpha)
/* XXX this underflow check may be needed for other systems*/
if (tmp < FLT_MIN_10_EXP)
tmp = FLT_MIN_10_EXP;
#endif
f = (GLfloat) exp(tmp);
f = CLAMP(f, 0.0F, 1.0F);
index[i] = (GLuint) ((GLfloat) index[i] + (1.0F - f) * fogIndex);
fogCoord += fogStep;
w += wStep;
}
}
break;
default:
_mesa_problem(ctx, "Bad fog mode in _swrast_fog_ci_span");
return;
}
}
else if (span->arrayMask & SPAN_FOG) {
/* The span's fog array values are blend factors.
* They were previously computed per-vertex.
*/
GLuint i;
for (i = 0; i < span->end; i++) {
const GLfloat f = span->array->fog[i];
index[i] = (GLuint) ((GLfloat) index[i] + (1.0F - f) * fogIndex);
}
}
else {
/* The span's fog start/step values are blend factors.
* They were previously computed per-vertex.
*/
const GLfloat fogStep = span->fogStep;
GLfloat fog = span->fog;
const GLfloat wStep = haveW ? span->dwdx : 0.0F;
GLfloat w = haveW ? span->w : 1.0F;
GLuint i;
ASSERT(span->interpMask & SPAN_FOG);
for (i = 0; i < span->end; i++) {
const GLfloat f = fog / w;
index[i] = (GLuint) ((GLfloat) index[i] + (1.0F - f) * fogIndex);
fog += fogStep;
w += wStep;
}
}
}
/**
* Apply stencil test to an array of stencil values (before depth buffering).
* Input: face - 0 or 1 for front or back-face polygons
* n - number of pixels in the array
* stencil - array of [n] stencil values
* mask - array [n] of flag: 0=skip the pixel, 1=stencil the pixel
* Output: mask - pixels which fail the stencil test will have their
* mask flag set to 0.
* stencil - updated stencil values (where the test passed)
* Return: GL_FALSE = all pixels failed, GL_TRUE = zero or more pixels passed.
*/
static GLboolean
do_stencil_test( GLcontext *ctx, GLuint face, GLuint n, GLstencil stencil[],
GLubyte mask[] )
{
GLubyte fail[MAX_WIDTH];
GLboolean allfail = GL_FALSE;
GLuint i;
GLstencil r, s;
const GLuint valueMask = ctx->Stencil.ValueMask[face];
ASSERT(n <= MAX_WIDTH);
/*
* Perform stencil test. The results of this operation are stored
* in the fail[] array:
* IF fail[i] is non-zero THEN
* the stencil fail operator is to be applied
* ELSE
* the stencil fail operator is not to be applied
* ENDIF
*/
switch (ctx->Stencil.Function[face]) {
case GL_NEVER:
/* never pass; always fail */
for (i=0;i<n;i++) {
if (mask[i]) {
mask[i] = 0;
fail[i] = 1;
}
else {
fail[i] = 0;
}
}
allfail = GL_TRUE;
break;
case GL_LESS:
r = (GLstencil) (ctx->Stencil.Ref[face] & valueMask);
for (i=0;i<n;i++) {
if (mask[i]) {
s = (GLstencil) (stencil[i] & valueMask);
if (r < s) {
/* passed */
fail[i] = 0;
}
else {
fail[i] = 1;
mask[i] = 0;
}
}
else {
fail[i] = 0;
}
}
break;
case GL_LEQUAL:
r = (GLstencil) (ctx->Stencil.Ref[face] & valueMask);
for (i=0;i<n;i++) {
if (mask[i]) {
s = (GLstencil) (stencil[i] & valueMask);
if (r <= s) {
/* pass */
fail[i] = 0;
}
else {
fail[i] = 1;
mask[i] = 0;
}
}
else {
fail[i] = 0;
}
}
break;
case GL_GREATER:
r = (GLstencil) (ctx->Stencil.Ref[face] & valueMask);
for (i=0;i<n;i++) {
if (mask[i]) {
s = (GLstencil) (stencil[i] & valueMask);
if (r > s) {
/* passed */
fail[i] = 0;
}
else {
fail[i] = 1;
mask[i] = 0;
}
}
else {
fail[i] = 0;
}
}
break;
case GL_GEQUAL:
r = (GLstencil) (ctx->Stencil.Ref[face] & valueMask);
for (i=0;i<n;i++) {
if (mask[i]) {
s = (GLstencil) (stencil[i] & valueMask);
if (r >= s) {
/* passed */
fail[i] = 0;
}
else {
fail[i] = 1;
mask[i] = 0;
}
}
else {
fail[i] = 0;
}
}
break;
case GL_EQUAL:
r = (GLstencil) (ctx->Stencil.Ref[face] & valueMask);
for (i=0;i<n;i++) {
if (mask[i]) {
s = (GLstencil) (stencil[i] & valueMask);
if (r == s) {
/* passed */
fail[i] = 0;
}
else {
fail[i] = 1;
mask[i] = 0;
}
}
else {
fail[i] = 0;
}
}
break;
case GL_NOTEQUAL:
r = (GLstencil) (ctx->Stencil.Ref[face] & valueMask);
for (i=0;i<n;i++) {
if (mask[i]) {
s = (GLstencil) (stencil[i] & valueMask);
if (r != s) {
/* passed */
fail[i] = 0;
}
else {
fail[i] = 1;
mask[i] = 0;
}
}
else {
fail[i] = 0;
}
}
break;
case GL_ALWAYS:
/* always pass */
for (i=0;i<n;i++) {
fail[i] = 0;
}
break;
default:
_mesa_problem(ctx, "Bad stencil func in gl_stencil_span");
return 0;
}
if (ctx->Stencil.FailFunc[face] != GL_KEEP) {
apply_stencil_op( ctx, ctx->Stencil.FailFunc[face], face, n, stencil, fail );
}
return !allfail;
}
/**
* Apply fog to a span of RGBA pixels.
* The fog value are either in the span->array->fog array or interpolated from
* the fog/fogStep values.
* They fog values are either fog coordinates (Z) or fog blend factors.
* _PreferPixelFog should be in sync with that state!
*/
void
_swrast_fog_rgba_span( const GLcontext *ctx, struct sw_span *span )
{
const SWcontext *swrast = SWRAST_CONTEXT(ctx);
const GLchan rFog = swrast->_FogColor[RCOMP];
const GLchan gFog = swrast->_FogColor[GCOMP];
const GLchan bFog = swrast->_FogColor[BCOMP];
const GLuint haveW = (span->interpMask & SPAN_W);
GLchan (*rgba)[4] = (GLchan (*)[4]) span->array->rgba;
ASSERT(swrast->_FogEnabled);
ASSERT((span->interpMask | span->arrayMask) & SPAN_FOG);
ASSERT(span->arrayMask & SPAN_RGBA);
/* NOTE: if haveW is true, that means the fog start/step values are
* perspective-corrected and we have to divide each fog coord by W.
*/
/* we need to compute fog blend factors */
if (swrast->_PreferPixelFog) {
/* The span's fog values are fog coordinates, now compute blend factors
* and blend the fragment colors with the fog color.
*/
switch (swrast->_FogMode) {
case GL_LINEAR:
{
const GLfloat fogEnd = ctx->Fog.End;
const GLfloat fogScale = (ctx->Fog.Start == ctx->Fog.End)
? 1.0F : 1.0F / (ctx->Fog.End - ctx->Fog.Start);
const GLfloat fogStep = span->fogStep;
GLfloat fogCoord = span->fog;
const GLfloat wStep = haveW ? span->dwdx : 0.0F;
GLfloat w = haveW ? span->w : 1.0F;
GLuint i;
for (i = 0; i < span->end; i++) {
GLfloat f, oneMinusF;
f = (fogEnd - fogCoord / w) * fogScale;
f = CLAMP(f, 0.0F, 1.0F);
oneMinusF = 1.0F - f;
rgba[i][RCOMP] = (GLchan) (f * rgba[i][RCOMP] + oneMinusF * rFog);
rgba[i][GCOMP] = (GLchan) (f * rgba[i][GCOMP] + oneMinusF * gFog);
rgba[i][BCOMP] = (GLchan) (f * rgba[i][BCOMP] + oneMinusF * bFog);
fogCoord += fogStep;
w += wStep;
}
}
break;
case GL_EXP:
{
const GLfloat density = -ctx->Fog.Density;
const GLfloat fogStep = span->fogStep;
GLfloat fogCoord = span->fog;
const GLfloat wStep = haveW ? span->dwdx : 0.0F;
GLfloat w = haveW ? span->w : 1.0F;
GLuint i;
for (i = 0; i < span->end; i++) {
GLfloat f, oneMinusF;
f = (GLfloat) exp(density * fogCoord / w);
f = CLAMP(f, 0.0F, 1.0F);
oneMinusF = 1.0F - f;
rgba[i][RCOMP] = (GLchan) (f * rgba[i][RCOMP] + oneMinusF * rFog);
rgba[i][GCOMP] = (GLchan) (f * rgba[i][GCOMP] + oneMinusF * gFog);
rgba[i][BCOMP] = (GLchan) (f * rgba[i][BCOMP] + oneMinusF * bFog);
fogCoord += fogStep;
w += wStep;
}
}
break;
case GL_EXP2:
{
const GLfloat negDensitySquared = -ctx->Fog.Density * ctx->Fog.Density;
const GLfloat fogStep = span->fogStep;
GLfloat fogCoord = span->fog;
const GLfloat wStep = haveW ? span->dwdx : 0.0F;
GLfloat w = haveW ? span->w : 1.0F;
GLuint i;
for (i = 0; i < span->end; i++) {
const GLfloat coord = fogCoord / w;
GLfloat tmp = negDensitySquared * coord * coord;
GLfloat f, oneMinusF;
#if defined(__alpha__) || defined(__alpha)
/* XXX this underflow check may be needed for other systems*/
if (tmp < FLT_MIN_10_EXP)
tmp = FLT_MIN_10_EXP;
#endif
f = (GLfloat) exp(tmp);
f = CLAMP(f, 0.0F, 1.0F);
oneMinusF = 1.0F - f;
rgba[i][RCOMP] = (GLchan) (f * rgba[i][RCOMP] + oneMinusF * rFog);
rgba[i][GCOMP] = (GLchan) (f * rgba[i][GCOMP] + oneMinusF * gFog);
rgba[i][BCOMP] = (GLchan) (f * rgba[i][BCOMP] + oneMinusF * bFog);
fogCoord += fogStep;
w += wStep;
}
}
break;
default:
_mesa_problem(ctx, "Bad fog mode in _swrast_fog_rgba_span");
return;
}
}
else if (span->arrayMask & SPAN_FOG) {
/* The span's fog array values are blend factors.
* They were previously computed per-vertex.
*/
GLuint i;
for (i = 0; i < span->end; i++) {
const GLfloat f = span->array->fog[i];
const GLfloat oneMinusF = 1.0F - f;
rgba[i][RCOMP] = (GLchan) (f * rgba[i][RCOMP] + oneMinusF * rFog);
rgba[i][GCOMP] = (GLchan) (f * rgba[i][GCOMP] + oneMinusF * gFog);
rgba[i][BCOMP] = (GLchan) (f * rgba[i][BCOMP] + oneMinusF * bFog);
}
}
else {
/* The span's fog start/step values are blend factors.
* They were previously computed per-vertex.
*/
const GLfloat fogStep = span->fogStep;
GLfloat fog = span->fog;
const GLfloat wStep = haveW ? span->dwdx : 0.0F;
GLfloat w = haveW ? span->w : 1.0F;
GLuint i;
ASSERT(span->interpMask & SPAN_FOG);
for (i = 0; i < span->end; i++) {
const GLfloat fact = fog / w;
const GLfloat oneMinusF = 1.0F - fact;
rgba[i][RCOMP] = (GLchan) (fact * rgba[i][RCOMP] + oneMinusF * rFog);
rgba[i][GCOMP] = (GLchan) (fact * rgba[i][GCOMP] + oneMinusF * gFog);
rgba[i][BCOMP] = (GLchan) (fact * rgba[i][BCOMP] + oneMinusF * bFog);
fog += fogStep;
w += wStep;
}
}
}
/**
* Apply stencil test to an array of pixels before depth buffering.
*
* \note Used for software stencil buffer only.
* Input: n - number of pixels in the span
* x, y - array of [n] pixels to stencil
* mask - array [n] of flag: 0=skip the pixel, 1=stencil the pixel
* Output: mask - pixels which fail the stencil test will have their
* mask flag set to 0.
* \return GL_FALSE = all pixels failed, GL_TRUE = zero or more pixels passed.
*/
static GLboolean
stencil_test_pixels( GLcontext *ctx, GLuint face, GLuint n,
const GLint x[], const GLint y[], GLubyte mask[] )
{
GLubyte fail[MAX_WIDTH];
GLstencil r, s;
GLuint i;
GLboolean allfail = GL_FALSE;
const GLuint valueMask = ctx->Stencil.ValueMask[face];
/* software stencil buffer only! */
ASSERT(ctx->DrawBuffer->UseSoftwareStencilBuffer);
ASSERT(!SWRAST_CONTEXT(ctx)->Driver.ReadStencilSpan);
ASSERT(!SWRAST_CONTEXT(ctx)->Driver.WriteStencilSpan);
/*
* Perform stencil test. The results of this operation are stored
* in the fail[] array:
* IF fail[i] is non-zero THEN
* the stencil fail operator is to be applied
* ELSE
* the stencil fail operator is not to be applied
* ENDIF
*/
switch (ctx->Stencil.Function[face]) {
case GL_NEVER:
/* always fail */
for (i=0;i<n;i++) {
if (mask[i]) {
mask[i] = 0;
fail[i] = 1;
}
else {
fail[i] = 0;
}
}
allfail = GL_TRUE;
break;
case GL_LESS:
r = (GLstencil) (ctx->Stencil.Ref[face] & valueMask);
for (i=0;i<n;i++) {
if (mask[i]) {
GLstencil *sptr = STENCIL_ADDRESS(x[i],y[i]);
s = (GLstencil) (*sptr & valueMask);
if (r < s) {
/* passed */
fail[i] = 0;
}
else {
fail[i] = 1;
mask[i] = 0;
}
}
else {
fail[i] = 0;
}
}
break;
case GL_LEQUAL:
r = (GLstencil) (ctx->Stencil.Ref[face] & valueMask);
for (i=0;i<n;i++) {
if (mask[i]) {
GLstencil *sptr = STENCIL_ADDRESS(x[i],y[i]);
s = (GLstencil) (*sptr & valueMask);
if (r <= s) {
/* pass */
fail[i] = 0;
}
else {
fail[i] = 1;
mask[i] = 0;
}
}
else {
fail[i] = 0;
}
}
break;
case GL_GREATER:
r = (GLstencil) (ctx->Stencil.Ref[face] & valueMask);
for (i=0;i<n;i++) {
if (mask[i]) {
GLstencil *sptr = STENCIL_ADDRESS(x[i],y[i]);
s = (GLstencil) (*sptr & valueMask);
if (r > s) {
/* passed */
fail[i] = 0;
}
else {
fail[i] = 1;
mask[i] = 0;
}
}
else {
fail[i] = 0;
}
}
break;
case GL_GEQUAL:
r = (GLstencil) (ctx->Stencil.Ref[face] & valueMask);
for (i=0;i<n;i++) {
if (mask[i]) {
GLstencil *sptr = STENCIL_ADDRESS(x[i],y[i]);
s = (GLstencil) (*sptr & valueMask);
if (r >= s) {
/* passed */
fail[i] = 0;
}
else {
fail[i] = 1;
mask[i] = 0;
}
}
else {
fail[i] = 0;
}
}
break;
case GL_EQUAL:
r = (GLstencil) (ctx->Stencil.Ref[face] & valueMask);
for (i=0;i<n;i++) {
if (mask[i]) {
GLstencil *sptr = STENCIL_ADDRESS(x[i],y[i]);
s = (GLstencil) (*sptr & valueMask);
if (r == s) {
/* passed */
fail[i] = 0;
}
else {
fail[i] = 1;
mask[i] = 0;
}
}
else {
fail[i] = 0;
}
}
break;
case GL_NOTEQUAL:
r = (GLstencil) (ctx->Stencil.Ref[face] & valueMask);
for (i=0;i<n;i++) {
if (mask[i]) {
GLstencil *sptr = STENCIL_ADDRESS(x[i],y[i]);
s = (GLstencil) (*sptr & valueMask);
if (r != s) {
/* passed */
fail[i] = 0;
}
else {
fail[i] = 1;
mask[i] = 0;
}
}
else {
fail[i] = 0;
}
}
break;
case GL_ALWAYS:
/* always pass */
for (i=0;i<n;i++) {
fail[i] = 0;
}
break;
default:
_mesa_problem(ctx, "Bad stencil func in gl_stencil_pixels");
return 0;
}
if (ctx->Stencil.FailFunc[face] != GL_KEEP) {
apply_stencil_op_to_pixels( ctx, n, x, y, ctx->Stencil.FailFunc[face],
face, fail );
}
return !allfail;
}
static void
setup_glsl_msaa_blit_shader(struct gl_context *ctx,
struct blit_state *blit,
const struct gl_framebuffer *drawFb,
struct gl_renderbuffer *src_rb,
GLenum target)
{
const char *vs_source;
char *fs_source;
void *mem_ctx;
enum blit_msaa_shader shader_index;
bool dst_is_msaa = false;
GLenum src_datatype;
const char *vec4_prefix;
const char *sampler_array_suffix = "";
char *name;
const char *texcoord_type = "vec2";
int samples;
int shader_offset = 0;
if (src_rb) {
samples = MAX2(src_rb->NumSamples, 1);
src_datatype = _mesa_get_format_datatype(src_rb->Format);
} else {
/* depth-or-color glCopyTexImage fallback path that passes a NULL rb and
* doesn't handle integer.
*/
samples = 1;
src_datatype = GL_UNSIGNED_NORMALIZED;
}
/* We expect only power of 2 samples in source multisample buffer. */
assert(samples > 0 && _mesa_is_pow_two(samples));
while (samples >> (shader_offset + 1)) {
shader_offset++;
}
/* Update the assert if we plan to support more than 16X MSAA. */
assert(shader_offset >= 0 && shader_offset <= 4);
if (drawFb->Visual.samples > 1) {
/* If you're calling meta_BlitFramebuffer with the destination
* multisampled, this is the only path that will work -- swrast and
* CopyTexImage won't work on it either.
*/
assert(ctx->Extensions.ARB_sample_shading);
dst_is_msaa = true;
/* We need shader invocation per sample, not per pixel */
_mesa_set_enable(ctx, GL_MULTISAMPLE, GL_TRUE);
_mesa_set_enable(ctx, GL_SAMPLE_SHADING, GL_TRUE);
_mesa_MinSampleShading(1.0);
}
switch (target) {
case GL_TEXTURE_2D_MULTISAMPLE:
case GL_TEXTURE_2D_MULTISAMPLE_ARRAY:
if (src_rb && (src_rb->_BaseFormat == GL_DEPTH_COMPONENT ||
src_rb->_BaseFormat == GL_DEPTH_STENCIL)) {
if (dst_is_msaa)
shader_index = BLIT_MSAA_SHADER_2D_MULTISAMPLE_DEPTH_COPY;
else
shader_index = BLIT_MSAA_SHADER_2D_MULTISAMPLE_DEPTH_RESOLVE;
} else {
if (dst_is_msaa)
shader_index = BLIT_MSAA_SHADER_2D_MULTISAMPLE_COPY;
else {
shader_index = BLIT_1X_MSAA_SHADER_2D_MULTISAMPLE_RESOLVE +
shader_offset;
}
}
if (target == GL_TEXTURE_2D_MULTISAMPLE_ARRAY) {
shader_index += (BLIT_1X_MSAA_SHADER_2D_MULTISAMPLE_ARRAY_RESOLVE -
BLIT_1X_MSAA_SHADER_2D_MULTISAMPLE_RESOLVE);
sampler_array_suffix = "Array";
texcoord_type = "vec3";
}
break;
default:
_mesa_problem(ctx, "Unkown texture target %s\n",
_mesa_enum_to_string(target));
shader_index = BLIT_2X_MSAA_SHADER_2D_MULTISAMPLE_RESOLVE;
}
/* We rely on the enum being sorted this way. */
STATIC_ASSERT(BLIT_1X_MSAA_SHADER_2D_MULTISAMPLE_RESOLVE_INT ==
BLIT_1X_MSAA_SHADER_2D_MULTISAMPLE_RESOLVE + 5);
STATIC_ASSERT(BLIT_1X_MSAA_SHADER_2D_MULTISAMPLE_RESOLVE_UINT ==
BLIT_1X_MSAA_SHADER_2D_MULTISAMPLE_RESOLVE + 10);
if (src_datatype == GL_INT) {
shader_index += 5;
vec4_prefix = "i";
} else if (src_datatype == GL_UNSIGNED_INT) {
shader_index += 10;
vec4_prefix = "u";
} else {
vec4_prefix = "";
}
if (blit->msaa_shaders[shader_index]) {
_mesa_UseProgram(blit->msaa_shaders[shader_index]);
return;
}
mem_ctx = ralloc_context(NULL);
if (shader_index == BLIT_MSAA_SHADER_2D_MULTISAMPLE_DEPTH_RESOLVE ||
shader_index == BLIT_MSAA_SHADER_2D_MULTISAMPLE_ARRAY_DEPTH_RESOLVE ||
shader_index == BLIT_MSAA_SHADER_2D_MULTISAMPLE_ARRAY_DEPTH_COPY ||
shader_index == BLIT_MSAA_SHADER_2D_MULTISAMPLE_DEPTH_COPY) {
char *sample_index;
const char *arb_sample_shading_extension_string;
if (dst_is_msaa) {
arb_sample_shading_extension_string = "#extension GL_ARB_sample_shading : enable";
sample_index = "gl_SampleID";
name = "depth MSAA copy";
} else {
/* Don't need that extension, since we're drawing to a single-sampled
* destination.
*/
arb_sample_shading_extension_string = "";
/* From the GL 4.3 spec:
*
* "If there is a multisample buffer (the value of SAMPLE_BUFFERS
* is one), then values are obtained from the depth samples in
* this buffer. It is recommended that the depth value of the
* centermost sample be used, though implementations may choose
* any function of the depth sample values at each pixel.
*
* We're slacking and instead of choosing centermost, we've got 0.
*/
sample_index = "0";
name = "depth MSAA resolve";
}
vs_source = ralloc_asprintf(mem_ctx,
"#version 130\n"
"in vec2 position;\n"
"in %s textureCoords;\n"
"out %s texCoords;\n"
"void main()\n"
"{\n"
" texCoords = textureCoords;\n"
" gl_Position = vec4(position, 0.0, 1.0);\n"
"}\n",
texcoord_type,
texcoord_type);
fs_source = ralloc_asprintf(mem_ctx,
"#version 130\n"
"#extension GL_ARB_texture_multisample : enable\n"
"%s\n"
"uniform sampler2DMS%s texSampler;\n"
"in %s texCoords;\n"
"out vec4 out_color;\n"
"\n"
"void main()\n"
"{\n"
" gl_FragDepth = texelFetch(texSampler, i%s(texCoords), %s).r;\n"
"}\n",
arb_sample_shading_extension_string,
sampler_array_suffix,
texcoord_type,
texcoord_type,
sample_index);
} else {
/* You can create 2D_MULTISAMPLE textures with 0 sample count (meaning 1
* sample). Yes, this is ridiculous.
*/
char *sample_resolve;
const char *arb_sample_shading_extension_string;
const char *merge_function;
name = ralloc_asprintf(mem_ctx, "%svec4 MSAA %s",
vec4_prefix,
dst_is_msaa ? "copy" : "resolve");
if (dst_is_msaa) {
arb_sample_shading_extension_string = "#extension GL_ARB_sample_shading : enable";
sample_resolve = ralloc_asprintf(mem_ctx, " out_color = texelFetch(texSampler, i%s(texCoords), gl_SampleID);", texcoord_type);
merge_function = "";
} else {
int i;
int step;
if (src_datatype == GL_INT || src_datatype == GL_UNSIGNED_INT) {
merge_function =
"gvec4 merge(gvec4 a, gvec4 b) { return (a >> gvec4(1)) + (b >> gvec4(1)) + (a & b & gvec4(1)); }\n";
} else {
/* The divide will happen at the end for floats. */
merge_function =
"vec4 merge(vec4 a, vec4 b) { return (a + b); }\n";
}
arb_sample_shading_extension_string = "";
/* We're assuming power of two samples for this resolution procedure.
*
* To avoid losing any floating point precision if the samples all
* happen to have the same value, we merge pairs of values at a time
* (so the floating point exponent just gets increased), rather than
* doing a naive sum and dividing.
*/
assert(_mesa_is_pow_two(samples));
/* Fetch each individual sample. */
sample_resolve = rzalloc_size(mem_ctx, 1);
for (i = 0; i < samples; i++) {
ralloc_asprintf_append(&sample_resolve,
" gvec4 sample_1_%d = texelFetch(texSampler, i%s(texCoords), %d);\n",
i, texcoord_type, i);
}
/* Now, merge each pair of samples, then merge each pair of those,
* etc.
*/
for (step = 2; step <= samples; step *= 2) {
for (i = 0; i < samples; i += step) {
ralloc_asprintf_append(&sample_resolve,
" gvec4 sample_%d_%d = merge(sample_%d_%d, sample_%d_%d);\n",
step, i,
step / 2, i,
step / 2, i + step / 2);
}
}
/* Scale the final result. */
if (src_datatype == GL_UNSIGNED_INT || src_datatype == GL_INT) {
ralloc_asprintf_append(&sample_resolve,
" out_color = sample_%d_0;\n",
samples);
} else {
ralloc_asprintf_append(&sample_resolve,
" gl_FragColor = sample_%d_0 / %f;\n",
samples, (float)samples);
}
}
vs_source = ralloc_asprintf(mem_ctx,
"#version 130\n"
"in vec2 position;\n"
"in %s textureCoords;\n"
"out %s texCoords;\n"
"void main()\n"
"{\n"
" texCoords = textureCoords;\n"
" gl_Position = vec4(position, 0.0, 1.0);\n"
"}\n",
texcoord_type,
texcoord_type);
fs_source = ralloc_asprintf(mem_ctx,
"#version 130\n"
"#extension GL_ARB_texture_multisample : enable\n"
"%s\n"
"#define gvec4 %svec4\n"
"uniform %ssampler2DMS%s texSampler;\n"
"in %s texCoords;\n"
"out gvec4 out_color;\n"
"\n"
"%s" /* merge_function */
"void main()\n"
"{\n"
"%s\n" /* sample_resolve */
"}\n",
arb_sample_shading_extension_string,
vec4_prefix,
vec4_prefix,
sampler_array_suffix,
texcoord_type,
merge_function,
sample_resolve);
}
_mesa_meta_compile_and_link_program(ctx, vs_source, fs_source, name,
&blit->msaa_shaders[shader_index]);
ralloc_free(mem_ctx);
}
static void texgen( GLcontext *ctx,
struct texgen_stage_data *store,
GLuint unit )
{
TNLcontext *tnl = TNL_CONTEXT(ctx);
struct vertex_buffer *VB = &tnl->vb;
GLvector4f *in = VB->AttribPtr[VERT_ATTRIB_TEX0 + unit];
GLvector4f *out = &store->texcoord[unit];
const struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit];
const GLvector4f *obj = VB->ObjPtr;
const GLvector4f *eye = VB->EyePtr;
const GLvector4f *normal = VB->AttribPtr[_TNL_ATTRIB_NORMAL];
const GLfloat *m = store->tmp_m;
const GLuint count = VB->Count;
GLfloat (*texcoord)[4] = (GLfloat (*)[4])out->data;
GLfloat (*f)[3] = store->tmp_f;
GLuint copy;
if (texUnit->_GenFlags & TEXGEN_NEED_M) {
build_m_tab[eye->size]( store->tmp_f, store->tmp_m, normal, eye );
} else if (texUnit->_GenFlags & TEXGEN_NEED_F) {
build_f_tab[eye->size]( (GLfloat *)store->tmp_f, 3, normal, eye );
}
out->size = MAX2(in->size, store->TexgenSize[unit]);
out->flags |= (in->flags & VEC_SIZE_FLAGS) | texUnit->TexGenEnabled;
out->count = count;
copy = (all_bits[in->size] & ~texUnit->TexGenEnabled);
if (copy)
_mesa_copy_tab[copy]( out, in );
if (texUnit->TexGenEnabled & S_BIT) {
GLuint i;
switch (texUnit->GenS.Mode) {
case GL_OBJECT_LINEAR:
_mesa_dotprod_tab[obj->size]( (GLfloat *)out->data,
sizeof(out->data[0]), obj,
texUnit->GenS.ObjectPlane );
break;
case GL_EYE_LINEAR:
_mesa_dotprod_tab[eye->size]( (GLfloat *)out->data,
sizeof(out->data[0]), eye,
texUnit->GenS.EyePlane );
break;
case GL_SPHERE_MAP:
for (i = 0; i < count; i++)
texcoord[i][0] = f[i][0] * m[i] + 0.5F;
break;
case GL_REFLECTION_MAP_NV:
for (i=0;i<count;i++)
texcoord[i][0] = f[i][0];
break;
case GL_NORMAL_MAP_NV: {
const GLfloat *norm = normal->start;
for (i=0;i<count;i++, STRIDE_F(norm, normal->stride)) {
texcoord[i][0] = norm[0];
}
break;
}
default:
_mesa_problem(ctx, "Bad S texgen");
}
}
if (texUnit->TexGenEnabled & T_BIT) {
GLuint i;
switch (texUnit->GenT.Mode) {
case GL_OBJECT_LINEAR:
_mesa_dotprod_tab[obj->size]( &(out->data[0][1]),
sizeof(out->data[0]), obj,
texUnit->GenT.ObjectPlane );
break;
case GL_EYE_LINEAR:
_mesa_dotprod_tab[eye->size]( &(out->data[0][1]),
sizeof(out->data[0]), eye,
texUnit->GenT.EyePlane );
break;
case GL_SPHERE_MAP:
for (i = 0; i < count; i++)
texcoord[i][1] = f[i][1] * m[i] + 0.5F;
break;
case GL_REFLECTION_MAP_NV:
for (i=0;i<count;i++)
texcoord[i][1] = f[i][1];
break;
case GL_NORMAL_MAP_NV: {
const GLfloat *norm = normal->start;
for (i=0;i<count;i++, STRIDE_F(norm, normal->stride)) {
texcoord[i][1] = norm[1];
}
break;
}
default:
_mesa_problem(ctx, "Bad T texgen");
}
}
if (texUnit->TexGenEnabled & R_BIT) {
GLuint i;
switch (texUnit->GenR.Mode) {
case GL_OBJECT_LINEAR:
_mesa_dotprod_tab[obj->size]( &(out->data[0][2]),
sizeof(out->data[0]), obj,
texUnit->GenR.ObjectPlane );
break;
case GL_EYE_LINEAR:
_mesa_dotprod_tab[eye->size]( &(out->data[0][2]),
sizeof(out->data[0]), eye,
texUnit->GenR.EyePlane );
break;
case GL_REFLECTION_MAP_NV:
for (i=0;i<count;i++)
texcoord[i][2] = f[i][2];
break;
case GL_NORMAL_MAP_NV: {
const GLfloat *norm = normal->start;
for (i=0;i<count;i++,STRIDE_F(norm, normal->stride)) {
texcoord[i][2] = norm[2];
}
break;
}
default:
_mesa_problem(ctx, "Bad R texgen");
}
}
if (texUnit->TexGenEnabled & Q_BIT) {
switch (texUnit->GenQ.Mode) {
case GL_OBJECT_LINEAR:
_mesa_dotprod_tab[obj->size]( &(out->data[0][3]),
sizeof(out->data[0]), obj,
texUnit->GenQ.ObjectPlane );
break;
case GL_EYE_LINEAR:
_mesa_dotprod_tab[eye->size]( &(out->data[0][3]),
sizeof(out->data[0]), eye,
texUnit->GenQ.EyePlane );
break;
default:
_mesa_problem(ctx, "Bad Q texgen");
}
}
}
static void
setup_glsl_msaa_blit_scaled_shader(struct gl_context *ctx,
struct blit_state *blit,
struct gl_renderbuffer *src_rb,
GLenum target, GLenum filter)
{
GLint loc_src_width, loc_src_height;
int i, samples;
int shader_offset = 0;
void *mem_ctx = ralloc_context(NULL);
char *fs_source;
char *name, *sample_number;
const uint8_t *sample_map;
char *sample_map_str = rzalloc_size(mem_ctx, 1);
char *sample_map_expr = rzalloc_size(mem_ctx, 1);
char *texel_fetch_macro = rzalloc_size(mem_ctx, 1);
const char *sampler_array_suffix = "";
float y_scale;
enum blit_msaa_shader shader_index;
assert(src_rb);
samples = MAX2(src_rb->NumSamples, 1);
y_scale = samples * 0.5;
/* We expect only power of 2 samples in source multisample buffer. */
assert(samples > 0 && _mesa_is_pow_two(samples));
while (samples >> (shader_offset + 1)) {
shader_offset++;
}
/* Update the assert if we plan to support more than 8X MSAA. */
assert(shader_offset > 0 && shader_offset < 4);
assert(target == GL_TEXTURE_2D_MULTISAMPLE ||
target == GL_TEXTURE_2D_MULTISAMPLE_ARRAY);
shader_index = BLIT_2X_MSAA_SHADER_2D_MULTISAMPLE_SCALED_RESOLVE +
shader_offset - 1;
if (target == GL_TEXTURE_2D_MULTISAMPLE_ARRAY) {
shader_index += BLIT_2X_MSAA_SHADER_2D_MULTISAMPLE_ARRAY_SCALED_RESOLVE -
BLIT_2X_MSAA_SHADER_2D_MULTISAMPLE_SCALED_RESOLVE;
sampler_array_suffix = "Array";
}
if (blit->msaa_shaders[shader_index]) {
_mesa_UseProgram(blit->msaa_shaders[shader_index]);
/* Update the uniform values. */
loc_src_width =
_mesa_GetUniformLocation(blit->msaa_shaders[shader_index], "src_width");
loc_src_height =
_mesa_GetUniformLocation(blit->msaa_shaders[shader_index], "src_height");
_mesa_Uniform1f(loc_src_width, src_rb->Width);
_mesa_Uniform1f(loc_src_height, src_rb->Height);
return;
}
name = ralloc_asprintf(mem_ctx, "vec4 MSAA scaled resolve");
/* Below switch is used to setup the shader expression, which computes
* sample index and map it to to a sample number on hardware.
*/
switch(samples) {
case 2:
sample_number = "sample_map[int(2 * fract(coord.x))]";
sample_map = ctx->Const.SampleMap2x;
break;
case 4:
sample_number = "sample_map[int(2 * fract(coord.x) + 4 * fract(coord.y))]";
sample_map = ctx->Const.SampleMap4x;
break;
case 8:
sample_number = "sample_map[int(2 * fract(coord.x) + 8 * fract(coord.y))]";
sample_map = ctx->Const.SampleMap8x;
break;
default:
sample_number = NULL;
sample_map = NULL;
_mesa_problem(ctx, "Unsupported sample count %d\n", samples);
unreachable("Unsupported sample count");
}
/* Create sample map string. */
for (i = 0 ; i < samples - 1; i++) {
ralloc_asprintf_append(&sample_map_str, "%d, ", sample_map[i]);
}
ralloc_asprintf_append(&sample_map_str, "%d", sample_map[samples - 1]);
/* Create sample map expression using above string. */
ralloc_asprintf_append(&sample_map_expr,
" const int sample_map[%d] = int[%d](%s);\n",
samples, samples, sample_map_str);
if (target == GL_TEXTURE_2D_MULTISAMPLE) {
ralloc_asprintf_append(&texel_fetch_macro,
"#define TEXEL_FETCH(coord) texelFetch(texSampler, ivec2(coord), %s);\n",
sample_number);
} else {
ralloc_asprintf_append(&texel_fetch_macro,
"#define TEXEL_FETCH(coord) texelFetch(texSampler, ivec3(coord, layer), %s);\n",
sample_number);
}
static const char vs_source[] =
"#version 130\n"
"in vec2 position;\n"
"in vec3 textureCoords;\n"
"out vec2 texCoords;\n"
"flat out int layer;\n"
"void main()\n"
"{\n"
" texCoords = textureCoords.xy;\n"
" layer = int(textureCoords.z);\n"
" gl_Position = vec4(position, 0.0, 1.0);\n"
"}\n"
;
fs_source = ralloc_asprintf(mem_ctx,
"#version 130\n"
"#extension GL_ARB_texture_multisample : enable\n"
"uniform sampler2DMS%s texSampler;\n"
"uniform float src_width, src_height;\n"
"in vec2 texCoords;\n"
"flat in int layer;\n"
"out vec4 out_color;\n"
"\n"
"void main()\n"
"{\n"
"%s"
" vec2 interp;\n"
" const vec2 scale = vec2(2.0f, %ff);\n"
" const vec2 scale_inv = vec2(0.5f, %ff);\n"
" const vec2 s_0_offset = vec2(0.25f, %ff);\n"
" vec2 s_0_coord, s_1_coord, s_2_coord, s_3_coord;\n"
" vec4 s_0_color, s_1_color, s_2_color, s_3_color;\n"
" vec4 x_0_color, x_1_color;\n"
" vec2 tex_coord = texCoords - s_0_offset;\n"
"\n"
" tex_coord *= scale;\n"
" clamp(tex_coord.x, 0.0f, scale.x * src_width - 1.0f);\n"
" clamp(tex_coord.y, 0.0f, scale.y * src_height - 1.0f);\n"
" interp = fract(tex_coord);\n"
" tex_coord = ivec2(tex_coord) * scale_inv;\n"
"\n"
" /* Compute the sample coordinates used for filtering. */\n"
" s_0_coord = tex_coord;\n"
" s_1_coord = tex_coord + vec2(scale_inv.x, 0.0f);\n"
" s_2_coord = tex_coord + vec2(0.0f, scale_inv.y);\n"
" s_3_coord = tex_coord + vec2(scale_inv.x, scale_inv.y);\n"
"\n"
" /* Fetch sample color values. */\n"
"%s"
" s_0_color = TEXEL_FETCH(s_0_coord)\n"
" s_1_color = TEXEL_FETCH(s_1_coord)\n"
" s_2_color = TEXEL_FETCH(s_2_coord)\n"
" s_3_color = TEXEL_FETCH(s_3_coord)\n"
"#undef TEXEL_FETCH\n"
"\n"
" /* Do bilinear filtering on sample colors. */\n"
" x_0_color = mix(s_0_color, s_1_color, interp.x);\n"
" x_1_color = mix(s_2_color, s_3_color, interp.x);\n"
" out_color = mix(x_0_color, x_1_color, interp.y);\n"
"}\n",
sampler_array_suffix,
sample_map_expr,
y_scale,
1.0f / y_scale,
1.0f / samples,
texel_fetch_macro);
_mesa_meta_compile_and_link_program(ctx, vs_source, fs_source, name,
&blit->msaa_shaders[shader_index]);
loc_src_width =
_mesa_GetUniformLocation(blit->msaa_shaders[shader_index], "src_width");
loc_src_height =
_mesa_GetUniformLocation(blit->msaa_shaders[shader_index], "src_height");
_mesa_Uniform1f(loc_src_width, src_rb->Width);
_mesa_Uniform1f(loc_src_height, src_rb->Height);
ralloc_free(mem_ctx);
}
/**
* Called via glRenderbufferStorageEXT() to set the format and allocate
* storage for a user-created renderbuffer.
*/
static GLboolean
intel_alloc_renderbuffer_storage(GLcontext * ctx, struct gl_renderbuffer *rb,
GLenum internalFormat,
GLuint width, GLuint height)
{
struct intel_context *intel = intel_context(ctx);
struct intel_renderbuffer *irb = intel_renderbuffer(rb);
int cpp;
GLuint pitch;
ASSERT(rb->Name != 0);
switch (internalFormat) {
case GL_R3_G3_B2:
case GL_RGB4:
case GL_RGB5:
rb->Format = MESA_FORMAT_RGB565;
rb->DataType = GL_UNSIGNED_BYTE;
break;
case GL_RGB:
case GL_RGB8:
case GL_RGB10:
case GL_RGB12:
case GL_RGB16:
rb->Format = MESA_FORMAT_XRGB8888;
rb->DataType = GL_UNSIGNED_BYTE;
break;
case GL_RGBA:
case GL_RGBA2:
case GL_RGBA4:
case GL_RGB5_A1:
case GL_RGBA8:
case GL_RGB10_A2:
case GL_RGBA12:
case GL_RGBA16:
rb->Format = MESA_FORMAT_ARGB8888;
rb->DataType = GL_UNSIGNED_BYTE;
break;
case GL_STENCIL_INDEX:
case GL_STENCIL_INDEX1_EXT:
case GL_STENCIL_INDEX4_EXT:
case GL_STENCIL_INDEX8_EXT:
case GL_STENCIL_INDEX16_EXT:
/* alloc a depth+stencil buffer */
rb->Format = MESA_FORMAT_S8_Z24;
rb->DataType = GL_UNSIGNED_INT_24_8_EXT;
break;
case GL_DEPTH_COMPONENT16:
rb->Format = MESA_FORMAT_Z16;
rb->DataType = GL_UNSIGNED_SHORT;
break;
case GL_DEPTH_COMPONENT:
case GL_DEPTH_COMPONENT24:
case GL_DEPTH_COMPONENT32:
rb->Format = MESA_FORMAT_S8_Z24;
rb->DataType = GL_UNSIGNED_INT_24_8_EXT;
break;
case GL_DEPTH_STENCIL_EXT:
case GL_DEPTH24_STENCIL8_EXT:
rb->Format = MESA_FORMAT_S8_Z24;
rb->DataType = GL_UNSIGNED_INT_24_8_EXT;
break;
default:
_mesa_problem(ctx,
"Unexpected format in intel_alloc_renderbuffer_storage");
return GL_FALSE;
}
rb->_BaseFormat = _mesa_base_fbo_format(ctx, internalFormat);
cpp = _mesa_get_format_bytes(rb->Format);
intelFlush(ctx);
/* free old region */
if (irb->region) {
intel_region_release(&irb->region);
}
/* allocate new memory region/renderbuffer */
/* Choose a pitch to match hardware requirements:
*/
pitch = ((cpp * width + 63) & ~63) / cpp;
/* alloc hardware renderbuffer */
DBG("Allocating %d x %d Intel RBO (pitch %d)\n", width, height, pitch);
irb->region = intel_region_alloc(intel, I915_TILING_NONE, cpp,
width, height, pitch, GL_TRUE);
if (!irb->region)
return GL_FALSE; /* out of memory? */
ASSERT(irb->region->buffer);
rb->Width = width;
rb->Height = height;
return GL_TRUE;
}
/**
* Disable the named extension.
* XXX is this really needed???
*/
void
_mesa_disable_extension( struct gl_context *ctx, const char *name )
{
if (!set_extension(ctx, name, GL_FALSE))
_mesa_problem(ctx, "Trying to disable unknown extension: %s", name);
}