/*
* Ensure that we are able to use the given physical memory range.
*
* We abort if the destination physical range overlaps us, or if it
* goes outside the bounds of memory.
*/
static void ensure_phys_range_valid(paddr_t paddr_min, paddr_t paddr_max)
{
/* Ensure that the kernel physical load address doesn't overwrite us. */
if (regions_overlap(paddr_min, paddr_max - 1, (word_t)_start, (word_t)_end - 1)) {
printf("Kernel load address would overlap ELF-loader!\n");
abort();
}
}
static void
builder_invalidate_region(struct builder *bld, struct eu_region *r)
{
for (int i = 0; i < ARRAY_LENGTH(bld->regs); i++) {
if ((bld->regs[i].contents & BUILDER_REG_CONTENTS_EU_REG) &&
regions_overlap(r, &bld->regs[i].region)) {
bld->regs[i].contents &= ~BUILDER_REG_CONTENTS_EU_REG;
ksim_trace(TRACE_AVX,
"*** invalidate g%d.%d (ymm%d)\n",
bld->regs[i].region.offset / 32,
bld->regs[i].region.offset & 31, i);
}
}
}
/**
* This isn't terribly efficient. If a driver really has combined
* depth/stencil buffers the driver should implement an optimized
* CopyPixels function.
*/
static void
copy_depth_stencil_pixels(GLcontext *ctx,
const GLint srcX, const GLint srcY,
const GLint width, const GLint height,
const GLint destX, const GLint destY)
{
struct gl_renderbuffer *stencilReadRb, *depthReadRb, *depthDrawRb;
GLint sy, dy, stepy;
GLint j;
GLstencil *tempStencilImage = NULL, *stencilPtr = NULL;
GLfloat *tempDepthImage = NULL, *depthPtr = NULL;
const GLfloat depthScale = ctx->DrawBuffer->_DepthMaxF;
const GLuint stencilMask = ctx->Stencil.WriteMask[0];
const GLboolean zoom = ctx->Pixel.ZoomX != 1.0F || ctx->Pixel.ZoomY != 1.0F;
const GLboolean shiftOrOffset
= ctx->Pixel.IndexShift || ctx->Pixel.IndexOffset;
const GLboolean scaleOrBias
= ctx->Pixel.DepthScale != 1.0 || ctx->Pixel.DepthBias != 0.0;
GLint overlapping;
depthDrawRb = ctx->DrawBuffer->_DepthBuffer;
depthReadRb = ctx->ReadBuffer->_DepthBuffer;
stencilReadRb = ctx->ReadBuffer->_StencilBuffer;
ASSERT(depthDrawRb);
ASSERT(depthReadRb);
ASSERT(stencilReadRb);
/* Determine if copy should be bottom-to-top or top-to-bottom */
if (srcY < destY) {
/* top-down max-to-min */
sy = srcY + height - 1;
dy = destY + height - 1;
stepy = -1;
}
else {
/* bottom-up min-to-max */
sy = srcY;
dy = destY;
stepy = 1;
}
if (ctx->DrawBuffer == ctx->ReadBuffer) {
overlapping = regions_overlap(srcX, srcY, destX, destY, width, height,
ctx->Pixel.ZoomX, ctx->Pixel.ZoomY);
}
else {
overlapping = GL_FALSE;
}
if (overlapping) {
GLint ssy = sy;
if (stencilMask != 0x0) {
tempStencilImage
= (GLstencil *) _mesa_malloc(width * height * sizeof(GLstencil));
if (!tempStencilImage) {
_mesa_error(ctx, GL_OUT_OF_MEMORY, "glCopyPixels");
return;
}
/* get copy of stencil pixels */
stencilPtr = tempStencilImage;
for (j = 0; j < height; j++, ssy += stepy) {
_swrast_read_stencil_span(ctx, stencilReadRb,
width, srcX, ssy, stencilPtr);
stencilPtr += width;
}
stencilPtr = tempStencilImage;
}
if (ctx->Depth.Mask) {
tempDepthImage
= (GLfloat *) _mesa_malloc(width * height * sizeof(GLfloat));
if (!tempDepthImage) {
_mesa_error(ctx, GL_OUT_OF_MEMORY, "glCopyPixels");
_mesa_free(tempStencilImage);
return;
}
/* get copy of depth pixels */
depthPtr = tempDepthImage;
for (j = 0; j < height; j++, ssy += stepy) {
_swrast_read_depth_span_float(ctx, depthReadRb,
width, srcX, ssy, depthPtr);
depthPtr += width;
}
depthPtr = tempDepthImage;
}
}
for (j = 0; j < height; j++, sy += stepy, dy += stepy) {
if (stencilMask != 0x0) {
GLstencil stencil[MAX_WIDTH];
/* Get stencil values */
if (overlapping) {
_mesa_memcpy(stencil, stencilPtr, width * sizeof(GLstencil));
stencilPtr += width;
}
else {
_swrast_read_stencil_span(ctx, stencilReadRb,
width, srcX, sy, stencil);
}
/* Apply shift, offset, look-up table */
if (shiftOrOffset) {
_mesa_shift_and_offset_stencil(ctx, width, stencil);
}
if (ctx->Pixel.MapStencilFlag) {
_mesa_map_stencil(ctx, width, stencil);
}
/* Write values */
if (zoom) {
_swrast_write_zoomed_stencil_span(ctx, destX, destY, width,
destX, dy, stencil);
}
else {
_swrast_write_stencil_span( ctx, width, destX, dy, stencil );
}
}
if (ctx->Depth.Mask) {
GLfloat depth[MAX_WIDTH];
GLuint zVals32[MAX_WIDTH];
GLushort zVals16[MAX_WIDTH];
GLvoid *zVals;
GLuint zBytes;
/* get depth values */
if (overlapping) {
_mesa_memcpy(depth, depthPtr, width * sizeof(GLfloat));
depthPtr += width;
}
else {
_swrast_read_depth_span_float(ctx, depthReadRb,
width, srcX, sy, depth);
}
/* scale & bias */
if (scaleOrBias) {
_mesa_scale_and_bias_depth(ctx, width, depth);
}
/* convert to integer Z values */
if (depthDrawRb->DataType == GL_UNSIGNED_SHORT) {
GLint k;
for (k = 0; k < width; k++)
zVals16[k] = (GLushort) (depth[k] * depthScale);
zVals = zVals16;
zBytes = 2;
}
else {
GLint k;
for (k = 0; k < width; k++)
zVals32[k] = (GLuint) (depth[k] * depthScale);
zVals = zVals32;
zBytes = 4;
}
/* Write values */
if (zoom) {
_swrast_write_zoomed_z_span(ctx, destX, destY, width,
destX, dy, zVals);
}
else {
_swrast_put_row(ctx, depthDrawRb, width, destX, dy, zVals, zBytes);
}
}
}
if (tempStencilImage)
_mesa_free(tempStencilImage);
if (tempDepthImage)
_mesa_free(tempDepthImage);
}
static void
copy_stencil_pixels( GLcontext *ctx, GLint srcx, GLint srcy,
GLint width, GLint height,
GLint destx, GLint desty )
{
struct gl_framebuffer *fb = ctx->ReadBuffer;
struct gl_renderbuffer *rb = fb->_StencilBuffer;
GLint sy, dy, stepy;
GLint j;
GLstencil *p, *tmpImage;
const GLboolean zoom = ctx->Pixel.ZoomX != 1.0F || ctx->Pixel.ZoomY != 1.0F;
const GLboolean shift_or_offset = ctx->Pixel.IndexShift || ctx->Pixel.IndexOffset;
GLint overlapping;
if (!rb) {
/* no readbuffer - OK */
return;
}
/* Determine if copy should be bottom-to-top or top-to-bottom */
if (srcy < desty) {
/* top-down max-to-min */
sy = srcy + height - 1;
dy = desty + height - 1;
stepy = -1;
}
else {
/* bottom-up min-to-max */
sy = srcy;
dy = desty;
stepy = 1;
}
if (ctx->DrawBuffer == ctx->ReadBuffer) {
overlapping = regions_overlap(srcx, srcy, destx, desty, width, height,
ctx->Pixel.ZoomX, ctx->Pixel.ZoomY);
}
else {
overlapping = GL_FALSE;
}
if (overlapping) {
GLint ssy = sy;
tmpImage = (GLstencil *) _mesa_malloc(width * height * sizeof(GLstencil));
if (!tmpImage) {
_mesa_error( ctx, GL_OUT_OF_MEMORY, "glCopyPixels" );
return;
}
p = tmpImage;
for (j = 0; j < height; j++, ssy += stepy) {
_swrast_read_stencil_span( ctx, rb, width, srcx, ssy, p );
p += width;
}
p = tmpImage;
}
else {
tmpImage = NULL; /* silence compiler warning */
p = NULL;
}
for (j = 0; j < height; j++, sy += stepy, dy += stepy) {
GLstencil stencil[MAX_WIDTH];
/* Get stencil values */
if (overlapping) {
_mesa_memcpy(stencil, p, width * sizeof(GLstencil));
p += width;
}
else {
_swrast_read_stencil_span( ctx, rb, width, srcx, sy, stencil );
}
/* Apply shift, offset, look-up table */
if (shift_or_offset) {
_mesa_shift_and_offset_stencil( ctx, width, stencil );
}
if (ctx->Pixel.MapStencilFlag) {
_mesa_map_stencil( ctx, width, stencil );
}
/* Write stencil values */
if (zoom) {
_swrast_write_zoomed_stencil_span(ctx, destx, desty, width,
destx, dy, stencil);
}
else {
_swrast_write_stencil_span( ctx, width, destx, dy, stencil );
}
}
if (overlapping)
_mesa_free(tmpImage);
}
/*
* TODO: Optimize!!!!
*/
static void
copy_depth_pixels( GLcontext *ctx, GLint srcx, GLint srcy,
GLint width, GLint height,
GLint destx, GLint desty )
{
SWcontext *swrast = SWRAST_CONTEXT(ctx);
struct gl_framebuffer *fb = ctx->ReadBuffer;
struct gl_renderbuffer *readRb = fb->_DepthBuffer;
const GLfloat depthMax = fb->_DepthMaxF;
GLfloat *p, *tmpImage;
GLint sy, dy, stepy;
GLint i, j;
const GLboolean zoom = ctx->Pixel.ZoomX != 1.0F || ctx->Pixel.ZoomY != 1.0F;
GLint overlapping;
struct sw_span span;
if (!readRb) {
/* no readbuffer - OK */
return;
}
INIT_SPAN(span, GL_BITMAP, 0, 0, SPAN_Z);
/* Determine if copy should be bottom-to-top or top-to-bottom */
if (srcy<desty) {
/* top-down max-to-min */
sy = srcy + height - 1;
dy = desty + height - 1;
stepy = -1;
}
else {
/* bottom-up min-to-max */
sy = srcy;
dy = desty;
stepy = 1;
}
if (ctx->DrawBuffer == ctx->ReadBuffer) {
overlapping = regions_overlap(srcx, srcy, destx, desty, width, height,
ctx->Pixel.ZoomX, ctx->Pixel.ZoomY);
}
else {
overlapping = GL_FALSE;
}
_swrast_span_default_color(ctx, &span);
if (swrast->_FogEnabled)
_swrast_span_default_fog(ctx, &span);
if (overlapping) {
GLint ssy = sy;
tmpImage = (GLfloat *) _mesa_malloc(width * height * sizeof(GLfloat));
if (!tmpImage) {
_mesa_error( ctx, GL_OUT_OF_MEMORY, "glCopyPixels" );
return;
}
p = tmpImage;
for (j = 0; j < height; j++, ssy += stepy) {
_swrast_read_depth_span_float(ctx, readRb, width, srcx, ssy, p);
p += width;
}
p = tmpImage;
}
else {
tmpImage = NULL; /* silence compiler warning */
p = NULL;
}
for (j = 0; j < height; j++, sy += stepy, dy += stepy) {
GLfloat depth[MAX_WIDTH];
/* get depth values */
if (overlapping) {
_mesa_memcpy(depth, p, width * sizeof(GLfloat));
p += width;
}
else {
_swrast_read_depth_span_float(ctx, readRb, width, srcx, sy, depth);
}
/* apply scale and bias */
for (i = 0; i < width; i++) {
GLfloat d = depth[i] * ctx->Pixel.DepthScale + ctx->Pixel.DepthBias;
span.array->z[i] = (GLuint) (CLAMP(d, 0.0F, 1.0F) * depthMax);
}
/* write depth values */
span.x = destx;
span.y = dy;
span.end = width;
if (fb->Visual.rgbMode) {
if (zoom)
_swrast_write_zoomed_rgba_span(ctx, destx, desty, &span,
(const GLchan (*)[4]) span.array->rgba);
else
_swrast_write_rgba_span(ctx, &span);
}
else {
if (zoom)
_swrast_write_zoomed_index_span(ctx, destx, desty, &span);
else
_swrast_write_index_span(ctx, &span);
}
}
if (overlapping)
_mesa_free(tmpImage);
}
static void
copy_ci_pixels( GLcontext *ctx, GLint srcx, GLint srcy,
GLint width, GLint height,
GLint destx, GLint desty )
{
SWcontext *swrast = SWRAST_CONTEXT(ctx);
GLuint *tmpImage,*p;
GLint sy, dy, stepy;
GLint j;
const GLboolean zoom = ctx->Pixel.ZoomX != 1.0F || ctx->Pixel.ZoomY != 1.0F;
const GLboolean shift_or_offset = ctx->Pixel.IndexShift || ctx->Pixel.IndexOffset;
GLint overlapping;
struct sw_span span;
if (!ctx->ReadBuffer->_ColorReadBuffer) {
/* no readbuffer - OK */
return;
}
INIT_SPAN(span, GL_BITMAP, 0, 0, SPAN_INDEX);
/* Determine if copy should be bottom-to-top or top-to-bottom */
if (srcy<desty) {
/* top-down max-to-min */
sy = srcy + height - 1;
dy = desty + height - 1;
stepy = -1;
}
else {
/* bottom-up min-to-max */
sy = srcy;
dy = desty;
stepy = 1;
}
if (ctx->DrawBuffer == ctx->ReadBuffer) {
overlapping = regions_overlap(srcx, srcy, destx, desty, width, height,
ctx->Pixel.ZoomX, ctx->Pixel.ZoomY);
}
else {
overlapping = GL_FALSE;
}
if (ctx->Depth.Test)
_swrast_span_default_z(ctx, &span);
if (swrast->_FogEnabled)
_swrast_span_default_fog(ctx, &span);
if (overlapping) {
GLint ssy = sy;
tmpImage = (GLuint *) _mesa_malloc(width * height * sizeof(GLuint));
if (!tmpImage) {
_mesa_error( ctx, GL_OUT_OF_MEMORY, "glCopyPixels" );
return;
}
/* read the image */
p = tmpImage;
for (j = 0; j < height; j++, ssy += stepy) {
_swrast_read_index_span( ctx, ctx->ReadBuffer->_ColorReadBuffer,
width, srcx, ssy, p );
p += width;
}
p = tmpImage;
}
else {
tmpImage = NULL; /* silence compiler warning */
p = NULL;
}
for (j = 0; j < height; j++, sy += stepy, dy += stepy) {
/* Get color indexes */
if (overlapping) {
_mesa_memcpy(span.array->index, p, width * sizeof(GLuint));
p += width;
}
else {
_swrast_read_index_span( ctx, ctx->ReadBuffer->_ColorReadBuffer,
width, srcx, sy, span.array->index );
}
/* Apply shift, offset, look-up table */
if (shift_or_offset) {
_mesa_shift_and_offset_ci( ctx, width, span.array->index );
}
if (ctx->Pixel.MapColorFlag) {
_mesa_map_ci( ctx, width, span.array->index );
}
/* write color indexes */
span.x = destx;
span.y = dy;
span.end = width;
if (zoom)
_swrast_write_zoomed_index_span(ctx, destx, desty, &span);
else
_swrast_write_index_span(ctx, &span);
}
if (overlapping)
_mesa_free(tmpImage);
}
/*
* RGBA copypixels
*/
static void
copy_rgba_pixels(GLcontext *ctx, GLint srcx, GLint srcy,
GLint width, GLint height, GLint destx, GLint desty)
{
SWcontext *swrast = SWRAST_CONTEXT(ctx);
struct gl_renderbuffer *drawRb;
GLchan *tmpImage,*p;
GLboolean quick_draw;
GLint sy, dy, stepy, j;
const GLboolean zoom = ctx->Pixel.ZoomX != 1.0F || ctx->Pixel.ZoomY != 1.0F;
GLint overlapping;
const GLuint transferOps = ctx->_ImageTransferState;
struct sw_span span;
if (!ctx->ReadBuffer->_ColorReadBuffer) {
/* no readbuffer - OK */
return;
}
INIT_SPAN(span, GL_BITMAP, 0, 0, SPAN_RGBA);
if (ctx->Pixel.Convolution2DEnabled || ctx->Pixel.Separable2DEnabled) {
copy_conv_rgba_pixels(ctx, srcx, srcy, width, height, destx, desty);
return;
}
/* Determine if copy should be done bottom-to-top or top-to-bottom */
if (srcy < desty) {
/* top-down max-to-min */
sy = srcy + height - 1;
dy = desty + height - 1;
stepy = -1;
}
else {
/* bottom-up min-to-max */
sy = srcy;
dy = desty;
stepy = 1;
}
if (ctx->DrawBuffer == ctx->ReadBuffer) {
overlapping = regions_overlap(srcx, srcy, destx, desty, width, height,
ctx->Pixel.ZoomX, ctx->Pixel.ZoomY);
}
else {
overlapping = GL_FALSE;
}
if (ctx->Depth.Test)
_swrast_span_default_z(ctx, &span);
if (swrast->_FogEnabled)
_swrast_span_default_fog(ctx, &span);
if (SWRAST_CONTEXT(ctx)->_RasterMask == 0
&& !zoom
&& destx >= 0
&& destx + width <= (GLint) ctx->DrawBuffer->Width) {
quick_draw = GL_TRUE;
drawRb = ctx->DrawBuffer->_ColorDrawBuffers[0][0];
}
else {
quick_draw = GL_FALSE;
drawRb = NULL;
}
if (overlapping) {
GLint ssy = sy;
tmpImage = (GLchan *) _mesa_malloc(width * height * sizeof(GLchan) * 4);
if (!tmpImage) {
_mesa_error( ctx, GL_OUT_OF_MEMORY, "glCopyPixels" );
return;
}
/* read the source image */
p = tmpImage;
for (j = 0; j < height; j++, ssy += stepy) {
_swrast_read_rgba_span( ctx, ctx->ReadBuffer->_ColorReadBuffer,
width, srcx, ssy, (GLchan (*)[4]) p );
p += width * 4;
}
p = tmpImage;
}
else {
tmpImage = NULL; /* silence compiler warnings */
p = NULL;
}
for (j = 0; j < height; j++, sy += stepy, dy += stepy) {
/* Get source pixels */
if (overlapping) {
/* get from buffered image */
ASSERT(width < MAX_WIDTH);
_mesa_memcpy(span.array->rgba, p, width * sizeof(GLchan) * 4);
p += width * 4;
}
else {
/* get from framebuffer */
ASSERT(width < MAX_WIDTH);
_swrast_read_rgba_span( ctx, ctx->ReadBuffer->_ColorReadBuffer,
width, srcx, sy, span.array->rgba );
}
if (transferOps) {
GLfloat rgbaFloat[MAX_WIDTH][4];
/* convert to float, transfer, convert back to chan */
chan_span_to_float(width, (CONST GLchan (*)[4]) span.array->rgba,
rgbaFloat);
_mesa_apply_rgba_transfer_ops(ctx, transferOps, width, rgbaFloat);
float_span_to_chan(width, (CONST GLfloat (*)[4]) rgbaFloat,
span.array->rgba);
}
/* Write color span */
if (quick_draw && dy >= 0 && dy < (GLint) ctx->DrawBuffer->Height) {
drawRb->PutRow(ctx, drawRb, width, destx, dy, span.array->rgba, NULL);
}
else {
span.x = destx;
span.y = dy;
span.end = width;
if (zoom) {
_swrast_write_zoomed_rgba_span(ctx, destx, desty, &span,
(CONST GLchan (*)[4]) span.array->rgba);
}
else {
_swrast_write_rgba_span(ctx, &span);
}
}
}
if (overlapping)
_mesa_free(tmpImage);
}
static void
copy_stencil_pixels( struct gl_context *ctx, GLint srcx, GLint srcy,
GLint width, GLint height,
GLint destx, GLint desty )
{
struct gl_framebuffer *fb = ctx->ReadBuffer;
struct gl_renderbuffer *rb = fb->Attachment[BUFFER_STENCIL].Renderbuffer;
GLint sy, dy, stepy;
GLint j;
GLubyte *p, *tmpImage, *stencil;
const GLboolean zoom = ctx->Pixel.ZoomX != 1.0F || ctx->Pixel.ZoomY != 1.0F;
GLint overlapping;
if (!rb) {
/* no readbuffer - OK */
return;
}
if (ctx->DrawBuffer == ctx->ReadBuffer) {
overlapping = regions_overlap(srcx, srcy, destx, desty, width, height,
ctx->Pixel.ZoomX, ctx->Pixel.ZoomY);
}
else {
overlapping = GL_FALSE;
}
/* Determine if copy should be bottom-to-top or top-to-bottom */
if (!overlapping && srcy < desty) {
/* top-down max-to-min */
sy = srcy + height - 1;
dy = desty + height - 1;
stepy = -1;
}
else {
/* bottom-up min-to-max */
sy = srcy;
dy = desty;
stepy = 1;
}
if (overlapping) {
GLint ssy = sy;
tmpImage = malloc(width * height * sizeof(GLubyte));
if (!tmpImage) {
_mesa_error( ctx, GL_OUT_OF_MEMORY, "glCopyPixels" );
return;
}
p = tmpImage;
for (j = 0; j < height; j++, ssy += stepy) {
_swrast_read_stencil_span( ctx, rb, width, srcx, ssy, p );
p += width;
}
p = tmpImage;
}
else {
tmpImage = NULL; /* silence compiler warning */
p = NULL;
}
stencil = malloc(width * sizeof(GLubyte));
if (!stencil) {
_mesa_error(ctx, GL_OUT_OF_MEMORY, "glCopyPixels()");
goto end;
}
for (j = 0; j < height; j++, sy += stepy, dy += stepy) {
/* Get stencil values */
if (overlapping) {
memcpy(stencil, p, width * sizeof(GLubyte));
p += width;
}
else {
_swrast_read_stencil_span( ctx, rb, width, srcx, sy, stencil );
}
_mesa_apply_stencil_transfer_ops(ctx, width, stencil);
/* Write stencil values */
if (zoom) {
_swrast_write_zoomed_stencil_span(ctx, destx, desty, width,
destx, dy, stencil);
}
else {
_swrast_write_stencil_span( ctx, width, destx, dy, stencil );
}
}
free(stencil);
end:
if (overlapping)
free(tmpImage);
}
/**
* Try to do a glCopyPixels for simple cases with a blit by calling
* pipe->resource_copy_region().
*
* We can do this when we're copying color pixels (depth/stencil
* eventually) with no pixel zoom, no pixel transfer ops, no
* per-fragment ops, the src/dest regions don't overlap and the
* src/dest pixel formats are the same.
*/
static GLboolean
blit_copy_pixels(struct gl_context *ctx, GLint srcx, GLint srcy,
GLsizei width, GLsizei height,
GLint dstx, GLint dsty, GLenum type)
{
struct st_context *st = st_context(ctx);
struct pipe_context *pipe = st->pipe;
struct gl_pixelstore_attrib pack, unpack;
GLint readX, readY, readW, readH;
if (type == GL_COLOR &&
ctx->Pixel.ZoomX == 1.0 &&
ctx->Pixel.ZoomY == 1.0 &&
ctx->_ImageTransferState == 0x0 &&
!ctx->Color.BlendEnabled &&
!ctx->Color.AlphaEnabled &&
!ctx->Depth.Test &&
!ctx->Fog.Enabled &&
!ctx->Stencil.Enabled &&
!ctx->FragmentProgram.Enabled &&
!ctx->VertexProgram.Enabled &&
!ctx->Shader.CurrentFragmentProgram &&
st_fb_orientation(ctx->ReadBuffer) == st_fb_orientation(ctx->DrawBuffer) &&
ctx->DrawBuffer->_NumColorDrawBuffers == 1 &&
!ctx->Query.CondRenderQuery) {
struct st_renderbuffer *rbRead, *rbDraw;
GLint drawX, drawY;
/*
* Clip the read region against the src buffer bounds.
* We'll still allocate a temporary buffer/texture for the original
* src region size but we'll only read the region which is on-screen.
* This may mean that we draw garbage pixels into the dest region, but
* that's expected.
*/
readX = srcx;
readY = srcy;
readW = width;
readH = height;
pack = ctx->DefaultPacking;
if (!_mesa_clip_readpixels(ctx, &readX, &readY, &readW, &readH, &pack))
return GL_TRUE; /* all done */
/* clip against dest buffer bounds and scissor box */
drawX = dstx + pack.SkipPixels;
drawY = dsty + pack.SkipRows;
unpack = pack;
if (!_mesa_clip_drawpixels(ctx, &drawX, &drawY, &readW, &readH, &unpack))
return GL_TRUE; /* all done */
readX = readX - pack.SkipPixels + unpack.SkipPixels;
readY = readY - pack.SkipRows + unpack.SkipRows;
rbRead = st_get_color_read_renderbuffer(ctx);
rbDraw = st_renderbuffer(ctx->DrawBuffer->_ColorDrawBuffers[0]);
if ((rbRead != rbDraw ||
!regions_overlap(readX, readY, drawX, drawY, readW, readH)) &&
rbRead->Base.Format == rbDraw->Base.Format) {
struct pipe_box srcBox;
/* flip src/dst position if needed */
if (st_fb_orientation(ctx->ReadBuffer) == Y_0_TOP) {
/* both buffers will have the same orientation */
readY = ctx->ReadBuffer->Height - readY - readH;
drawY = ctx->DrawBuffer->Height - drawY - readH;
}
u_box_2d(readX, readY, readW, readH, &srcBox);
pipe->resource_copy_region(pipe,
rbDraw->texture,
rbDraw->rtt_level, drawX, drawY, 0,
rbRead->texture,
rbRead->rtt_level, &srcBox);
return GL_TRUE;
}
}
return GL_FALSE;
}
/**
* RGBA copypixels
*/
static void
copy_rgba_pixels(GLcontext *ctx, GLint srcx, GLint srcy,
GLint width, GLint height, GLint destx, GLint desty)
{
SWcontext *swrast = SWRAST_CONTEXT(ctx);
GLfloat *tmpImage, *p;
GLint sy, dy, stepy, row;
const GLboolean zoom = ctx->Pixel.ZoomX != 1.0F || ctx->Pixel.ZoomY != 1.0F;
GLint overlapping;
GLuint transferOps = ctx->_ImageTransferState;
SWspan span;
if (!ctx->ReadBuffer->_ColorReadBuffer) {
/* no readbuffer - OK */
return;
}
if (ctx->Pixel.Convolution2DEnabled || ctx->Pixel.Separable2DEnabled) {
copy_conv_rgba_pixels(ctx, srcx, srcy, width, height, destx, desty);
return;
}
else if (ctx->Pixel.Convolution1DEnabled) {
/* make sure we don't apply 1D convolution */
transferOps &= ~(IMAGE_CONVOLUTION_BIT |
IMAGE_POST_CONVOLUTION_SCALE_BIAS);
}
if (ctx->DrawBuffer == ctx->ReadBuffer) {
overlapping = regions_overlap(srcx, srcy, destx, desty, width, height,
ctx->Pixel.ZoomX, ctx->Pixel.ZoomY);
}
else {
overlapping = GL_FALSE;
}
/* Determine if copy should be done bottom-to-top or top-to-bottom */
if (!overlapping && srcy < desty) {
/* top-down max-to-min */
sy = srcy + height - 1;
dy = desty + height - 1;
stepy = -1;
}
else {
/* bottom-up min-to-max */
sy = srcy;
dy = desty;
stepy = 1;
}
INIT_SPAN(span, GL_BITMAP, 0, 0, SPAN_RGBA);
if (ctx->Depth.Test)
_swrast_span_default_z(ctx, &span);
if (swrast->_FogEnabled)
_swrast_span_default_fog(ctx, &span);
_swrast_span_default_secondary_color(ctx, &span);
if (overlapping) {
tmpImage = (GLfloat *) _mesa_malloc(width * height * sizeof(GLfloat) * 4);
if (!tmpImage) {
_mesa_error( ctx, GL_OUT_OF_MEMORY, "glCopyPixels" );
return;
}
/* read the source image as RGBA/float */
p = tmpImage;
for (row = 0; row < height; row++) {
_swrast_read_rgba_span( ctx, ctx->ReadBuffer->_ColorReadBuffer,
width, srcx, sy + row, GL_FLOAT, p );
p += width * 4;
}
p = tmpImage;
}
else {
tmpImage = NULL; /* silence compiler warnings */
p = NULL;
}
ASSERT(width < MAX_WIDTH);
for (row = 0; row < height; row++, sy += stepy, dy += stepy) {
GLvoid *rgba = span.array->attribs[FRAG_ATTRIB_COL0];
/* Get row/span of source pixels */
if (overlapping) {
/* get from buffered image */
_mesa_memcpy(rgba, p, width * sizeof(GLfloat) * 4);
p += width * 4;
}
else {
/* get from framebuffer */
_swrast_read_rgba_span( ctx, ctx->ReadBuffer->_ColorReadBuffer,
width, srcx, sy, GL_FLOAT, rgba );
}
if (transferOps) {
_mesa_apply_rgba_transfer_ops(ctx, transferOps, width,
(GLfloat (*)[4]) rgba);
}
/* Write color span */
span.x = destx;
span.y = dy;
span.end = width;
span.array->ChanType = GL_FLOAT;
if (zoom) {
_swrast_write_zoomed_rgba_span(ctx, destx, desty, &span, rgba);
}
else {
_swrast_write_rgba_span(ctx, &span);
}
}
span.array->ChanType = CHAN_TYPE; /* restore */
if (overlapping)
_mesa_free(tmpImage);
}
/*
* TODO: Optimize!!!!
*/
static void
copy_depth_pixels( GLcontext *ctx, GLint srcx, GLint srcy,
GLint width, GLint height,
GLint destx, GLint desty )
{
struct gl_framebuffer *fb = ctx->ReadBuffer;
struct gl_renderbuffer *readRb = fb->_DepthBuffer;
GLfloat *p, *tmpImage;
GLint sy, dy, stepy;
GLint j;
const GLboolean zoom = ctx->Pixel.ZoomX != 1.0F || ctx->Pixel.ZoomY != 1.0F;
GLint overlapping;
SWspan span;
if (!readRb) {
/* no readbuffer - OK */
return;
}
INIT_SPAN(span, GL_BITMAP);
_swrast_span_default_attribs(ctx, &span);
span.arrayMask = SPAN_Z;
if (ctx->DrawBuffer == ctx->ReadBuffer) {
overlapping = regions_overlap(srcx, srcy, destx, desty, width, height,
ctx->Pixel.ZoomX, ctx->Pixel.ZoomY);
}
else {
overlapping = GL_FALSE;
}
/* Determine if copy should be bottom-to-top or top-to-bottom */
if (!overlapping && srcy < desty) {
/* top-down max-to-min */
sy = srcy + height - 1;
dy = desty + height - 1;
stepy = -1;
}
else {
/* bottom-up min-to-max */
sy = srcy;
dy = desty;
stepy = 1;
}
if (overlapping) {
GLint ssy = sy;
tmpImage = (GLfloat *) malloc(width * height * sizeof(GLfloat));
if (!tmpImage) {
_mesa_error( ctx, GL_OUT_OF_MEMORY, "glCopyPixels" );
return;
}
p = tmpImage;
for (j = 0; j < height; j++, ssy += stepy) {
_swrast_read_depth_span_float(ctx, readRb, width, srcx, ssy, p);
p += width;
}
p = tmpImage;
}
else {
tmpImage = NULL; /* silence compiler warning */
p = NULL;
}
for (j = 0; j < height; j++, sy += stepy, dy += stepy) {
GLfloat depth[MAX_WIDTH];
/* get depth values */
if (overlapping) {
memcpy(depth, p, width * sizeof(GLfloat));
p += width;
}
else {
_swrast_read_depth_span_float(ctx, readRb, width, srcx, sy, depth);
}
/* apply scale and bias */
scale_and_bias_z(ctx, width, depth, span.array->z);
/* write depth values */
span.x = destx;
span.y = dy;
span.end = width;
if (zoom)
_swrast_write_zoomed_depth_span(ctx, destx, desty, &span);
else
_swrast_write_rgba_span(ctx, &span);
}
if (overlapping)
free(tmpImage);
}
/**
* Copy pixel block from src surface to dst surface.
* Overlapping regions are acceptable.
* Flipping and stretching are supported.
* \param filter one of PIPE_TEX_MIPFILTER_NEAREST/LINEAR
* \param writemask controls which channels in the dest surface are sourced
* from the src surface. Disabled channels are sourced
* from (0,0,0,1).
*/
void
util_blit_pixels(struct blit_state *ctx,
struct pipe_resource *src_tex,
unsigned src_level,
int srcX0, int srcY0,
int srcX1, int srcY1,
int srcZ0,
struct pipe_surface *dst,
int dstX0, int dstY0,
int dstX1, int dstY1,
float z, uint filter,
uint writemask, uint zs_writemask)
{
struct pipe_context *pipe = ctx->pipe;
struct pipe_screen *screen = pipe->screen;
enum pipe_format src_format, dst_format;
struct pipe_sampler_view *sampler_view = NULL;
struct pipe_sampler_view sv_templ;
struct pipe_surface *dst_surface;
struct pipe_framebuffer_state fb;
const int srcW = abs(srcX1 - srcX0);
const int srcH = abs(srcY1 - srcY0);
unsigned offset;
boolean overlap;
float s0, t0, s1, t1;
boolean normalized;
boolean is_stencil, is_depth, blit_depth, blit_stencil;
const struct util_format_description *src_desc =
util_format_description(src_tex->format);
assert(filter == PIPE_TEX_MIPFILTER_NEAREST ||
filter == PIPE_TEX_MIPFILTER_LINEAR);
assert(src_level <= src_tex->last_level);
/* do the regions overlap? */
overlap = src_tex == dst->texture &&
dst->u.tex.level == src_level &&
dst->u.tex.first_layer == srcZ0 &&
regions_overlap(srcX0, srcY0, srcX1, srcY1,
dstX0, dstY0, dstX1, dstY1);
src_format = util_format_linear(src_tex->format);
dst_format = util_format_linear(dst->texture->format);
/* See whether we will blit depth or stencil. */
is_depth = util_format_has_depth(src_desc);
is_stencil = util_format_has_stencil(src_desc);
blit_depth = is_depth && (zs_writemask & BLIT_WRITEMASK_Z);
blit_stencil = is_stencil && (zs_writemask & BLIT_WRITEMASK_STENCIL);
assert((writemask && !zs_writemask && !is_depth && !is_stencil) ||
(!writemask && (blit_depth || blit_stencil)));
/*
* Check for simple case: no format conversion, no flipping, no stretching,
* no overlapping, same number of samples.
* Filter mode should not matter since there's no stretching.
*/
if (formats_compatible(src_format, dst_format) &&
src_tex->nr_samples == dst->texture->nr_samples &&
is_stencil == blit_stencil &&
is_depth == blit_depth &&
srcX0 < srcX1 &&
dstX0 < dstX1 &&
srcY0 < srcY1 &&
dstY0 < dstY1 &&
(dstX1 - dstX0) == (srcX1 - srcX0) &&
(dstY1 - dstY0) == (srcY1 - srcY0) &&
!overlap) {
struct pipe_box src_box;
src_box.x = srcX0;
src_box.y = srcY0;
src_box.z = srcZ0;
src_box.width = srcW;
src_box.height = srcH;
src_box.depth = 1;
pipe->resource_copy_region(pipe,
dst->texture, dst->u.tex.level,
dstX0, dstY0, dst->u.tex.first_layer,/* dest */
src_tex, src_level,
&src_box);
return;
}
/* XXX Reading multisample textures is unimplemented. */
assert(src_tex->nr_samples <= 1);
if (src_tex->nr_samples > 1) {
return;
}
/* It's a mistake to call this function with a stencil format and
* without shader stencil export. We don't do software fallbacks here.
* Ignore stencil and only copy depth.
*/
if (blit_stencil && !ctx->has_stencil_export) {
blit_stencil = FALSE;
if (!blit_depth)
return;
}
if (dst_format == dst->format) {
dst_surface = dst;
} else {
struct pipe_surface templ = *dst;
templ.format = dst_format;
dst_surface = pipe->create_surface(pipe, dst->texture, &templ);
}
/* Create a temporary texture when src and dest alias.
*/
if (src_tex == dst_surface->texture &&
dst_surface->u.tex.level == src_level &&
dst_surface->u.tex.first_layer == srcZ0) {
/* Make a temporary texture which contains a copy of the source pixels.
* Then we'll sample from the temporary texture.
*/
struct pipe_resource texTemp;
struct pipe_resource *tex;
struct pipe_sampler_view sv_templ;
struct pipe_box src_box;
const int srcLeft = MIN2(srcX0, srcX1);
const int srcTop = MIN2(srcY0, srcY1);
if (srcLeft != srcX0) {
/* left-right flip */
int tmp = dstX0;
dstX0 = dstX1;
dstX1 = tmp;
}
if (srcTop != srcY0) {
/* up-down flip */
int tmp = dstY0;
dstY0 = dstY1;
dstY1 = tmp;
}
/* create temp texture */
memset(&texTemp, 0, sizeof(texTemp));
texTemp.target = ctx->internal_target;
texTemp.format = src_format;
texTemp.last_level = 0;
texTemp.width0 = srcW;
texTemp.height0 = srcH;
texTemp.depth0 = 1;
texTemp.array_size = 1;
texTemp.bind = PIPE_BIND_SAMPLER_VIEW;
tex = screen->resource_create(screen, &texTemp);
if (!tex)
return;
src_box.x = srcLeft;
src_box.y = srcTop;
src_box.z = srcZ0;
src_box.width = srcW;
src_box.height = srcH;
src_box.depth = 1;
/* load temp texture */
pipe->resource_copy_region(pipe,
tex, 0, 0, 0, 0, /* dest */
src_tex, src_level, &src_box);
normalized = tex->target != PIPE_TEXTURE_RECT;
if(normalized) {
s0 = 0.0f;
s1 = 1.0f;
t0 = 0.0f;
t1 = 1.0f;
}
else {
s0 = 0;
s1 = srcW;
t0 = 0;
t1 = srcH;
}
u_sampler_view_default_template(&sv_templ, tex, tex->format);
if (!blit_depth && blit_stencil) {
/* set a stencil-only format, e.g. Z24S8 --> X24S8 */
sv_templ.format = util_format_stencil_only(tex->format);
assert(sv_templ.format != PIPE_FORMAT_NONE);
}
sampler_view = pipe->create_sampler_view(pipe, tex, &sv_templ);
if (!sampler_view) {
pipe_resource_reference(&tex, NULL);
return;
}
pipe_resource_reference(&tex, NULL);
}
else {
/* Directly sample from the source resource/texture */
u_sampler_view_default_template(&sv_templ, src_tex, src_format);
if (!blit_depth && blit_stencil) {
/* set a stencil-only format, e.g. Z24S8 --> X24S8 */
sv_templ.format = util_format_stencil_only(src_format);
assert(sv_templ.format != PIPE_FORMAT_NONE);
}
sampler_view = pipe->create_sampler_view(pipe, src_tex, &sv_templ);
if (!sampler_view) {
return;
}
s0 = srcX0;
s1 = srcX1;
t0 = srcY0;
t1 = srcY1;
normalized = sampler_view->texture->target != PIPE_TEXTURE_RECT;
if(normalized)
{
s0 /= (float)(u_minify(sampler_view->texture->width0, src_level));
s1 /= (float)(u_minify(sampler_view->texture->width0, src_level));
t0 /= (float)(u_minify(sampler_view->texture->height0, src_level));
t1 /= (float)(u_minify(sampler_view->texture->height0, src_level));
}
}
assert(screen->is_format_supported(screen, sampler_view->format,
ctx->internal_target, sampler_view->texture->nr_samples,
PIPE_BIND_SAMPLER_VIEW));
assert(screen->is_format_supported(screen, dst_format, ctx->internal_target,
dst_surface->texture->nr_samples,
is_depth || is_stencil ? PIPE_BIND_DEPTH_STENCIL :
PIPE_BIND_RENDER_TARGET));
/* save state (restored below) */
cso_save_blend(ctx->cso);
cso_save_depth_stencil_alpha(ctx->cso);
cso_save_rasterizer(ctx->cso);
cso_save_sample_mask(ctx->cso);
cso_save_samplers(ctx->cso, PIPE_SHADER_FRAGMENT);
cso_save_sampler_views(ctx->cso, PIPE_SHADER_FRAGMENT);
cso_save_stream_outputs(ctx->cso);
cso_save_viewport(ctx->cso);
cso_save_framebuffer(ctx->cso);
cso_save_fragment_shader(ctx->cso);
cso_save_vertex_shader(ctx->cso);
cso_save_geometry_shader(ctx->cso);
cso_save_vertex_elements(ctx->cso);
cso_save_aux_vertex_buffer_slot(ctx->cso);
cso_save_render_condition(ctx->cso);
/* set misc state we care about */
if (writemask)
cso_set_blend(ctx->cso, &ctx->blend_write_color);
else
cso_set_blend(ctx->cso, &ctx->blend_keep_color);
cso_set_sample_mask(ctx->cso, ~0);
cso_set_rasterizer(ctx->cso, &ctx->rasterizer);
cso_set_vertex_elements(ctx->cso, 2, ctx->velem);
cso_set_stream_outputs(ctx->cso, 0, NULL, 0);
cso_set_render_condition(ctx->cso, NULL, 0);
/* default sampler state */
ctx->sampler.normalized_coords = normalized;
ctx->sampler.min_img_filter = filter;
ctx->sampler.mag_img_filter = filter;
ctx->sampler.min_lod = src_level;
ctx->sampler.max_lod = src_level;
/* Depth stencil state, fragment shader and sampler setup depending on what
* we blit.
*/
if (blit_depth && blit_stencil) {
cso_single_sampler(ctx->cso, PIPE_SHADER_FRAGMENT, 0, &ctx->sampler);
/* don't filter stencil */
ctx->sampler.min_img_filter = PIPE_TEX_FILTER_NEAREST;
ctx->sampler.mag_img_filter = PIPE_TEX_FILTER_NEAREST;
cso_single_sampler(ctx->cso, PIPE_SHADER_FRAGMENT, 1, &ctx->sampler);
cso_set_depth_stencil_alpha(ctx->cso, &ctx->dsa_write_depthstencil);
set_depthstencil_fragment_shader(ctx, sampler_view->texture->target);
}
else if (blit_depth) {
cso_single_sampler(ctx->cso, PIPE_SHADER_FRAGMENT, 0, &ctx->sampler);
cso_set_depth_stencil_alpha(ctx->cso, &ctx->dsa_write_depth);
set_depth_fragment_shader(ctx, sampler_view->texture->target);
}
else if (blit_stencil) {
/* don't filter stencil */
ctx->sampler.min_img_filter = PIPE_TEX_FILTER_NEAREST;
ctx->sampler.mag_img_filter = PIPE_TEX_FILTER_NEAREST;
cso_single_sampler(ctx->cso, PIPE_SHADER_FRAGMENT, 0, &ctx->sampler);
cso_set_depth_stencil_alpha(ctx->cso, &ctx->dsa_write_stencil);
set_stencil_fragment_shader(ctx, sampler_view->texture->target);
}
else { /* color */
cso_single_sampler(ctx->cso, PIPE_SHADER_FRAGMENT, 0, &ctx->sampler);
cso_set_depth_stencil_alpha(ctx->cso, &ctx->dsa_keep_depthstencil);
set_fragment_shader(ctx, writemask, sampler_view->texture->target);
}
cso_single_sampler_done(ctx->cso, PIPE_SHADER_FRAGMENT);
/* textures */
if (blit_depth && blit_stencil) {
/* Setup two samplers, one for depth and the other one for stencil. */
struct pipe_sampler_view templ;
struct pipe_sampler_view *views[2];
templ = *sampler_view;
templ.format = util_format_stencil_only(templ.format);
assert(templ.format != PIPE_FORMAT_NONE);
views[0] = sampler_view;
views[1] = pipe->create_sampler_view(pipe, views[0]->texture, &templ);
cso_set_sampler_views(ctx->cso, PIPE_SHADER_FRAGMENT, 2, views);
pipe_sampler_view_reference(&views[1], NULL);
}
else {
cso_set_sampler_views(ctx->cso, PIPE_SHADER_FRAGMENT, 1, &sampler_view);
}
/* viewport */
ctx->viewport.scale[0] = 0.5f * dst_surface->width;
ctx->viewport.scale[1] = 0.5f * dst_surface->height;
ctx->viewport.scale[2] = 0.5f;
ctx->viewport.scale[3] = 1.0f;
ctx->viewport.translate[0] = 0.5f * dst_surface->width;
ctx->viewport.translate[1] = 0.5f * dst_surface->height;
ctx->viewport.translate[2] = 0.5f;
ctx->viewport.translate[3] = 0.0f;
cso_set_viewport(ctx->cso, &ctx->viewport);
set_vertex_shader(ctx);
cso_set_geometry_shader_handle(ctx->cso, NULL);
/* drawing dest */
memset(&fb, 0, sizeof(fb));
fb.width = dst_surface->width;
fb.height = dst_surface->height;
if (blit_depth || blit_stencil) {
fb.zsbuf = dst_surface;
} else {
fb.nr_cbufs = 1;
fb.cbufs[0] = dst_surface;
}
cso_set_framebuffer(ctx->cso, &fb);
/* draw quad */
offset = setup_vertex_data_tex(ctx,
(float) dstX0 / dst_surface->width * 2.0f - 1.0f,
(float) dstY0 / dst_surface->height * 2.0f - 1.0f,
(float) dstX1 / dst_surface->width * 2.0f - 1.0f,
(float) dstY1 / dst_surface->height * 2.0f - 1.0f,
s0, t0,
s1, t1,
z);
if (ctx->vbuf) {
util_draw_vertex_buffer(ctx->pipe, ctx->cso, ctx->vbuf,
cso_get_aux_vertex_buffer_slot(ctx->cso),
offset,
PIPE_PRIM_TRIANGLE_FAN,
4, /* verts */
2); /* attribs/vert */
}
/* restore state we changed */
cso_restore_blend(ctx->cso);
cso_restore_depth_stencil_alpha(ctx->cso);
cso_restore_rasterizer(ctx->cso);
cso_restore_sample_mask(ctx->cso);
cso_restore_samplers(ctx->cso, PIPE_SHADER_FRAGMENT);
cso_restore_sampler_views(ctx->cso, PIPE_SHADER_FRAGMENT);
cso_restore_viewport(ctx->cso);
cso_restore_framebuffer(ctx->cso);
cso_restore_fragment_shader(ctx->cso);
cso_restore_vertex_shader(ctx->cso);
cso_restore_geometry_shader(ctx->cso);
cso_restore_vertex_elements(ctx->cso);
cso_restore_aux_vertex_buffer_slot(ctx->cso);
cso_restore_stream_outputs(ctx->cso);
cso_restore_render_condition(ctx->cso);
pipe_sampler_view_reference(&sampler_view, NULL);
if (dst_surface != dst)
pipe_surface_reference(&dst_surface, NULL);
}
static bool
opt_cmod_propagation_local(const gen_device_info *devinfo, bblock_t *block)
{
bool progress = false;
int ip = block->end_ip + 1;
foreach_inst_in_block_reverse_safe(fs_inst, inst, block) {
ip--;
if ((inst->opcode != BRW_OPCODE_AND &&
inst->opcode != BRW_OPCODE_CMP &&
inst->opcode != BRW_OPCODE_MOV) ||
inst->predicate != BRW_PREDICATE_NONE ||
!inst->dst.is_null() ||
inst->src[0].file != VGRF ||
inst->src[0].abs)
continue;
/* Only an AND.NZ can be propagated. Many AND.Z instructions are
* generated (for ir_unop_not in fs_visitor::emit_bool_to_cond_code).
* Propagating those would require inverting the condition on the CMP.
* This changes both the flag value and the register destination of the
* CMP. That result may be used elsewhere, so we can't change its value
* on a whim.
*/
if (inst->opcode == BRW_OPCODE_AND &&
!(inst->src[1].is_one() &&
inst->conditional_mod == BRW_CONDITIONAL_NZ &&
!inst->src[0].negate))
continue;
if (inst->opcode == BRW_OPCODE_CMP && !inst->src[1].is_zero())
continue;
if (inst->opcode == BRW_OPCODE_MOV &&
inst->conditional_mod != BRW_CONDITIONAL_NZ)
continue;
bool read_flag = false;
foreach_inst_in_block_reverse_starting_from(fs_inst, scan_inst, inst) {
if (regions_overlap(scan_inst->dst, scan_inst->size_written,
inst->src[0], inst->size_read(0))) {
if (scan_inst->is_partial_write() ||
scan_inst->dst.offset != inst->src[0].offset ||
scan_inst->exec_size != inst->exec_size)
break;
/* CMP's result is the same regardless of dest type. */
if (inst->conditional_mod == BRW_CONDITIONAL_NZ &&
scan_inst->opcode == BRW_OPCODE_CMP &&
(inst->dst.type == BRW_REGISTER_TYPE_D ||
inst->dst.type == BRW_REGISTER_TYPE_UD)) {
inst->remove(block);
progress = true;
break;
}
/* If the AND wasn't handled by the previous case, it isn't safe
* to remove it.
*/
if (inst->opcode == BRW_OPCODE_AND)
break;
/* Comparisons operate differently for ints and floats */
if (scan_inst->dst.type != inst->dst.type &&
(scan_inst->dst.type == BRW_REGISTER_TYPE_F ||
inst->dst.type == BRW_REGISTER_TYPE_F))
break;
/* If the instruction generating inst's source also wrote the
* flag, and inst is doing a simple .nz comparison, then inst
* is redundant - the appropriate value is already in the flag
* register. Delete inst.
*/
if (inst->conditional_mod == BRW_CONDITIONAL_NZ &&
!inst->src[0].negate &&
scan_inst->flags_written()) {
inst->remove(block);
progress = true;
break;
}
/* The conditional mod of the CMP/CMPN instructions behaves
* specially because the flag output is not calculated from the
* result of the instruction, but the other way around, which
* means that even if the condmod to propagate and the condmod
* from the CMP instruction are the same they will in general give
* different results because they are evaluated based on different
* inputs.
*/
if (scan_inst->opcode == BRW_OPCODE_CMP ||
scan_inst->opcode == BRW_OPCODE_CMPN)
break;
/* Otherwise, try propagating the conditional. */
enum brw_conditional_mod cond =
inst->src[0].negate ? brw_swap_cmod(inst->conditional_mod)
: inst->conditional_mod;
if (scan_inst->can_do_cmod() &&
((!read_flag && scan_inst->conditional_mod == BRW_CONDITIONAL_NONE) ||
scan_inst->conditional_mod == cond)) {
scan_inst->conditional_mod = cond;
inst->remove(block);
progress = true;
}
break;
}
if (scan_inst->flags_written())
break;
read_flag = read_flag || scan_inst->flags_read(devinfo);
}
}
/**
* RGBA copypixels
*/
static void
copy_rgba_pixels(struct gl_context *ctx, GLint srcx, GLint srcy,
GLint width, GLint height, GLint destx, GLint desty)
{
GLfloat *tmpImage, *p;
GLint sy, dy, stepy, row;
const GLboolean zoom = ctx->Pixel.ZoomX != 1.0F || ctx->Pixel.ZoomY != 1.0F;
GLint overlapping;
GLuint transferOps = ctx->_ImageTransferState;
SWspan span;
if (!ctx->ReadBuffer->_ColorReadBuffer) {
/* no readbuffer - OK */
return;
}
if (ctx->DrawBuffer == ctx->ReadBuffer) {
overlapping = regions_overlap(srcx, srcy, destx, desty, width, height,
ctx->Pixel.ZoomX, ctx->Pixel.ZoomY);
}
else {
overlapping = GL_FALSE;
}
/* Determine if copy should be done bottom-to-top or top-to-bottom */
if (!overlapping && srcy < desty) {
/* top-down max-to-min */
sy = srcy + height - 1;
dy = desty + height - 1;
stepy = -1;
}
else {
/* bottom-up min-to-max */
sy = srcy;
dy = desty;
stepy = 1;
}
INIT_SPAN(span, GL_BITMAP);
_swrast_span_default_attribs(ctx, &span);
span.arrayMask = SPAN_RGBA;
span.arrayAttribs = VARYING_BIT_COL0; /* we'll fill in COL0 attrib values */
if (overlapping) {
tmpImage = malloc(width * height * sizeof(GLfloat) * 4);
if (!tmpImage) {
_mesa_error( ctx, GL_OUT_OF_MEMORY, "glCopyPixels" );
return;
}
/* read the source image as RGBA/float */
p = tmpImage;
for (row = 0; row < height; row++) {
_swrast_read_rgba_span( ctx, ctx->ReadBuffer->_ColorReadBuffer,
width, srcx, sy + row, p );
p += width * 4;
}
p = tmpImage;
}
else {
tmpImage = NULL; /* silence compiler warnings */
p = NULL;
}
ASSERT(width < SWRAST_MAX_WIDTH);
for (row = 0; row < height; row++, sy += stepy, dy += stepy) {
GLvoid *rgba = span.array->attribs[VARYING_SLOT_COL0];
/* Get row/span of source pixels */
if (overlapping) {
/* get from buffered image */
memcpy(rgba, p, width * sizeof(GLfloat) * 4);
p += width * 4;
}
else {
/* get from framebuffer */
_swrast_read_rgba_span( ctx, ctx->ReadBuffer->_ColorReadBuffer,
width, srcx, sy, rgba );
}
if (transferOps) {
_mesa_apply_rgba_transfer_ops(ctx, transferOps, width,
(GLfloat (*)[4]) rgba);
}
/* Write color span */
span.x = destx;
span.y = dy;
span.end = width;
span.array->ChanType = GL_FLOAT;
if (zoom) {
_swrast_write_zoomed_rgba_span(ctx, destx, desty, &span, rgba);
}
else {
_swrast_write_rgba_span(ctx, &span);
}
}
span.array->ChanType = CHAN_TYPE; /* restore */
if (overlapping)
free(tmpImage);
}
static bool
opt_saturate_propagation_local(fs_visitor *v, bblock_t *block)
{
bool progress = false;
int ip = block->end_ip + 1;
foreach_inst_in_block_reverse(fs_inst, inst, block) {
ip--;
if (inst->opcode != BRW_OPCODE_MOV ||
!inst->saturate ||
inst->dst.file != VGRF ||
inst->dst.type != inst->src[0].type ||
inst->src[0].file != VGRF ||
inst->src[0].abs)
continue;
int src_var = v->live_intervals->var_from_reg(inst->src[0]);
int src_end_ip = v->live_intervals->end[src_var];
bool interfered = false;
foreach_inst_in_block_reverse_starting_from(fs_inst, scan_inst, inst) {
if (regions_overlap(scan_inst->dst, scan_inst->size_written,
inst->src[0], inst->size_read(0))) {
if (scan_inst->is_partial_write() ||
(scan_inst->dst.type != inst->dst.type &&
!scan_inst->can_change_types()))
break;
if (scan_inst->saturate) {
inst->saturate = false;
progress = true;
} else if (src_end_ip == ip || inst->dst.equals(inst->src[0])) {
if (scan_inst->can_do_saturate()) {
if (scan_inst->dst.type != inst->dst.type) {
scan_inst->dst.type = inst->dst.type;
for (int i = 0; i < scan_inst->sources; i++) {
scan_inst->src[i].type = inst->dst.type;
}
}
if (inst->src[0].negate) {
if (scan_inst->opcode == BRW_OPCODE_MUL) {
scan_inst->src[0].negate = !scan_inst->src[0].negate;
inst->src[0].negate = false;
} else if (scan_inst->opcode == BRW_OPCODE_MAD) {
scan_inst->src[0].negate = !scan_inst->src[0].negate;
scan_inst->src[1].negate = !scan_inst->src[1].negate;
inst->src[0].negate = false;
} else if (scan_inst->opcode == BRW_OPCODE_ADD) {
if (scan_inst->src[1].file == IMM) {
if (!brw_negate_immediate(scan_inst->src[1].type,
&scan_inst->src[1].as_brw_reg())) {
break;
}
} else {
scan_inst->src[1].negate = !scan_inst->src[1].negate;
}
scan_inst->src[0].negate = !scan_inst->src[0].negate;
inst->src[0].negate = false;
} else {
break;
}
}
scan_inst->saturate = true;
inst->saturate = false;
progress = true;
}
}
break;
}
for (int i = 0; i < scan_inst->sources; i++) {
if (scan_inst->src[i].file == VGRF &&
scan_inst->src[i].nr == inst->src[0].nr &&
scan_inst->src[i].offset / REG_SIZE ==
inst->src[0].offset / REG_SIZE) {
if (scan_inst->opcode != BRW_OPCODE_MOV ||
!scan_inst->saturate ||
scan_inst->src[0].abs ||
scan_inst->src[0].negate ||
scan_inst->src[0].abs != inst->src[0].abs ||
scan_inst->src[0].negate != inst->src[0].negate) {
interfered = true;
break;
}
}
}
if (interfered)
break;
}
}
/**
* Copy pixel block from src surface to dst surface.
* Overlapping regions are acceptable.
* Flipping and stretching are supported.
* \param filter one of PIPE_TEX_MIPFILTER_NEAREST/LINEAR
* \param writemask controls which channels in the dest surface are sourced
* from the src surface. Disabled channels are sourced
* from (0,0,0,1).
*/
void
util_blit_pixels(struct blit_state *ctx,
struct pipe_resource *src_tex,
unsigned src_level,
int srcX0, int srcY0,
int srcX1, int srcY1,
int srcZ0,
struct pipe_surface *dst,
int dstX0, int dstY0,
int dstX1, int dstY1,
float z, uint filter,
uint writemask, uint zs_writemask)
{
struct pipe_context *pipe = ctx->pipe;
enum pipe_format src_format, dst_format;
const int srcW = abs(srcX1 - srcX0);
const int srcH = abs(srcY1 - srcY0);
boolean overlap;
boolean is_stencil, is_depth, blit_depth, blit_stencil;
const struct util_format_description *src_desc =
util_format_description(src_tex->format);
struct pipe_blit_info info;
assert(filter == PIPE_TEX_MIPFILTER_NEAREST ||
filter == PIPE_TEX_MIPFILTER_LINEAR);
assert(src_level <= src_tex->last_level);
/* do the regions overlap? */
overlap = src_tex == dst->texture &&
dst->u.tex.level == src_level &&
dst->u.tex.first_layer == srcZ0 &&
regions_overlap(srcX0, srcY0, srcX1, srcY1,
dstX0, dstY0, dstX1, dstY1);
src_format = util_format_linear(src_tex->format);
dst_format = util_format_linear(dst->texture->format);
/* See whether we will blit depth or stencil. */
is_depth = util_format_has_depth(src_desc);
is_stencil = util_format_has_stencil(src_desc);
blit_depth = is_depth && (zs_writemask & BLIT_WRITEMASK_Z);
blit_stencil = is_stencil && (zs_writemask & BLIT_WRITEMASK_STENCIL);
assert((writemask && !zs_writemask && !is_depth && !is_stencil) ||
(!writemask && (blit_depth || blit_stencil)));
/*
* XXX: z parameter is deprecated. dst->u.tex.first_layer
* specificies the destination layer.
*/
assert(z == 0.0f);
/*
* Check for simple case: no format conversion, no flipping, no stretching,
* no overlapping, same number of samples.
* Filter mode should not matter since there's no stretching.
*/
if (formats_compatible(src_format, dst_format) &&
src_tex->nr_samples == dst->texture->nr_samples &&
is_stencil == blit_stencil &&
is_depth == blit_depth &&
srcX0 < srcX1 &&
dstX0 < dstX1 &&
srcY0 < srcY1 &&
dstY0 < dstY1 &&
(dstX1 - dstX0) == (srcX1 - srcX0) &&
(dstY1 - dstY0) == (srcY1 - srcY0) &&
!overlap) {
struct pipe_box src_box;
src_box.x = srcX0;
src_box.y = srcY0;
src_box.z = srcZ0;
src_box.width = srcW;
src_box.height = srcH;
src_box.depth = 1;
pipe->resource_copy_region(pipe,
dst->texture, dst->u.tex.level,
dstX0, dstY0, dst->u.tex.first_layer,/* dest */
src_tex, src_level,
&src_box);
return;
}
memset(&info, 0, sizeof info);
info.dst.resource = dst->texture;
info.dst.level = dst->u.tex.level;
info.dst.box.x = dstX0;
info.dst.box.y = dstY0;
info.dst.box.z = dst->u.tex.first_layer;
info.dst.box.width = dstX1 - dstX0;
info.dst.box.height = dstY1 - dstY0;
assert(info.dst.box.width >= 0);
assert(info.dst.box.height >= 0);
info.dst.box.depth = 1;
info.dst.format = dst->texture->format;
info.src.resource = src_tex;
info.src.level = src_level;
info.src.box.x = srcX0;
info.src.box.y = srcY0;
info.src.box.z = srcZ0;
info.src.box.width = srcX1 - srcX0;
info.src.box.height = srcY1 - srcY0;
info.src.box.depth = 1;
info.src.format = src_tex->format;
info.mask = writemask | (zs_writemask << 4);
info.filter = filter;
info.scissor_enable = 0;
pipe->blit(pipe, &info);
}
static void
copy_ci_pixels( GLcontext *ctx, GLint srcx, GLint srcy,
GLint width, GLint height,
GLint destx, GLint desty )
{
GLuint *tmpImage,*p;
GLint sy, dy, stepy;
GLint j;
const GLboolean zoom = ctx->Pixel.ZoomX != 1.0F || ctx->Pixel.ZoomY != 1.0F;
GLint overlapping;
SWspan span;
if (!ctx->ReadBuffer->_ColorReadBuffer) {
/* no readbuffer - OK */
return;
}
INIT_SPAN(span, GL_BITMAP);
_swrast_span_default_attribs(ctx, &span);
span.arrayMask = SPAN_INDEX;
if (ctx->DrawBuffer == ctx->ReadBuffer) {
overlapping = regions_overlap(srcx, srcy, destx, desty, width, height,
ctx->Pixel.ZoomX, ctx->Pixel.ZoomY);
}
else {
overlapping = GL_FALSE;
}
/* Determine if copy should be bottom-to-top or top-to-bottom */
if (!overlapping && srcy < desty) {
/* top-down max-to-min */
sy = srcy + height - 1;
dy = desty + height - 1;
stepy = -1;
}
else {
/* bottom-up min-to-max */
sy = srcy;
dy = desty;
stepy = 1;
}
if (overlapping) {
GLint ssy = sy;
tmpImage = (GLuint *) _mesa_malloc(width * height * sizeof(GLuint));
if (!tmpImage) {
_mesa_error( ctx, GL_OUT_OF_MEMORY, "glCopyPixels" );
return;
}
/* read the image */
p = tmpImage;
for (j = 0; j < height; j++, ssy += stepy) {
_swrast_read_index_span( ctx, ctx->ReadBuffer->_ColorReadBuffer,
width, srcx, ssy, p );
p += width;
}
p = tmpImage;
}
else {
tmpImage = NULL; /* silence compiler warning */
p = NULL;
}
for (j = 0; j < height; j++, sy += stepy, dy += stepy) {
/* Get color indexes */
if (overlapping) {
_mesa_memcpy(span.array->index, p, width * sizeof(GLuint));
p += width;
}
else {
_swrast_read_index_span( ctx, ctx->ReadBuffer->_ColorReadBuffer,
width, srcx, sy, span.array->index );
}
if (ctx->_ImageTransferState)
_mesa_apply_ci_transfer_ops(ctx, ctx->_ImageTransferState,
width, span.array->index);
/* write color indexes */
span.x = destx;
span.y = dy;
span.end = width;
if (zoom)
_swrast_write_zoomed_index_span(ctx, destx, desty, &span);
else
_swrast_write_index_span(ctx, &span);
}
if (overlapping)
_mesa_free(tmpImage);
}