Eric Anholt [Fri, 20 Jan 2012 23:23:42 +0000 (15:23 -0800)]
mesa: Add missing integer R/RG cases to _mesa_is_color_format().
This is part of fixing Intel oglconform
negative.typeFormatMismatch.copyteximage.
NOTE: This is a candidate for the 8.0 branch.
Reviewed-by: Brian Paul <brianp@vmware.com>
Eric Anholt [Fri, 20 Jan 2012 21:59:37 +0000 (13:59 -0800)]
meta: Detect currently-unsupported integer CopyTexSubImage and complain.
This code is unprepared for handling integer (particularly, the
baseFormat of the TexFormat comes out as GL_RGBA, not GL_RGBA_INTEGER,
so the direct call of Driver.ReadPixels crashes due to the int vs
non-int error checking not having happened). I'm frankly tempted to
convert this code to MapRenderbuffer/MapTexImage rather than doing it
as meta ops, now that we have that support.
Improves the remaining crash in Intel oglconform for int-textures to
just a rendering failure.
NOTE: This is a candidate for the 8.0 branch.
Reviewed-by: Brian Paul <brianp@vmware.com>
Eric Anholt [Fri, 20 Jan 2012 20:50:34 +0000 (12:50 -0800)]
mesa: Add the remaining from/to types for GL_EXT_texture_integer (and R/RG).
This aborts and crashes in intel oglconform's int-textures into being
just rendering failures. Clamping isn't handled yet.
v2: Add missing "break".
v3: Drop the int/uint distinction, since they don't need different clamping.
NOTE: This is a candidate for the 8.0 branch.
Reviewed-by: Brian Paul <brianp@vmware.com> (v2)
Eric Anholt [Fri, 20 Jan 2012 23:03:12 +0000 (15:03 -0800)]
mesa: Add support for glGetTexImage on GL_TEXTURE_1D_ARRAY
Similarly to how we handle this in texstore, we have to remap height
to depth so that we MapTextureImage each image layer individually.
Fixes part of Intel oglconform's int-textures advanced.fbo.rtt
NOTE: This is a candidate for the 8.0 branch.
Reviewed-by: Brian Paul <brianp@vmware.com>
Eric Anholt [Mon, 23 Jan 2012 23:27:36 +0000 (15:27 -0800)]
mesa: s/GLuint col/GLint col/ to avoid signed/unsigned comparison warnings.
Suggested-by: Brian Paul <brianp@vmware.com>
Eric Anholt [Thu, 19 Jan 2012 17:46:24 +0000 (09:46 -0800)]
mesa: Add support for glGetTexImage() from integer textures.
This is a step toward fixing Intel oglconform's
int-textures advanced.fbo.rtt.
NOTE: This is a candidate for the 8.0 branch.
Reviewed-by: Brian Paul <brianp@vmware.com>
Eric Anholt [Fri, 20 Jan 2012 21:19:10 +0000 (13:19 -0800)]
i965/gen6+: Work around GPU hangs with logic ops on integer textures.
This doesn't result in correct rendering -- GL requires that logic ops
work, while the hardware specs say it doesn't do them. I'm not sure
how we would want to handle this.
NOTE: This is a candidate for the 8.0 branch.
Eric Anholt [Thu, 19 Jan 2012 23:41:12 +0000 (15:41 -0800)]
i965/gen6+: Disable blending, alpha test, and dither on integer FBOs.
Fixes GPU hangs and some rendering failures in piglit
EXT_texture_integer/fbo-blending
NOTE: This is a candidate for the 8.0 branch.
Brian Paul [Tue, 24 Jan 2012 22:13:19 +0000 (15:13 -0700)]
mesa/gdi: remove wmesa_set_renderbuffer_funcs() function
The code is no longer relevant.
Note: this driver is probably broken now. There's no implementation
of ctx->Driver.Map/UnmapRenderbuffer().
Brian Paul [Tue, 24 Jan 2012 22:08:20 +0000 (15:08 -0700)]
dri/swrast: remove obsolete swrast_span.c file from source list
Brian Paul [Tue, 24 Jan 2012 20:32:13 +0000 (13:32 -0700)]
swrast: fix uninitialized variable warning
Brian Paul [Tue, 24 Jan 2012 20:29:05 +0000 (13:29 -0700)]
swrast: make rowStride variable signed in put_z32_values()
As with commit
aed5c8299fe47b8e1728f8140d069bc89d3fa947
Brian Paul [Thu, 19 Jan 2012 00:47:32 +0000 (17:47 -0700)]
intel: use swrast code to map/unmap renderbuffers for swrast rendering
Brian Paul [Thu, 19 Jan 2012 00:39:33 +0000 (17:39 -0700)]
mesa: update comments, fix whitespace in dd.h
Brian Paul [Thu, 19 Jan 2012 00:39:33 +0000 (17:39 -0700)]
swrast: remove unused StoreTexel code
No longer needed since we do all rendering to texture with the buffer
mapping and pixel packing functions.
Brian Paul [Thu, 19 Jan 2012 00:39:33 +0000 (17:39 -0700)]
swrast: use Map/UnmapTextureImage() in framebuffer map/unmap code
When we're actually rendering into a texture, map the texture image
instead of the corresponding renderbuffer. Before, we just copied
a pointer from the texture image to the renderbuffer. This change
will make the code usable by hardware drivers.
Brian Paul [Mon, 16 Jan 2012 20:06:03 +0000 (13:06 -0700)]
mesa: remove ctx->Driver.Map/UnmapTexture() hooks
No longer used anywhere.
Brian Paul [Mon, 16 Jan 2012 20:05:59 +0000 (13:05 -0700)]
nouveau: stop calling ctx->Driver.Map/UnmapTexture()
And remove unused nouveau_texture_map/unmap()
Brian Paul [Mon, 16 Jan 2012 20:05:55 +0000 (13:05 -0700)]
tnl: remove ctx->Driver.Map/UnmapTexture() calls
ctx->Driver.MapTexture() always points to _swrast_map_texture().
We're already reaching into swrast from t_vb_program.c anyway.
This will let us remove the ctx->Driver.Map/UnmapTexture() functions.
Brian Paul [Mon, 16 Jan 2012 20:05:49 +0000 (13:05 -0700)]
swrast: move some renderbuffer functions to s_renderbuffer.c
Brian Paul [Mon, 16 Jan 2012 20:05:44 +0000 (13:05 -0700)]
intel: remove intel_span_supports_format()
It always returned True.
Brian Paul [Mon, 16 Jan 2012 20:05:41 +0000 (13:05 -0700)]
swrast: remove a few extra _mesa_get_format_bytes() calls
Brian Paul [Mon, 16 Jan 2012 20:05:36 +0000 (13:05 -0700)]
mesa: update comments for gl_renderbuffer
Brian Paul [Mon, 16 Jan 2012 20:05:07 +0000 (13:05 -0700)]
mesa/swrast/drivers: remove obsolete gl_renderbuffer fields
This removes the last of the legacy fields from gl_renderbuffer.
Brian Paul [Mon, 16 Jan 2012 20:03:28 +0000 (13:03 -0700)]
dri/swrast: use swrast_renderbuffer type
Brian Paul [Mon, 16 Jan 2012 20:02:19 +0000 (13:02 -0700)]
swrast: use swrast_renderbuffer instead of gl_renderbuffer
Brian Paul [Mon, 16 Jan 2012 20:01:35 +0000 (13:01 -0700)]
osmesa: use swrast_renderbuffer
Brian Paul [Mon, 16 Jan 2012 19:52:21 +0000 (12:52 -0700)]
radeon: derive radeon_renderbuffer from swrast_renderbuffer
Brian Paul [Mon, 16 Jan 2012 19:40:01 +0000 (12:40 -0700)]
intel: derive intel_renderbuffer from swrast_renderbuffer
Drivers that rely on swrast need to do this, as with swrast_texture_image.
Brian Paul [Mon, 16 Jan 2012 19:35:49 +0000 (12:35 -0700)]
xlib: derive xmesa_renderbuffer from swrast_renderbuffer
Brian Paul [Mon, 16 Jan 2012 19:35:45 +0000 (12:35 -0700)]
swrast: allocate swrast_renderbuffers instead of gl_renderbuffers
Brian Paul [Mon, 16 Jan 2012 19:29:15 +0000 (12:29 -0700)]
swrast: new swrast_renderbuffer type
This will let us move the swrast-specific fields out of gl_renderbuffer.
Brian Paul [Mon, 16 Jan 2012 19:29:11 +0000 (12:29 -0700)]
intel: use intel_rb_format() to get renderbuffer format
This will make future changes cleaner and less invasive.
Brian Paul [Mon, 16 Jan 2012 19:29:04 +0000 (12:29 -0700)]
dri/swrast: rename swrast_renderbuffer to dri_swrast_renderbuffer
To prevent name collision with future swrast_renderbuffer in the swrast
module.
Brian Paul [Mon, 16 Jan 2012 19:28:59 +0000 (12:28 -0700)]
swrast: use stencil packing function in s_stencil.c
Brian Paul [Mon, 16 Jan 2012 19:28:56 +0000 (12:28 -0700)]
swrast: use color packing functions in s_span.c
Brian Paul [Mon, 16 Jan 2012 19:28:51 +0000 (12:28 -0700)]
swrast: remove s_spantemp.h
Brian Paul [Mon, 16 Jan 2012 19:28:47 +0000 (12:28 -0700)]
xlib: remove xm_span.c and related code
Brian Paul [Mon, 16 Jan 2012 19:28:42 +0000 (12:28 -0700)]
mesa: remove gl_renderbuffer::Wrapped
There's no such thing as renderbuffer wrappers anymore.
Brian Paul [Mon, 16 Jan 2012 19:27:50 +0000 (12:27 -0700)]
swrast: rewrite, simplify the the render-to-texture code
Brian Paul [Mon, 16 Jan 2012 19:10:46 +0000 (12:10 -0700)]
mesa: rename gl_renderbuffer::Data to Buffer
To better indicate that this pointer to the malloc'd memory.
Brian Paul [Mon, 16 Jan 2012 19:15:40 +0000 (12:15 -0700)]
mesa: move freeing of software renderbuffers into swrast
Brian Paul [Mon, 16 Jan 2012 19:03:09 +0000 (12:03 -0700)]
mesa: remove gl_renderbuffer::DataType
Brian Paul [Mon, 16 Jan 2012 18:48:18 +0000 (11:48 -0700)]
mesa: remove gl_renderbuffer:RowStride field
Brian Paul [Mon, 16 Jan 2012 18:32:12 +0000 (11:32 -0700)]
mesa: finally, remove the GetRow/PutRow/etc functions
Brian Paul [Mon, 16 Jan 2012 18:30:03 +0000 (11:30 -0700)]
dri: remove all the obsolete spantmp files
Brian Paul [Mon, 16 Jan 2012 18:29:58 +0000 (11:29 -0700)]
radeon: remove obsolete GetRow/PutRow code
Brian Paul [Mon, 16 Jan 2012 18:29:53 +0000 (11:29 -0700)]
nouveau: remove obsolete GetRow/PutRow code
Brian Paul [Mon, 16 Jan 2012 18:29:35 +0000 (11:29 -0700)]
intel: remove most of the span Get/PutRow code
Brian Paul [Mon, 16 Jan 2012 18:28:15 +0000 (11:28 -0700)]
dri/swrast: remove obsolete GetRow/PutRow code
Brian Paul [Mon, 16 Jan 2012 18:28:02 +0000 (11:28 -0700)]
osmesa: remove obsolete GetRow/PutRow code
Brian Paul [Mon, 16 Jan 2012 18:24:18 +0000 (11:24 -0700)]
xlib: remove obsolete GetRow/PutRow code
Brian Paul [Mon, 16 Jan 2012 18:24:09 +0000 (11:24 -0700)]
mesa: remove obsolete PutRow, etc assignments
Brian Paul [Fri, 13 Jan 2012 01:33:09 +0000 (18:33 -0700)]
swrast: remove Get/PutRow()-related code
Brian Paul [Mon, 16 Jan 2012 18:38:46 +0000 (11:38 -0700)]
st/mesa: remove gl_renderbuffer::GetPointer stuff
Brian Paul [Mon, 16 Jan 2012 17:55:39 +0000 (10:55 -0700)]
swrast: stop using Put/GetRow/Values() in swrast code
All color buffer rendering is now done by accessing mapped renderbuffer
memory. We're now able to get rid of all the GetRow/PutRow stuff.
Brian Paul [Mon, 16 Jan 2012 17:55:32 +0000 (10:55 -0700)]
swrast: use gl_renderbuffer::StrideInBytes in depth/stencil code
Brian Paul [Mon, 16 Jan 2012 17:55:25 +0000 (10:55 -0700)]
mesa: use gl_renderbuffer::Map for all depth/stencil accesses
Instead of using the obsolete gl_renderbuffer::Data field.
Color buffer are still accessed through GetRow/PutRow().
Brian Paul [Mon, 16 Jan 2012 17:55:20 +0000 (10:55 -0700)]
intel: make intel_renderbuffer_map/unmap() static
Brian Paul [Mon, 16 Jan 2012 17:55:15 +0000 (10:55 -0700)]
mesa: add new gl_renderbuffer fields
These are temporary, actually, but they'll make follow-on work easier to
implement in a step-by-step manner. Eventually the Map and RowStrideBytes
fields will go into a new swrast_renderbuffer type, but adding that type
now would involve touching a _lot_ of code that'll eventually be removed.
The fields marked as obsolete will go away completely at some point.
Brian Paul [Mon, 16 Jan 2012 17:55:10 +0000 (10:55 -0700)]
swrast: flush pending rendering before unmapping buffers
Brian Paul [Mon, 16 Jan 2012 17:54:53 +0000 (10:54 -0700)]
swrast: new assertions in _swrast_pixel_address()
Brian Paul [Mon, 16 Jan 2012 17:54:47 +0000 (10:54 -0700)]
swrast: use _swrast_pixel_address() in more places
Brian Paul [Mon, 16 Jan 2012 17:54:41 +0000 (10:54 -0700)]
swrast: s/Data/Map/ in swrast_texture_image
To indicate that it points to mapped texture memory.
Brian Paul [Mon, 16 Jan 2012 17:54:36 +0000 (10:54 -0700)]
swrast: remove gl_renderbuffer::DataType check in DrawPixels()
The field will be going away so update this code.
Brian Paul [Mon, 16 Jan 2012 17:54:31 +0000 (10:54 -0700)]
swrast: remove gl_renderbuffer::DataType assertions
This field will go away, so remove some uses of it.
Brian Paul [Mon, 16 Jan 2012 17:54:23 +0000 (10:54 -0700)]
st/mesa: remove gl_renderbuffer:DataType assignments
That field is only used by swrast code so there's no reason to mess
with it in the gallium state tracker.
This also lets us remove the unused st_format_data() type function and
related code.
Brian Paul [Mon, 16 Jan 2012 17:54:19 +0000 (10:54 -0700)]
swrast: make _swrast_get_values(), _swrast_get_row() static
They were only called from in s_span.c
Brian Paul [Mon, 16 Jan 2012 17:54:13 +0000 (10:54 -0700)]
swrast: remove dstType param from _swrast_read_rgba_span()
It was always GL_FLOAT.
Brian Paul [Mon, 16 Jan 2012 17:54:01 +0000 (10:54 -0700)]
swrast: remove unused _swrast_put_row()
Ian Romanick [Sat, 21 Jan 2012 01:23:02 +0000 (17:23 -0800)]
mesa: Don't resurrect deleted ARB VAOs in glPopClientAttrib
When ARB VAOs are used, glPopClientAttrib does not resurrect a deleted
VAO or VBO. This difference between the two spec is, unfortunately,
not very well spelled out in the specs.
Fixes oglc vao(advanced.pushPop.deleteVAO) and
vao(advanced.pushPop.deleteVBO) tests.
NOTE: This is a candidate for release branches.
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
Ian Romanick [Mon, 23 Jan 2012 22:22:38 +0000 (14:22 -0800)]
mesa: Rename gl_array_object::VBOonly to ::ARBsemantics
There are more differences between Apple and ARB than just requiring
that all arrays be stored in VBOs. Additional uses will be added in
following commits.
Also, set the flag at Bind time instead of Gen time. The ARB_vao spec
specifies that behavior.
NOTE: This is a candidate for release branches.
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
Ian Romanick [Fri, 20 Jan 2012 01:29:37 +0000 (17:29 -0800)]
swrast: Use fixed-function processing instead _TexEnvProgram for DrawPixels
This is a hack to work around drivers such as i965 that:
- Set _MaintainTexEnvProgram to generate GLSL IR for
fixed-function fragment processing.
- Don't call _mesa_ir_link_shader to generate Mesa IR from the
GLSL IR.
- May use swrast to handle glDrawPixels.
Since _mesa_ir_link_shader is never called, there is no Mesa IR to
execute. Instead do regular fixed-function processing.
Even on platforms that don't need this, the software fixed-function
code is much faster than the software shader code.
NOTE: This is a candidate for the 8.0 branch.
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=44749
Ian Romanick [Fri, 20 Jan 2012 01:23:51 +0000 (17:23 -0800)]
mesa: Make sure _TexEnvProgram points at the current ff fragment program
At least one place, the _mesa_need_secondary_color function in
state.h, uses this to make decisions. The next patch in this series
will add another dependency. Ideally, this field would go away and be
replace by a flag or something.
NOTE: This is a candidate for the 8.0 branch.
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Brian Paul [Tue, 24 Jan 2012 18:15:43 +0000 (11:15 -0700)]
softpipe: move var initialization to silence warning
Brian Paul [Tue, 24 Jan 2012 18:15:23 +0000 (11:15 -0700)]
r600g: remove unused variable
Alexander von Gluck [Sun, 22 Jan 2012 18:42:38 +0000 (12:42 -0600)]
glsl: Don't use newlocale on Haiku
NOTE: This is a candidate for the 8.0 branch.
Signed-off-by: Brian Paul <brianp@vmware.com>
Alexander von Gluck [Sun, 22 Jan 2012 18:42:32 +0000 (12:42 -0600)]
mesa: Don't use newlocale on Haiku
NOTE: This is a candidate for the 8.0 branch.
Signed-off-by: Brian Paul <brianp@vmware.com>
Alexander von Gluck [Sun, 22 Jan 2012 18:42:24 +0000 (12:42 -0600)]
svga: fix typedef conflicts on Haiku
NOTE: This is a candidate for the 8.0 branch.
Signed-off-by: Brian Paul <brianp@vmware.com>
Alexander von Gluck [Sun, 22 Jan 2012 18:42:17 +0000 (12:42 -0600)]
llvmpipe: fix symbol conflict on Haiku
NOTE: This is a candidate for the 8.0 branch.
Signed-off-by: Brian Paul <brianp@vmware.com>
Alexander von Gluck [Sun, 22 Jan 2012 18:42:10 +0000 (12:42 -0600)]
scons: Remove Haiku one-offs for gallium drivers
NOTE: This is a candidate for the 8.0 branch.
Signed-off-by: Brian Paul <brianp@vmware.com>
Michel Dänzer [Tue, 24 Jan 2012 16:57:56 +0000 (17:57 +0100)]
gallium/postprocess: Proper reference counting of pp_jimenezmlaa depth buffer.
Fixes https://bugs.freedesktop.org/show_bug.cgi?id=40776
NOTE: This is a candidate for the stable branches.
Matt Turner [Tue, 24 Jan 2012 16:34:42 +0000 (11:34 -0500)]
Revert "Always build shared glapi"
This reverts commit
adefee50d954151f76150af80207081ae3c247d9.
Shared glapi was never tested with --enable-xlib-glx and turns out
to cause a lot of problems.
Conflicts:
configure.ac
Matt Turner [Tue, 24 Jan 2012 16:14:42 +0000 (11:14 -0500)]
gbm: install libgbm.so into lib
This partially reverts commit
90e256853418eaaba3717f930cc6a331e4099056.
Adam Jackson [Tue, 24 Jan 2012 15:51:15 +0000 (10:51 -0500)]
llvmpipe: Fix math error in LP_DEBUG=counters output
Signed-off-by: Adam Jackson <ajax@redhat.com>
Vadim Girlin [Tue, 24 Jan 2012 14:22:20 +0000 (18:22 +0400)]
r600g: fix inconsistency with INTERP* opcode definitions
Signed-off-by: Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin [Tue, 24 Jan 2012 11:32:52 +0000 (15:32 +0400)]
r600g: replace raw opcodes with names in the is_alu_trans/vector
Signed-off-by: Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin [Tue, 24 Jan 2012 11:32:51 +0000 (15:32 +0400)]
r600g: add missing opcode definitions
Signed-off-by: Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Vadim Girlin [Mon, 23 Jan 2012 09:23:19 +0000 (13:23 +0400)]
r600g: fix loop break/continue operations
Signed-off-by: Vadim Girlin <vadimgirlin@gmail.com>
Vadim Girlin [Mon, 23 Jan 2012 10:12:46 +0000 (14:12 +0400)]
r600g: fix fragcoord.w
Signed-off-by: Vadim Girlin <vadimgirlin@gmail.com>
Benjamin Franzke [Tue, 24 Jan 2012 09:12:59 +0000 (10:12 +0100)]
targets/gbm: Fix install path
GBM_BACKEND_INSTALL_DIR was deleted by commit
06ad64ad29e7aa9e2d001f6bd1f8c1c1f77050b8.
Since we dont need this configurable, use $(INSTALL_LIB_DIR)/gbm now.
Benjamin Franzke [Tue, 24 Jan 2012 08:29:43 +0000 (09:29 +0100)]
gbm/Makefile.template: Remove builtin-compile path
Builtins are handled by the main gbm Makefile since
06ad64ad29e7aa9e2d001f6bd1f8c1c1f77050b8.
Benjamin Franzke [Tue, 24 Jan 2012 08:24:06 +0000 (09:24 +0100)]
egl,gbm_gallium: Fix linkage against gbm from automake
Add src/gbm/.libs to ldflags.
The gbm lib is src/gbm/.libs/ instead of lib/
as of commit
06ad64ad29e7aa9e2d001f6bd1f8c1c1f77050b8.
Vadim Girlin [Mon, 23 Jan 2012 11:58:31 +0000 (15:58 +0400)]
r600g: fix interpolation with clipvertex
Signed-off-by: Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Chad Versace [Wed, 18 Jan 2012 23:56:58 +0000 (15:56 -0800)]
swrast: Fix unsigned promotion in pointer arithmetic
When rowstride was negatie, unsigned promotion caused a segfault here:
299│ if (rb->Format == MESA_FORMAT_S8) {
300│ const GLuint rowStride = rb->RowStride;
301│ for (i = 0; i < count; i++) {
302│ if (x[i] >= 0 && y[i] >= 0 && x[i] < w && y[i] < h) {
303├> stencil[i] = *(map + y[i] * rowStride + x[i]);
304│ }
305│ }
306│ }
Fixes segfault in oglconform
separatestencil-neu(NonPolygon.BothFacesBitmapCoreAPI),
though test still fails.
Note: This is a candidate for the stable branches.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=43327
Reviewed-by: Brian Paul <brianp@vmware.com>
Signed-off-by: Chad Versace <chad.versace@linux.intel.com>
Ian Romanick [Sat, 21 Jan 2012 00:08:01 +0000 (16:08 -0800)]
meta: Fallback for glBlitFramebuffer from a multisample surface
NOTE: This is a candidate for the 8.0 branch.
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=44818
Paul Berry [Thu, 19 Jan 2012 23:49:43 +0000 (15:49 -0800)]
i965/vs: Fix bogus assertion in emit_block_move()
i965 processes assignments of whole structures using
vec4_visitor::emit_block_move, a recursive function which visits each
element of a structure or array (to arbitrary nesting depth) and
copies it from the source to the destination. Then it increments the
source and destination register numbers so that further recursive
invocations will copy the rest of the structure. In addition, it sets
the swizzle field for the source register to an appropriate value of
swizzle_for_size(...) for the size of each element being copied, so
that later optimization passes won't be fooled into thinking that
unused vector elements are live.
This all works fine. However, emit_block_move also contains an
assertion to verify, before setting the swizzle field for the source
register, that the source register doesn't already contain a
nontrivial swizzle. The intention is to make sure that the caller of
emit_block_move hasn't already done some swizzling of the data before
the call, which emit_block_move would then counteract when it
overwrites the swizzle field. But the assertion is at the lowest
level of nesting of emit_block_move, which means that after the first
element is copied, instead of checking the swizzle field set by the
caller, it checks the swizzle field used when moving the previous
element. That means that if the structure contains elements of
different vector sizes (which therefore require different swizzles),
the assertion will erroneously fire.
This patch moves the assertion from emit_block_move to the calling
function, vec4_visitor::visit(ir_assignment *). Since the caller is
non-recursive, the assertion will only happen once, and won't be
fooled by emit_block_move's modification of the swizzle field.
This patch also reverts commit fe006a7 (i965/vs: Fix swizzle related
assertion), which attempted to fix the bug by making the assertion
more lenient, but only worked properly for structures, arrays, and
matrices in which each constituent vector is the same size.
This fixes the problem described in comment 9 of
https://bugs.freedesktop.org/show_bug.cgi?id=40865. Unfortunately, it
doesn't fix the whole bug, since the test in question is also failing
due to lack of register spilling support in the VS.
Fixes piglit test vs-assign-varied-struct. No piglit regressions on
Sandy Bridge.
This is a candidate for the 8.0 release branch.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=40865#c9
Reviewed-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Alex Deucher [Mon, 23 Jan 2012 21:41:52 +0000 (16:41 -0500)]
r600g: clean up register headers
- CP_INTERRUPT packet doesn't exist
- remove lots of r6xx copy/paste remnants from evergreen reg header
- fix some cayman specific registers
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Eric Anholt [Wed, 18 Jan 2012 22:42:26 +0000 (14:42 -0800)]
i965/vs: Enable workaround-free math on gen7.
This is similar to a commit that did the same for the FS.
Shaves several more instructions off of the VS in Lightsmark, but no
statistically significant performance difference (n=5).
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Eric Anholt [Wed, 18 Jan 2012 20:58:45 +0000 (12:58 -0800)]
i965/vs: Use the embedded-comparison SEL on gen6+, like the FS does.
Shaves a few instructions off of the VS in Lightsmark, but no
statistically significant performance difference on gen7 (n=5).
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>