Major changes in Mesa 10.5, from the release announcement:
- The new IR - NIR. A flat, typeless IR, supporting SSA.
- GLX/DRI3 fixes and improvements.
- A lot of preparation work for GL_ARB_direct_state_access.
- Piglit and drawElements Quality Program (dEQP) fixes.
- Initial support for the new Intel Skylake GPUs and Adreno 4xx GPUs.
- Support for GL_EXT_polygon_offset_clamp on i965, nv50, nvc0, r600,
radeonsi, llvmpipe
Besides that, Direct3D9 state tracker ("nine") has been enabled.
It is now provided by lib(64)d3dtracker1 package, the development
files are in lib(64)d3dtracker-devel.
From the log:
egl: Pass the correct X visual depth to xcb_put_image().
gallium/util: Don't use __builtin_clrsb in util_last_bit().
glsl: Link glsl_test with pthreads library.
glsl: Reduce memory consumption of copy propagation passes.
glx/dri3: Request non-vsynced Present for swapinterval zero. (v3)
i965: Fix integer border color on Haswell.
i965: Fix max_wm_threads for CHV
i965: Override swizzles for integer luminance formats.
i965: Respect the no_8 flag on Gen6, not just Gen7+.
i965: Use a gl_color_union for sampler border color.
i965: Work around mysterious Gen4 GPU hangs with minimal state changes.
main: Fixed _mesa_GetCompressedTexImage_sw to copy slices correctly.
mesa: Fix clamping to -1.0 in snorm_to_float
mesa: fix AtomicBuffer typo in _mesa_DeleteBuffers
mesa: fix display list 8-byte alignment issue
nvc0: allow holes in xfb target lists
nvc0: bail out of 2d blits with non-A8_UNORM alpha formats
r600g,radeonsi: don't append to streamout buffers that haven't been used
yet
radeonsi: fix a crash if a stencil ref state is set before a DSA state
radeonsi: fix instanced arrays with non-zero start instance
radeonsi: small fix in SPI state
st/mesa: Don't use PIPE_USAGE_STREAM for GL_PIXEL_UNPACK_BUFFER_ARB
st/mesa: treat resource-less xfb buffers as if they weren't there
st/osmesa: Fix osbuffer->textures indexing
Mesa 10.4.2 provides lots of bug fixes and implements a few more OpenGL
features:
* GL_ARB_conditional_render_inverted on nv50
* GL_ARB_sample_shading on r600
* GL_ARB_texture_view on nv50, nvc0
* GL_ARB_clip_control on nv50, nvc0, r300, r600, radeonsi, llvmpipe,
softpipe
* GL_KHR_context_flush_control on all drivers
1. libXvMCgallium.* are now also installed although they were not meant
to before. Other libXvMC* libraries are hardlinks to it and that
confused the build system.
2. lib(64)dri-drivers has been renamed to dri-drivers: it is not a
regular shared library but rather a collection of special shared
objects.
3. Added *.rpmlintrc file to silence the complaints about
libXvMCgallium.* and dri-drivers.
Apart from the actual update to 10.3.3, the following changes were made:
* removed (broken) XvMCConfig, provided instructions how to create it;
* removed the no longer used patches and files.
From the announcement:
Changes from 10.2.4 to 10.2.5:
Adel Gadllah (1):
i915: Fix up intelInitScreen2 for DRI3
Anuj Phogat (2):
i965: Fix z_offset computation in
intel_miptree_unmap_depthstencil()
mesa: Don't use memcpy() in _mesa_texstore() for float depth
texture data
Carl Worth (4):
docs: Add SHA256 checksums for the 10.2.4 release
cherry-ignore: Ignore a few patches picked in the previous stable
release
Update version to 10.2.5
docs: Add release notes for 10.2.5
Christian König (1):
radeonsi: fix order of r600_need_dma_space and
r600_context_bo_reloc
Ian Romanick (2):
mesa: Don't allow GL_TEXTURE_BORDER queries outside compat profile
mesa: Don't allow GL_TEXTURE_{LUMINANCE,INTENSITY}_* queries
outside compat profile
Ilia Mirkin (1):
nvc0: make sure that the local memory allocation is aligned to
0x10
Jason Ekstrand (2):
main/format_pack: Fix a wrong datatype in pack_ubyte_R8G8_UNORM
main/get_hash_params: Add GL_SAMPLE_SHADING_ARB
José Fonseca (1):
st/wgl: Clamp wglChoosePixelFormatARB's output nNumFormats to
nMaxFormats.
Kenneth Graunke (4):
i965: Add missing persample_shading field to
brw_wm_debug_recompile.
i965/fs: Fix gl_SampleID for 2x MSAA and SIMD16 mode.
i965/fs: Fix gl_SampleMask handling for SIMD16 on Gen8+.
i965/fs: Set LastRT on the final FB write on Broadwell.
Marek Olšák (10):
r600g: switch SNORM conversion to DX and GLES behavior
radeonsi: fix CMASK and HTILE calculations for Hawaii
gallium/util: add a helper for calculating primitive count from
vertex count
radeonsi: fix a hang with instancing on Hawaii
radeonsi: fix a hang with streamout on Hawaii
winsys/radeon: fix vram_size overflow with Hawaii
radeonsi: fix occlusion queries on Hawaii
r600g,radeonsi: switch all occurences of array_size to
util_max_layer
radeonsi: fix build because of lack of draw_indirect
infrastructure in 10.2
radeonsi: use DRAW_PREAMBLE on CIK
Thorsten Glaser (1):
nv50: fix build failure on m68k due to invalid struct alignment
assumptions
Tom Stellard (1):
clover: Call end_query before getting timestamp result v2
This is a bugfix release. From the announcement:
Changes from 10.2.3 to 10.2.4:
Abdiel Janulgue (3):
i965/fs: Refactor check for potential copy propagated
instructions.
i965/fs: skip copy-propate for logical instructions with negated
src entries
i965/vec4: skip copy-propate for logical instructions with negated
src entries
Brian Paul (3):
mesa: fix geometry shader memory leaks
st/mesa: fix geometry shader memory leak
gallium/u_blitter: fix some shader memory leaks
Carl Worth (3):
docs: Add sha256 checksums for the 10.2.3 release
Update VERSION to 10.2.4
Add release notes for 10.2.4
Eric Anholt (1):
i965: Generalize the pixel_x/y workaround for all UW types.
Ilia Mirkin (4):
nv50/ir: retrieve shadow compare from first arg
nv50/ir: ignore bias for samplerCubeShadow on nv50
nvc0/ir: do quadops on the right texture coordinates for TXD
nvc0/ir: use manual TXD when offsets are involved
Jordan Justen (1):
i965: Add auxiliary surface field #defines for Broadwell.
Kenneth Graunke (9):
i965: Don't copy propagate abs into Broadwell logic instructions.
i965: Set execution size to 8 for instructions with force_sechalf
set.
i965/fs: Set force_uncompressed and force_sechalf on samplepos
setup.
i965/fs: Use WE_all for gl_SampleID header register munging.
i965: Add plumbing for Broadwell's auxiliary surface support.
i965: Drop SINT workaround for CMS layout on Broadwell.
i965: Hook up the MCS buffers in SURFACE_STATE on Broadwell.
i965: Add 2x MSAA support to the MCS allocation function.
i965: Enable compressed multisample support (CMS) on Broadwell.
Marek Olšák (4):
gallium: fix u_default_transfer_inline_write for textures
st/mesa: fix samplerCubeShadow with bias
radeonsi: fix samplerCubeShadow with bias
radeonsi: add support for TXB2
Matt Turner (8):
i965/vec4: Don't return void from a void function.
i965/vec4: Don't fix_math_operand() on Gen >= 8.
i965/fs: Don't fix_math_operand() on Gen >= 8.
i965/fs: Make try_constant_propagate() static.
i965/fs: Constant propagate into 2-src math instructions on Gen8.
i965/vec4: Constant propagate into 2-src math instructions on
Gen8.
i965/fs: Don't use brw_imm_* unnecessarily.
i965/fs: Set correct number of regs_written for MCS fetches.
This is a bugfux release.
Changes from 10.2.2 to 10.2.3, from the announcement:
Aaron Watry (1):
radeon/llvm: Allocate space for kernel metadata operands
Carl Worth (3):
docs: Add sha256 sums for the 10.2.2 release
cherry-ignore: Add a patch that's been rejected
docs: Add release notes for the 10.2.3 release.
Ilia Mirkin (4):
nouveau: dup fd before passing it to device
nv50: disable dedicated ubo upload method
nv50: do an explicit flush on draw when there are persistent
buffers
nvc0: add a memory barrier when there are persistent UBOs
Jasper St. Pierre (1):
glxext: Send the Drawable's ID in the GLX_BufferSwapComplete event
Kenneth Graunke (3):
i965: Don't emit SURFACE_STATEs for gather workarounds on
Broadwell.
i965: Include marketing names for Broadwell GPUs.
i965/disasm: Fix INTEL_DEBUG=fs on Broadwell for ARB_fp
applications.
Michel Dänzer (1):
radeon/llvm: Use the llvm.rsq.clamped intrinsic for RSQ
Rob Clark (9):
xa: fix segfault
freedreno: use OUT_RELOCW when buffer is written
freedreno/a3xx: fix depth/stencil GMEM positioning
freedreno/a3xx: fix depth/stencil gmem restore
freedreno/a3xx: fix blend opcode
freedreno: few caps fixes
freedreno/a3xx: texture fixes
freedreno: fix for null textures
freedreno/a3xx: vtx formats
Roland Scheidegger (1):
draw: (trivial) fix clamping of viewport index
Takashi Iwai (1):
llvmpipe: Fix zero-division in llvmpipe_texture_layout()
Thomas Hellstrom (1):
st/xa: Don't close the drm fd on failure v2
Tobias Klausmann (1):
nv50/ir: allow gl_ViewportIndex to work on non-provoking vertices
Notable changes in this release (from the announcement):
Adrian Negreanu:
add megadriver_stub_FILES
Daniel Manjarres (1):
glx: Don't crash on swap event for a Window (non-GLXWindow)
Emil Velikov (3):
targets/xa: limit the amount of exported symbols
configure: error out when building opencl without LLVM
configure: correctly autodetect xvmc/vdpau/omx
Grigori Goronzy (1):
radeon/uvd: disable VC-1 simple/main on UVD 2.x
Iago Toral Quiroga (1):
mesa: Copy Geom.UsesEndPrimitive when cloning a geometry program.
Ian Romanick:
meta: Respect the driver's maximum number of draw buffers
Ilia Mirkin (7):
gk110/ir: emit saturate flag on fadd when needed
gk110/ir: fix emitting constbuf file index
gk110/ir: fix bfind emission
nv50: make sure to mark first scissor dirty after blit
nv30: plug some memory leaks on screen destroy and shader compile
nv30: avoid dangling references to deleted contexts
nv30: hack to avoid errors on unexpected color/zeta combinations
Jason Ekstrand (1):
meta_blit: properly compute texture width for the CopyTexSubImage
fallback
José Fonseca (1):
mesa/main: Prevent sefgault on
glGetIntegerv(GL_ATOMIC_COUNTER_BUFFER_BINDING).
Kenneth Graunke (9):
i965: Don't use the head sentinel as an fs_inst in Gen4 workaround
code.
i965: Invalidate live intervals when inserting Gen4 SEND
workarounds.
i965/vec4: Fix dead code elimination for VGRFs of size > 1.
i965: Add missing MOCS setup for 3DSTATE_INDEX_BUFFER on
Broadwell.
i965: Drop Broadwell perf_debugs about missing MOCS that aren't
missing.
i965: Add missing newlines to a few perf_debug messages.
i965/vec4: Use the sampler for pull constant loads on Broadwell.
i965: Use 8x4 aligned rectangles for HiZ operations on Broadwell.
i965: Save meta stencil blit programs in the context.
Kristian Høgsberg (1):
mesa: Remove glClear optimization based on drawable size
Michel Dänzer (1):
configure: Only check for OpenCL without LLVM when the latter is
certain
Neil Roberts (1):
i965: Set the fast clear color value for texture surfaces
Tom Stellard (2):
clover: Prevent Clang from printing number of errors and warnings
to stderr.
clover: Don't use llvm's global context
Ville Syrjälä (1):
i915: Fix gen2 texblend setup
1. Restored --enable-r600-llvm-compiler option. The support for LLVM
compiler for r600 will still remain disabled by default but the users
may enable it in runtime by setting R600_DEBUG=llvm environment
variable.
2. Made it clear that XvMC backends are provided for r600+ and
nouveau only.
3. Other fixes.
