New in version 10.4.3

January 24th, 2015
  • st/nine: Add new texture format strings
  • st/nine: Correctly advertise D3DPMISCCAPS_CLIPTLVERTS
  • st/nine: NineBaseTexture9: fix setting of last_layer
  • st/nine: CubeTexture: fix GetLevelDesc
  • st/nine: Fix crash when deleting non-implicit swapchain
  • st/nine: Return D3DERR_INVALIDCALL when trying to create a texture of bad format
  • st/nine: NineBaseTexture9: update sampler view creation
  • st/nine: Check if srgb format is supported before trying to use it.
  • st/nine: Add ATI1 and ATI2 support
  • st/nine: Rework of boolean constants
  • st/nine: Convert integer constants to floats before storing them when cards don't support integers
  • st/nine: Remove some shader unused code
  • st/nine: Saturate oFog and oPts vs outputs
  • st/nine: Correctly declare NineTranslateInstruction_Mkxn inputs
  • st/nine: Fix typo for M4x4
  • st/nine: Fix POW implementation
  • st/nine: Handle RSQ special cases
  • st/nine: Handle NRM with input of null norm
  • st/nine: Correct LOG on negative values
  • st/nine: Rewrite LOOP implementation, and a0 aL handling
  • st/nine: Fix CND implementation
  • st/nine: Clamp ps 1.X constants
  • st/nine: Fix some fixed function pipeline operation
  • st/nine: Implement TEXCOORD special behaviours
  • st/nine: Fill missing dst and src number for some instructions.
  • st/nine: Fix TEXM3x3 and implement TEXM3x3VSPEC
  • st/nine: implement TEXM3x2DEPTH
  • st/nine: Implement TEXM3x2TEX
  • st/nine: Implement TEXM3x3SPEC
  • st/nine: Implement TEXDEPTH
  • st/nine: Implement TEXDP3
  • st/nine: Implement TEXDP3TEX
  • st/nine: Implement TEXREG2AR, TEXREG2GB and TEXREG2RGB
  • st/nine: Correct rules for relative adressing and constants.
  • st/nine: Remove unused code for ps
  • st/nine: Fix sm3 relative addressing for non-debug build
  • st/nine: Add variables containing the size of the constant buffers
  • st/nine: Allocate the correct size for the user constant buffer
  • st/nine: Allocate vs constbuf buffer for indirect addressing once.
  • docs: Add sha256 sums for the 10.4.2 release
  • mesa: Fix clamping to -1.0 in snorm_to_float
  • glsl: Link glsl_test with pthreads library.
  • nine: Drop use of TGSI_OPCODE_CND.
  • i965: Respect the no_8 flag on Gen6, not just Gen7+.
  • i965: Work around mysterious Gen4 GPU hangs with minimal state changes.
  • st/nine: Hack to generate resource if it doesn't exist when getting view
  • st/nine: Additional defines to d3dtypes.h
  • st/nine: Add missing c++ declaration for IDirect3DVolumeTexture9
  • st/nine: Fix D3DRS_POINTSPRITE support

New in version 10.4.2 (January 12th, 2015)

  • i965: Use safer pointer arithmetic in intel_texsubimage_tiled_memcpy()
  • i965: Use safer pointer arithmetic in gather_oa_results()
  • Revert "r600g/sb: fix issues cause by GLSL switching to loops for switch"
  • r600g: fix regression since UCMP change
  • r600g/sb: implement r600 gpr index workaround. (v3.1)
  • docs: Add sha256 sums for the 10.4.1 release
  • nv50,nvc0: set vertex id base to index_bias
  • nv50/ir: fix texture offsets in release builds
  • i965: Add missing BRW_NEW_*_PROG_DATA to texture/renderbuffer atoms.
  • i965: Fix start/base_vertex_location for >1 prims but !BRW_NEW_VERTICES.
  • gallium/util: make sure cache line size is not zero
  • glsl_to_tgsi: fix a bug in copy propagation
  • vbo: ignore primitive restart if FixedIndex is enabled in DrawArrays
  • st/mesa: fix GL_PRIMITIVE_RESTART_FIXED_INDEX
  • radeonsi: fix VertexID for OpenGL
  • radeonsi: Don't modify PA_SC_RASTER_CONFIG register value if rb_mask == 0
  • gallium/util: fix crash with daz detection on x86
  • nv50,nvc0: implement half_pixel_center
  • r600g/sb: fix issues with loops created for switch

New in version 10.4.1 (December 30th, 2014)

  • i965/brw_reg: struct constructor now needs explicit negate and abs values.
  • i965: Require pixel alignment for GPU copy blit
  • docs: Add 10.4 sha256 sums, news item and link release notes
  • Revert "glx/dri3: Request non-vsynced Present for swapinterval zero. (v3)"
  • linker: Wrap access of producer_var with a NULL check
  • linker: Assign varying locations geometry shader inputs for SSO
  • glx/dri3: Fix glXWaitForSbcOML() to handle targetSBC==0 correctly. (v2)
  • glx/dri3: Track separate (ust, msc) for PresentPixmap vs. PresentNotifyMsc (v2)
  • glx/dri3: Request non-vsynced Present for swapinterval zero. (v3)
  • glx/dri3: Don't fail on glXSwapBuffersMscOML(dpy, window, 0, 0, 0) (v2)
  • glsl: Add gl_MaxViewports to available builtin constants

New in version 10.4.0 (December 14th, 2014)

  • ir_to_mesa: Remove sat to clamp lowering pass
  • mesa/drivers: Add missing mesautil lib to Haiku swrast
  • i965/gs: Avoid DW * DW mul
  • i965/Gen6-7: Fix point sprites with PolygonMode(GL_POINT)
  • r600g: only init GS_VERT_ITEMSIZE on r600
  • util/primconvert: pass index bias through
  • util/primconvert: support instanced rendering
  • util/primconvert: take ib offset into account
  • util/primconvert: Avoid point arithmetic; apply offset on all cases
  • docs/relnotes: document the removal of GALLIUM_MSAA
  • i965: Disable unlit-centroid workaround on Gen < 6
  • llvmpipe: fix lp_test_arit denorm handling
  • mesa: use build flag to ensure stack is realigned on x86
  • radeonsi: Program RASTER_CONFIG for harvested GPUs v5

New in version 10.4.0 RC3 (November 29th, 2014)

  • i965: Handle nested uniform array indexing
  • glsl: Generate unique names for each const array lowered to uniforms
  • mesa: Fix Get(GL_TRANSPOSE_CURRENT_MATRIX_ARB) to transpose
  • i965/Gen6-7: Do not replace texcoords with point coord if not drawing points
  • r600: fix texture gradients instruction emission (v2)
  • r600g: do all CUBE ALU operations before gradient texture operations (v2.1)
  • r600g: merge the TXQ and BUFFER constant buffers (v1.1)
  • nine: the .pc file should not follow mesa version
  • nv50/ir: set neg modifiers on min/max args
  • nv50,nvc0: actually check constbufs for invalidation
  • nv50,nvc0: buffer resources can be bound as other things down the line
  • freedreno/ir3: don't pass consts to madsh.m16 in MOD logic
  • util/u_snprintf: Don't redefine HAVE_STDINT_H as 0.
  • mapi/glapi: Fix dll linkage of GLES1 symbols.
  • st/wgl: Don't export wglGetExtensionsStringARB.
  • glsl: Make lower_constant_arrays_to_uniforms require dereferences.

New in version 10.4.0 RC2 (November 22nd, 2014)

  • st/mesa: copy sampler_array_size field when copying instructions
  • i965: Fix segfault in WebGL Conformance on Ivybridge
  • r600g/cayman: fix integer multiplication output overwrite (v2)
  • r600g/cayman: fix texture gather tests
  • r600g/cayman: handle empty vertex shaders
  • r600g: geom shaders: always load texture src regs from inputs
  • r600g: limit texture offset application to specific types (v2)
  • configure.ac: roll up a program for the sse4.1 check
  • nv50,nvc0: use clip_halfz setting when creating rasterizer state
  • st/mesa: add a fallback for clear_with_quad when no vs_layer
  • radeonsi: support per-sample gl_FragCoord
  • radeonsi: Disable asynchronous DMA except for PIPE_BUFFER
  • scons: Require glproto >= 1.4.13 for X11.

New in version 10.3.4 (November 22nd, 2014)

  • st/mesa: copy sampler_array_size field when copying instructions
  • i965: Fix segfault in WebGL Conformance on Ivybridge
  • r600g/cayman: fix integer multiplication output overwrite (v2)
  • r600g/cayman: fix texture gather tests
  • r600g/cayman: handle empty vertex shaders
  • r600g: geom shaders: always load texture src regs from inputs
  • r600g: limit texture offset application to specific types (v2)
  • docs: Add sha256 sums for the 10.3.3 release
  • configure.ac: roll up a program for the sse4.1 check
  • get-pick-list.sh: Require explicit "10.3" for nominating stable patches
  • st/mesa: add a fallback for clear_with_quad when no vs_layer
  • llvmpipe: Avoid deadlock when unloading opengl32.dll
  • i915g: we also have more than 0 viewports!
  • radeonsi: Disable asynchronous DMA except for PIPE_BUFFER

New in version 10.3.3 (November 8th, 2014)

  • Anuj Phogat (2):
  • glsl: Fix crash due to negative array index
  • glsl: Use signed array index in update_max_array_access()
  • Brian Paul (1):
  • mesa: fix UNCLAMPED_FLOAT_TO_UBYTE() macro for MSVC
  • Emil Velikov (3):
  • docs: Add sha256 sums for the 10.3.2 release
  • Update version to 10.3.3
  • Add release notes for the 10.3.3 release
  • Ilia Mirkin (27):
  • freedreno/ir3: fix FSLT/etc handling to return 0/-1 instead of 0/1.0
  • freedreno/ir3: INEG operates on src0, not src1
  • freedreno/ir3: add UARL support
  • freedreno/ir3: negate result of USLT/etc
  • freedreno/ir3: use unsigned comparison for UIF
  • freedreno/ir3: add TXL support
  • freedreno/ir3: fix UCMP handling
  • freedreno/ir3: implement UMUL correctly
  • freedreno: add default .dir-locals.el for emacs settings
  • freedreno/ir3: make texture instruction construction more dynamic
  • freedreno/ir3: fix TXB/TXL to actually pull the bias/lod argument
  • freedreno/ir3: add TXQ support
  • freedreno/ir3: add TXB2 support
  • freedreno: dual-source render targets are not supported
  • freedreno: instanced drawing/compute not yet supported
  • freedreno/ir3: avoid fan-in sources referring to same instruction
  • freedreno/ir3: add IDIV/UDIV support
  • freedreno/ir3: add UMOD support, based on UDIV
  • freedreno/ir3: add MOD support
  • freedreno/ir3: add ISSG support
  • freedreno/ir3: add UMAD support
  • freedreno/ir3: make TXQ return integers, not floats
  • freedreno/ir3: shadow comes before array
  • freedreno/ir3: add texture offset support
  • freedreno/ir3: add TXD support and expose ARB_shader_texture_lod
  • freedreno/ir3: add TXF support
  • freedreno: positions come out as integers, not half-integers
  • Jan Vesely (1):
  • configure: include llvm systemlibs when using static llvm
  • Marek Olšák (5):
  • r600g: fix polygon mode for points and lines and point/line fill modes
  • radeonsi: fix polygon mode for points and lines and point/line fill modes
  • radeonsi: fix incorrect index buffer max size for lowered 8-bit indices
  • Revert "st/mesa: set MaxUnrollIterations = 255"
  • r300g: remove enabled/disabled hyperz and AA compression messages
  • Mauro Rossi (1):
  • gallium/nouveau: fully build the driver under android
  • Michel Dänzer (1):
  • radeon/llvm: Dynamically allocate branch/loop stack arrays
  • Rob Clark (62):
  • freedreno/ir3: detect scheduler fail
  • freedreno/ir3: add TXB
  • freedreno/ir3: add DDX/DDY
  • freedreno/ir3: bit of debug
  • freedreno/ir3: fix error in bail logic
  • freedreno/ir3: fix constlen with relative addressing
  • freedreno/ir3: add no-copy-propagate fallback step
  • freedreno: don't overflow cmdstream buffer so much
  • freedreno/ir3: fix potential segfault in RA
  • freedreno: update generated headers
  • freedreno/a3xx: enable hw primitive-restart
  • freedreno/a3xx: handle rendering to layer != 0
  • freedreno: update generated headers
  • freedreno/a3xx: format fixes
  • util/u_format: add _is_alpha()
  • freedreno/a3xx: alpha render-target shenanigans
  • freedreno/ir3: catch incorrect usage of tmp-dst
  • freedreno/ir3: add missing put_dst
  • freedreno: "fix" problems with excessive flushes
  • freedreno: update generated headers
  • freedreno/a3xx: 3d/array textures
  • freedreno: add DRM_CONF_SHARE_FD
  • freedreno/a3xx: more texture array fixes
  • freedreno/a3xx: initial texture border-color
  • freedreno: fix compiler warning
  • freedreno: don't advertise mirror-clamp support
  • freedreno: update generated headers
  • freedreno: we have more than 0 viewports!
  • freedreno: turn missing caps into compile warnings
  • freedreno/a3xx: add LOD_BIAS
  • freedreno/a3xx: add flat interpolation mode
  • freedreno/a3xx: add 32bit integer vtx formats
  • freedreno/a3xx: fix border color order
  • freedreno: move bind_sampler_states to per-generation
  • freedreno: add texcoord clamp support to lowering
  • freedreno/a3xx: add support to emulate GL_CLAMP
  • freedreno/a3xx: re-emit shaders on variant change
  • freedreno/lowering: fix token calculation for lowering
  • freedreno: destroy transfer pool after blitter
  • freedreno: max-texture-lod-bias should be 15.0f
  • freedreno: update generated headers
  • freedreno/a3xx: handle large shader program sizes
  • freedreno/a3xx: emit all immediates in one shot
  • freedreno/ir3: fix lockups with lame FRAG shaders
  • freedreno/a3xx: handle VS only outputting BCOLOR
  • freedreno: query fixes
  • freedreno/a3xx: refactor vertex state emit
  • freedreno/a3xx: refactor/optimize emit
  • freedreno/ir3: optimize shader key comparision
  • freedreno: inline fd_draw_emit()
  • freedreno: fix layer_stride
  • freedreno: update generated headers
  • freedreno/ir3: large const support
  • freedreno/a3xx: more layer/level fixes
  • freedreno/ir3: comment + better fxn name
  • freedreno/ir3: fix potential gpu lockup with kill
  • freedreno/a3xx: disable early-z when we have kill's
  • freedreno/ir3: add debug flag to disable cp
  • freedreno: clear vs scissor
  • freedreno: mark scissor state dirty when enable bit changes
  • freedreno/a3xx: fix viewport state during clear
  • freedreno/a3xx: fix depth/stencil restore format
  • Tapani Pälli (2):
  • glsl: fix uniform location count used for glsl types
  • mesa: check that uniform exists in glUniform* functions

New in version 10.3.2 (October 25th, 2014)

  • mesa: fix spurious wglGetProcAddress / GL_INVALID_OPERATION error
  • st/wgl: add WINAPI qualifiers on wgl function typedefs
  • glsl: fix several use-after-free bugs
  • glx: Fix glxUseXFont for glxWindow and glxPixmaps
  • mesa: fix GetTexImage for 1D array depth textures
  • docs: Add sha256 sums for the 10.3.1 release
  • gm107/ir: add dnz emission for fmul
  • gk110/ir: add dnz flag emission for fmul/fmad
  • nouveau: 3d textures are unsupported, limit 3d levels to 1
  • st/gbm: fix order of arguments passed to is_format_supported
  • i965: Add a BRW_MOCS_PTE #define.
  • i965: Use BDW_MOCS_PTE for renderbuffers.
  • i965: Fix register write checks.
  • st/mesa: use pipe_sampler_view_release for releasing sampler views
  • glsl_to_tgsi: fix the value of gl_FrontFacing with native integers
  • radeonsi: Clear sampler view flags when binding a buffer
  • r600g,radeonsi: Always use GTT again for PIPE_USAGE_STREAM buffers
  • winsys/radeon: Use separate caching buffer manager for each set of flags
  • r600g: Drop references to destroyed blend state