summaryrefslogtreecommitdiffstats
path: root/src
Commit message (Collapse)AuthorAgeFilesLines
* i965/fs: Lower the MULH virtual instruction.Francisco Jerez2015-08-061-0/+55
| | | | | | | | | | | | Translate MULH into the MUL/MACH sequence. This does roughly the same thing that nir_emit_alu() used to do but we can now handle 16-wide by taking advantage of the SIMD lowering pass. The force_sechalf workaround near the bottom is required because the SIMD lowering pass will emit instructions with non-zero quarter control and we need to make sure we avoid that on integer arithmetic instructions with implicit accumulator access due to a known hardware bug on IVB. Reviewed-by: Matt Turner <[email protected]>
* i965/fs: Indent the implementation of 32x32-bit MUL lowering by one level.Francisco Jerez2015-08-061-130/+134
| | | | | | | | | In order to make room for the code that will lower the MULH virtual instruction. Also move the hardware generation and execution type checks into the same branch, they are going to have to be different for MULH. Reviewed-by: Matt Turner <[email protected]>
* i965/fs: Lower 32x32 bit multiplication on BXT.Francisco Jerez2015-08-061-2/+2
| | | | | | | AFAIK BXT has the same annoying alignment limitation as CHV on the source register regions of 32x32 bit MULs, give it the same treatment. Reviewed-by: Matt Turner <[email protected]>
* i965: Define virtual instruction to calculate the high 32 bits of a multiply.Francisco Jerez2015-08-066-0/+13
| | | | | | | | | | | This instruction will translate to the MUL/MACH sequence that computes the high 32-bits of the result of a 64-bit multiply. Before Gen8 integer operations that used the accumulator were limited to 8-wide, but the SIMD lowering pass can easily be hooked up to sidestep this limitation, we just need a virtual opcode to represent the MUL/MACH sequence in the IR. Reviewed-by: Matt Turner <[email protected]>
* glsl: Initialize patch member of glsl_struct_fieldMichel Dänzer2015-08-061-0/+2
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | There is apparently a subtle difference in C++ between F f; and F f(); The former will use the default constructor. If there is no default constructor specified, the compiler provides one that simply invokes the default constructor for each field. For built-in basic types, the default constructor does nothing. The later will, according to http://stackoverflow.com/questions/2417065/does-the-default-constructor-initialize-built-in-types) perform value-initialization of the type. For built-in types this means initializing to zero. The per_vertex_accumulator constructor is: per_vertex_accumulator::per_vertex_accumulator() : fields(), num_fields(0) { } This is the second form of constructor, so the glsl_struct_field objects were previously zero initialized. With the addition of an empty default constructor in commit 7ac946e5, per_vertex_accumulator::fields receive no initialization. Fixes a bunch of random (mostly tessellation related) piglit failures since commit 7ac946e5 ("glsl: Add constuctors for the common cases of glsl_struct_field"). Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=91544 Reviewed-by: Ian Romanick <[email protected]>
* nir: add missing type to type_size_vec4()Timothy Arceri2015-08-051-0/+2
| | | | Reviewed-by: Iago Toral Quiroga <[email protected]>
* mesa: Add missing check of format and type in glTexSubImageXD on GLES 3.0Eduardo Lima Mitev2015-08-051-47/+69
| | | | | | | | | | | | | | | | Argument validation for glTexSubImageXD is missing a check of format and type against texture object's internal format when profile is OpenGL-ES 3.0+. This patch also groups together all format and type checks on GLES into a new function texture_format_error_check_gles(), to factorize similar code in texture_format_error_check(). Fixes 2 dEQP tests: * dEQP-GLES3.functional.negative_api.texture.texsubimage2d * dEQP-GLES3.functional.negative_api.texture.texsubimage3d Reviewed-by: Ian Romanick <[email protected]> Reviewed-by: Brian Paul <[email protected]>
* mesa: Fix error returned by glCopyTexImage2D() upon an invalid internal formatEduardo Lima Mitev2015-08-051-9/+9
| | | | | | | | | | | | | | | | | | | | | | | Page 161 of the OpenGL-ES 3.1 (PDF) spec, and page 207 of the OpenGL 4.5 (PDF), both on section '8.6. ALTERNATE TEXTURE IMAGE SPECIFICATION COMMANDS', states: "An INVALID_ENUM error is generated if an invalid value is specified for internalformat". It is currently returning INVALID_OPERATION error because _mesa_get_read_renderbuffer_for_format() is called before the internalformat argument has been validated. To fix this, we move this call down the validation process, after _mesa_base_tex_format() has been called. _mesa_base_tex_format() effectively serves as a validator for the internal format. Fixes 1 dEQP test: * dEQP-GLES3.functional.negative_api.texture.copyteximage2d_invalid_format Fixes 1 piglit test: * spec@oes_compressed_etc1_rgb8_texture@basic Reviewed-by: Ian Romanick <[email protected]> Cc: 10.6 <[email protected]>
* mesa: Validate target before resolving tex obj in glTex(ture)SubImageXDEduardo Lima Mitev2015-08-051-15/+14
| | | | | | | | | | | | | | | Currently, glTexSubImageXD attempt to resolve the texture object (by calling _mesa_get_current_tex_object()) before validating the given target. However, that method explicitly states that target must have been validated before calling it, so it never returns a user error. The target validation occurs later when texsubimage_error_check() is called. This patch reorganizes target validation, taking it out from the error check function and into a point before the texture object is resolved. Reviewed-by: Ian Romanick <[email protected]> Cc: 10.6 <[email protected]>
* mesa: Fix errors values returned by glShaderBinary()Eduardo Lima Mitev2015-08-051-3/+14
| | | | | | | | | | | | | | | | | Page 68, section 7.2 'Shader Binaries" of the of the OpenGL ES 3.1, and page 88 of the OpenGL 4.5 specs state: "An INVALID_VALUE error is generated if count or length is negative. An INVALID_ENUM error is generated if binaryformat is not a supported format returned in SHADER_BINARY_FORMATS." Currently, an INVALID_OPERATION error is returned for all cases. Fixes 1 dEQP test: * dEQP-GLES3.functional.negative_api.shader.shader_binary Reviewed-by: Ian Romanick <[email protected]> Cc: 10.6 <[email protected]>
* mesa: do not modify args when errors with GetProgramResourceNameTapani Pälli2015-08-051-6/+0
| | | | | | | | | | | | Original purpose of these lines was to be more friendly against GUI tools using the extension. However conformance suite explicitly checks that buffers are not modified in error conditions. Fixes: ES31-CTS.program_interface_query.buff-length Signed-off-by: Tapani Pälli <[email protected]> Reviewed-by: Timothy Arceri <[email protected]>
* glsl: add variable mode check to build_stagerefTapani Pälli2015-08-051-4/+13
| | | | | | | | | | | | | Currently stage reference mask is built using the variable name only. However it can happen that input of one stage has same name as output from another stage. Adding check of variable mode makes sure we do not pick wrong variable. Fixes some subcases from ES31-CTS.program_interface_query.no-locations Signed-off-by: Tapani Pälli <[email protected]> Reviewed-by: Timothy Arceri <[email protected]>
* dri: set the __DRI_API_OPENGL bit based on max gl compat versionFrank Binns2015-08-041-1/+3
| | | | | | | | This matches similar behaviour for the __DRI_API_OPENGL_CORE bit. Signed-off-by: Frank Binns <[email protected]> Reviewed-by: Emil Velikov <[email protected]> Reviewed-by: Eric Anholt <[email protected]>
* egl: Add eglQuerySurface surface type check for EGL_LARGEST_PBUFFER attribFrank Binns2015-08-041-1/+2
| | | | | | | | | | | | | | | | | | | Calling eglQuerySurface on a window or pixmap with the EGL_LARGEST_PBUFFER attribute resulted in the contents of the 'value' parameter being modified. This is the wrong behaviour according to the EGL spec, which states: "Querying EGL_LARGEST_PBUFFER for a pbuffer surface returns the same attribute value specified when the surface was created with eglCreatePbufferSurface. For a window or pixmap surface, the contents of value are not modified." Avoid this from happening by checking that the surface type is EGL_PBUFFER_BIT before modifying the contents of the parameter. Cc: <[email protected]> Signed-off-by: Frank Binns <[email protected]> Reviewed-by: Emil Velikov <[email protected]> Reviewed-by: Eric Anholt <[email protected]>
* egl/dri: Add error info needed for EGL_EXT_image_dma_buf_import extensionFrank Binns2015-08-041-0/+4
| | | | | | | | | | | | | Update the DRI image interface error codes to reflect the needs of the EGL_EXT_image_dma_buf_import extension. This means updating the existing error code documentation and adding a new __DRI_IMAGE_ERROR_BAD_ACCESS error code so that drivers can correctly reject unsupported pitches and offsets. Hook the new error code up in EGL to return EGL_BAD_ACCESS. Cc: <[email protected]> Signed-off-by: Frank Binns <[email protected]> Reviewed-by: Emil Velikov <[email protected]> Reviewed-by: Eric Anholt <[email protected]>
* vc4: Use nir_lower_load_const_to_scalar().Eric Anholt2015-08-041-0/+1
|
* nir: Add a nir_lower_load_const_to_scalar() pass.Eric Anholt2015-08-043-0/+105
| | | | | | | | This is useful to increase the CSE opportunities for a scalar backend. It avoids regressions when dropping vc4's custom CSE implementation. v2: Cleanups by Matt (decl in the for loop, and unreachable()). Reviewed-by: Matt Turner <[email protected]>
* vc4: Don't bother de-SSAing values that aren't part of phi webs.Eric Anholt2015-08-041-15/+44
| | | | We can just support them the same way we do load_const's SSA values.
* nir: Add algebraic opt for no-op iand.Eric Anholt2015-08-041-0/+1
| | | | | | I lazily generated some of these in VC4 NIR lowering. Reviewed-by: Iago Toral Quiroga <[email protected]>
* vc4: Don't bother saturating the dst color for blending.Eric Anholt2015-08-041-8/+2
| | | | | | | | | Since we just pulled it out of the destination as 8-bit unorm, we know it's in [0, 1] already. shader-db: total instructions in shared programs: 100040 -> 98208 (-1.83%) instructions in affected programs: 14084 -> 12252 (-13.01%)
* vc4: Make r4-writes implicitly move to a temp, and allocate temps to r4.Eric Anholt2015-08-048-107/+106
| | | | | | | | | | | Previously, SFU values always moved to a temporary, and TLB color reads and texture reads always lived in r4. Instead, we can have these results just be normal temporaries, and the register allocator can leave the values in r4 when they don't interfere with anything else using r4. shader-db results: total instructions in shared programs: 100809 -> 100040 (-0.76%) instructions in affected programs: 42383 -> 41614 (-1.81%)
* vc4: Drop a dead prototype.Eric Anholt2015-08-041-8/+0
|
* Revert "nir: Use a single bit for the dual-source blend index"Eric Anholt2015-08-041-6/+2
| | | | | This reverts commit ab5b7a0fe659ff6f9c1885d5cb047b6531959506. We use more than one bit of value in tgsi_to_nir.
* freedreno/a4xx: add independent blend function supportRob Clark2015-08-042-8/+10
| | | | | | needed for MRT Signed-off-by: Rob Clark <[email protected]>
* freedreno/a4xx: MRT supportRob Clark2015-08-0412-132/+212
| | | | Signed-off-by: Rob Clark <[email protected]>
* freedreno: move the half-precision logic into coreRob Clark2015-08-044-31/+38
| | | | | | | | Both a3xx and a4xx need the same logic to decide if half-precision can be used for blit shaders. So move it to core and simplify things a bit with a helper that considers all render targets. Signed-off-by: Rob Clark <[email protected]>
* freedreno: simplify/cleanup resource status trackingRob Clark2015-08-044-48/+71
| | | | | | | | | Collapse dirty/reading bools into status bitmask (and drop writing which should really be the same as dirty). And use 'used_resources' list for all tracking, including zsbuf/cbufs, rather than special casing the color and depth/stencil buffers. Signed-off-by: Rob Clark <[email protected]>
* freedreno: fix stream-out caps vec4->componentsRob Clark2015-08-041-2/+2
| | | | | | Should be in units of components, not vec4's Signed-off-by: Rob Clark <[email protected]>
* freedreno: small bit of cleanup about max rendertargetsRob Clark2015-08-0413-17/+40
| | | | | | | | We hard-coded 4 or 8 as the max in various places. Switch it all to a define since the limit will go up with a4xx (and maybe even again in the future?) Signed-off-by: Rob Clark <[email protected]>
* glx: Use _mesa_lroundevenf() in glPixelStoref().Matt Turner2015-08-041-12/+13
| | | | | | Functional change in which way half-way cases are rounded from towards positive-infinity to even. The spec says "the passed value is rounded to the nearest integer". Removes another case of bad half-up rounding.
* util: Use SSE intrinsics in _mesa_lroundeven{f,}.Matt Turner2015-08-041-0/+22
| | | | | | | | | | | | | | gcc actually generates this for us now that we use -fno-math-errno (which is weird, since lrintf()/lrint() don't set errno) but clang still does not. Presumably helps MSVC as well. Reduced .text size by 8.5k with gcc before -fno-math-errno. text data bss dec hex filename 4935850 195136 26192 5157178 4eb13a i965_dri.so before 4927225 195128 26192 5148545 4e8f81 i965_dri.so after Reviewed-by: Roland Scheidegger <[email protected]>
* mesa: Use _mesa_lroundevenf() in some more places.Matt Turner2015-08-042-12/+6
|
* vl/mpeg12: Silence GCC unused-variable warning.Vinson Lee2015-08-031-0/+1
| | | | | | | | | | vl/vl_mpeg12_bitstream.c: In function 'decode_slice': vl/vl_mpeg12_bitstream.c:928:19: warning: unused variable 'extra' [-Wunused-variable] unsigned extra = vl_vlc_get_uimsbf(&bs->vlc, 1); ^ Signed-off-by: Vinson Lee <[email protected]> Reviewed-by: Christian König <[email protected]>
* glsl: replace old hash table with new and faster oneAlejandro Seguí2015-08-041-7/+11
| | | | | | | | The util/hash_table was intended to be a fast hash table replacement for the program/hash_table see 35fd61bd99c1 and 72e55bb6888ff. Reviewed-by: Timothy Arceri <[email protected]>
* glsl: Add constuctors for the common cases of glsl_struct_fieldIan Romanick2015-08-034-56/+53
| | | | | | | | | | | | | | | | | | | | | | | | Fixes a giant pile of GCC warnings: builtin_types.cpp:60:1: warning: missing initializer for member 'glsl_struct_field::stream' [-Wmissing-field-initializers] I had to add a default constructor because a non-default constructor was added. Otherwise the only constructor would be the one with parameters, and all the plases like glsl_struct_field foo; would fail to compile. I wanted to do this in two patches. All of the initializers of glsl_struct_field structures had to be converted to use the constructor because C++ apparently forces you to do one or the other: builtin_types.cpp:61:1: error: could not convert '{glsl_type::float_type, "near", -1, 0, 0, 0, GLSL_MATRIX_LAYOUT_INHERITED, 0, -1}' from '<brace-enclosed initializer list>' to 'glsl_struct_field' Signed-off-by: Ian Romanick <[email protected]> Reviewed-by: Francisco Jerez <[email protected]> Reviewed-by: Juha-Pekka Heikkila <[email protected]>
* i965: Make gen7_upload_ps_state staticIan Romanick2015-08-032-10/+1
| | | | | | | | It is only ever called from within the same file. Signed-off-by: Ian Romanick <[email protected]> Acked-by: Jason Ekstrand <[email protected]> Reviewed-by: Juha-Pekka Heikkila <[email protected]>
* i965: Remove extern declaration for nonexistent state atomIan Romanick2015-08-031-1/+0
| | | | | | Signed-off-by: Ian Romanick <[email protected]> Acked-by: Jason Ekstrand <[email protected]> Reviewed-by: Juha-Pekka Heikkila <[email protected]>
* i965: Trivial formatting changes in gen7_vs_state.cIan Romanick2015-08-031-5/+5
| | | | | | Signed-off-by: Ian Romanick <[email protected]> Acked-by: Jason Ekstrand <[email protected]> Reviewed-by: Juha-Pekka Heikkila <[email protected]>
* i965: Trivial formatting changes in gen6_multisample_state.cIan Romanick2015-08-031-5/+2
| | | | | | Signed-off-by: Ian Romanick <[email protected]> Acked-by: Jason Ekstrand <[email protected]> Reviewed-by: Juha-Pekka Heikkila <[email protected]>
* i965: Trivial formatting changes in brw_misc_state.cIan Romanick2015-08-031-26/+23
| | | | | | Signed-off-by: Ian Romanick <[email protected]> Acked-by: Jason Ekstrand <[email protected]> Reviewed-by: Juha-Pekka Heikkila <[email protected]>
* i965: Trivial formatting changes in brw_draw_upload.cIan Romanick2015-08-031-3/+6
| | | | | | Signed-off-by: Ian Romanick <[email protected]> Acked-by: Jason Ekstrand <[email protected]> Reviewed-by: Juha-Pekka Heikkila <[email protected]>
* i965: Trivial formatting changes in brw_draw.cIan Romanick2015-08-031-51/+51
| | | | | | Signed-off-by: Ian Romanick <[email protected]> Acked-by: Jason Ekstrand <[email protected]> Reviewed-by: Juha-Pekka Heikkila <[email protected]>
* i965: Trivial formatting changes in brw_wm.cIan Romanick2015-08-031-43/+41
| | | | | | Signed-off-by: Ian Romanick <[email protected]> Acked-by: Jason Ekstrand <[email protected]> Reviewed-by: Juha-Pekka Heikkila <[email protected]>
* i965/nir: Do not scalarize phis in non-scalar setupsIago Toral Quiroga2015-08-031-2/+6
| | | | | | Significantly reduces register pressure in some piglit tests. Reviewed-by: Jason Ekstrand <[email protected]>
* i965/vec4: Handle uniform and GRF array access on vertex programs (NIR)Antia Puentes2015-08-031-1/+1
| | | | | | | | | | | | | | | | When the NIR-vec4 pass is enabled, handles uniform and GRF array access on ARB_vertex_program like it is done on vertex shaders. When the old IR-vec4 pass is used, emit_program_code() emits pull constant loads directly instead of using relative addressing, hence to call to move_uniform_array_access_to_pull_constants() is not needed and it is enough to call to split_uniform_registers(). The patch also calls to move_grf_array_access_to_scratch() like it is done for shaders, however I suspect this is a no-op for vertex programs and we could remove it. Reviewed-by: Jason Ekstrand <[email protected]>
* i965/nir/vec4: Handle uniforms on vertex programsAntia Puentes2015-08-031-2/+32
| | | | | | | The implementation takes into account that on ARB_vertex_program only a single nir variable is generated to support all the uniform data. Reviewed-by: Jason Ekstrand <[email protected]>
* i965/vec4: Enable NIR-vec4 pass on ARB_vertex_programsAntia Puentes2015-08-031-23/+24
| | | | Reviewed-by: Jason Ekstrand <[email protected]>
* i965/nir/gs: Implement support for gl_InvocationID system valueIago Toral Quiroga2015-08-032-0/+29
| | | | Reviewed-by: Jason Ekstrand <[email protected]>
* i965/gs/gen6: Refactor ir_emit_vertex and ir_end_primitive for gen6Samuel Iglesias Gonsalvez2015-08-032-1/+14
| | | | | | | | | So the implementation is independent of GLSL IR and the visit methods of the gen6 GS visitor. This way we will be able to reuse that implementation directly from the NIR vec4 backend. Signed-off-by: Samuel Iglesias Gonsalvez <[email protected]> Reviewed-by: Jason Ekstrand <[email protected]>
* i965/nir/gs: Implement EmitVertex and EndPrimitiveIago Toral Quiroga2015-08-032-0/+23
| | | | Reviewed-by: Jason Ekstrand <[email protected]>
generated by cgit v1.2.3 (git 2.39.1) at 2024-12-29 01:27:07 +0000