Lines Matching full:gallivm
222 <li>gallivm: expand AoS sampling to cover all filtering modes</li>
223 <li>gallivm: fix incorrect vector shuffle datatype</li>
224 <li>gallivm: move i32_vec_type inside the #ifdef</li>
226 <li>gallivm: fix wrong return value in bitwise functions</li>
228 <li>gallivm: added missing case for PIPE_TEXTURE_RECT</li>
237 <li>gallivm: remove debug code</li>
250 <li>gallivm: fix lp_build_sample_compare()</li>
261 <li>gallivm: fix repeat() function for NPOT textures</li>
276 <li>gallivm: check for level=0 case in lp_build_minify()</li>
277 <li>gallivm: added some comments</li>
285 <li>gallivm: remove newlines</li>
289 <li>gallivm: work-around trilinear mipmap filtering regression with LLVM 2.8</li>
291 <li>gallivm: add compile-time option to emit inst addrs and/or line numbers</li>
293 <li>gallivm: added lp_build_print_vec4()</li>
294 <li>gallivm: added lp_build_load_volatile()</li>
300 <li>gallivm: use util_snprintf()</li>
303 <li>gallivm: fix incorrect type for zero vector in emit_kilp()</li>
406 <li>gallivm: implement scatter stores into temp register file</li>
407 <li>gallivm: add some LLVM var labels</li>
408 <li>gallivm: added debug code to dump temp registers</li>
409 <li>gallivm: add pixel offsets in scatter stores</li>
410 <li>gallivm: added lp_elem_type()</li>
411 <li>gallivm: implement execution mask for scatter stores</li>
414 <li>gallivm: alloca() was called too often for temporary arrays</li>
415 <li>gallivm: add const qualifiers, fix comment string</li>
466 <li>gallivm/llvmpipe: squash merge of the llvm-context branch</li>
475 <li>gallivm/llvmpipe: remove lp_build_context::builder</li>
476 <li>gallivm: fix null builder pointers</li>
1676 <li>gallivm: Fix address register swizzle.</li>
1677 <li>gallivm: Start collecting bitwise arithmetic helpers in a new module.</li>
1678 <li>gallivm: Clamp indirect register indices to file_max.</li>
1685 <li>gallivm: Add unorm support to lp_build_lerp()</li>
1690 <li>gallivm: Fetch the lod from the dynamic state when min_lod == max_lod.</li>
1691 <li>gallivm: Remove dead experimental code.</li>
1697 <li>gallivm: Use SSE4.1's ROUNDSS/ROUNDSD for scalar rounding.</li>
1698 <li>gallivm: More comprehensive border usage logic.</li>
1703 <li>gallivm: Take the type signedness in consideration in round/ceil/floor.</li>
1704 <li>gallivm: Use a faster (and less accurate) log2 in lod computation.</li>
1705 <li>gallivm: Fast implementation of iround(log2(x))</li>
1706 <li>gallivm: Combined ifloor & fract helper.</li>
1707 <li>gallivm: Only apply min/max_lod when necessary.</li>
1708 <li>gallivm: Compute lod as integer whenever possible.</li>
1711 <li>gallivm: Vectorize the rho computation.</li>
1712 gallivm: Do not do mipfiltering when magnifying.</li>
1713 <li>gallivm: Simplify lp_build_mipmap_level_sizes' interface.</li>
1714 <li>gallivm: Don't compute the second mipmap level when frac(lod) == 0</li>
1715 <li>gallivm: Use lp_build_ifloor_fract for lod computation.</li>
1716 <li>gallivm: Clamp mipmap level and zero mip weight simultaneously.</li>
1717 <li>gallivm: Fix copy'n'paste typo in previous commit.</li>
1718 <li>gallivm: Implement brilinear filtering.</li>
1719 <li>gallivm: Use the wrappers for SSE pack intrinsics.</li>
1720 <li>gallivm: Avoid control flow for two-sided stencil test.</li>
1721 <li>gallivm: Warn when doing inefficient integer comparisons.</li>
1722 <li>gallivm: Move into the as much of the second level code as possible.</li>
1724 <li>gallivm: Help for combined extraction and broadcasting.</li>
1725 <li>gallivm: Do size computations simultanously for all dimensions (AoS).</li>
1728 <li>gallivm: Handle code have ret correctly.</li>
1730 <li>gallivm: Less code duplication in log computation.</li>
1731 <li>gallivm: Special bri-linear computation path for unmodified rho.</li>
1732 <li>gallivm: Don't generate Phis for execution mask.</li>
1733 <li>gallivm: Use varilables instead of Phis for cubemap selection.</li>
1734 <li>gallivm: Remove support for Phi generation.</li>
1735 <li>gallivm: Factor out the SI->FP texture size conversion for SoA path too</li>
1736 <li>gallivm: Simplify if/then/else implementation.</li>
1737 <li>gallivm: Cleanup the rest of the flow module.</li>
1738 <li>gallivm: Fix a long standing bug with nested if-then-else emission.</li>
1739 <li>gallivm: Allow to disable bri-linear filtering with GALLIVM_DEBUG=no_brilinear runtime option</li>
1740 <li>gallivm: Use variables instead of Phis in loops.</li>
1741 <li>gallivm: Pass texture coords derivates as scalars.</li>
1743 <li>gallivm: Eliminate unsigned integer arithmetic from texture coordinates.</li>
1746 <li>gallivm: More detailed analysis of tgsi shaders.</li>
1751 <li>gallivm: Name anonymous union.</li>
1753 <li>gallivm: More accurate float -> 24bit & 32bit unorm conversion.</li>
1757 <li>gallivm: Fix SoA cubemap derivative computation.</li>
1760 <li>gallivm: Comment lp_build_insert_new_block().</li>
1761 <li>gallivm: Add a note about SSE4.1's nearest mode rounding.</li>
1764 <li>gallivm: always enable LLVMAddInstructionCombiningPass()</li>
1765 <li>gallivm: Remove the EMMS opcodes.</li>
1781 <li>gallivm: Allocate TEMP/OUT arrays only once.</li>
1816 <li>gallivm: make lp_build_sample_nop public</li>
1826 <li>gallivm: don't apply zero lod_bias</li>
1829 <li>gallivm: special case conversion 4x4f to 1x16ub</li>
1830 <li>gallivm: round rather than truncate in new 4x4f->1x16ub conversion path</li>
1838 <li>gallivm: specialized x8z24 depthtest path</li>
1839 <li>gallivm: prefer blendvb for integer arguments</li>
1840 <li>gallivm: simpler uint8->float conversions</li>
1850 <li>gallivm: don't branch on KILLs near end of shader</li>
2305 <li>gallivm: fix copy&paste bug</li>
2306 <li>gallivm: don't use URem/UDiv when calculating offsets for blocks</li>
2307 <li>gallivm: optimize yuv decoding</li>
2308 <li>gallivm: fix trunc/itrunc comment</li>
2309 <li>gallivm: faster iround implementation for sse2</li>
2310 <li>gallivm: replace sub/floor/ifloor combo with ifloor_fract</li>
2311 <li>gallivm: optimize some tex wrap mode calculations a bit</li>
2312 <li>gallivm: more linear tex wrap mode calculation simplification</li>
2313 <li>gallivm: avoid unnecessary URem in linear wrap repeat case</li>
2314 <li>gallivm: optimize soa linear clamp to edge wrap mode a bit</li>
2315 <li>gallivm: make use of new iround code in lp_bld_conv.</li>
2316 <li>gallivm: fix different handling of [non]normalized coords in linear soa path</li>
2317 <li>gallivm: only use lp_build_conv 4x4f -> 1x16 ub fastpath with sse2</li>
2478 <li>gallivm: Remove unnecessary headers.</li>
2483 <li>gallivm: Remove unnecessary header.</li>
2556 <li>gallivm: Remove unnecessary header.</li>
2567 <li>gallivm: Remove unnecessary header.</li>
2581 <li>gallivm: Silence uninitialized variable warnings.</li>
2582 <li>gallivm: Silence uninitialized variable warnings.</li>
2583 <li>gallivm: Silence uninitialized variable warning.</li>
2709 <li>gallivm: implement indirect addressing of the output registers</li>
2710 <li>gallivm: implement indirect addressing over inputs</li>
2711 <li>gallivm: fix storing of the addr register</li>
