summaryrefslogtreecommitdiffstats
path: root/src/gallium/auxiliary/gallivm
Commit message (Collapse)AuthorAgeFilesLines
* gallivm: More accurate float -> 24bit & 32bit unorm conversion.José Fonseca2010-10-131-40/+86
|
* gallivm: work-around trilinear mipmap filtering regression with LLVM 2.8Brian Paul2010-10-131-0/+20
| | | | The bug only happens on the AOS / fixed-pt path.
* gallivm: Remove unnecessary header.Vinson Lee2010-10-131-1/+0
|
* gallivm: only use lp_build_conv 4x4f -> 1x16 ub fastpath with sse2Roland Scheidegger2010-10-131-19/+5
| | | | | | | | This is relying on lp_build_pack2 using the sse2 pack intrinsics which handle clamping. (Alternatively could have make it use lp_build_packs2 but it might not even produce more efficient code than not using the fastpath in the first place.)
* gallivm: remove newlinesBrian Paul2010-10-121-2/+0
|
* gallivm: fix different handling of [non]normalized coords in linear soa pathRoland Scheidegger2010-10-131-16/+6
| | | | | There seems to be no reason for it, so do same math for both (except the scale mul, of course).
* gallivm: Name anonymous union.José Fonseca2010-10-122-7/+7
|
* gallivm: don't branch on KILLs near end of shaderKeith Whitwell2010-10-121-10/+47
|
* gallivm: More detailed analysis of tgsi shaders.José Fonseca2010-10-112-0/+557
| | | | To allow more optimizations, in particular for direct textures.
* gallivm: Eliminate unsigned integer arithmetic from texture coordinates.José Fonseca2010-10-113-42/+32
| | | | | | | | | | | SSE support for 32bit and 16bit unsigned arithmetic is not complete, and can easily result in inefficient code. In most cases signed/unsigned doesn't make a difference, such as for integer texture coordinates. So remove uint_coord_type and uint_coord_bld to avoid inefficient operations to sneak in the future.
* gallivm: Pass texture coords derivates as scalars.José Fonseca2010-10-104-26/+38
| | | | | We end up treating them as scalars in the end, and it saves some instructions.
* gallivm: Use variables instead of Phis in loops.José Fonseca2010-10-102-42/+23
| | | | With this commit all explicit Phi emission is now gone.
* gallivm: Allow to disable bri-linear filtering with ↵José Fonseca2010-10-103-9/+10
| | | | GALLIVM_DEBUG=no_brilinear runtime option
* gallivm: Fix a long standing bug with nested if-then-else emission.José Fonseca2010-10-101-17/+6
| | | | | | | | | | | | We can't patch true-block at end-if time, as there is no guarantee that the block at the beginning of the true stanza is the same at the end of the true stanza -- other control flow elements may have been emitted half way the true stanza. Although this bug surfaced recently with the commit to skip mip filtering when lod is an integer the bug was always there, although probably it was avoided until now: e.g., cubemap selection nests if-then-else on the else stanza, which does not suffer from the same problem.
* gallivm: Cleanup the rest of the flow module.José Fonseca2010-10-092-201/+37
|
* gallivm: Simplify if/then/else implementation.José Fonseca2010-10-095-91/+34
| | | | No need for for a flow stack anymore.
* gallivm: Factor out the SI->FP texture size conversion for SoA path tooJosé Fonseca2010-10-091-34/+56
|
* gallivm: Remove support for Phi generation.José Fonseca2010-10-092-221/+0
| | | | Simply rely on mem2reg pass. It's easier and more reliable.
* gallivm: Use varilables instead of Phis for cubemap selection.José Fonseca2010-10-091-36/+26
|
* gallivm: Don't generate Phis for execution mask.José Fonseca2010-10-092-8/+25
|
* gallivm: Special bri-linear computation path for unmodified rho.José Fonseca2010-10-091-20/+91
|
* gallivm: Less code duplication in log computation.José Fonseca2010-10-092-35/+88
|
* gallivm: Handle code have ret correctly.José Fonseca2010-10-091-2/+4
| | | | Stop disassembling on unconditional backwards jumps.
* llvmpipe: try to be sensible about whether to branch after mask updatesKeith Whitwell2010-10-093-6/+14
| | | | | Don't branch more than once in quick succession. Don't branch at the end of the shader.
* gallivm: simpler uint8->float conversionsKeith Whitwell2010-10-091-0/+10
| | | | | LLVM seems to finds it easier to reason about these than our mantissa-manipulation code.
* gallivm: prefer blendvb for integer argumentsKeith Whitwell2010-10-091-2/+4
|
* llvmpipe: store zero into all alloca'd valuesKeith Whitwell2010-10-091-0/+1
| | | | Fixes slowdown in isosurf with earlier versions of llvm.
* gallivm: Do size computations simultanously for all dimensions (AoS).José Fonseca2010-10-094-92/+177
| | | | | | | | | | Operate simultanouesly on <width, height, depth> vector as much as possible, instead of doing the operations on vectors with broadcasted scalars. Also do the 24.8 fixed point scalar with integer shift of the texture size, for unnormalized coordinates. AoS path only for now -- the same thing can be done for SoA.
* gallivm: make use of new iround code in lp_bld_conv.Roland Scheidegger2010-10-091-4/+4
| | | | Only requires sse2 now.
* gallivm: optimize soa linear clamp to edge wrap mode a bitRoland Scheidegger2010-10-091-23/+30
| | | | | | | | | Clamp against 0 instead of -0.5, which simplifies things. The former version would have resulted in both int coords being zero (in case of coord being smaller than 0) and some "unused" weight value, whereas now the int coords will be 0 and 1, but weight will be 0, hence the lerp should produce the same value. Still not happy about differences between normalized and non-normalized...
* gallivm: avoid unnecessary URem in linear wrap repeat caseRoland Scheidegger2010-10-091-3/+7
| | | | | | Haven't looked at what code this exactly generates but URem can't be fast. Instead of using two URem only use one and replace the second one with select/add (this is what the corresponding aos code already does).
* gallivm: more linear tex wrap mode calculation simplificationRoland Scheidegger2010-10-091-12/+9
| | | | | | | Rearrange order of operations a bit to make some clamps easier. All calculations should be equivalent. Note there seems to be some inconsistency in the clamp to edge case wrt normalized/non-normalized coords, could potentially simplify this too.
* gallivm: optimize some tex wrap mode calculations a bitRoland Scheidegger2010-10-091-6/+11
| | | | | | | | | Sometimes coords are clamped to positive numbers before doing conversion to int, or clamped to 0 afterwards, in this case can use itrunc instead of ifloor which is easier. This is only the case for nearest calculations unfortunately, except linear MIRROR_CLAMP_TO_EDGE which for the same reason can use a unsigned float build context so the ifloor_fract helper can reduce this to itrunc in the ifloor helper itself.
* gallivm: replace sub/floor/ifloor combo with ifloor_fractRoland Scheidegger2010-10-091-5/+1
|
* gallivm: faster iround implementation for sse2Roland Scheidegger2010-10-091-1/+53
| | | | | | sse2 supports round to nearest directly (or rather, assuming default nearest rounding mode in MXCSR). Use intrinsic to use this rather than round (sse41) or bit manipulation whenever possible.
* gallivm: fix trunc/itrunc commentRoland Scheidegger2010-10-091-6/+6
| | | | trunc of -1.5 is -1.0 not 1.0...
* gallivm: Remove unnecessary header.Vinson Lee2010-10-081-1/+0
|
* gallivm: Help for combined extraction and broadcasting.José Fonseca2010-10-083-15/+102
| | | | Doesn't change generated code quality, but saves some typing.
* llvmpipe: First minify the texture size, then broadcast.José Fonseca2010-10-083-32/+42
|
* gallivm: Move into the as much of the second level code as possible.José Fonseca2010-10-085-219/+120
| | | | | Also, pass more stuff trhough the sample build context, instead of arguments.
* gallivm: Warn when doing inefficient integer comparisons.José Fonseca2010-10-081-2/+18
|
* gallivm: round rather than truncate in new 4x4f->1x16ub conversion pathKeith Whitwell2010-10-081-17/+42
|
* gallivm: Use the wrappers for SSE pack intrinsics.José Fonseca2010-10-081-15/+3
| | | | Fixes assertion failures on LLVM 2.6.
* gallivm: special case conversion 4x4f to 1x16ubKeith Whitwell2010-10-081-0/+84
| | | | | Nice reduction in the number of operations required for final color output in many shaders.
* gallivm: Implement brilinear filtering.José Fonseca2010-10-081-1/+89
|
* gallivm: Fix copy'n'paste typo in previous commit.José Fonseca2010-10-081-2/+2
|
* gallivm: Clamp mipmap level and zero mip weight simultaneously.José Fonseca2010-10-084-14/+52
|
* gallivm: Use lp_build_ifloor_fract for lod computation.José Fonseca2010-10-081-5/+3
| | | | Forgot this one before.
* gallivm: Don't compute the second mipmap level when frac(lod) == 0José Fonseca2010-10-082-139/+175
|
* gallivm: Simplify lp_build_mipmap_level_sizes' interface.José Fonseca2010-10-084-67/+42
|
generated by cgit v1.2.3 (git 2.39.1) at 2024-12-26 09:36:33 +0000