Commit graph

9 commits

Author SHA1 Message Date
Johan Engelen
e2edc3e9f4
Add __ir to ldc.llvmasm and deprecate ldc.simd.inlineIR. (#2931) 2018-12-14 21:21:07 +01:00
Martin Kinkelin
f46e7e70a9 Adapt lit-tests to new FileCheck CHECK-DAG semantics
See https://reviews.llvm.org/D47106.
2018-09-26 17:21:35 +02:00
Johan Engelen
f98adfa8a3 Fix inlineIR_math.d test. (#2063)
I don't know why the original test tests that fused mult-add is not emitted for these pieces, while at the same time specifying "fast" math on the operations.
2017-04-09 09:23:00 +02:00
Johan Engelen
00dfb4d138 Apply TargetMachine options as function attributes in IR. Needed for LTO. 2016-10-06 19:06:30 +02:00
Martin
fc2523aa46 Fix issue #1711 2016-08-21 14:26:23 +02:00
Johan Engelen
e50f15af7f Tests: use -mtriple= instead of without equals sign. 2016-05-27 23:49:56 +02:00
Johan Engelen
91d308fa00 Apply "unsafe-fp-math"="false" per default on all functions.
This is needed because of: see comment in applyDefaultMathAttributes.
2016-05-22 19:56:57 +02:00
Johan Engelen
7b8f5d375e [LIT] Add features for the targets that LLVM can target.
Examples: target_X86, target_ARM.
Apart from future need, this is currently needed for a few tests that specifically test X86 capabilities.
2016-05-11 23:08:27 +02:00
Johan Engelen
9955d8e4e4 For LDC_inline_ir functions: instantiate a new "alwaysinline" function for each call, which is the current behavior for inline ASM too.
When certain attributes are applied to the calling function, like "unsafe-fp-math", the inlined inlineIR function has to have the same attributes otherwise the calling function attribute will be reset to the safe merge of the two: "false". Because the same inlineIR function can be called in different functions using an alias definition, a new function (with possibly different attributes) has to be instantiated upon every call.

Related GH issue #1438
2016-04-21 10:16:33 +02:00