emscripten-fastcomp/lib/Transforms, branch 1.13.2 LLVM with the emscripten fastcomp javascript backend https://git.amat.us/emscripten-fastcomp/atom/lib/Transforms?h=1.13.2 2014-03-08T01:35:43Z fix the case where exceptions are lowered and we already had getHigh/preInvoke/postInvoke defined in the module 2014-03-08T01:35:43Z Alon Zakai alonzakai@gmail.com 2014-03-08T01:35:43Z urn:sha1:06b7ce14ded3d39a8e4703d22b360fa782c1522a Implement integer promotion for urem, udiv, srem, and sdiv 2014-03-06T20:56:53Z Dan Gohman sunfish@mozilla.com 2014-03-06T20:23:34Z urn:sha1:68081ee62375a1d900d46f969e38330f14fb8775 The optimizer sometimes thinks it's beneficial to truncate all manner of i64 operators to narrower types, even when still wider than the platform's widest legal type. Fix a use-after-free error in GlobalOpt CleanupConstantGlobalUsers 2014-03-06T19:54:11Z Dan Gohman sunfish@mozilla.com 2014-03-06T19:51:23Z urn:sha1:d644d4a9ca114ceb5d6e69f0c0df6cd571bb33c4 This is a backport of r197178 from LLVM trunk. Disable the ResolveAliases and GlobalCleanup passes. 2014-03-06T03:42:53Z Dan Gohman sunfish@mozilla.com 2014-03-06T03:40:14Z urn:sha1:805f34e607fadc603eefa3120ff514a2f3432392 Also add a test for handling of global aliases. Disable the LowerExpectIntrinsic pass. 2014-03-06T03:04:42Z Dan Gohman sunfish@mozilla.com 2014-03-06T03:04:42Z urn:sha1:0e03f9e9cecf388df2e3d19139be0a9faca63672 Don't run a pass to strip dead prototypes; just ignore them. 2014-03-06T02:51:17Z Dan Gohman sunfish@mozilla.com 2014-03-06T01:51:45Z urn:sha1:8346647dd5ddc2d1f379cc530b20494a41e6ef31 Also, fix the code for supressing declarations for no-op intrinsics. Clean up the handling of inline asm. 2014-03-06T01:48:22Z Dan Gohman sunfish@mozilla.com 2014-03-06T01:48:22Z urn:sha1:9bef807b3bfa2397fb24f4f9c4df5a633a9717b3 Make inline asm a report_fatal_error instead of an assertion failure so that it's a little friendlier, and add a test to make sure llc -march=js rejects inline asm. Also disable the PNaCl inline asm("":::"memory") lowering pass. If people are using this, it's best that we diagnose it as it likely isn't portable. There are usually better alternatives. Fix alignment of vararg buffer allocas. 2014-03-03T23:33:21Z Dan Gohman sunfish@mozilla.com 2014-03-03T23:11:09Z urn:sha1:fa1c90b56aa3697fb0868ccea0befc0be3f89921 Since we do 8-byte loads for va_arg of double, make sure the vararg buffer alloca is 8-byte aligned. Use range metadata instead of introducing selects. 2014-02-28T20:39:49Z Dan Gohman sunfish@mozilla.com 2014-02-28T20:39:49Z urn:sha1:dc59054db06864099c1b1051b8ef28a30f13e6c7 When GlobalOpt has determined that a GlobalVariable only ever has two values, it would convert the GlobalVariable to a boolean, and introduce SelectInsts at every load, to choose between the two possible values. These SelectInsts introduce overhead and other unpleasantness. This patch makes GlobalOpt just add range metadata to loads from such GlobalVariables instead. This enables the same main optimization (as seen in test/Transforms/GlobalOpt/integer-bool.ll), without introducing selects. The main downside is that it doesn't get the memory savings of shrinking such GlobalVariables, but this is expected to be negligible. Move ExpandI64 into lib/Target/JSBackend. 2014-02-28T19:22:37Z Dan Gohman sunfish@mozilla.com 2014-02-26T16:47:51Z urn:sha1:0dac79df771f7708a03e724126dded4f64531bc5