Setup: * GEN9 HW * Latest Mesa from git Test-case: 1. Run something with SIMD32 fragment shaders $ INTEL_DEBUG=do32 glxgears Expected outcome: * Everything works fine, like it did earlier, and it does with SIMD16 Actual outcome: --------------------------------------------- glxgears: src/intel/compiler/brw_fs.cpp:1726: void fs_visitor::assign_urb_setup(): Assertion `inst->src[i].offset < REG_SIZE / 2' failed. Aborted (core dumped) --------------------------------------------- This has broken somewhere between: * 2019-04-22 e983a975c6843c307380d7caa083eee89e02bd3c: gallivm: disable NEON instructions if they are not supported * 2019-04-23 951d60f8cdc886adff09201ff65002e3ee1a4c61: radeonsi: delay adding BOs at the beginning of IBs until the first draw (Not sure what severity this should have. SIMD32 is important performance feature, e.g. on BXT it improves perf in one test-case by ~35%, and on average several percents with suitable heuristic.)
Looks like a duplicate of https://bugs.freedesktop.org/show_bug.cgi?id=110507 I took a look at it some time ago but failed.
*** This bug has been marked as a duplicate of bug 110507 ***
Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.