Scott Todd

It's been a few months with no action taken here. Is anyone using the traces from the benchmark CI? At this point I'd vote to remove Tracy from that path...

> Maybe as a first step disable on PR runs as discussed above [#16856 (comment)](https://github.com/iree-org/iree/issues/16856#issuecomment-2012396348). Then after a month of that, ask whether to drop altogether. Unless that benchmarking has...

> Ran the experiment at #16857... found that wholesale dropping all the related code would be a -345 lines of code shrink, and found the following timings from comparing that...

Oh, I was looking at android-cpu + android-gpu (the default config for postsubmit), while that table was generated on a pull request with just android-cpu only. The timings make sense...

> try --trace_execution? ``` [module.__init+00000000] [module.__init+00000001] %r0 = vm.const.ref.zero [module.__init+00000004] %i0 = vm.const.i32 2 // 0x00000002 [module.__init+0000000B] %i1 = vm.const.i32 7 // 0x00000007 [module.__init+00000012] %i2 = vm.const.i32 1 // 0x00000001...

That was my feeling too, or a divergent / infinite loop based on the input values (wondering about the ui8 inputs).

I no longer see these hangs locally. Weird, since I want to debug an unrelated hang and thought these could be helpful with debugging :/

Starting to narrow down the poor failure mode here. Seems like pytest-timeout and pytest-retry are (sometimes?) incompatible with one another. Still not sure why the tests are actually hanging though.

Related: https://github.com/nod-ai/SHARK-TestSuite/issues/47

Tried with `--iree-vulkan-target-triple=turing-unknown-unknown` and got a similar error: ``` open_llama_3b_v2_f16_vulkan\configured_module_prefill_bs4$async_dispatch_1.mlir:9:6: error: 'func.func' op uses -127270912 bytes of shared memory; exceeded the limit of 49152 bytes ```