Re: [Mesa-dev] [PATCH 1/1] nir: Use a freelist in nir_opt_dce to avoid spamming ralloc

Yup, most definitely. I just have one more thing to test before sending out a V2. I've toyed around with arrays and sets and stuff to see if there are better options than a linked list. At least for now the answer is: "no, there isn't", but I'm gonna test u_vector for this use later today to see if

Re: [Mesa-dev] [PATCH 1/1] nir: Use a freelist in nir_opt_dce to avoid spamming ralloc

Hello Thomas, is this useful even after '[Mesa-dev] [PATCH 0/2] V2: Use hash table cloning in copy propagation' landed? I've running both together with Dave's '[Mesa-dev] [PATCH] radv/winsys: replace bo list searchs with a hash table.' patch. Dieter Am 24.01.2018 08:33, schrieb Thomas Hell

Re: [Mesa-dev] [PATCH 1/1] nir: Use a freelist in nir_opt_dce to avoid spamming ralloc

2018-01-21 23:58 GMT+01:00 Eric Anholt : > Thomas Helland writes: > >> Also, allocate worklist_elem in groups of 20, to reduce the burden of >> allocation. Do not use rzalloc, as there is no need. This lets us drop >> the number of calls to ralloc from aproximately 10% of all calls to >> ralloc(13

Re: [Mesa-dev] [PATCH 1/1] nir: Use a freelist in nir_opt_dce to avoid spamming ralloc

Thomas Helland writes: > Also, allocate worklist_elem in groups of 20, to reduce the burden of > allocation. Do not use rzalloc, as there is no need. This lets us drop > the number of calls to ralloc from aproximately 10% of all calls to > ralloc(130 000 calls), down to a mere 2000 calls to ra

Re: [Mesa-dev] [PATCH 1/1] nir: Use a freelist in nir_opt_dce to avoid spamming ralloc

Hello to all of you and a Happy New Year! ;-) Any thoughts about this, Marek, Nicolai? I'm running this for four weeks without any hickup, now. Thanks, Dieter Am 06.12.2017 09:56, schrieb Dieter Nützel: Tested-by: Dieter Nützel Dieter Am 02.12.2017 15:49, schrieb Thomas Helland: Also, all

Re: [Mesa-dev] [PATCH 1/1] nir: Use a freelist in nir_opt_dce to avoid spamming ralloc

Tested-by: Dieter Nützel Dieter Am 02.12.2017 15:49, schrieb Thomas Helland: Also, allocate worklist_elem in groups of 20, to reduce the burden of allocation. Do not use rzalloc, as there is no need. This lets us drop the number of calls to ralloc from aproximately 10% of all calls to ralloc(1

[Mesa-dev] [PATCH 1/1] nir: Use a freelist in nir_opt_dce to avoid spamming ralloc

Also, allocate worklist_elem in groups of 20, to reduce the burden of allocation. Do not use rzalloc, as there is no need. This lets us drop the number of calls to ralloc from aproximately 10% of all calls to ralloc(130 000 calls), down to a mere 2000 calls to ralloc_array_size. This cuts the r