intel/fs: Write the address register with NoMask for MOV_INDIRECT

This fixes a hang in the following Vulkan CTS test on TGL-LP:

    dEQP-VK.descriptor_indexing.storage_buffer_dynamic_in_loop

Cc: mesa-stable@lists.freedesktop.org
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Tested-by: Marge Bot <https://gitlab.freedesktop.org/mesa/mesa/merge_requests/3642>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/merge_requests/3642>
(cherry picked from commit f93dfb509cbf9474befae9919dd8c135bbd67093)

diff --git a/.pick_status.json b/.pick_status.json
index 37c0d97..1140a73 100644 (file)
--- a/.pick_status.json
+++ b/.pick_status.json
@@ -382,7 +382,7 @@
         "description": "intel/fs: Write the address register with NoMask for MOV_INDIRECT",
         "nominated": true,
         "nomination_type": 0,
-        "resolution": 0,
+        "resolution": 1,
         "master_sha": null,
         "because_sha": null
     },

diff --git a/src/intel/compiler/brw_fs_generator.cpp b/src/intel/compiler/brw_fs_generator.cpp
index 036aa58..61ac089 100644 (file)
--- a/src/intel/compiler/brw_fs_generator.cpp
+++ b/src/intel/compiler/brw_fs_generator.cpp
@@ -452,8 +452,17 @@ fs_generator::generate_mov_indirect(fs_inst *inst,
        * In the end, while base_offset is nice to look at in the generated
        * code, using it saves us 0 instructions and would require quite a bit
        * of case-by-case work.  It's just not worth it.
+       *
+       * There's some sort of HW bug on Gen12 which causes issues if we write
+       * to the address register in control-flow.  Since we only ever touch
+       * the address register from the generator, we can easily enough work
+       * around it by setting NoMask on the add.
        */
+      brw_push_insn_state(p);
+      if (devinfo->gen == 12)
+         brw_set_default_mask_control(p, BRW_MASK_DISABLE);
       brw_ADD(p, addr, indirect_byte_offset, brw_imm_uw(imm_byte_offset));
+      brw_pop_insn_state(p);
       brw_set_default_swsb(p, tgl_swsb_regdist(1));
 
       if (type_sz(reg.type) > 4 &&