Mesa (staging/21.3): aco: only load streamout buffers if streamout is enabled

[GitLab Mirror]

Module: Mesa
Branch: staging/21.3
Commit: 5635071adc2c578e90b35ceaff65a16349c39a20
URL:    http://cgit.freedesktop.org/mesa/mesa/commit/?id=5635071adc2c578e90b35ceaff65a16349c39a20

Author: Samuel Pitoiset <samuel.pitoiset at gmail.com>
Date:   Mon Oct 25 13:58:20 2021 +0200

aco: only load streamout buffers if streamout is enabled

The streamout_config SGPR is used to determine if streamout is enabled.

This fixes a GPU hang with various transform feedback tests:
 - dEQP-GLES3.functional.transform_feedback.*
 - KHR-GL46.transform_feedback.api_errors_test
 - KHR-GL46.draw_indirect.basic-draw*-xfbPaused
 - KHR-GL46.geometry_shader.api.draw_calls_while_tf_is_paused

Cc: 21.3 mesa-stable
Signed-off-by: Samuel Pitoiset <samuel.pitoiset at gmail.com>
Reviewed-by: Rhys Perry <pendingchaos02 at gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13514>
(cherry picked from commit dc74285d32df9c09eb896bc4571066402d32d10b)

---

 .pick_status.json                              |  2 +-
 src/amd/compiler/aco_instruction_selection.cpp | 16 +++++-----------
 2 files changed, 6 insertions(+), 12 deletions(-)

diff --git a/.pick_status.json b/.pick_status.json
index 4e8c787513c..429ae82a566 100644
--- a/.pick_status.json
+++ b/.pick_status.json
@@ -13,7 +13,7 @@
         "description": "aco: only load streamout buffers if streamout is enabled",
         "nominated": true,
         "nomination_type": 0,
-        "resolution": 0,
+        "resolution": 1,
         "main_sha": null,
         "because_sha": null
     },
diff --git a/src/amd/compiler/aco_instruction_selection.cpp b/src/amd/compiler/aco_instruction_selection.cpp
index 808e004edfa..ebc8555d40d 100644
--- a/src/amd/compiler/aco_instruction_selection.cpp
+++ b/src/amd/compiler/aco_instruction_selection.cpp
@@ -11218,17 +11218,6 @@ emit_streamout(isel_context* ctx, unsigned stream)
 {
    Builder bld(ctx->program, ctx->block);
 
-   Temp so_buffers[4];
-   Temp buf_ptr = convert_pointer_to_64_bit(ctx, get_arg(ctx, ctx->args->streamout_buffers));
-   for (unsigned i = 0; i < 4; i++) {
-      unsigned stride = ctx->program->info->so.strides[i];
-      if (!stride)
-         continue;
-
-      Operand off = bld.copy(bld.def(s1), Operand::c32(i * 16u));
-      so_buffers[i] = bld.smem(aco_opcode::s_load_dwordx4, bld.def(s4), buf_ptr, off);
-   }
-
    Temp so_vtx_count =
       bld.sop2(aco_opcode::s_bfe_u32, bld.def(s1), bld.def(s1, scc),
                get_arg(ctx, ctx->args->ac.streamout_config), Operand::c32(0x70010u));
@@ -11245,13 +11234,18 @@ emit_streamout(isel_context* ctx, unsigned stream)
    Temp so_write_index =
       bld.vadd32(bld.def(v1), get_arg(ctx, ctx->args->ac.streamout_write_index), tid);
 
+   Temp so_buffers[4];
    Temp so_write_offset[4];
+   Temp buf_ptr = convert_pointer_to_64_bit(ctx, get_arg(ctx, ctx->args->streamout_buffers));
 
    for (unsigned i = 0; i < 4; i++) {
       unsigned stride = ctx->program->info->so.strides[i];
       if (!stride)
          continue;
 
+      so_buffers[i] = bld.smem(aco_opcode::s_load_dwordx4, bld.def(s4), buf_ptr,
+                               bld.copy(bld.def(s1), Operand::c32(i * 16u)));
+
       if (stride == 1) {
          Temp offset = bld.sop2(aco_opcode::s_add_i32, bld.def(s1), bld.def(s1, scc),
                                 get_arg(ctx, ctx->args->ac.streamout_write_index),