diff options
| author | Ian Romanick <ian.d.romanick@intel.com> | 2018-06-05 15:04:24 -0700 |
|---|---|---|
| committer | Dylan Baker <dylan@pnwbakers.com> | 2018-06-15 13:55:04 -0700 |
| commit | f3ec346ab19d25de43a3fd14f533fbc40522d8ea (patch) | |
| tree | df1bfa1663e2b29728b713c16cbff7b09869f818 | |
| parent | ae3949683182dfa9d20de2e4d7e59dce02586260 (diff) | |
glsl: Don't copy propagate from SSBO or shared variables either
Since SSBOs can be written by other GPU threads, copy propagating a read
can cause the value to magically change. SSBO reads are also very
expensive, so doing it twice will be slower.
Haswell, Broadwell, and Skylake had similar results. (Skylake shown)
total instructions in shared programs: 14399120 -> 14399119 (<.01%)
instructions in affected programs: 684 -> 683 (-0.15%)
helped: 1
HURT: 0
total cycles in shared programs: 532978931 -> 532973113 (<.01%)
cycles in affected programs: 530484 -> 524666 (-1.10%)
helped: 1
HURT: 0
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Cc: mesa-stable@lists.freedesktop.org
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=106774
(cherry picked from commit 461a5c899c08064467abb635536381a5a5659280)
| -rw-r--r-- | src/compiler/glsl/opt_copy_propagation.cpp | 2 |
1 files changed, 2 insertions, 0 deletions
diff --git a/src/compiler/glsl/opt_copy_propagation.cpp b/src/compiler/glsl/opt_copy_propagation.cpp index 6220aa86da..206dffe4f1 100644 --- a/src/compiler/glsl/opt_copy_propagation.cpp +++ b/src/compiler/glsl/opt_copy_propagation.cpp @@ -347,6 +347,8 @@ ir_copy_propagation_visitor::add_copy(ir_assignment *ir) if (lhs_var != NULL && rhs_var != NULL && lhs_var != rhs_var) { if (lhs_var->data.mode != ir_var_shader_storage && lhs_var->data.mode != ir_var_shader_shared && + rhs_var->data.mode != ir_var_shader_storage && + rhs_var->data.mode != ir_var_shader_shared && lhs_var->data.precise == rhs_var->data.precise) { _mesa_hash_table_insert(acp, lhs_var, rhs_var); } |