Mesa (main): aco: Remember when exec mask is const, and restore the const then.

[GitLab Mirror]

Module: Mesa
Branch: main
Commit: c850af936acacc0ae0c780a41ecf451edf82ce91
URL:    http://cgit.freedesktop.org/mesa/mesa/commit/?id=c850af936acacc0ae0c780a41ecf451edf82ce91

Author: Timur Kristóf <timur.kristof at gmail.com>
Date:   Fri May  7 15:02:58 2021 +0200

aco: Remember when exec mask is const, and restore the const then.

Previously, we would store even the constant -1 exec mask from the
beginning of every merged shader. With this change it is no longer
necessary because we can restore to constant exec mask directly.

Hence, this frees up a register pair (single register for Wave32)
in every merged shader.

No Fossil DB changes.

Signed-off-by: Timur Kristóf <timur.kristof at gmail.com>
Reviewed-by: Tony Wasserka <tony.wasserka at gmx.de>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10691>

---

 src/amd/compiler/aco_insert_exec_mask.cpp | 15 ++++++++++-----
 1 file changed, 10 insertions(+), 5 deletions(-)

diff --git a/src/amd/compiler/aco_insert_exec_mask.cpp b/src/amd/compiler/aco_insert_exec_mask.cpp
index 9741c12dca3..e39b5a40153 100644
--- a/src/amd/compiler/aco_insert_exec_mask.cpp
+++ b/src/amd/compiler/aco_insert_exec_mask.cpp
@@ -75,7 +75,7 @@ struct loop_info {
 };
 
 struct block_info {
-   std::vector<std::pair<Operand, uint8_t>> exec;
+   std::vector<std::pair<Operand, uint8_t>> exec; /* Vector of exec masks. Either a temporary or const -1. */
    std::vector<WQMState> instr_needs;
    uint8_t block_needs;
    uint8_t ever_again_needs;
@@ -399,13 +399,16 @@ unsigned add_coupling_code(exec_ctx& ctx, Block* block,
       assert(startpgm->opcode == aco_opcode::p_startpgm);
       bld.insert(std::move(startpgm));
 
+      Operand start_exec(bld.lm);
+
       /* exec seems to need to be manually initialized with combined shaders */
       if (ctx.program->stage.num_sw_stages() > 1 || ctx.program->stage.hw == HWStage::NGG) {
-         bld.copy(Definition(exec, bld.lm), Operand(UINT32_MAX, bld.lm == s2));
+         start_exec = Operand(-1u, bld.lm == s2);
+         bld.copy(Definition(exec, bld.lm), start_exec);
       }
 
       if (ctx.handle_wqm) {
-         ctx.info[0].exec.emplace_back(Operand(bld.lm), mask_type_global | mask_type_exact);
+         ctx.info[0].exec.emplace_back(start_exec, mask_type_global | mask_type_exact);
          /* if this block only needs WQM, initialize already */
          if (ctx.info[0].block_needs == WQM)
             transition_to_WQM(ctx, bld, 0);
@@ -417,7 +420,7 @@ unsigned add_coupling_code(exec_ctx& ctx, Block* block,
          } else {
             mask |= mask_type_exact;
          }
-         ctx.info[0].exec.emplace_back(Operand(bld.lm), mask);
+         ctx.info[0].exec.emplace_back(start_exec, mask);
       }
 
       return 1;
@@ -641,7 +644,9 @@ unsigned add_coupling_code(exec_ctx& ctx, Block* block,
    if (block->kind & block_kind_merge && !ctx.info[idx].exec.back().first.isUndefined()) {
       Operand restore = ctx.info[idx].exec.back().first;
       assert(restore.size() == bld.lm.size());
-      ctx.info[idx].exec.back().first = bld.pseudo(aco_opcode::p_parallelcopy, Definition(exec, bld.lm), restore);
+      bld.pseudo(aco_opcode::p_parallelcopy, Definition(exec, bld.lm), restore);
+      if (!restore.isConstant())
+         ctx.info[idx].exec.back().first = Operand(bld.lm);
    }
 
    return i;