diff options
| author | Timur Kristóf <[email protected]> | 2020-03-09 17:07:41 +0100 |
|---|---|---|
| committer | Marge Bot <[email protected]> | 2020-03-11 08:34:10 +0000 |
| commit | 655c050119719e185ae41bdafb1e62d71ccc3069 (patch) | |
| tree | ac486c65c7ff687fb3d6ae8c83ca774a8c8c297b /src | |
| parent | c70b0d0267234716e94aeaf0e585f27c8a8e21fc (diff) | |
aco: Fix combining DS additions in the optimizer.
Previously, it was calculated incorrectly for 64-bit writes and reads.
Signed-off-by: Timur Kristóf <[email protected]>
Reviewed-by: Rhys Perry <[email protected]>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/3964>
Diffstat (limited to 'src')
| -rw-r--r-- | src/amd/compiler/aco_optimizer.cpp | 13 |
1 files changed, 8 insertions, 5 deletions
diff --git a/src/amd/compiler/aco_optimizer.cpp b/src/amd/compiler/aco_optimizer.cpp index 95773c1205c..a18060f485b 100644 --- a/src/amd/compiler/aco_optimizer.cpp +++ b/src/amd/compiler/aco_optimizer.cpp @@ -816,12 +816,15 @@ void label_instruction(opt_ctx &ctx, Block& block, aco_ptr<Instruction>& instr) instr->opcode != aco_opcode::ds_swizzle_b32) { if (instr->opcode == aco_opcode::ds_write2_b32 || instr->opcode == aco_opcode::ds_read2_b32 || instr->opcode == aco_opcode::ds_write2_b64 || instr->opcode == aco_opcode::ds_read2_b64) { - if (offset % 4 == 0 && - ds->offset0 + (offset >> 2) <= 255 && - ds->offset1 + (offset >> 2) <= 255) { + unsigned mask = (instr->opcode == aco_opcode::ds_write2_b64 || instr->opcode == aco_opcode::ds_read2_b64) ? 0x7 : 0x3; + unsigned shifts = (instr->opcode == aco_opcode::ds_write2_b64 || instr->opcode == aco_opcode::ds_read2_b64) ? 3 : 2; + + if ((offset & mask) == 0 && + ds->offset0 + (offset >> shifts) <= 255 && + ds->offset1 + (offset >> shifts) <= 255) { instr->operands[i].setTemp(base); - ds->offset0 += offset >> 2; - ds->offset1 += offset >> 2; + ds->offset0 += offset >> shifts; + ds->offset1 += offset >> shifts; } } else { if (ds->offset0 + offset <= 65535) { |