A significant perf increase is expected for shaders doing a lot of sqrt
operations.
shader-db changes for complete series:
total instructions in shared programs : 5472103 -> 5454825 (-0.32%)
total gprs used in shared programs : 647530 -> 647551 (0.00%)
total shared used in shared programs : 389120 -> 389120 (0.00%)
total local used in shared programs : 21064 -> 21064 (0.00%)
total bytes used in shared programs : 58459304 -> 58274448 (-0.32%)
local shared gpr inst bytes
helped 0 0 0 8526 8526
hurt 0 0 21 83 83
Karol Herbst (2):
gm200/ir: add native OP_SQRT support
gm200/ir: optimize rcp(sqrt) to rsq
.../drivers/nouveau/codegen/nv50_ir_emit_gm107.cpp | 4 +++-
.../drivers/nouveau/codegen/nv50_ir_lowering_nvc0.cpp | 3 +++
.../drivers/nouveau/codegen/nv50_ir_peephole.cpp | 11 ++++++++++-
.../drivers/nouveau/codegen/nv50_ir_target_gm107.cpp | 8 +++++++-
.../drivers/nouveau/codegen/nv50_ir_target_nvc0.cpp | 1 +
5 files changed, 24 insertions(+), 3 deletions(-)
--
2.17.1
_______________________________________________
mesa-dev mailing list
mesa-dev@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/mesa-dev
