review.tizen.org Git - contrib/beignet.git/commit

author	Ruiling Song <ruiling.song@intel.com>
	Tue, 15 Apr 2014 08:53:17 +0000 (16:53 +0800)
committer	Zhigang Gong <zhigang.gong@intel.com>
	Mon, 21 Apr 2014 01:09:40 +0000 (09:09 +0800)
commit	ae1e342c5bf7e0c6e325313aefffa3bd168e5678
tree	4d4c5c14f133730e9602ac33621825481b84f1dc	tree \| snapshot
parent	6e08fb89c08e3ff11ed9b7222a1e1bfe8cb6fb9d	commit \| diff

GBE: Implement instruction compact.

A native GEN ASM would takes 2*64bit, but GEN also support compact instruction
which only takes 64bit. To make code easily understood, GenInstruction now only
stands for 64bit memory, and use GenNativeInstruction & GenCompactInstruction
to represent normal(native) and compact instruction.

After this change, it is not easily to map SelectionInstruction distance to ASM distance.
As the instructions in the distance maybe compacted. To not introduce too much
complexity, JMP, IF, ENDIF, NOP will NEVER be compacted.

Some experiment in luxMark shows it could reduce about 20% instruction memory.
But it is sad that no performance improvement observed.

Signed-off-by: Ruiling Song <ruiling.song@intel.com>
Reviewed-by: Zhigang Gong <zhigang.gong@linux.intel.com>

backend/src/CMakeLists.txt		diff \| blob \| history
backend/src/backend/gen/gen_mesa_disasm.c		diff \| blob \| history
backend/src/backend/gen_context.cpp		diff \| blob \| history
backend/src/backend/gen_defs.hpp		diff \| blob \| history
backend/src/backend/gen_encoder.cpp		diff \| blob \| history
backend/src/backend/gen_encoder.hpp		diff \| blob \| history
backend/src/backend/gen_insn_compact.cpp	[new file with mode: 0644]	blob
backend/src/backend/gen_insn_selection.cpp		diff \| blob \| history
backend/src/backend/gen_reg_allocation.cpp		diff \| blob \| history
backend/src/backend/gen_register.hpp		diff \| blob \| history