Commit Graph

50 Commits

Author SHA1 Message Date
tevador
6e3136b37f Fixed cache alignment
Performance tuning
2019-04-06 17:07:40 +02:00
tevador
77dbe14658 SuperscalarHash JIT compiler
(unfinished)
2019-04-06 12:00:56 +02:00
tevador
59bbb572c2 WIP 2019-03-28 15:27:10 +01:00
tevador
107270d93d Reduced Dataset size to 2 GiB with 8 memory accesses per block
Disabled Dataset growth
2019-03-22 12:53:16 +01:00
tevador
233af9f14f Minor fixes for non-x86 platforms 2019-03-22 12:03:39 +01:00
tevador
4c1ae951de Merge branch 'feature/branches' into dev
Conflicts:
	src/JitCompilerX86.cpp
	src/JitCompilerX86.hpp
	src/main.cpp
2019-03-22 11:53:48 +01:00
tevador
73a11f5c01 CompiledLightVirtualMachine 2019-03-21 20:44:59 +01:00
tevador
007f8599b9 Implemented branches in the interpreter
Fixed x86 immediate encoding
2019-03-20 23:38:37 +01:00
tevador
174754cb2b Added branches - ASM and JIT only 2019-03-17 23:09:11 +01:00
tevador
958d2bdc15 Fixed non-portable deserialization 2019-03-11 23:04:34 +01:00
tevador
e65d9da66c Configurable parameters separated into configuration.h 2019-03-08 15:34:34 +01:00
tevador
d9bc6cfeda Updated JIT compiler and assembly generator for new int -> float conversion 2019-02-24 17:24:06 +01:00
tevador
f3b114af88 Replaced division instructions with IMUL_RCP 2019-02-22 17:48:26 +01:00
tevador
a145caa185 Fixed JIT compiler not producing the same code as genAsm and genNative 2019-02-15 16:43:52 +01:00
tevador
447e8a1d4f Simplified division in interpreted mode
Fixed incorrect condition code in JitCompilerX86
Refactoring
2019-02-15 10:41:02 +01:00
tevador
f76e8c2e20 Reworked "FNEG" instruction to make ASIC optimizations more difficult 2019-02-13 00:01:34 +01:00
tevador
376c868ca0 Fixed wrong REX prefix in FDIV_M code 2019-02-12 23:20:10 +01:00
tevador
5a89c9b28e Use allocExecutableMemory 2019-02-12 18:18:02 +01:00
tevador
b8ce504be6 Added comments to hashAes1Rx4 and fillAes1Rx4
Fixed gcc compilation
Added performance numbers
2019-02-09 19:32:53 +01:00
tevador
32d827d0a6 Interpreter with bytecode
Fixed some undefined behavior with signed types
Fixed different results on big endian systems
Removed unused code files
Restored FNEG_R instructions
Updated documentation
2019-02-09 15:45:26 +01:00
tevador
a586751f6b Removed FPNEG instruction
Optimized instruction frequencies
Increased the range for A registers from [1,65536) to [1, 4294967296)
2019-02-07 16:11:27 +01:00
tevador
ac4462ad42 Renamed floating point instructions
Fixed negative source operand for FMUL_M and FDIV_M
2019-02-05 23:43:57 +01:00
tevador
b417fd08ea 16 -> 8 chained programs
constant address loads are always from L3
2019-02-05 23:06:44 +01:00
tevador
1ee94bef2a Added ISWAP instruction
Scratchpad -> 2 MiB
New scratchpad initialization
New dataset initialization
2019-02-04 17:07:00 +01:00
tevador
20eb549725 Merged load/store of integer and FP registers 2019-01-27 19:33:55 +01:00
tevador
8f2abd6c05 NOP instruction
register load/store from L3
2019-01-27 18:19:49 +01:00
tevador
005c67f64c Added explicit STORE instructions
JIT compiler
2019-01-27 10:52:30 +01:00
tevador
d2cb086221 ASM code generator for "small" programs that fit into the uOP cache 2019-01-24 19:29:59 +01:00
tevador
bd0dba88a8 4 scratchpad segments 2019-01-20 00:44:01 +01:00
tevador
16db607025 Scratchpad size increased to 1 MiB
New AES-based scratchpad hashing function
2019-01-18 23:51:18 +01:00
tevador
4fb168e249 Large page support for cache
Bug fixes
2019-01-18 17:57:47 +01:00
tevador
8b1102ee05 Interpreter + async mode 2019-01-15 00:01:11 +01:00
tevador
a7ffe8c19a Mix dataset cacheline with registers r0-r7 2019-01-13 21:14:59 +01:00
tevador
48d85643de Dataset intialization algorithm (AES) 2019-01-13 13:47:25 +01:00
tevador
67e741ff22 Reduced x86 code size by 512 bytes (and ecx -> and eax) 2019-01-12 20:27:35 +01:00
tevador
1426fcbab5 Print average program code size
Fixed assembly for MUL_64 and IMUL_32
Division weight 4 -> 8
2019-01-12 16:05:09 +01:00
tevador
2756bcdcfe Added magic division to JIT compiler
New B operand selection rules
2019-01-11 16:53:52 +01:00
tevador
c02ee4291d FPROUND - variable flag offset 2019-01-11 10:52:12 +01:00
tevador
e487092f07 Simplified CALL and RET 2019-01-11 10:18:24 +01:00
tevador
557241cd95 JUMP instruction 2019-01-11 09:58:06 +01:00
tevador
d1a808643d Random accesses - JIT compiler 2019-01-10 22:04:55 +01:00
tevador
3caecc7646 Vector FPU instructions
JitCompilerX86 - static code written in asm
Updated ALU/FPU tests
Updated instruction weights
2018-12-31 19:06:45 +01:00
tevador
a09bee8d60 js -> jz to enable macro-op fusion on Intel CPUs (~1% speed-up) 2018-12-28 14:18:41 +01:00
tevador
76b6b05cf2 Unconditional RET 2018-12-28 12:09:37 +01:00
tevador
39c569ae44 Fixed a potential crash in JitCompilerX86 2018-12-27 21:42:38 +01:00
tevador
c05947db09 Bug fixes 2018-12-23 14:25:22 +01:00
tevador
ca59925495 JitCompilerX86: use mmap to allocate an executable buffer
compile as c++11
2018-12-23 14:09:09 +01:00
tevador
740c40b218 8 branch conditions for CALL/RET 2018-12-21 22:41:35 +01:00
tevador
1db7dd6e8b Renamed immediate constants 2018-12-20 18:36:09 +01:00
tevador
ed0bc906d6 JIT compiler for x86 2018-12-18 22:00:58 +01:00