This is the mail archive of the
gcc-help@gcc.gnu.org
mailing list for the GCC project.
Q:How to dev SimpleMPP compiler with gcc
- From: Shigeyuki Takano <syouyu at nifty dot com>
- To: gcc-help at gcc dot gnu dot org
- Date: Wed, 4 Jul 2012 18:44:12 +0900 (JST)
- Subject: Q:How to dev SimpleMPP compiler with gcc
Dear All,
Sorry to mistake to post email to gcc, about previous email.
----
I try to develop a gcc based compiler (I am beginner to gcc) for very simple m
assively parallel processing (MPP) system.
So, I consider the RTL to MD part in now.
Does anyone know a site about introduction to the part (focusing to the MPP),
excepting for gcc internal site?
Or, do you have any idea to implement following concept?
I want to develop is as follows;
----
1) Processing Elements (PEs)
PE has an ID (ID),
2-word Instruction Register (IR),
1-word Data Register (DR),
2 input-operands (OPR-1&2), note OPR-1&2 is ID or immediate (IMM),
Operation: DR <= IR(OPR-1(DR), OPR-2(DR)); operation is defined by IR,
2) Memory Elements (MEs)
ME has an ID (ID),
2-word Instruction Register (IR),
1-word Data Register (DR),
2 input-operands (OPR-1&2), note OPR-1&2 is ID or immediate (IMM),
Load Operation: DR <= Mem[IR(OPR-1(DR), OPR-2(DR))] operation is defined by I
R,
Store Operation: DR => Mem[IR(OPR-1(DR), OPR-2(DR))] operation is defined by I
R,
3) Instruction Format
IR[63:48]: ID of PE (sink or source for store)
IR[47:32]: etc
IR[31:16]: ID/IMM of OPR-1
IR[15:0]: ID/IMM of OPR-2
4) Sequence of Entire Computation
The instruction stream is fetched to global control that chains between elemen
ts having dependency represented by the ID.
Best Regards,