1. © 2009, Renesas Technology America, Inc., All Rights Reserved 1 Course Introduction  Purpose:  This course provides an overview of the SH-2 32-bit RISC CPU core used in the popular SH-2 series microcontrollers  Objectives:  Learn about error detection and address errors  Explore the pipeline in the SH-2 CPU  Understand branching  Obtain some helpful coding tips  Content:  16 pages  3 questions  Learning Time:  15 minutes

2. © 2009, Renesas Technology America, Inc., All Rights Reserved 2 CPU Error Detection  SH-2 CPU can detect and react to various error conditions:  Address errors  Illegal instructions (invalid op codes) –Undefined instruction executed  Slot illegal instructions –Certain instructions, such as one that changes the PC, cannot be located in a delay slot (i.e., a slot after a delayed branch)

3. © 2009, Renesas Technology America, Inc., All Rights Reserved Address Errors  Instruction fetch from odd address  Instruction fetch from on-chip peripheral module space  Instruction fetch from external memory space in single-chip mode  Word data accessed from odd address  Longword data accessed from other than a longword boundary  Longword accessed in 8-bit on-chip peripheral module space  External memory space accessed when in single-chip mode  Stack access to address that is not multiple of four (longword alignment required)  Access using vector base register when VBR is not a multiple of four (longword alignment required)

5. © 2009, Renesas Technology America, Inc., All Rights Reserved IFIDALUMAWBIFIDALUMAWBIF IDALUMAWB Instruction Why Use a Pipeline?  CISC CPUs perform all operations sequentially, then execute instructions in series, one at a time; instruction execution can take several cycles  RISC CPUs use a pipeline to try to overlap “independent parts" of instructions, so that one instruction may be able to execute every cycle Instruction CISC Ideal pipeline speedup: T pipeline = T unpipelined / # of Pipeline stages RISC Time saved Instruction Fetch Inst. Dec./ Reg. Acc. ALU Mem Access/ Data Cycle Write Back Register Content of a typical instruction CISC: Variable length SH-2: 16 bits long

6. © 2009, Renesas Technology America, Inc., All Rights Reserved Ideal SH-2 Pipeline Flow  IF (Instruction Fetch): Fetches an instruction from the memory in which the program is stored  ID (Instruction Decode): Decodes instruction fetched  EX (Instruction Execution): Performs data operations and address calculations according to the results of decoding  MA(Memory Access): Accesses data in memory; generated by instructions that involve memory access (with some exceptions)  WB(Write Back): Returns results of memory access (data to a register) - Generated by instructions that involve memory loads, with some exceptions Instruction 1IFIDEXMAWB Instruction 2IFIDEXMAWB Instruction 3IFIDEXMAWB Instruction 4IFIDEXMAWB Instruction 5IFIDEXMAWB Instruction 6IFIDEXMAWB Time Slots Instruction stream Ideal execution – One instruction every slot (1 cycle per slot)

7. © 2009, Renesas Technology America, Inc., All Rights Reserved Hazards Prevent Ideal Flow Pipeline can be “stalled” by:  Structural hazards  Arise from resource conflicts that occur when the hardware cannot support all possible combinations of instructions in simultaneous, overlapped execution  Data hazards  Arise when an instruction depends on the result of a previous instruction in a way that is exposed by the overlapping of instructions in the pipeline  Control hazards  Arise from the pipelining of branches and other instructions that change the PC

9. © 2009, Renesas Technology America, Inc., All Rights Reserved 9 Slots Requiring Multiple Cycles MA stage takes three cycles IF stage takes two cycles Execution time of Instruction 1: 4 cycles

10. © 2009, Renesas Technology America, Inc., All Rights Reserved Pipeline and Flow Control  Due to the pipeline, instructions are fetched before previous instructions are totally completed  Program timing is affected when fetched instructions are not used; this situation can occur for:  Conditional branching  Unconditional branching Understand the difference between  “Virtual time” for instruction execution (ideal case)  Total time for an instruction to execute (actual case) SuperH ® compilers organize code to minimize pipeline stalls

11. © 2009, Renesas Technology America, Inc., All Rights Reserved Conditional Branch Taken When branch is taken...  Conditional branch instructions:  BF  BT Destination address is known at this point

12. © 2009, Renesas Technology America, Inc., All Rights Reserved Conditional Branch Not Taken When branch is NOT taken...  Conditional branch instructions:  BF  BT IFIDEX IFIDEXMAWB MAWB IFIDEXMAWB IFIDEXMAWB BR Inst Next Inst Execution continues normally

13. © 2009, Renesas Technology America, Inc., All Rights Reserved Conditional Delayed Branch When branch is taken...  Instructions:  BF/S  BT/S Delayed Branch / Instruction Re-order Traditional CPU –ADD.W R1,R0 –BGT target_address SuperH CPU –BF/S target_address –ADD.W R1,R0 Delay Delay Introduced

15. © 2009, Renesas Technology America, Inc., All Rights Reserved Programming Tips  Use local variables wherever possible to improve execution (locals use less ROM and RAM)  Use modular programming to reduce far branches  Be careful with constants; use 8-bit wherever possible  Avoid operations on the MAC that will stall the pipeline  Place functions that call each other close together  Try to align instructions on 32-bit boundaries, especially load/store instructions  Convert byte and word values to signed long integers, the most efficient data type for the SH-2 architecture  Make sure that instructions that immediately follow an instruction that loads from memory, do not use the same destination register as the load instruction

16. © 2009, Renesas Technology America, Inc., All Rights Reserved 16 Course Summary  Error detection  Address errors  Pipeline  Branching  Recommendations for coding