Generating Rectify( ) Test driven development approach to TigerSHARC

Transcription

1 Generating Rectify( ) Test driven development approach to TigerSHARC assembly code production Assembly code examples Part 1 of 3 Concepts Concepts of C++ stubs Forcing the test to fail test of test Generating valid C++ code to satisfy the tests Need for name mangling for overloaded functions How do you find out the name mangled name so it can be used in assembly code Learning just enough TigerSHARC assembly code to make things work 2 / 38 Software AM radio concept Antenna Pickup Local Oscillator RF STAGE Mixer IF STAGE Low pass Filter Rectifier AUDIO STAGE Most stages handled with high speed software Low pass Filter + amplifier Audio out Standard development method Waterfall method Heavy on documentation, light on testing Describe Requirements Design Solution Build Solution Test Solution Write Analysis Document Write Design Document Write Test Plan Document TLD -- IDAA Test Last Development (if done at all) 3 / 38 4 / 31

2 Test Driven Development EmbeddedUnit Test.cpp files have four main components. Many error messages if not present Work with customer to check that the tests properly express what the customer wants done. Iterative process with customer heavily involved Agile methodology. #include <EmbeddedUnit/EmbeddedUnit.h> #include <EmbeddedUnit/CommonTests.h> #include <EmbeddedUnit/EmbeddedTests.h> Include Files cut-and-paste (always the same) TEST_CONNECT (TestFileInfo) CUSTOMER Describe Requirements Write Acceptance Tests TEST(testNAME, testtype) DEVELOPER Design Solution Write Unit Tests NOTE: Tests execute from LAST in file to FIRST. As normally the LAST test is the most recently added test, this is good. Build Solution Test Solution You test if new code works and then check (regression test) that you did not break anything 5 / 38 LINK_TEST(TestFileInfo, testtype) 6 / 38 Now expand the Customer Tests to do what the customer has requested Add test for If N <= 0, return NULL otherwise return the start of the output array Name mangled names needed in your assembly code can be seen from linker error messages Tests are working by mistake as We are not resetting the output array to 0 between function calls 7 / 38 C++ name as used The name mangled name generated by in C++ code by the C++ compiler in response to function overloading. These are the assembly code names 8 / 38

3 Next step: Write just enough code to satisfy the linker C++ stubs Write the assembly language stub 9 / 38 ERROR REPORTED BY VDSP 4.0 We lost control of the processors in the debug environment. 10 / 38 Build the code incrementally to satisfy tests Note Special marker Compiler optimization speed changes FLOATS THREE FOLD INTS SIX FOLD See speed change now we Are executing code but why failures Note: -- what if N < = 0 Why the difference in float and int? Can we do better by writing in assembly code? code Meaning, what is best possible speed? 11 / 38 Note the failures what are they 12 / 38

4 Fix Tests to only show FAILURES Generate assembly code Do the code in steps, attempting to satisfy one test at a time Learn the assembler in steps Get some idea of the issues we need to learn about as we go along Just enough knowledge to get things to work Worry about full details later 13 / / 38 What we need to know based on experiences from other processors Can we return from an assembly language routine without crashing the processor? Return a parameter from assembly language routine Pass parameters into assembly language Do IF THEN ELSE statements Read and write values to memory Read and write values in a loop Do some mathematics on the values fetched from memory All this stuff is demonstrated by coding HalfWaveRectifyASM( ) TigerSHARC assembly code file HalfWaveRectifyASM.asm Discuss in detail in a couple of slides 15 / / 38

5 Write tests about passing values back from an assembly code routine What we have learned We passed the very general test Managed to call and return from an assembly code and did not crash the system We passed some specific tests in the test file by accident 17 / 38 CJMP is the way to return from an assembly code function to C++ Instruction format is interesting nop; nop; nop;; ; separate instructions executed together CJMP (ABS);; ;; indicates the end of an grouped instruction CJMP must be like RTS meaning there is a CJMP register (or memory location) storing the address to return to after this COMPARE TO Blackfin P0 = [FP + 4]; Place storing return address UNLINK; JUMP (P0); 18 / 38 More detailed look at the code As with 68K and Blackfin needs a.section But name and format different Start function label End function label Used for profiling code As with 68K need.align statement Is the 4 in bytes (8 bits) or words (32 bits)??? As with 68K need.global to tell other code that this function exists Label format similar to 68K Needs leading underscore and final colon Single semi-colons Double semi-colons 19 / 38 Need to know How do we return an integer pointer Need to look at C++ manual for coding conventions As with 68K, MIPS and Blackfin expect to have Volatile registers function variate registers, that DON T need to be conserved when developing a function Non-volatile, preserved registers function invariate registers, that DO need to be conserved when developing a function 20 / 38

6 Return registers Using J8 for returned int * value There are many, depending on what you need to return Here we need to use J8 to return an integer pointer value Very slow for integer operations good for integer pointer ops Many registers available need ability to control usage J0 to J31 registers (integers and pointers) (SISD mode) XR0 to XR31 registers (integers) (SISD mode) XFR0 to XFR31 registers (floats) (SISD mode) Did I also mention I0 to I31 registers (integers and pointers) (SISD mode) YR0 to YR31, YFR0 to YFR31 (SIMD mode) XYR, YXR and R registers (SIMD mode) And also the MIMD modes And the double registers and the quad registers. #define return_pt_j8 J8 // J8 is a VOLATILE, NON-PRESERVED register 21 / 38 Now passing this test by accident Should be conditionally passing back NULL 22 / 38 Conditional tests Parameter passing Need to code returning a NULL or the starting address of the final array int *HalfWaveRectifyRelease(int initial_array[ ], int final_array[ ], int N) if ( N < 1) return_pt = NULL; else /* after some calculations */ return_pt = &final[ 0]; Questions to ask the instruction manual How are parameters passed to us? On the stack (as with 68K) or in registers / stack (as with MIPS and Blackfin)? answer turns out to be more like MIPS and Blackfin How do you do an IF? How do you do conditional jumps? 23 / 38 Spaces for first four parameters ARE ALWAYS present on the stack (as with 68K) But the first four parameters are passed in registers (J4, J5, J6 and J7 most of the time) (as with MIPS) The parameters passed in registers are often then stored into the spaces on the stack (like the MIPS) for safe keeping when assembly code functions call assembly code functions J4, J5, J6 and J7 are volatile, non-preserved registers 24 / 38

7 Coding convention // int *HalfWaveRectifyRelease(int initial_array[ ], // int final_array[ ], int N) #define initial_pt_inpar1 J4 incoming parameter -- pointer #define final_pt_inpar2 J5 incoming parameter -- pointer #define N_J6_inpar3 J6 incoming parameter -- integer Can we pass back the start of the final array Still passing tests by accident and the start of the array needs to be conditional return value #define return_pt_j8 J8 return value -- pointer 25 / / 38 What we need to know based on experiences from other processors Can we return from an assembly language routine without crashing the processor? Return a parameter from assembly language routine Pass parameters into assembly language Do IF THEN ELSE statements Read and write values to memory Read and write values in a loop Do some mathematics on the values fetched from memory All this stuff is demonstrated by coding HalfWaveRectifyASM( ) Doing an IF (N < 1) JUMP type of instruction 68K version CMP.L #1, D1 ; Performs subtraction (D1 1) and sets ; condition code flag BLT ELSE ; Branch if result of (D1 1) < 0 ; BLE is a branch if less than ; zero instruction NOT on whether D1 < 1 TigerSHARC version COMP(N_inpar3, 1);; // Perform N < 1 test IF JLT, JUMP ELSE;; // NOTE: Use of comma, and semi-colons ;; Same possible error on BOTH processors 68K -- which test BLE, BLT or BGT should be used? TigerSHARC which test JLE, JLT or NJLE should be used? 27 / / 38

8 ELSE is a TigerSHARC keyword Should have guessed as editor turned in blue ELSE is a KEYWORD Fix that error first Why is ELSE a keyword FOUR PART ELSE INSTRUCTION IS LEGAL IF JLT; ELSE, J1 = J2 + J3; // Conditional execution if true ELSE, XR1 = XR2 + XR3; // Conditional if true YFR1 = YFR2 + YFR3;; // Unconditional -- always IF JLT; DO, J1 = J2 + J3; // Conditional execution -- if true DO, XR1 = XR2 + XR3; // Conditional -- if true YFR1 = YFR2 + YFR3;; // Unconditional -- always Having this sort of format means that the instruction pipeline is not disrupted by jumps when we do IF statements 29 / / 38 Fix ELSE keyword error Label name is not the problem GREATER a keyword? Not blue Just change it to something else rather than wasting time worrying if it s causing the problem. CHANGE IT RATHER THAN WORRY ABOUT IT NOTE: This is C-like syntax, But it is not C Statement must end in ;; Not ; 31 / / 38

9 Should learn to read looking at wrong error. Click on error line Still not got the correct syntax Because of missing ;; (dual semicolons) Processor thinks we want return_pt = 0; JUMP END_IF; return_pt = INPAR3 ;; Apparently such a complicated instruction IS LEGAL provided the jump is at the start of the multiple issue instruction Missing ;; 33 / / 38 Add dual-semicolons everywhere Worry about multiple issues later This dual semi-colon Is so important that you MUST code review for it all the time or else you waste so much time in the Lab. Key in exams / quizzes Well I thought I understood it!!! Speed issue JUMPS can t be too close together. Not normally a problem when if code is larger At last an error I know how to fix35 / / 38

10 Add a single instruction of 4 NOPs nop; nop; nop; nop;; Fix the last error as part of Assignment Fix the remaining error 1 in handling the IF THEN ELSE as part of assignment 1 Worry about code efficiency later (refactor) when all code working What we need to know based on experiences from other processors Can we return from an assembly language routine without crashing the processor? Return a parameter from assembly language routine Pass parameters into assembly language Do IF THEN ELSE statements Read and write values to memory Read and write values in a loop Do some mathematics on the values fetched from memory All this stuff is demonstrated by coding HalfWaveRectifyASM( ) 37 / / 38 Assignment 1 code the following as a software loop follow 68K approach extern C int CalculateSum(void) { int sum = 0; for (int count = 0; count < 6; count++) { sum = sum + count; } return sum; } extern C means that this function is C compatible rather than C++. No overloading (requiring name-mangling) permitted Reminder software for-loop becomes while loop with initial test extern C int CalculateSum(void) { } int sum = 0; int count = 0; while (count < 6) { } sum = sum + count; count++; return sum; Do line by line translation 39 / / 38