TMS2837xD is a real 32 bits processor?

A GIR,

In your example, each instruction is 32-bits long.  The processor is using an immediate addressing in which the data is embedded into the instruction.   The opcode occupies the upper 16-bits, and the data the lower 16-bits.  This is pretty efficient, since if any other addressing mode had been used a pointer would have had to be initialised with the data address before the move.

If there had been an instruction such as "MOV32 ??, #447a0001" it would occupy more than 32-bits and could not have been fetched in a single cycle.

Probably you have declared "Test1" as a local variable, which is why the compiler is using the stack.  If you make it global, you'll see a different series of instructions to load an internal register and ending in a 32-bit move to the variable data address.

If you assign Test1 to another float variable, say "Test2", you should see the move accomplished in a pair of 32-bit moves, probably through the accumulator.

Chapter 5 of the user's guide Toby recommended should help.

Regards,

Richard

Ok,

So you said to have a 32 bit instruction is faster to have an instruction coded on many 32bit right?

Because in comparaison, on old 8 bit controller, for exemple we had instruction
MOV x, #0x1 : Copy the immediate value0x01 to the adresse pointed by register x. This instruction take 2 bytes to copy a 8 bits.

Now, in 32 µP bits, i'm exepcted to have:
"Mov, x, #0xnnnnnnnn" which can be coded with more than 32bits, but which copy in one instruction the immediate value 0xnnnnnnnn to the adress pointed by x register (stack in ou case) because the bus has 32 bits..

But it is ok,

Thank for your explanation.

Could you know where I can find the instruction set?