prestaacked annotation can create custom calling conv

riscv-total-embedded.adoc

@@ -611,16 +611,17 @@ Simplified ranges (e.g. shadow register file):
 
 `\\__attribute__\((prestacked("x8-x15")))`
 
-TEIC irq, range0 + shadow regs of half regfile (where bit 2 is set, covers range1+2):
+TEIC irq, range0 + shadow regs of half integer regfile (where bit 2 of operand is set, covers range1+2)
+and F + P extensions:
 
-`\\__attribute__\((prestacked("x4-x7,x10,x11,x12-x15,x20-x23,x28-x31")))`
+`\\__attribute__\((prestacked("x4-x7,x10,x11,x12-x15,x20-x23,x28-x31,fcsr,vxsat")))`
 
 ch32v003 irq (ilp32e + PFIC HW stacker, assuming `ra` doesn't have some undocumented use)
 
 `\\__attribute__\((interrupt, prestacked("x1,x5-x7,x10-x15")))`
 
 NOTE: unannotated `ra` is assumed as a valid return address, otherwise a special return mechanism must be
-used
+used (e.g. return by `mret` in `\\__attribute__\((interrupt))`
 
 ===== optimization for `noreturn` functions
 
@@ -645,10 +646,15 @@ and consequently underflow the stack.
 
 ===== functions with partially custom calling conventions
 
-can be
-for IPRA
-
+It can be additionally abused to:
 
+- define IPRA clobbers of assembly functions in its C function declarations 
+(see <<applying IPRA to assembly functions>>)
+- certain (premature) optimizations (manually solving 2way IPRA recursion etc.)
+- dynamic linked functions with a subset of clobbers. 
+e.g. functions like `memcpy()`,`strcmp()` etc.  don't need to clobber all caller saved registers 
+so only common clobbers for straightforward, unrolled (?) and vectorized implementations need to be 
+applied. Requires standardization of cannoical clobbers. (quite unrealistic)
 
 ==== IPRA - Inter procedural register allocation
 
@@ -678,7 +684,7 @@ possible to just move `ra` and `s0/s1` into caller saved regs.
 
 NOTE: This is also non IPRA optimization (-Oz kind)
 
-- need special assembly directive to annotate such exports from pure assembly code (workaround exist
+- need special assembly directive to annotate such exports from pure assembly code (workaround exists
 <<applying IPRA to assembly functions>>)
 
 [NOTE]