binutils, gcc, glibc: new calling convention

v0-1 are removed; arguments are now returned in a0-1.  Registers are
renumbered so that s0-1 and a0-5 map to x8-15 to simplify the RVC ISA.
(These are the most popular 8 registers besides x0 and sp.)

6 files changed, 99 insertions, 130 deletions

diff --git a/newlib/libgloss/riscv/crt0.S b/newlib/libgloss/riscv/crt0.S
index da7db15..fda6ad2 100644
--- a/newlib/libgloss/riscv/crt0.S
+++ b/newlib/libgloss/riscv/crt0.S
@@ -28,9 +28,7 @@ _start:
   addi    a1, sp, _RISCV_SZPTR/8  # a1 = argv
   li      a2, 0                   # a2 = envp = NULL
   call    main
-
-  move    a0, v0                  # Call exit function with return value
-  jump    exit                    #  from main as exit code
+  jump    exit
 
   .global _init
   .global _fini
diff --git a/newlib/libgloss/riscv/machine/syscall.h b/newlib/libgloss/riscv/machine/syscall.h
index 029b1b0..c8e759f 100644
--- a/newlib/libgloss/riscv/machine/syscall.h
+++ b/newlib/libgloss/riscv/machine/syscall.h
@@ -46,17 +46,17 @@
 static inline long
 __internal_syscall(long n, long _a0, long _a1, long _a2, long _a3)
 {
-  register long v0 asm("v0") = n;
   register long a0 asm("a0") = _a0;
   register long a1 asm("a1") = _a1;
   register long a2 asm("a2") = _a2;
   register long a3 asm("a3") = _a3;
+  register long a7 asm("a7") = n;
 
   asm volatile ("scall\n"
-		"bltz v0, __syscall_error"
-		: "+r"(v0) : "r"(a0), "r"(a1), "r"(a2), "r"(a3));
+		"bltz a0, __syscall_error"
+		: "+r"(a0) : "r"(a1), "r"(a2), "r"(a3), "r"(a7));
 
-  return v0;
+  return a0;
 }
 
 #endif
diff --git a/newlib/libgloss/riscv/syscalls.c b/newlib/libgloss/riscv/syscalls.c
index 5d7b9d5..c838c18 100644
--- a/newlib/libgloss/riscv/syscalls.c
+++ b/newlib/libgloss/riscv/syscalls.c
@@ -91,10 +91,9 @@
 #define syscall_errno(n, a, b, c, d) \
         __internal_syscall(n, (long)(a), (long)(b), (long)(c), (long)(d))
 
-int __syscall_error()
+long __syscall_error(long a0)
 {
-  register int v0 asm("v0");
-  errno = -v0;
+  errno = -a0;
   return -1;
 }
 
diff --git a/newlib/newlib/libc/machine/riscv/memset.S b/newlib/newlib/libc/machine/riscv/memset.S
index 85a16ce..4205511 100644
--- a/newlib/newlib/libc/machine/riscv/memset.S
+++ b/newlib/newlib/libc/machine/riscv/memset.S
@@ -5,73 +5,64 @@
 .text
 .global memset
 memset:
-  li a6, 7
-  move v0, a0
-  bleu a2, a6, .tiny
-  and a5, a0, 7
-  bnez a5, .misaligned
+  li a6, 15
+  move a4, a0
+  bleu a2, a6, .Ltiny
+  and a5, a4, 15
+  bnez a5, .Lmisaligned
 
-.aligned:
-  bnez a1, .wordify
+.Laligned:
+  bnez a1, .Lwordify
 
-.wordified:
-  add a3, a2, -16
-  bltu a3, a2, .block16
-
-.8to15:
-#ifdef __riscv64
-  sd a1, 0(a0)
-#else
-  sw a1, 0(a0)
-  sw a1, 4(a0)
-#endif
-  add a0, a0, 8
-  add a2, a2, -8
-  bnez a2, .tiny
-  ret
+.Lwordified:
+  and a3, a2, ~15
+  and a2, a2, 15
+  add a3, a3, a4
 
-.block16:
-  add a5, a0, a2
-  add a3, a3, a0
 #ifdef __riscv64
-1:sd a1, 0(a0)
-  sd a1, 8(a0)
+1:sd a1, 0(a4)
+  sd a1, 8(a4)
 #else
-1:sw a1, 0(a0)
-  sw a1, 4(a0)
-  sw a1, 8(a0)
-  sw a1, 12(a0)
+1:sw a1, 0(a4)
+  sw a1, 4(a4)
+  sw a1, 8(a4)
+  sw a1, 12(a4)
 #endif
-  add a0, a0, 16
-  bleu a0, a3, 1b
+  add a4, a4, 16
+  bltu a4, a3, 1b
 
-  bne a0, a5, .more
+  bnez a2, .Ltiny
   ret
-.more:
-  sub a2, a5, a0
-  bgtu a2, a6, .8to15
-  /* Fall through */
 
-.tiny:
+.Ltiny:
   sub a3, a6, a2
   sll a3, a3, 2
+1:auipc t0, %pcrel_hi(.Ltable)
+  add a3, a3, t0
 .option push
 .option norvc
-  lui a4, %hi(1f)
-  add a3, a3, a4
-  jalr x0, a3, %lo(1f)
-
-1:sb a1, 6(a0)
-  sb a1, 5(a0)
-  sb a1, 4(a0)
-  sb a1, 3(a0)
-  sb a1, 2(a0)
-  sb a1, 1(a0)
-  sb a1, 0(a0)
-  ret
+.Ltable_misaligned:
+  jr a3, %pcrel_lo(1b)
+.Ltable:
+  sb a1,14(a4)
+  sb a1,13(a4)
+  sb a1,12(a4)
+  sb a1,11(a4)
+  sb a1,10(a4)
+  sb a1, 9(a4)
+  sb a1, 8(a4)
+  sb a1, 7(a4)
+  sb a1, 6(a4)
+  sb a1, 5(a4)
+  sb a1, 4(a4)
+  sb a1, 3(a4)
+  sb a1, 2(a4)
+  sb a1, 1(a4)
+  sb a1, 0(a4)
 .option pop
+  ret
 
-.wordify:
+.Lwordify:
   and a1, a1, 0xFF
   sll a3, a1, 8
   or  a1, a1, a3
@@ -81,26 +72,18 @@ memset:
   sll a3, a1, 32
   or  a1, a1, a3
 #endif
-  j .wordified
+  j .Lwordified
 
-.misaligned:
+.Lmisaligned:
   sll a3, a5, 2
-.option push
-.option norvc
-  lui a4, %hi(1f)
-  add a3, a3, a4
-1:jalr x0, a3, %lo(1b)
+1:auipc t0, %pcrel_hi(.Ltable_misaligned)
+  add a3, a3, t0
+  mv t0, ra
+  jalr a3, %pcrel_lo(1b)
+  mv ra, t0
 
-  sb a1, 6(a0)
-  sb a1, 5(a0)
-  sb a1, 4(a0)
-  sb a1, 3(a0)
-  sb a1, 2(a0)
-  sb a1, 1(a0)
-  sb a1, 0(a0)
-  add a5, a5, -8
-  sub a0, a0, a5
+  add a5, a5, -16
+  sub a4, a4, a5
   add a2, a2, a5
-  bleu a2, a6, .tiny
-  j .aligned
-.option pop
+  bleu a2, a6, .Ltiny
+  j .Laligned
diff --git a/newlib/newlib/libc/machine/riscv/setjmp.S b/newlib/newlib/libc/machine/riscv/setjmp.S
index 07d04ac..2ddc59f 100644
--- a/newlib/newlib/libc/machine/riscv/setjmp.S
+++ b/newlib/newlib/libc/machine/riscv/setjmp.S
@@ -42,16 +42,12 @@ setjmp:
 	fsd   fs9, 16*SZREG+ 9*8(a0)
 	fsd   fs10,16*SZREG+10*8(a0)
 	fsd   fs11,16*SZREG+11*8(a0)
-	fsd   fs12,16*SZREG+12*8(a0)
-	fsd   fs13,16*SZREG+13*8(a0)
-	fsd   fs14,16*SZREG+14*8(a0)
-	fsd   fs15,16*SZREG+15*8(a0)
 
 	REG_S a3, 15*SZREG(a0)
 #endif
 
-  li v0, 0
-  ret
+	li    a0, 0
+	ret
 
 /* volatile void longjmp (jmp_buf, int);  */
   .globl  longjmp
@@ -75,26 +71,21 @@ longjmp:
 #ifdef __riscv_hard_float
 	REG_L a3, 15*SZREG(a0)
 
-	fld   fs0, 16*SZREG+ 0*8(a0)
-	fld   fs1, 16*SZREG+ 1*8(a0)
-	fld   fs2, 16*SZREG+ 2*8(a0)
-	fld   fs3, 16*SZREG+ 3*8(a0)
-	fld   fs4, 16*SZREG+ 4*8(a0)
-	fld   fs5, 16*SZREG+ 5*8(a0)
-	fld   fs6, 16*SZREG+ 6*8(a0)
-	fld   fs7, 16*SZREG+ 7*8(a0)
-	fld   fs8, 16*SZREG+ 8*8(a0)
-	fld   fs9, 16*SZREG+ 9*8(a0)
-	fld   fs10,16*SZREG+10*8(a0)
-	fld   fs11,16*SZREG+11*8(a0)
-	fld   fs12,16*SZREG+12*8(a0)
-	fld   fs13,16*SZREG+13*8(a0)
-	fld   fs14,16*SZREG+14*8(a0)
-	fld   fs15,16*SZREG+15*8(a0)
+	fld  fs0, 16*SZREG+ 0*8(a0)
+	fld  fs1, 16*SZREG+ 1*8(a0)
+	fld  fs2, 16*SZREG+ 2*8(a0)
+	fld  fs3, 16*SZREG+ 3*8(a0)
+	fld  fs4, 16*SZREG+ 4*8(a0)
+	fld  fs5, 16*SZREG+ 5*8(a0)
+	fld  fs6, 16*SZREG+ 6*8(a0)
+	fld  fs7, 16*SZREG+ 7*8(a0)
+	fld  fs8, 16*SZREG+ 8*8(a0)
+	fld  fs9, 16*SZREG+ 9*8(a0)
+	fld  fs10,16*SZREG+10*8(a0)
 
 	fssr a3
 #endif
 
-  sltiu v0, a1, 1    # v0 = (a1 == 0)
-  add   v0, v0, a1   # v0 = (a1 == 0) ? 1 : a1
-  ret
+	seqz a0, a1
+	add  a0, a0, a1   # a0 = (a1 == 0) ? 1 : a1
+	ret
diff --git a/newlib/newlib/libc/machine/riscv/strcmp.S b/newlib/newlib/libc/machine/riscv/strcmp.S
index 65cbdf3..6004d9b 100644
--- a/newlib/newlib/libc/machine/riscv/strcmp.S
+++ b/newlib/newlib/libc/machine/riscv/strcmp.S
@@ -9,11 +9,10 @@
 .text
 .globl strcmp
 strcmp:
-  or    v1, a0, a1
+  or    a4, a0, a1
   li    t2, -1
-  and   v1, v1, SZREG-1
-  li    v0, 0
-  bnez  v1, .Lmisaligned
+  and   a4, a4, SZREG-1
+  bnez  a4, .Lmisaligned
 
 #if SZREG == 4
   li t3, 0x7f7f7f7f
@@ -50,16 +49,16 @@ strcmp:
         .Lnull0:
       .endif
       bne   a2, a3, .Lmisaligned
+      li    a0, 0
       ret
     .endif
   .endm
 
 .Lloop:
   # examine full words
-  check_one_word 0 4
-  check_one_word 1 4
-  check_one_word 2 4
-  check_one_word 3 4
+  check_one_word 0 3
+  check_one_word 1 3
+  check_one_word 2 3
   # backwards branch to .Lloop contained above
 
 .Lmismatch:
@@ -78,42 +77,41 @@ strcmp:
 
   srl   a0, a2, 8*SZREG-16
   srl   a1, a3, 8*SZREG-16
-  sub   v0, a0, a1
-  and   v1, v0, 0xff
-  bnez  v1, 1f
+  sub   a0, a0, a1
+  and   a4, a0, 0xff
+  bnez  a4, 1f
   ret
 
 .Lmismatch_upper:
   srl   a0, a0, 8*SZREG-16
   srl   a1, a1, 8*SZREG-16
-  sub   v0, a0, a1
-  and   v1, v0, 0xff
-  bnez  v1, 1f
+  sub   a2, a0, a1
+  and   a3, a2, 0xff
+  bnez  a3, 1f
   ret
 
 1:and   a0, a0, 0xff
   and   a1, a1, 0xff
-  sub   v0, a0, a1
+  sub   a0, a0, a1
   ret
 
 .Lmisaligned:
   # misaligned
-  lbu   v0, 0(a0)
-  lbu   v1, 0(a1)
+  lbu   a2, 0(a0)
+  lbu   a3, 0(a1)
   add   a0, a0, 1
   add   a1, a1, 1
-  bne   v0, v1, 1f
-  bnez  v0, .Lmisaligned
+  bne   a2, a3, 1f
+  bnez  a2, .Lmisaligned
 
 1:
-  sub   v0, v0, v1
+  sub   a0, a2, a3
   ret
 
   # cases in which a null byte was detected
-  foundnull 0, 4
-  foundnull 1, 4
-  foundnull 2, 4
-  foundnull 3, 4
+  foundnull 0, 3
+  foundnull 1, 3
+  foundnull 2, 3
 
 #if SZREG == 8
 .section .srodata.cst8,"aM",@progbits,8