runtime: fix uint64 division on 386

The uint64 divide function calls _mul64x32 to do a 64x32-bit multiply
and then compares the result against the 64-bit numerator.
If the result is bigger than the numerator, must use the slow path.

Unfortunately, the 64x32 produces a 96-bit product, and only the
low 64 bits were being used in the comparison. Return all 96 bits,
the bottom 64 via the original uint64* pointer, and the top 32
as the function's return value.

Fixes 386 build (broken by ARM division tests).

R=golang-dev, iant
CC=golang-dev
https://golang.org/cl/13722044

diff --git a/src/pkg/runtime/vlop_386.s b/src/pkg/runtime/vlop_386.s
index edc659b8487f89ca269b7c37656cbb452d652321..9783fdc936aeba7180e7d925be57bd804946b082 100644 (file)
--- a/src/pkg/runtime/vlop_386.s
+++ b/src/pkg/runtime/vlop_386.s
@@ -29,6 +29,8 @@
  * C runtime for 64-bit divide.
  */
 
+// _mul64x32(r *uint64, a uint64, b uint32)
+// sets *r = low 64 bits of 96-bit product a*b; returns high 32 bits.
 TEXT _mul64by32(SB), NOSPLIT, $0
        MOVL    r+0(FP), CX
        MOVL    a+4(FP), AX
@@ -38,7 +40,9 @@ TEXT _mul64by32(SB), NOSPLIT, $0
        MOVL    a+8(FP), AX
        MULL    b+12(FP)
        ADDL    AX, BX
+       ADCL    $0, DX
        MOVL    BX, 4(CX)
+       MOVL    DX, AX
        RET
 
 TEXT _div64by32(SB), NOSPLIT, $0

diff --git a/src/pkg/runtime/vlrt_386.c b/src/pkg/runtime/vlrt_386.c
index d8bc94bd94a5d813db8fc22ac036e493125a6695..8d965c086e4df1b5029fe3011d515a646bfacc55 100644 (file)
--- a/src/pkg/runtime/vlrt_386.c
+++ b/src/pkg/runtime/vlrt_386.c
@@ -147,7 +147,7 @@ _v2f(Vlong x)
 }
 
 ulong  _div64by32(Vlong, ulong, ulong*);
-void   _mul64by32(Vlong*, Vlong, ulong);
+int    _mul64by32(Vlong*, Vlong, ulong);
 
 static void
 slowdodiv(Vlong num, Vlong den, Vlong *q, Vlong *r)
@@ -232,8 +232,7 @@ dodiv(Vlong num, Vlong den, Vlong *qp, Vlong *rp)
        if(den.hi != 0){
                q.hi = 0;
                n = num.hi/den.hi;
-               _mul64by32(&x, den, n);
-               if(x.hi > num.hi || (x.hi == num.hi && x.lo > num.lo))
+               if(_mul64by32(&x, den, n) || x.hi > num.hi || (x.hi == num.hi && x.lo > num.lo))
                        slowdodiv(num, den, &q, &r);
                else {
                        q.lo = n;