diff options
| author |   millert <millert@openbsd.org> | 2003-11-01 22:19:33 +0000 |
|---|---|---|
| committer | millert <millert@openbsd.org> | 2003-11-01 22:19:33 +0000 |
| commit | 0596ab59e2dcf28f889c4af9857b9bb397b7db25 (patch) | |
| tree | f766cb6a0b68adf82b5bc05cbf5d5e4c948f5ebf | |
| parent | process signals at the right time. also handle stdin failure better; (diff) | |
| download | wireguard-openbsd-0596ab59e2dcf28f889c4af9857b9bb397b7db25.tar.xz wireguard-openbsd-0596ab59e2dcf28f889c4af9857b9bb397b7db25.zip | |
Historically, scalb() in 4BSD had the actual semantics of scalbn(). We changed
scalb() to scalbn() some time ago in the tree to fix this. However, it turns
out the vax scalb() had already been fixed to support standard scalb()
semantics. This commit undoes that change, making the vax scalbn() the same
as the 4BSD scalb() (but with the correct name) and fixes the comment.
We still could use wrappers for scalb(), scalbf() and scalbnf().
Fixes exp() and a number of perl tests; OK hugh@
| -rw-r--r-- | lib/libm/arch/vax/n_support.S | 9 |
1 files changed, 4 insertions, 5 deletions
diff --git a/lib/libm/arch/vax/n_support.S b/lib/libm/arch/vax/n_support.S index f0f7644da20..2bf94cfe465 100644 --- a/lib/libm/arch/vax/n_support.S +++ b/lib/libm/arch/vax/n_support.S @@ -100,19 +100,18 @@ _finite: /* * double scalbn(x,N) - * double x; double N; + * double x; int N; */ .globl _scalbn .set ERANGE,34 .text .align 1 _scalbn: - .word 0x3c + .word 0xc movq 4(ap),r0 bicl3 $0xffff807f,r0,r3 beql ret1 # 0 or reserved operand - movq 12(ap),r4 - cvtdl r4, r2 + movl 12(ap),r2 cmpl r2,$0x12c bgeq ovfl cmpl r2,$-0x12c @@ -215,7 +214,7 @@ E2: movw $0,r8 #clear r8 movq $0,r0 #x underflow to zero C5: - bisw2 r8,r0 /* put r8 into x's exponent field */ + bisw2 r8,r0 #put r8 into x's exponent field ret Rop: #Reserved operand pushl $EDOM |