[BR 2887108]: Use overflow_ helper to catch inapropriate imm optimization
|
View:
New views
3 Messages
—
Rating Filter:
Alert me
|
 |
[BR 2887108]: Use overflow_ helper to catch inapropriate imm optimization
by Cyrill Gorcunov
::
Rate this Message:
Before any further action I would like to get some comments
on this patch. Hmm? We have a number of is_sbyte16/32 calls which have the same issue I suppose. -- Cyrill --- From: Cyrill Gorcunov <gorcunov@...> Date: Sat, 31 Oct 2009 01:12:50 +0300 Subject: [PATCH] [BR 2887108]: Use overflow_ helper to catch inapropriate imm optimization Use proper method to find minimal size of storage for immediate operand (and as result -- to find respective instruction template): we should use overflow_ helpers before smaller storage unit tests since we may loose high bits. The bug example is assembling the following instruction: imul eax,eax,0x10000 immediate operand will be "optimized" up to imm8 which is wrong since it must be imm32. The bug is triggered with optimization specified at command line only (-Ox). Signed-off-by: Cyrill Gorcunov <gorcunov@...> --- parser.c | 15 ++++++++------- 1 files changed, 8 insertions(+), 7 deletions(-) diff --git a/parser.c b/parser.c index d0d660b..3fabcee 100644 --- a/parser.c +++ b/parser.c @@ -855,11 +855,10 @@ restart_parse: result->oprs[operand].segment = NO_SEG; /* don't care again */ result->oprs[operand].wrt = NO_SEG; /* still don't care */ + /* Be optimistic */ if(optimizing >= 0 && !(result->oprs[operand].type & STRICT)) - { - /* Be optimistic */ result->oprs[operand].type |= SBYTE16 | SBYTE32 | SBYTE64; - } + } else if (is_reloc(value)) { /* it's immediate */ result->oprs[operand].type |= IMMEDIATE; result->oprs[operand].offset = reloc_value(value); @@ -876,10 +875,12 @@ restart_parse: if (v64 >= -128 && v64 <= 127) result->oprs[operand].type |= SBYTE64; - if (v32 >= -128 && v32 <= 127) - result->oprs[operand].type |= SBYTE32; - if (v16 >= -128 && v16 <= 127) - result->oprs[operand].type |= SBYTE16; + if (!overflow_signed(v64, sizeof(v32))) + if (v32 >= -128 && v32 <= 127) + result->oprs[operand].type |= SBYTE32; + if (!overflow_signed(v64, sizeof(v16))) + if (v16 >= -128 && v16 <= 127) + result->oprs[operand].type |= SBYTE16; } } } else { /* it's a register */ -- 1.6.5 ------------------------------------------------------------------------------ Come build with us! The BlackBerry(R) Developer Conference in SF, CA is the only developer event you need to attend this year. Jumpstart your developing skills, take BlackBerry mobile applications to market and stay ahead of the curve. Join us from November 9 - 12, 2009. Register now! http://p.sf.net/sfu/devconference _______________________________________________ Nasm-devel mailing list Nasm-devel@... https://lists.sourceforge.net/lists/listinfo/nasm-devel |
|
|
Re: [BR 2887108]: Use overflow_ helper to catch inapropriate imm optimization
by Cyrill Gorcunov
::
Rate this Message:
[Cyrill Gorcunov - Sat, Oct 31, 2009 at 01:29:14AM +0300]
| Before any further action I would like to get some comments | on this patch. Hmm? | | We have a number of is_sbyte16/32 calls which have the same issue | I suppose. | | -- Cyrill ... Just pushed it out into main repo. -- Cyrill ------------------------------------------------------------------------------ Come build with us! The BlackBerry(R) Developer Conference in SF, CA is the only developer event you need to attend this year. Jumpstart your developing skills, take BlackBerry mobile applications to market and stay ahead of the curve. Join us from November 9 - 12, 2009. Register now! http://p.sf.net/sfu/devconference _______________________________________________ Nasm-devel mailing list Nasm-devel@... https://lists.sourceforge.net/lists/listinfo/nasm-devel |
|
|
Re: [BR 2887108]: Use overflow_ helper to catch inapropriate imm optimization
by H. Peter Anvin
::
Rate this Message:
On 10/31/2009 07:29 AM, Cyrill Gorcunov wrote:
> Before any further action I would like to get some comments > on this patch. Hmm? > > We have a number of is_sbyte16/32 calls which have the same issue > I suppose. > > The bug example is assembling the following instruction: > > imul eax,eax,0x10000 0x10000 should be SBYTE16, but not SBYTE32, so it getting promoted to an sbyte is an indication of an invalid instruction pattern. -hpa ------------------------------------------------------------------------------ Come build with us! The BlackBerry(R) Developer Conference in SF, CA is the only developer event you need to attend this year. Jumpstart your developing skills, take BlackBerry mobile applications to market and stay ahead of the curve. Join us from November 9 - 12, 2009. Register now! http://p.sf.net/sfu/devconference _______________________________________________ Nasm-devel mailing list Nasm-devel@... https://lists.sourceforge.net/lists/listinfo/nasm-devel |
| Free embeddable forum powered by Nabble | Forum Help |
