Bug 15018

Summary: [PATCH] Use x86 BSR instruction in Log2 functions
Product: xorg Reporter: Matt Turner <mattst88>
Component: Driver/RadeonAssignee: xf86-video-ati maintainers <xorg-driver-ati>
Status: RESOLVED FIXED QA Contact: Xorg Project Team <xorg-team>
Severity: enhancement    
Priority: medium    
Version: unspecified   
Hardware: Other   
OS: All   
Whiteboard:
i915 platform: i915 features:
Attachments:
Description Flags
Enable BSR in Log2 functions none

Description Matt Turner 2008-03-14 19:19:30 UTC 
Created attachment 15133 [details] [review]
Enable BSR in Log2 functions

This patch edits RADEONLog2 and ATILog2 to use the x86 BSR instruction instead of looping through bits. It should provide a somewhat of a speed increase in this function on x86 and AMD64 architectures.

Note: the BSR instruction was added with the 80386 CPU and is therefore not compatible with earlier CPUs, though I highly doubt it's even possible to use a 286 in conjunction with a Radeon.

The inline assembly also works with Intel's compiler (icc).

Assembly output for current RADEONLog2:
RADEONLog2:
        testl   %edi, %edi
        movl    $-1, %eax
        je      .L4
        xorl    %eax, %eax
        .p2align 4,,7
.L5:
        addl    $1, %eax
        sarl    %edi
        jne     .L5
        subl    $1, %eax
.L4:
        rep ; ret

Assembly output for BSR-enabled RADEONLog2:
RADEONLog2:
        movl    %edi, %ecx
        bsrl    %ecx, %eax
        ret
Comment 1 Alex Deucher 2008-03-19 15:18:31 UTC 
committed: 17cd42ed31814ba329a6a68edd0d75390a7da40e

Thanks

Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.