[x265] [PATCH 3 of 4] asm: fix bug in macro vpbroadcastd for case ymm, xmm

[chen]

the macro for yasm workaround only, so we'd better keep origin instruction format.



At 2015-08-13 19:22:30,"Dnyaneshwar Gorade" <dnyaneshwar at multicorewareinc.com> wrote:

Hi Min,

This still generates wrong code for case ymm, xmm (as %ifidni %2,xm will be false always).

How about the below code ?

%macro vpbroadcastd 2-3 ;; incresed one argument for case ymm, xmm
  %ifid %3                          ; case vpbroadcastd ymm, ymm, xmm
    vpbroadcastd %1, %3
  %elifid %2
    movd         %1 %+ xmm, %2      ; case vpbroadcastd ymm, rN
    vpbroadcastd %1, %1 %+ xmm
  %else
    vpbroadcastd %1, %2             ; case vpbroadcastd ymm, [memory addr]
  %endif
%endmacro




Thanks,

Dnyaneshwar G


On Thu, Aug 13, 2015 at 8:52 AM, Min Chen <chenm003 at 163.com> wrote:
# HG changeset patch
# User Min Chen <chenm003 at 163.com>
# Date 1439424913 25200
# Node ID caf9562dc947f93e8ee237574575e9b67d494fc8
# Parent  09846d1566428a73d70d2fcf2d50324c0dfbbb7f
asm: fix bug in macro vpbroadcastd for case ymm,xmm
---
 source/common/x86/x86inc.asm |    6 ++++--
 1 files changed, 4 insertions(+), 2 deletions(-)

diff -r 09846d156642 -r caf9562dc947 source/common/x86/x86inc.asm
--- a/source/common/x86/x86inc.asm      Wed Aug 12 16:46:57 2015 -0700
+++ b/source/common/x86/x86inc.asm      Wed Aug 12 17:15:13 2015 -0700
@@ -1486,10 +1486,12 @@

 ; workaround: vpbroadcastd with register, the yasm will generate wrong code
 %macro vpbroadcastd 2
-  %ifid %2
+  %ifidni %2,xm                     ; case ymm,xmm
+    vpbroadcastd %1, %2
+  %elifid %2                        ; case ymm,rN
     movd         %1 %+ xmm, %2
     vpbroadcastd %1, %1 %+ xmm
   %else
-    vpbroadcastd %1, %2
+    vpbroadcastd %1, %2             ; case ymm,[address]
   %endif
 %endmacro

_______________________________________________
x265-devel mailing list
x265-devel at videolan.org
https://mailman.videolan.org/listinfo/x265-devel


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.videolan.org/pipermail/x265-devel/attachments/20150814/29b5acd1/attachment-0001.html>