[Commit-gnuradio] r8684 - gnuradio/trunk/gnuradio-core/src/lib/filter

[eb]

Author: eb
Date: 2008-06-23 16:49:20 -0600 (Mon, 23 Jun 2008)
New Revision: 8684

Modified:
   gnuradio/trunk/gnuradio-core/src/lib/filter/complex_dotprod_sse.S
   gnuradio/trunk/gnuradio-core/src/lib/filter/complex_dotprod_sse64.S
   gnuradio/trunk/gnuradio-core/src/lib/filter/fcomplex_dotprod_sse.S
   gnuradio/trunk/gnuradio-core/src/lib/filter/fcomplex_dotprod_sse64.S
   gnuradio/trunk/gnuradio-core/src/lib/filter/float_dotprod_sse.S
   gnuradio/trunk/gnuradio-core/src/lib/filter/float_dotprod_sse64.S
Log:
in SIMD assmebler, changed .cleanup to a local symbol

Modified: gnuradio/trunk/gnuradio-core/src/lib/filter/complex_dotprod_sse.S
===================================================================
--- gnuradio/trunk/gnuradio-core/src/lib/filter/complex_dotprod_sse.S   
2008-06-23 22:25:42 UTC (rev 8683)
+++ gnuradio/trunk/gnuradio-core/src/lib/filter/complex_dotprod_sse.S   
2008-06-23 22:49:20 UTC (rev 8684)
@@ -105,7 +105,7 @@
        movaps  %xmm5, %xmm7 
 
        shrl    $2, %ecx                # n_2_complex_blocks / 4
-       je      .cleanup                # if zero, take short path
+       je      .Lcleanup               # if zero, take short path
 
        # finish setup and loop priming
 
@@ -188,7 +188,7 @@
        # At this point, xmm4 contains 2x2 partial sums.  We need
        # to compute a "horizontal complex add" across xmm4.  
        
-.cleanup:                              # xmm4 = r1 i2 r3 i4
+.Lcleanup:                             # xmm4 = r1 i2 r3 i4
        movl    20(%ebp), %eax          # @result
        movhlps %xmm4, %xmm0            # xmm0 = ?? ?? r1 r2
        addps   %xmm4, %xmm0            # xmm0 = ?? ?? r1+r3 i2+i4

Modified: gnuradio/trunk/gnuradio-core/src/lib/filter/complex_dotprod_sse64.S
===================================================================
--- gnuradio/trunk/gnuradio-core/src/lib/filter/complex_dotprod_sse64.S 
2008-06-23 22:25:42 UTC (rev 8683)
+++ gnuradio/trunk/gnuradio-core/src/lib/filter/complex_dotprod_sse64.S 
2008-06-23 22:49:20 UTC (rev 8684)
@@ -103,7 +103,7 @@
        shr     $2, %rdx                # n_2_complex_blocks / 4
        movaps  %xmm5, %xmm7 
 
-       je      .cleanup                # if zero, take short path
+       je      .Lcleanup               # if zero, take short path
 
        # finish setup and loop priming
 
@@ -186,7 +186,7 @@
        # At this point, xmm4 contains 2x2 partial sums.  We need
        # to compute a "horizontal complex add" across xmm4.  
        
-.cleanup:                              # xmm4 = r1 i2 r3 i4
+.Lcleanup:                             # xmm4 = r1 i2 r3 i4
        movhlps %xmm4, %xmm0            # xmm0 = ?? ?? r1 r2
        addps   %xmm4, %xmm0            # xmm0 = ?? ?? r1+r3 i2+i4
        movlps  %xmm0, (%rcx)           # store low 2x32 bits (complex) to 
memory

Modified: gnuradio/trunk/gnuradio-core/src/lib/filter/fcomplex_dotprod_sse.S
===================================================================
--- gnuradio/trunk/gnuradio-core/src/lib/filter/fcomplex_dotprod_sse.S  
2008-06-23 22:25:42 UTC (rev 8683)
+++ gnuradio/trunk/gnuradio-core/src/lib/filter/fcomplex_dotprod_sse.S  
2008-06-23 22:49:20 UTC (rev 8684)
@@ -102,7 +102,7 @@
        movaps  %xmm5, %xmm7 
 
        shrl    $2, %ecx                # n_2_complex_blocks / 4
-       je      .cleanup                # if zero, take short path
+       je      .Lcleanup               # if zero, take short path
 
        # finish setup and loop priming
 
@@ -171,7 +171,7 @@
        # At this point, xmm4 contains 2x2 partial sums.  We need
        # to compute a "horizontal complex add" across xmm4.  
        
-.cleanup:                              # xmm4 = r1 i2 r3 i4
+.Lcleanup:                             # xmm4 = r1 i2 r3 i4
        movl    20(%ebp), %eax          # @result
        movhlps %xmm4, %xmm0            # xmm0 = ?? ?? r1 r2
        addps   %xmm4, %xmm0            # xmm0 = ?? ?? r1+r3 i2+i4

Modified: gnuradio/trunk/gnuradio-core/src/lib/filter/fcomplex_dotprod_sse64.S
===================================================================
--- gnuradio/trunk/gnuradio-core/src/lib/filter/fcomplex_dotprod_sse64.S        
2008-06-23 22:25:42 UTC (rev 8683)
+++ gnuradio/trunk/gnuradio-core/src/lib/filter/fcomplex_dotprod_sse64.S        
2008-06-23 22:49:20 UTC (rev 8684)
@@ -99,7 +99,7 @@
        movaps  %xmm5, %xmm7 
 
        shr     $2, %rdx                # n_2_complex_blocks / 4
-       je      .cleanup                # if zero, take short path
+       je      .Lcleanup               # if zero, take short path
 
        # finish setup and loop priming
 
@@ -168,7 +168,7 @@
        # At this point, xmm4 contains 2x2 partial sums.  We need
        # to compute a "horizontal complex add" across xmm4.  
        
-.cleanup:                              # xmm4 = r1 i2 r3 i4
+.Lcleanup:                             # xmm4 = r1 i2 r3 i4
        movhlps %xmm4, %xmm0            # xmm0 = ?? ?? r1 r2
        addps   %xmm4, %xmm0            # xmm0 = ?? ?? r1+r3 i2+i4
        movlps  %xmm0, (%rcx)           # store low 2x32 bits (complex) to 
memory

Modified: gnuradio/trunk/gnuradio-core/src/lib/filter/float_dotprod_sse.S
===================================================================
--- gnuradio/trunk/gnuradio-core/src/lib/filter/float_dotprod_sse.S     
2008-06-23 22:25:42 UTC (rev 8683)
+++ gnuradio/trunk/gnuradio-core/src/lib/filter/float_dotprod_sse.S     
2008-06-23 22:49:20 UTC (rev 8684)
@@ -97,7 +97,7 @@
        movaps  %xmm5, %xmm7 
 
        shrl    $2, %ecx                # n_4_float_blocks / 4
-       je      .cleanup                # if zero, take short path
+       je      .Lcleanup               # if zero, take short path
 
        # finish setup and loop priming
 
@@ -150,7 +150,7 @@
        # to compute a "horizontal add" across xmm4.  
        # This is a fairly nasty operation...
        
-.cleanup:                              # xmm4 = d1 d2 d3 d4
+.Lcleanup:                             # xmm4 = d1 d2 d3 d4
        xorps   %xmm0, %xmm0            # xmm0 = 0 0 0 0  (may be unnecessary)
        movhlps %xmm4, %xmm0            # xmm0 = 0 0 d1 d2
        addps   %xmm4, %xmm0            # xmm0 = d1 d2 d1+d3 d2+d4

Modified: gnuradio/trunk/gnuradio-core/src/lib/filter/float_dotprod_sse64.S
===================================================================
--- gnuradio/trunk/gnuradio-core/src/lib/filter/float_dotprod_sse64.S   
2008-06-23 22:25:42 UTC (rev 8683)
+++ gnuradio/trunk/gnuradio-core/src/lib/filter/float_dotprod_sse64.S   
2008-06-23 22:49:20 UTC (rev 8684)
@@ -94,7 +94,7 @@
        movaps  %xmm5, %xmm7 
 
        shr     $2, %rdx                # n_4_float_blocks / 4
-       je      .cleanup                # if zero, take short path
+       je      .Lcleanup               # if zero, take short path
 
        # finish setup and loop priming
 
@@ -147,7 +147,7 @@
        # to compute a "horizontal add" across xmm4.  
        # This is a fairly nasty operation...
        
-.cleanup:                              # xmm4 = d1 d2 d3 d4
+.Lcleanup:                             # xmm4 = d1 d2 d3 d4
        xorps   %xmm0, %xmm0            # xmm0 = 0 0 0 0  (may be unnecessary)
        movhlps %xmm4, %xmm0            # xmm0 = 0 0 d1 d2
        addps   %xmm4, %xmm0            # xmm0 = d1 d2 d1+d3 d2+d4