Fix some misattributions in profiling

Cycles spent in load_hadamard and the avg2 w16 ssse3 cacheline split code were misattributed.

diff --git a/common/x86/mc-a.asm b/common/x86/mc-a.asm
index 482c2bd654c95f5d3f9fec97fa029e12cae034c5..5faf33ca12aff67e461f4d3c0b6cbcaf6cd8e847 100644 (file)
--- a/common/x86/mc-a.asm
+++ b/common/x86/mc-a.asm
@@ -754,15 +754,6 @@ avg_w16_align%1_%2_ssse3:
     rep ret
 %endmacro
 
-%assign j 1
-%assign k 2
-%rep 15
-AVG16_CACHELINE_LOOP_SSSE3 j, j
-AVG16_CACHELINE_LOOP_SSSE3 j, k
-%assign j j+1
-%assign k k+1
-%endrep
-
 cglobal x264_pixel_avg2_w16_cache64_ssse3
     mov    eax, r2m
     and    eax, 0x3f
@@ -790,6 +781,14 @@ cglobal x264_pixel_avg2_w16_cache64_ssse3
     RET
 %endif
 
+%assign j 1
+%assign k 2
+%rep 15
+AVG16_CACHELINE_LOOP_SSSE3 j, j
+AVG16_CACHELINE_LOOP_SSSE3 j, k
+%assign j j+1
+%assign k k+1
+%endrep
 
 ;=============================================================================
 ; pixel copy

diff --git a/common/x86/pixel-a.asm b/common/x86/pixel-a.asm
index c1ec948a13dc755b90b709783c878987921b6b50..2c8284d39efe7b5dac9c00e0ffd52287270a8cff 100644 (file)
--- a/common/x86/pixel-a.asm
+++ b/common/x86/pixel-a.asm
@@ -1247,8 +1247,8 @@ cglobal x264_intra_sa8d_x3_8x8_core_%1, 3,3,16
 ; in: r0 = fenc
 ; out: m0..m3 = hadamard coefs
 INIT_MMX
-ALIGN 16
-load_hadamard:
+cglobal x264_hadamard_load
+; not really a global, but otherwise cycles get attributed to the wrong function in profiling
     pxor        m7, m7
     movd        m0, [r0+0*FENC_STRIDE]
     movd        m1, [r0+1*FENC_STRIDE]
@@ -1259,7 +1259,7 @@ load_hadamard:
     punpcklbw   m2, m7
     punpcklbw   m3, m7
     HADAMARD4_2D 0, 1, 2, 3, 4
-    SAVE_MM_PERMUTATION load_hadamard
+    SAVE_MM_PERMUTATION x264_hadamard_load
     ret
 
 %macro SCALAR_SUMSUB 4
@@ -1393,7 +1393,7 @@ cglobal x264_intra_satd_x3_4x4_%1, 2,6
     %define  t0 r2
 %endif
 
-    call load_hadamard
+    call x264_hadamard_load
     SCALAR_HADAMARD_LEFT 0, r0, r3, r4, r5
     mov         t0d, r0d
     SCALAR_HADAMARD_TOP  0, r0, r3, r4, r5
@@ -1466,7 +1466,7 @@ cglobal x264_intra_satd_x3_16x16_%1, 0,7
 .loop_y:
     xor         r4d, r4d
 .loop_x:
-    call load_hadamard
+    call x264_hadamard_load
 
     SUM3x4 %1
     SUM4x3 t2d, [left_1d+8*r3], [top_1d+8*r4]
@@ -1555,7 +1555,7 @@ cglobal x264_intra_satd_x3_8x8c_%1, 0,6
 .loop_y:
     xor         r4d, r4d
 .loop_x:
-    call load_hadamard
+    call x264_hadamard_load
 
     SUM3x4 %1
     SUM4x3 [r5+4*r4], [left_1d+8*r3], [top_1d+8*r4]