store quant_shift as an unsigned char

in encodframe.c, quant_shift is set to 0 or 1 in vp8cx_invert_quant

only use 8 bits to store this, instead of 16. will allow saving an
xmm register in an updated version of the regular quantize

Change-Id: Ie88c47fe2aff5af0283dab1147fb2791e4b12f90

diff --git a/vp8/encoder/block.h b/vp8/encoder/block.h
index 5a2568d..6de4c85 100644 (file)
--- a/vp8/encoder/block.h
+++ b/vp8/encoder/block.h
@@ -34,7 +34,7 @@ typedef struct
     // 16 Y blocks, 4 U blocks, 4 V blocks each with 16 entries
     short *quant;
     short *quant_fast;
-    short *quant_shift;
+    unsigned char *quant_shift;
     short *zbin;
     short *zrun_zbin_boost;
     short *round;

diff --git a/vp8/encoder/encodeframe.c b/vp8/encoder/encodeframe.c
index 1dbf46e..66718b5 100644 (file)
--- a/vp8/encoder/encodeframe.c
+++ b/vp8/encoder/encodeframe.c
@@ -147,7 +147,7 @@ static const int qzbin_factors_y2[129] =
 #define EXACT_QUANT
 #ifdef EXACT_QUANT
 static void vp8cx_invert_quant(int improved_quant, short *quant,
-                               short *shift, short d)
+                               unsigned char *shift, short d)
 {
     if(improved_quant)
     {

diff --git a/vp8/encoder/onyx_int.h b/vp8/encoder/onyx_int.h
index 0e53f68..c11c097 100644 (file)
--- a/vp8/encoder/onyx_int.h
+++ b/vp8/encoder/onyx_int.h
@@ -251,17 +251,17 @@ typedef struct
 {
 
     DECLARE_ALIGNED(16, short, Y1quant[QINDEX_RANGE][16]);
-    DECLARE_ALIGNED(16, short, Y1quant_shift[QINDEX_RANGE][16]);
+    DECLARE_ALIGNED(16, unsigned char, Y1quant_shift[QINDEX_RANGE][16]);
     DECLARE_ALIGNED(16, short, Y1zbin[QINDEX_RANGE][16]);
     DECLARE_ALIGNED(16, short, Y1round[QINDEX_RANGE][16]);
 
     DECLARE_ALIGNED(16, short, Y2quant[QINDEX_RANGE][16]);
-    DECLARE_ALIGNED(16, short, Y2quant_shift[QINDEX_RANGE][16]);
+    DECLARE_ALIGNED(16, unsigned char, Y2quant_shift[QINDEX_RANGE][16]);
     DECLARE_ALIGNED(16, short, Y2zbin[QINDEX_RANGE][16]);
     DECLARE_ALIGNED(16, short, Y2round[QINDEX_RANGE][16]);
 
     DECLARE_ALIGNED(16, short, UVquant[QINDEX_RANGE][16]);
-    DECLARE_ALIGNED(16, short, UVquant_shift[QINDEX_RANGE][16]);
+    DECLARE_ALIGNED(16, unsigned char, UVquant_shift[QINDEX_RANGE][16]);
     DECLARE_ALIGNED(16, short, UVzbin[QINDEX_RANGE][16]);
     DECLARE_ALIGNED(16, short, UVround[QINDEX_RANGE][16]);
 

diff --git a/vp8/encoder/quantize.c b/vp8/encoder/quantize.c
index 803e3a5..86ed267 100644 (file)
--- a/vp8/encoder/quantize.c
+++ b/vp8/encoder/quantize.c
@@ -27,7 +27,7 @@ void vp8_fast_quantize_b_c(BLOCK *b, BLOCKD *d)
     short *zbin_ptr        = b->zbin;
     short *round_ptr       = b->round;
     short *quant_ptr       = b->quant_fast;
-    short *quant_shift_ptr = b->quant_shift;
+    unsigned char *quant_shift_ptr = b->quant_shift;
     short *qcoeff_ptr      = d->qcoeff;
     short *dqcoeff_ptr     = d->dqcoeff;
     short *dequant_ptr     = d->dequant;
@@ -112,7 +112,7 @@ void vp8_regular_quantize_b(BLOCK *b, BLOCKD *d)
     short *zbin_ptr        = b->zbin;
     short *round_ptr       = b->round;
     short *quant_ptr       = b->quant;
-    short *quant_shift_ptr = b->quant_shift;
+    unsigned char *quant_shift_ptr = b->quant_shift;
     short *qcoeff_ptr      = d->qcoeff;
     short *dqcoeff_ptr     = d->dqcoeff;
     short *dequant_ptr     = d->dequant;
@@ -166,7 +166,7 @@ void vp8_strict_quantize_b(BLOCK *b, BLOCKD *d)
     int sz;
     short *coeff_ptr;
     short *quant_ptr;
-    short *quant_shift_ptr;
+    unsigned char *quant_shift_ptr;
     short *qcoeff_ptr;
     short *dqcoeff_ptr;
     short *dequant_ptr;

diff --git a/vp8/encoder/x86/quantize_sse2.asm b/vp8/encoder/x86/quantize_sse2.asm
index 5e40dc7..7b7ae70 100644 (file)
--- a/vp8/encoder/x86/quantize_sse2.asm
+++ b/vp8/encoder/x86/quantize_sse2.asm
@@ -142,7 +142,7 @@ sym(vp8_regular_quantize_b_sse2):
     movsx       edi, WORD PTR[rsp + temp_qcoeff + %1 * 2]
 
     ; downshift by quant_shift[rc]
-    movsx       ecx, WORD PTR[rax + %1 * 2] ; quant_shift_ptr[rc]
+    movsx       cx, BYTE PTR[rax + %1]      ; quant_shift_ptr[rc]
     sar         edi, cl                     ; also sets Z bit
     je          rq_zigzag_loop_%1           ; !y
     mov         WORD PTR[rsp + qcoeff + %1 * 2], di ;qcoeff_ptr[rc] = temp_qcoeff[rc]