dimmus
/
efl
forked from enlightenment/efl
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

fix neon (thnx nash) to not round down incorrectly on blend 

SVN revision: 55115
This commit is contained in:
Carsten Haitzler 2010-12-01 09:46:49 +00:00
parent afa0050f32
commit 10bfc098fb
5 changed files with 52 additions and 50 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

16
legacy/evas/src/lib/engines/common/evas_op_blend/op_blend_color_neon.c
View File

@ -28,7 +28,7 @@ _op_blend_c_dp_neon(DATA32 *s __UNUSED__, DATA8 *m __UNUSED__, DATA32 c, DATA32
"vld1.32 d0[0], [%[d]] \n\t"
// Only touch d1
"vmull.u8 q0, d0, d14 \n\t"
"vshrn.u16 d0, q0, #8 \n\t"
"vqrshrn.u16 d0, q0, #8 \n\t"
"vadd.u8 d0, d12, d0 \n\t"
"vst1.32 d0[0], [%[d]] \n\t"
@ -47,7 +47,7 @@ _op_blend_c_dp_neon(DATA32 *s __UNUSED__, DATA8 *m __UNUSED__, DATA32 c, DATA32
AP "dualloopint: \n\t"
"vldr.32 d0, [%[d]] \n\t"
"vmull.u8 q1, d0, d14 \n\t"
"vshrn.u16 d0, q1, #8 \n\t"
"vqrshrn.u16 d0, q1, #8 \n\t"
"vqadd.u8 d0, d0, d12 \n\t"
"vstm %[d]!, {d0} \n\t"
@ -70,10 +70,10 @@ _op_blend_c_dp_neon(DATA32 *s __UNUSED__, DATA8 *m __UNUSED__, DATA32 c, DATA32
"vmull.u8 q4, d2, d14 \n\t"
"vmull.u8 q5, d3, d15 \n\t"
"vshrn.u16 d0, q2, #8 \n\t"
"vshrn.u16 d1, q3, #8 \n\t"
"vshrn.u16 d2, q4, #8 \n\t"
"vshrn.u16 d3, q5, #8 \n\t"
"vqrshrn.u16 d0, q2, #8 \n\t"
"vqrshrn.u16 d1, q3, #8 \n\t"
"vqrshrn.u16 d2, q4, #8 \n\t"
"vqrshrn.u16 d3, q5, #8 \n\t"
"vqadd.u8 q0, q6, q0 \n\t"
"vqadd.u8 q1, q6, q1 \n\t"
@ -95,7 +95,7 @@ _op_blend_c_dp_neon(DATA32 *s __UNUSED__, DATA8 *m __UNUSED__, DATA32 c, DATA32
AP "dualloop2int: \n\t"
"vldr.64 d0, [%[d]] \n\t"
"vmull.u8 q1, d0, d14 \n\t"
"vshrn.u16 d0, q1, #8 \n\t"
"vqrshrn.u16 d0, q1, #8 \n\t"
"vqadd.u8 d0, d0, d12 \n\t"
"vstr.64 d0, [%[d]] \n\t"
@ -111,7 +111,7 @@ _op_blend_c_dp_neon(DATA32 *s __UNUSED__, DATA8 *m __UNUSED__, DATA32 c, DATA32
AP "singleloop2: \n\t"
"vld1.32 d0[0], [%[d]] \n\t"
"vmull.u8 q1, d0, d14 \n\t"
"vshrn.u16 d0, q1, #8 \n\t"
"vqrshrn.u16 d0, q1, #8 \n\t"
"vqadd.u8 d0, d0, d12 \n\t"
"vst1.32 d0[0], [%[d]] \n\t"

24
legacy/evas/src/lib/engines/common/evas_op_blend/op_blend_mask_color_neon.c
View File

@ -40,12 +40,12 @@ _op_blend_mas_c_dp_neon(DATA32 *s __UNUSED__, DATA8 *m, DATA32 c, DATA32 *d, int
" vld1.32 d4[0], [%[d]] \n\t"
" vdup.u8 d0, d0[0] \n\t"
" vmull.u8 q4, d0, d30 \n\t"
" vshrn.u16 d12, q4, #8 \n\t"
" vqrshrn.u16 d12, q4, #8 \n\t"
" vmvn.u16 d14, d12 \n\t"
" vshr.u32 d16, d14, #24 \n\t"
" vmul.u32 d16, d16, d28 \n\t"
" vmull.u8 q7, d16, d4 \n\t"
" vshrn.u16 d0, q7, #8 \n\t"
" vqrshrn.u16 d0, q7, #8 \n\t"
" vqadd.u8 d0, d0, d12 \n\t"
" vst1.32 d0[0], [%[d]]! \n\t"
@ -64,12 +64,12 @@ _op_blend_mas_c_dp_neon(DATA32 *s __UNUSED__, DATA8 *m, DATA32 c, DATA32 *d, int
" vmovl.u8 q0, d0 \n\t"
" vmul.u32 q0, q14 \n\t"
" vmull.u8 q4, d0, d30 \n\t"
" vshrn.u16 d12, q4, #8 \n\t"
" vqrshrn.u16 d12, q4, #8 \n\t"
" vmvn.u16 d14, d12 \n\t"
" vshr.u32 d16, d14, #24 \n\t"
" vmul.u32 d16, d16, d28 \n\t"
" vmull.u8 q7, d16, d4 \n\t"
" vshrn.u16 d0, q7, #8 \n\t"
" vqrshrn.u16 d0, q7, #8 \n\t"
" vqadd.u8 q0, q0, q6 \n\t"
" vstm %[d]!, {d0} \n\t"
@ -104,8 +104,8 @@ _op_blend_mas_c_dp_neon(DATA32 *s __UNUSED__, DATA8 *m, DATA32 c, DATA32 *d, int
" vmull.u8 q5, d1, d31 \n\t"
// Shorten
" vshrn.u16 d12, q4, #8 \n\t"
" vshrn.u16 d13, q5, #8 \n\t"
" vqrshrn.u16 d12, q4, #8 \n\t"
" vqrshrn.u16 d13, q5, #8 \n\t"
// extract negated alpha
" vmvn.u16 q7, q6 \n\t"
@ -116,8 +116,8 @@ _op_blend_mas_c_dp_neon(DATA32 *s __UNUSED__, DATA8 *m, DATA32 c, DATA32 *d, int
" vmull.u8 q7, d16, d4 \n\t"
" vmull.u8 q8, d17, d5 \n\t"
" vshrn.u16 d0, q7, #8 \n\t"
" vshrn.u16 d1, q8, #8 \n\t"
" vqrshrn.u16 d0, q7, #8 \n\t"
" vqrshrn.u16 d1, q8, #8 \n\t"
// Add
" vqadd.u8 q0, q0, q6 \n\t"
@ -151,12 +151,12 @@ _op_blend_mas_c_dp_neon(DATA32 *s __UNUSED__, DATA8 *m, DATA32 c, DATA32 *d, int
" vmovl.u8 q0, d0 \n\t"
" vmul.u32 q0, q14 \n\t"
" vmull.u8 q4, d0, d30 \n\t"
" vshrn.u16 d12, q4, #8 \n\t"
" vqrshrn.u16 d12, q4, #8 \n\t"
" vmvn.u16 d14, d12 \n\t"
" vshr.u32 d16, d14, #24 \n\t"
" vmul.u32 d16, d16, d28 \n\t"
" vmull.u8 q7, d16, d4 \n\t"
" vshrn.u16 d0, q7, #8 \n\t"
" vqrshrn.u16 d0, q7, #8 \n\t"
" vqadd.u8 q0, q0, q6 \n\t"
" vstm %[d]!, {d0} \n\t"
@ -168,12 +168,12 @@ _op_blend_mas_c_dp_neon(DATA32 *s __UNUSED__, DATA8 *m, DATA32 c, DATA32 *d, int
" vld1.32 d4[0], [%[d]] \n\t"
" vdup.u8 d0, d0[0] \n\t"
" vmull.u8 q4, d0, d30 \n\t"
" vshrn.u16 d12, q4, #8 \n\t"
" vqrshrn.u16 d12, q4, #8 \n\t"
" vmvn.u16 d14, d12 \n\t"
" vshr.u32 d16, d14, #24 \n\t"
" vmul.u32 d16, d16, d28 \n\t"
" vmull.u8 q7, d16, d4 \n\t"
" vshrn.u16 d0, q7, #8 \n\t"
" vqrshrn.u16 d0, q7, #8 \n\t"
" vqadd.u8 q0, q0, q6 \n\t"
" vst1.32 d0[0], [%[d]]! \n\t"

24
legacy/evas/src/lib/engines/common/evas_op_blend/op_blend_pixel_color_neon.c
View File

@ -23,7 +23,7 @@ _op_blend_p_c_dp_neon(DATA32 *s, DATA8 *m __UNUSED__, DATA32 c, DATA32 *d, int l
// Mulitply s * c (= sc)
"vmull.u8 q4, d0,d14 \n\t"
// sc in d8
"vshrn.u16 d4, q4, #8 \n\t"
"vqrshrn.u16 d4, q4, #8 \n\t"
// sca in d9
"vmvn.u32 d6, d4 \n\t"
@ -33,7 +33,7 @@ _op_blend_p_c_dp_neon(DATA32 *s, DATA8 *m __UNUSED__, DATA32 c, DATA32 *d, int l
/* d * alpha */
"vmull.u8 q4, d6, d2 \n\t"
"vshrn.u16 d0, q4, #8 \n\t"
"vqrshrn.u16 d0, q4, #8 \n\t"
"vqadd.u8 d2, d0, d4 \n\t"
@ -57,7 +57,7 @@ _op_blend_p_c_dp_neon(DATA32 *s, DATA8 *m __UNUSED__, DATA32 c, DATA32 *d, int l
// Mulitply s * c (= sc)
"vmull.u8 q4, d0,d14 \n\t"
// sc in d8
"vshrn.u16 d4, q4, #8 \n\t"
"vqrshrn.u16 d4, q4, #8 \n\t"
// sca in d9
"vmvn.u32 d6, d4 \n\t"
@ -67,7 +67,7 @@ _op_blend_p_c_dp_neon(DATA32 *s, DATA8 *m __UNUSED__, DATA32 c, DATA32 *d, int l
/* d * alpha */
"vmull.u8 q4, d6, d2 \n\t"
"vshrn.u16 d0, q4, #8 \n\t"
"vqrshrn.u16 d0, q4, #8 \n\t"
"vqadd.u8 d2, d0, d4 \n\t"
@ -90,8 +90,8 @@ _op_blend_p_c_dp_neon(DATA32 *s, DATA8 *m __UNUSED__, DATA32 c, DATA32 *d, int l
"vmull.u8 q5, d1,d14 \n\t"
// Get sc & sc alpha
"vshrn.u16 d4, q4, #8 \n\t"
"vshrn.u16 d5, q5, #8 \n\t"
"vqrshrn.u16 d4, q4, #8 \n\t"
"vqrshrn.u16 d5, q5, #8 \n\t"
// sc is now in q2, 8bpp
// Shift out, then spread alpha for q2
"vmvn.u32 q3, q2 \n\t"
@ -102,8 +102,8 @@ _op_blend_p_c_dp_neon(DATA32 *s, DATA8 *m __UNUSED__, DATA32 c, DATA32 *d, int l
"vmull.u8 q4, d6,d2 \n\t"
"vmull.u8 q5, d7,d3 \n\t"
"vshrn.u16 d0, q4, #8 \n\t"
"vshrn.u16 d1, q5, #8 \n\t"
"vqrshrn.u16 d0, q4, #8 \n\t"
"vqrshrn.u16 d1, q5, #8 \n\t"
"vqadd.u8 q1, q0, q2 \n\t"
@ -131,7 +131,7 @@ _op_blend_p_c_dp_neon(DATA32 *s, DATA8 *m __UNUSED__, DATA32 c, DATA32 *d, int l
// Mulitply s * c (= sc)
"vmull.u8 q4, d0,d14 \n\t"
// sc in d8
"vshrn.u16 d4, q4, #8 \n\t"
"vqrshrn.u16 d4, q4, #8 \n\t"
// sca in d9
// XXX: I can probably squash one of these 3
@ -141,7 +141,7 @@ _op_blend_p_c_dp_neon(DATA32 *s, DATA8 *m __UNUSED__, DATA32 c, DATA32 *d, int l
/* d * alpha */
"vmull.u8 q4, d6, d2 \n\t"
"vshrn.u16 d0, q4, #8 \n\t"
"vqrshrn.u16 d0, q4, #8 \n\t"
"vqadd.u8 d2, d0, d4 \n\t"
@ -160,7 +160,7 @@ _op_blend_p_c_dp_neon(DATA32 *s, DATA8 *m __UNUSED__, DATA32 c, DATA32 *d, int l
// Mulitply s * c (= sc)
"vmull.u8 q4, d0,d14 \n\t"
// sc in d8
"vshrn.u16 d4, q4, #8 \n\t"
"vqrshrn.u16 d4, q4, #8 \n\t"
// sca in d6
"vmvn.u32 d6, d4 \n\t"
@ -169,7 +169,7 @@ _op_blend_p_c_dp_neon(DATA32 *s, DATA8 *m __UNUSED__, DATA32 c, DATA32 *d, int l
/* d * alpha */
"vmull.u8 q4, d6, d2 \n\t"
"vshrn.u16 d0, q4, #8 \n\t"
"vqrshrn.u16 d0, q4, #8 \n\t"
"vqadd.u8 d2, d0, d4 \n\t"

2
legacy/evas/src/lib/engines/common/evas_op_blend/op_blend_pixel_i386.c
View File

@ -24,6 +24,8 @@ _op_blend_p_dp_mmx(DATA32 *s, DATA8 *m __UNUSED__, DATA32 c __UNUSED__, DATA32 *
static void
_op_blend_pas_dp_mmx(DATA32 *s, DATA8 *m __UNUSED__, DATA32 c __UNUSED__, DATA32 *d, int l) {
_op_blend_p_dp_mmx(s, m, c, d, l);
return;
DATA32 *e = d + l;
pxor_r2r(mm0, mm0);
MOV_A2R(ALPHA_256, mm6)

36
legacy/evas/src/lib/engines/common/evas_op_blend/op_blend_pixel_neon.c
View File

@ -30,7 +30,7 @@ _op_blend_p_dp_neon(DATA32 *s, DATA8 *m, DATA32 c, DATA32 *d, int l) {
"vmul.u32 d8, d16, d8 \n\t"
"vmull.u8 q6, d4,d8 \n\t"
"vshrn.u16 d8, q6, #8 \n\t"
"vqrshrn.u16 d8, q6, #8 \n\t"
// Add to 's'
"vqadd.u8 q2, q4,q0 \n\t"
@ -61,7 +61,7 @@ _op_blend_p_dp_neon(DATA32 *s, DATA8 *m, DATA32 c, DATA32 *d, int l) {
"vmul.u32 d8, d16, d8 \n\t"
"vmull.u8 q6, d4,d8 \n\t"
"vshrn.u16 d8, q6, #8 \n\t"
"vqrshrn.u16 d8, q6, #8 \n\t"
// Add to 's'
"vqadd.u8 d4, d8,d0 \n\t"
"vstr d4, [%[d]] \n\t"
@ -91,8 +91,8 @@ _op_blend_p_dp_neon(DATA32 *s, DATA8 *m, DATA32 c, DATA32 *d, int l) {
"vmull.u8 q2, d5,d9 \n\t"
// Shift & narrow it
"vshrn.u16 d8, q6, #8 \n\t"
"vshrn.u16 d9, q2, #8 \n\t"
"vqrshrn.u16 d8, q6, #8 \n\t"
"vqrshrn.u16 d9, q2, #8 \n\t"
// Add to s
"vqadd.u8 q2, q4,q0 \n\t"
@ -134,10 +134,10 @@ _op_blend_p_dp_neon(DATA32 *s, DATA8 *m, DATA32 c, DATA32 *d, int l) {
"cmp %[tmp], %[d]\n\t"
// Shift & narrow it
"vshrn.u16 d8, q6, #8 \n\t"
"vshrn.u16 d9, q2, #8 \n\t"
"vshrn.u16 d10, q7, #8 \n\t"
"vshrn.u16 d11, q3, #8 \n\t"
"vqrshrn.u16 d8, q6, #8 \n\t"
"vqrshrn.u16 d9, q2, #8 \n\t"
"vqrshrn.u16 d10, q7, #8 \n\t"
"vqrshrn.u16 d11, q3, #8 \n\t"
// Add to s
@ -174,7 +174,7 @@ _op_blend_p_dp_neon(DATA32 *s, DATA8 *m, DATA32 c, DATA32 *d, int l) {
"vmul.u32 d8, d16, d8 \n\t"
"vmull.u8 q6, d4,d8 \n\t"
"vshrn.u16 d8, q6, #8 \n\t"
"vqrshrn.u16 d8, q6, #8 \n\t"
// Add to 's'
"vqadd.u8 d4, d8,d0 \n\t"
@ -198,7 +198,7 @@ _op_blend_p_dp_neon(DATA32 *s, DATA8 *m, DATA32 c, DATA32 *d, int l) {
"vmul.u32 d8, d8, d16 \n\t"
"vmull.u8 q6, d8,d4 \n\t"
"vshrn.u16 d8, q6, #8 \n\t"
"vqrshrn.u16 d8, q6, #8 \n\t"
// Add to 's'
"vqadd.u8 d0, d0,d8 \n\t"
"vst1.32 d0[0], [%[d]] \n\t"
@ -251,7 +251,7 @@ _op_blend_pas_dp_neon(DATA32 *s, DATA8 *m, DATA32 c, DATA32 *d, int l) {
// Multiply out
"vmull.u8 q6, d8, d4 \n\t"
"vshrn.u16 d8, q6, #8 \n\t"
"vqrshrn.u16 d8, q6, #8 \n\t"
// Add to s
"vqadd.u8 d0, d0,d8 \n\t"
@ -282,7 +282,7 @@ _op_blend_pas_dp_neon(DATA32 *s, DATA8 *m, DATA32 c, DATA32 *d, int l) {
// Multiply out
"vmull.u8 q6, d8, d4 \n\t"
"vshrn.u16 d8, q6, #8 \n\t"
"vqrshrn.u16 d8, q6, #8 \n\t"
// Add to s
"vqadd.u8 d0, d0,d8 \n\t"
@ -325,10 +325,10 @@ _op_blend_pas_dp_neon(DATA32 *s, DATA8 *m, DATA32 c, DATA32 *d, int l) {
"add %[pl], %[d], #32\n\t"
"vshrn.u16 d8, q6, #8 \n\t"
"vshrn.u16 d10, q7, #8 \n\t"
"vshrn.u16 d9, q2, #8 \n\t"
"vshrn.u16 d11, q3, #8 \n\t"
"vqrshrn.u16 d8, q6, #8 \n\t"
"vqrshrn.u16 d10, q7, #8 \n\t"
"vqrshrn.u16 d9, q2, #8 \n\t"
"vqrshrn.u16 d11, q3, #8 \n\t"
"pld [%[pl]]\n\t"
"cmp %[tmp], %[pl] \n\t"
@ -364,7 +364,7 @@ _op_blend_pas_dp_neon(DATA32 *s, DATA8 *m, DATA32 c, DATA32 *d, int l) {
// Multiply out
"vmull.u8 q6, d8, d4 \n\t"
"vshrn.u16 d8, q6, #8 \n\t"
"vqrshrn.u16 d8, q6, #8 \n\t"
// Add to s
"vqadd.u8 d0, d0,d8 \n\t"
@ -392,7 +392,7 @@ _op_blend_pas_dp_neon(DATA32 *s, DATA8 *m, DATA32 c, DATA32 *d, int l) {
// Multiply out
"vmull.u8 q6, d8, d4 \n\t"
"vshrn.u16 d8, q6, #8 \n\t"
"vqrshrn.u16 d8, q6, #8 \n\t"
// Add to s
"vqadd.u8 d0, d0,d8 \n\t"