Aren
/
haiku
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

the DMA acc sub-cmd-lists are now issued to a single engine command (except for blit because it's hardware can't do that), instead of to multiple issued engine commands. (one could say 'parallel processing' but that's not the case: it remains serial: it's just faster). 

DMA acceleration speed is up another 10% for rect and span commands.
commencing more testing now: code is complete (I hope).


git-svn-id: file:///srv/svn/repos/haiku/trunk/current@11433 a95241bf-73f2-0310-859d-f6bbb57e9c96
This commit is contained in:
Rudolf Cornelissen 2005-02-21 16:20:40 +00:00
parent dffca5825b
commit 7540371dbb
2 changed files with 18 additions and 10 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

26
src/add-ons/accelerants/nvidia/engine/nv_acc_dma.c
View File

@ -1058,6 +1058,11 @@ void nv_acc_assert_fifo_dma(void)
note also:
splitting up each command list into sublists (see routines below) prevents
a lot more nested calls, further increasing the speed with upto 70%.
finally:
sending the sublist to just one single engine command even further increases
speed with upto another 10%. This can't be done for blits though, as this engine-
command's hardware does not support multiple objects.
*/
/* screen to screen blit - i.e. move windows around and scroll within them. */
@ -1135,12 +1140,13 @@ void FILL_RECTANGLE_DMA(engine_token *et, uint32 colorIndex, fill_rect_params *l
count -= subcnt;
/* wait for room in fifo for bitmap cmd if needed. */
if (nv_acc_fifofree_dma(3 * subcnt) != B_OK) return;
if (nv_acc_fifofree_dma(1 + (2 * subcnt)) != B_OK) return;
/* issue fill command once... */
nv_acc_cmd_dma(NV4_GDI_RECTANGLE_TEXT, NV4_GDI_RECTANGLE_TEXT_UCR0_LEFTTOP, (2 * subcnt));
/* ... and send multiple rects (engine cmd supports 32 max) */
while (subcnt--)
{
/* now setup fill (writing 3 32bit words) */
nv_acc_cmd_dma(NV4_GDI_RECTANGLE_TEXT, NV4_GDI_RECTANGLE_TEXT_UCR0_LEFTTOP, 2);
si->engine.dma.cmdbuffer[si->engine.dma.current++] =
(((list[i].left) << 16) | ((list[i].top) & 0x0000ffff)); /* Unclipped Rect 0 LeftTop */
si->engine.dma.cmdbuffer[si->engine.dma.current++] =
@ -1183,12 +1189,13 @@ void FILL_SPAN_DMA(engine_token *et, uint32 colorIndex, uint16 *list, uint32 cou
count -= subcnt;
/* wait for room in fifo for bitmap cmd if needed. */
if (nv_acc_fifofree_dma(3 * subcnt) != B_OK) return;
if (nv_acc_fifofree_dma(1 + (2 * subcnt)) != B_OK) return;
/* issue fill command once... */
nv_acc_cmd_dma(NV4_GDI_RECTANGLE_TEXT, NV4_GDI_RECTANGLE_TEXT_UCR0_LEFTTOP, (2 * subcnt));
/* ... and send multiple rects (spans) (engine cmd supports 32 max) */
while (subcnt--)
{
/* now setup fill (writing 3 32bit words) */
nv_acc_cmd_dma(NV4_GDI_RECTANGLE_TEXT, NV4_GDI_RECTANGLE_TEXT_UCR0_LEFTTOP, 2);
si->engine.dma.cmdbuffer[si->engine.dma.current++] =
(((list[i+1]) << 16) | ((list[i]) & 0x0000ffff)); /* Unclipped Rect 0 LeftTop */
si->engine.dma.cmdbuffer[si->engine.dma.current++] =
@ -1230,12 +1237,13 @@ void INVERT_RECTANGLE_DMA(engine_token *et, fill_rect_params *list, uint32 count
count -= subcnt;
/* wait for room in fifo for bitmap cmd if needed. */
if (nv_acc_fifofree_dma(3 * subcnt) != B_OK) return;
if (nv_acc_fifofree_dma(1 + (2 * subcnt)) != B_OK) return;
/* issue fill command once... */
nv_acc_cmd_dma(NV4_GDI_RECTANGLE_TEXT, NV4_GDI_RECTANGLE_TEXT_UCR0_LEFTTOP, (2 * subcnt));
/* ... and send multiple rects (engine cmd supports 32 max) */
while (subcnt--)
{
/* now setup fill (writing 3 32bit words) */
nv_acc_cmd_dma(NV4_GDI_RECTANGLE_TEXT, NV4_GDI_RECTANGLE_TEXT_UCR0_LEFTTOP, 2);
si->engine.dma.cmdbuffer[si->engine.dma.current++] =
(((list[i].left) << 16) | ((list[i].top) & 0x0000ffff)); /* Unclipped Rect 0 LeftTop */
si->engine.dma.cmdbuffer[si->engine.dma.current++] =

2
src/add-ons/accelerants/nvidia/engine/nv_general.c
View File

@ -90,7 +90,7 @@ status_t nv_general_powerup()
{
status_t status;
LOG(1,("POWERUP: Haiku nVidia Accelerant 0.40 running.\n"));
LOG(1,("POWERUP: Haiku nVidia Accelerant 0.41 running.\n"));
/* preset no laptop */
si->ps.laptop = false;