diff options
| author | Chris Wilson <chris@chris-wilson.co.uk> | 2012-06-24 19:28:03 +0100 |
|---|---|---|
| committer | Chris Wilson <chris@chris-wilson.co.uk> | 2012-06-25 12:27:57 +0100 |
| commit | 15c0ee445f603033c82f357fedfc7737d198d7b3 (patch) | |
| tree | 5f01d87dc5e7493a56f554426f3f45fdea104078 /src/sna | |
| parent | fa10005ce31483827547b7f71eae066899f0026c (diff) | |
sna/gen5: Tweak thread allocations
Bump the alloted number of threads to their max. Using more threads than
cores helps hide the stalls due to sampler fetch, math functions and urb
write. Specifying too many threads seems to not incur a performance
regression, suggesting that the hardware scheduler is sane enough not to
overpopulate the EU.
A small but significant boost, peak x11perf -aa10text on an i3-330m is
raised from 1.93Mglyphs/s to 2.35Mglyphs/s.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Diffstat (limited to 'src/sna')
| -rw-r--r-- | src/sna/gen5_render.c | 8 |
1 files changed, 4 insertions, 4 deletions
diff --git a/src/sna/gen5_render.c b/src/sna/gen5_render.c index 27ba04d4..0fd51d0d 100644 --- a/src/sna/gen5_render.c +++ b/src/sna/gen5_render.c @@ -65,7 +65,7 @@ #define URB_CS_ENTRIES 0 #define URB_VS_ENTRY_SIZE 1 -#define URB_VS_ENTRIES 128 /* minimum of 8 */ +#define URB_VS_ENTRIES 256 /* minimum of 8 */ #define URB_GS_ENTRY_SIZE 0 #define URB_GS_ENTRIES 0 @@ -74,7 +74,7 @@ #define URB_CLIP_ENTRIES 0 #define URB_SF_ENTRY_SIZE 2 -#define URB_SF_ENTRIES 32 +#define URB_SF_ENTRIES 64 /* * this program computes dA/dx and dA/dy for the texture coordinates along @@ -82,10 +82,10 @@ */ #define SF_KERNEL_NUM_GRF 16 -#define SF_MAX_THREADS 2 +#define SF_MAX_THREADS 48 #define PS_KERNEL_NUM_GRF 32 -#define PS_MAX_THREADS 48 +#define PS_MAX_THREADS 72 static const uint32_t sf_kernel[][4] = { #include "exa_sf.g5b" |