Message ID | 1438078795-14360-2-git-send-email-thuth@redhat.com (mailing list archive) |
---|---|
State | Not Applicable |
Headers | show |
On Tue, Jul 28, 2015 at 12:19:54PM +0200, Thomas Huth wrote: > : invert-region ( addr len -- ) > - 0 ?DO dup dup rb@ -1 xor swap rb! 1+ LOOP drop > -; > - > -: invert-region-x ( addr len -- ) > - /x / 0 ?DO dup dup rx@ -1 xor swap rx! xa1+ LOOP drop > + 2dup or 7 and CASE > + 0 OF 3 rshift 0 ?DO dup dup rx@ -1 xor swap rx! xa1+ LOOP ENDOF > + 2 OF 1 rshift 0 ?DO dup dup rw@ -1 xor swap rw! wa1+ LOOP ENDOF > + 4 OF 2 rshift 0 ?DO dup dup rl@ -1 xor swap rl! la1+ LOOP ENDOF > + 6 OF 1 rshift 0 ?DO dup dup rw@ -1 xor swap rw! wa1+ LOOP ENDOF > + dup OF 0 ?DO dup dup rb@ -1 xor swap rb! 1+ LOOP ENDOF > + ENDCASE > + drop > ; Can you access device memory as 64 bits for all supported devices? You can get a bigger speedup by writing some of the core blitting functions in C, btw. A small simplification: 2dup or 7 and CASE 0 OF 3 rshift 0 ?DO dup dup rx@ -1 xor swap rx! xa1+ LOOP ENDOF 4 OF 2 rshift 0 ?DO dup dup rl@ -1 xor swap rl! la1+ LOOP ENDOF 3 and 2 OF 1 rshift 0 ?DO dup dup rw@ -1 xor swap rw! wa1+ LOOP ENDOF dup OF 0 ?DO dup dup rb@ -1 xor swap rb! 1+ LOOP ENDOF ENDCASE If this code is often called unaligned, it makes more sense to special- case the begin and end probably. Segher
Hi Segher, On 28/07/15 19:04, Segher Boessenkool wrote: > On Tue, Jul 28, 2015 at 12:19:54PM +0200, Thomas Huth wrote: >> : invert-region ( addr len -- ) >> - 0 ?DO dup dup rb@ -1 xor swap rb! 1+ LOOP drop >> -; >> - >> -: invert-region-x ( addr len -- ) >> - /x / 0 ?DO dup dup rx@ -1 xor swap rx! xa1+ LOOP drop >> + 2dup or 7 and CASE >> + 0 OF 3 rshift 0 ?DO dup dup rx@ -1 xor swap rx! xa1+ LOOP ENDOF >> + 2 OF 1 rshift 0 ?DO dup dup rw@ -1 xor swap rw! wa1+ LOOP ENDOF >> + 4 OF 2 rshift 0 ?DO dup dup rl@ -1 xor swap rl! la1+ LOOP ENDOF >> + 6 OF 1 rshift 0 ?DO dup dup rw@ -1 xor swap rw! wa1+ LOOP ENDOF >> + dup OF 0 ?DO dup dup rb@ -1 xor swap rb! 1+ LOOP ENDOF >> + ENDCASE >> + drop >> ; > > Can you access device memory as 64 bits for all supported devices? Yes, should be fine since 64 bit access was already used in the original code, see fb8-invert-screen in https://github.com/aik/SLOF/commit/99c534ecc7a8566bd9ca6346915d9ac1bfacae1e > You can get a bigger speedup by writing some of the core blitting > functions in C, btw. Well, the above code is for js2x only ... so this is likely not worth the effort anymore. The code for qemu-spapr calls into a hypercall already, so this is already accelerated. > A small simplification: > > 2dup or 7 and CASE > 0 OF 3 rshift 0 ?DO dup dup rx@ -1 xor swap rx! xa1+ LOOP ENDOF > 4 OF 2 rshift 0 ?DO dup dup rl@ -1 xor swap rl! la1+ LOOP ENDOF > 3 and > 2 OF 1 rshift 0 ?DO dup dup rw@ -1 xor swap rw! wa1+ LOOP ENDOF > dup OF 0 ?DO dup dup rb@ -1 xor swap rb! 1+ LOOP ENDOF > ENDCASE Ok, nice idea, makes sense! I'll include it in v2 (after waiting a little bit to see if there's other feedback) > If this code is often called unaligned, it makes more sense to special- > case the begin and end probably. It's only used for drawing the cursor, so it always should be aligned. Thomas
diff --git a/board-js2x/slof/helper.fs b/board-js2x/slof/helper.fs index 6030330..5941315 100644 --- a/board-js2x/slof/helper.fs +++ b/board-js2x/slof/helper.fs @@ -28,9 +28,12 @@ ; : invert-region ( addr len -- ) - 0 ?DO dup dup rb@ -1 xor swap rb! 1+ LOOP drop -; - -: invert-region-x ( addr len -- ) - /x / 0 ?DO dup dup rx@ -1 xor swap rx! xa1+ LOOP drop + 2dup or 7 and CASE + 0 OF 3 rshift 0 ?DO dup dup rx@ -1 xor swap rx! xa1+ LOOP ENDOF + 2 OF 1 rshift 0 ?DO dup dup rw@ -1 xor swap rw! wa1+ LOOP ENDOF + 4 OF 2 rshift 0 ?DO dup dup rl@ -1 xor swap rl! la1+ LOOP ENDOF + 6 OF 1 rshift 0 ?DO dup dup rw@ -1 xor swap rw! wa1+ LOOP ENDOF + dup OF 0 ?DO dup dup rb@ -1 xor swap rb! 1+ LOOP ENDOF + ENDCASE + drop ; diff --git a/board-qemu/slof/helper.fs b/board-qemu/slof/helper.fs index c807bc6..6613782 100644 --- a/board-qemu/slof/helper.fs +++ b/board-qemu/slof/helper.fs @@ -33,10 +33,16 @@ swap - ; -: invert-region ( addr len -- ) - over swap 0 swap 1 hv-logical-memop drop +: invert-region-cs ( addr len cellsize -- ) + >r over swap r@ rshift r> swap 1 hv-logical-memop drop ; -: invert-region-x ( addr len -- ) - over swap /x / 3 swap 1 hv-logical-memop drop +: invert-region ( addr len -- ) + 2dup or 7 and CASE + 0 OF 3 invert-region-cs ENDOF + 2 OF 1 invert-region-cs ENDOF + 4 OF 2 invert-region-cs ENDOF + 6 OF 1 invert-region-cs ENDOF + dup OF 0 invert-region-cs ENDOF + ENDCASE ; diff --git a/slof/fs/fbuffer.fs b/slof/fs/fbuffer.fs index fcdd2fa..0128c07 100644 --- a/slof/fs/fbuffer.fs +++ b/slof/fs/fbuffer.fs @@ -170,7 +170,7 @@ CREATE bitmap-buffer 400 4 * allot ; : fb8-invert-screen ( -- ) - frame-buffer-adr screen-height screen-width * screen-depth * invert-region-x + frame-buffer-adr screen-height screen-width * screen-depth * invert-region ; : fb8-blink-screen ( -- ) fb8-invert-screen fb8-invert-screen ;
The introduction of invert-region already speeded up the cursor drawing very much. But there is still space for improvement: So far invert-region is accessing the memory only byte by byte, but with some additional logic that checks the alignment of the address and the length of the area, we can also make this function to access the memory with half-word, word or long-word accesses. With this additional logic, invert-region-x is also no longer necessary and thus can be removed. Signed-off-by: Thomas Huth <thuth@redhat.com> --- board-js2x/slof/helper.fs | 13 ++++++++----- board-qemu/slof/helper.fs | 14 ++++++++++---- slof/fs/fbuffer.fs | 2 +- 3 files changed, 19 insertions(+), 10 deletions(-)