summaryrefslogtreecommitdiff
path: root/mfb/fastblt.h
diff options
context:
space:
mode:
Diffstat (limited to 'mfb/fastblt.h')
-rw-r--r--mfb/fastblt.h118
1 files changed, 118 insertions, 0 deletions
diff --git a/mfb/fastblt.h b/mfb/fastblt.h
new file mode 100644
index 000000000..6567a8936
--- /dev/null
+++ b/mfb/fastblt.h
@@ -0,0 +1,118 @@
+/* $Xorg: fastblt.h,v 1.4 2001/02/09 02:05:17 xorgcvs Exp $ */
+/*
+
+Copyright 1989, 1998 The Open Group
+
+Permission to use, copy, modify, distribute, and sell this software and its
+documentation for any purpose is hereby granted without fee, provided that
+the above copyright notice appear in all copies and that both that
+copyright notice and this permission notice appear in supporting
+documentation.
+
+The above copyright notice and this permission notice shall be included
+in all copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+IN NO EVENT SHALL THE OPEN GROUP BE LIABLE FOR ANY CLAIM, DAMAGES OR
+OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+OTHER DEALINGS IN THE SOFTWARE.
+
+Except as contained in this notice, the name of The Open Group shall
+not be used in advertising or otherwise to promote the sale, use or
+other dealings in this Software without prior written authorization
+from The Open Group.
+
+*/
+
+/*
+ * Fast bitblt macros for certain hardware. If your machine has an addressing
+ * mode of small constant + register, you'll probably want this magic specific
+ * code. It's 25% faster for the R2000. I haven't studied the Sparc
+ * instruction set, but I suspect it also has this addressing mode. Also,
+ * unrolling the loop by 32 is possibly excessive for mfb. The number of times
+ * the loop is actually looped through is pretty small.
+ */
+
+/*
+ * WARNING: These macros make *a lot* of assumptions about
+ * the environment they are invoked in. Plenty of implicit
+ * arguments, lots of side effects. Don't use them casually.
+ */
+
+#define SwitchOdd(n) case n: BodyOdd(n)
+#define SwitchEven(n) case n: BodyEven(n)
+
+/* to allow mfb and cfb to share code... */
+#ifndef BitRight
+#define BitRight(a,b) SCRRIGHT(a,b)
+#define BitLeft(a,b) SCRLEFT(a,b)
+#endif
+
+#ifdef LARGE_INSTRUCTION_CACHE
+#define UNROLL 8
+#define PackedLoop \
+ switch (nl & (UNROLL-1)) { \
+ SwitchOdd( 7) SwitchEven( 6) SwitchOdd( 5) SwitchEven( 4) \
+ SwitchOdd( 3) SwitchEven( 2) SwitchOdd( 1) \
+ } \
+ while ((nl -= UNROLL) >= 0) { \
+ LoopReset \
+ BodyEven( 8) \
+ BodyOdd( 7) BodyEven( 6) BodyOdd( 5) BodyEven( 4) \
+ BodyOdd( 3) BodyEven( 2) BodyOdd( 1) \
+ }
+#else
+#define UNROLL 4
+#define PackedLoop \
+ switch (nl & (UNROLL-1)) { \
+ SwitchOdd( 3) SwitchEven( 2) SwitchOdd( 1) \
+ } \
+ while ((nl -= UNROLL) >= 0) { \
+ LoopReset \
+ BodyEven( 4) \
+ BodyOdd( 3) BodyEven( 2) BodyOdd( 1) \
+ }
+#endif
+
+#if PPW == 32
+#define DuffL(counter,label,body) \
+ switch (counter & 3) { \
+ label: \
+ body \
+ case 3: \
+ body \
+ case 2: \
+ body \
+ case 1: \
+ body \
+ case 0: \
+ if ((counter -= 4) >= 0) \
+ goto label; \
+ }
+#else /* PPW == 64 */
+#define DuffL(counter,label,body) \
+ switch (counter & 7) { \
+ label: \
+ body \
+ case 7: \
+ body \
+ case 6: \
+ body \
+ case 5: \
+ body \
+ case 4: \
+ body \
+ case 3: \
+ body \
+ case 2: \
+ body \
+ case 1: \
+ body \
+ case 0: \
+ if ((counter -= 8) >= 0) \
+ goto label; \
+ }
+#endif /* PPW */