Module Name: src
Committed By: macallan
Date: Sun Jan 26 05:20:57 UTC 2025
Modified Files:
src/sys/dev/ic: summitreg.h
Log Message:
document FIFO behaviour and cursor register access
NFC
To generate a diff of this commit:
cvs rdiff -u -r1.13 -r1.14 src/sys/dev/ic/summitreg.h
Please note that diffs are not public domain; they are subject to the
copyright notices on the relevant files.
Modified files:
Index: src/sys/dev/ic/summitreg.h
diff -u src/sys/dev/ic/summitreg.h:1.13 src/sys/dev/ic/summitreg.h:1.14
--- src/sys/dev/ic/summitreg.h:1.13 Fri Jan 24 07:43:42 2025
+++ src/sys/dev/ic/summitreg.h Sun Jan 26 05:20:57 2025
@@ -1,4 +1,4 @@
-/* $NetBSD: summitreg.h,v 1.13 2025/01/24 07:43:42 macallan Exp $ */
+/* $NetBSD: summitreg.h,v 1.14 2025/01/26 05:20:57 macallan Exp $ */
/*
* Copyright (c) 2024 Michael Lorenz
@@ -42,6 +42,14 @@
#define CONTROL_WFC 0x00000200 // FIFO when 0, direct when 1
#define VISFX_FC 0x641040 // Fault Control
#define VISFX_STATUS 0x641400 // zero when idle
+/*
+ * about the FIFO register:
+ * - on FX4, there are 0x800 FIFO slots, quite a lot
+ * - based on observation, every register write seems to occupy *two* slots
+ * - we need to write 0 to VISFX_CONTROL to enable FIFO pacing
+ * - the FIFO is quite difficult to overrun but things like x11perf copywinwin
+ * will do it if we're not careful
+ */
#define VISFX_FIFO 0x641440
#define VISFX_FOE 0x920404 // Fragment Operation Enable
#define FOE_TEXTURE 0x00000001
@@ -144,9 +152,7 @@
*/
/*
- * Turns out 0x40xxxx and 0x80xxxx access the same registers, one difference
- * is that through 0x80xxxx we can read back at least some values, so use
- * that one
+ * use unbuffered space for cursor registers
* The _POS, _INDEX and _DATA registers work exactly like on HCRX
*/
