arch/tile: finish enabling support for TILE-Gx 64-bit chip

This support was partially present in the existing code (look for
"__tilegx__" ifdefs) but with this change you can build a working
kernel using the TILE-Gx toolchain and ARCH=tilegx.

Most of these files are new, generally adding a foo_64.c file
where previously there was just a foo_32.c file.

The ARCH=tilegx directive redirects to arch/tile, not arch/tilegx,
using the existing SRCARCH mechanism in the top-level Makefile.

Changes to existing files:

- <asm/bitops.h> and <asm/bitops_32.h> changed to factor the
  include of <asm-generic/bitops/non-atomic.h> in the common header.

- <asm/compat.h> and arch/tile/kernel/compat.c changed to remove
  the "const" markers I had put on compat_sys_execve() when trying
  to match some recent similar changes to the non-compat execve.
  It turns out the compat version wasn't "upgraded" to use const.

- <asm/opcode-tile_64.h> and <asm/opcode_constants_64.h> were
  previously included accidentally, with the 32-bit contents.  Now
  they have the proper 64-bit contents.

Finally, I had to hack the existing hacky drivers/input/input-compat.h
to add yet another "#ifdef" for INPUT_COMPAT_TEST (same as x86_64).

Signed-off-by: Chris Metcalf <cmetcalf@tilera.com>
Acked-by: Dmitry Torokhov <dmitry.torokhov@gmail.com> [drivers/input]

1 files changed, 145 insertions, 0 deletions

diff --git a/arch/tile/lib/memset_64.c b/arch/tile/lib/memset_64.c
new file mode 100644
index 00000000000..3873085711d
--- /dev/null
+++ b/arch/tile/lib/memset_64.c
@@ -0,0 +1,145 @@
+/*
+ * Copyright 2011 Tilera Corporation. All Rights Reserved.
+ *
+ *   This program is free software; you can redistribute it and/or
+ *   modify it under the terms of the GNU General Public License
+ *   as published by the Free Software Foundation, version 2.
+ *
+ *   This program is distributed in the hope that it will be useful, but
+ *   WITHOUT ANY WARRANTY; without even the implied warranty of
+ *   MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE, GOOD TITLE or
+ *   NON INFRINGEMENT.  See the GNU General Public License for
+ *   more details.
+ */
+
+#include <arch/chip.h>
+
+#include <linux/types.h>
+#include <linux/string.h>
+#include <linux/module.h>
+
+#undef memset
+
+void *memset(void *s, int c, size_t n)
+{
+	uint64_t *out64;
+	int n64, to_align64;
+	uint64_t v64;
+	uint8_t *out8 = s;
+
+	/* Experimentation shows that a trivial tight loop is a win up until
+	 * around a size of 20, where writing a word at a time starts to win.
+	 */
+#define BYTE_CUTOFF 20
+
+#if BYTE_CUTOFF < 7
+	/* This must be at least at least this big, or some code later
+	 * on doesn't work.
+	 */
+#error "BYTE_CUTOFF is too small"
+#endif
+
+	if (n < BYTE_CUTOFF) {
+		/* Strangely, this turns out to be the tightest way to
+		 * write this loop.
+		 */
+		if (n != 0) {
+			do {
+				/* Strangely, combining these into one line
+				 * performs worse.
+				 */
+				*out8 = c;
+				out8++;
+			} while (--n != 0);
+		}
+
+		return s;
+	}
+
+	/* Align 'out8'. We know n >= 7 so this won't write past the end. */
+	while (((uintptr_t) out8 & 7) != 0) {
+		*out8++ = c;
+		--n;
+	}
+
+	/* Align 'n'. */
+	while (n & 7)
+		out8[--n] = c;
+
+	out64 = (uint64_t *) out8;
+	n64 = n >> 3;
+
+	/* Tile input byte out to 64 bits. */
+	/* KLUDGE */
+	v64 = 0x0101010101010101ULL * (uint8_t)c;
+
+	/* This must be at least 8 or the following loop doesn't work. */
+#define CACHE_LINE_SIZE_IN_DOUBLEWORDS (CHIP_L2_LINE_SIZE() / 8)
+
+	/* Determine how many words we need to emit before the 'out32'
+	 * pointer becomes aligned modulo the cache line size.
+	 */
+	to_align64 = (-((uintptr_t)out64 >> 3)) &
+		(CACHE_LINE_SIZE_IN_DOUBLEWORDS - 1);
+
+	/* Only bother aligning and using wh64 if there is at least
+	 * one full cache line to process.  This check also prevents
+	 * overrunning the end of the buffer with alignment words.
+	 */
+	if (to_align64 <= n64 - CACHE_LINE_SIZE_IN_DOUBLEWORDS) {
+		int lines_left;
+
+		/* Align out64 mod the cache line size so we can use wh64. */
+		n64 -= to_align64;
+		for (; to_align64 != 0; to_align64--) {
+			*out64 = v64;
+			out64++;
+		}
+
+		/* Use unsigned divide to turn this into a right shift. */
+		lines_left = (unsigned)n64 / CACHE_LINE_SIZE_IN_DOUBLEWORDS;
+
+		do {
+			/* Only wh64 a few lines at a time, so we don't
+			 * exceed the maximum number of victim lines.
+			 */
+			int x = ((lines_left < CHIP_MAX_OUTSTANDING_VICTIMS())
+				  ? lines_left
+				  : CHIP_MAX_OUTSTANDING_VICTIMS());
+			uint64_t *wh = out64;
+			int i = x;
+			int j;
+
+			lines_left -= x;
+
+			do {
+				__insn_wh64(wh);
+				wh += CACHE_LINE_SIZE_IN_DOUBLEWORDS;
+			} while (--i);
+
+			for (j = x * (CACHE_LINE_SIZE_IN_DOUBLEWORDS / 4);
+			     j != 0; j--) {
+				*out64++ = v64;
+				*out64++ = v64;
+				*out64++ = v64;
+				*out64++ = v64;
+			}
+		} while (lines_left != 0);
+
+		/* We processed all full lines above, so only this many
+		 * words remain to be processed.
+		 */
+		n64 &= CACHE_LINE_SIZE_IN_DOUBLEWORDS - 1;
+	}
+
+	/* Now handle any leftover values. */
+	if (n64 != 0) {
+		do {
+			*out64 = v64;
+			out64++;
+		} while (--n64 != 0);
+	}
+
+	return s;
+}
+EXPORT_SYMBOL(memset);