Unify CAR so the same compiled code does the right thing on both

K8 and Fam10+ CPUs.

What this patch does:
1. Enable SSE (to get some more registers to play with)
2. Determine CPUID, and stash it in an XMM register, and reference
   value for comparison in another XMM register (mangled somewhat to
   simplify inequality comparisons)
3. Add a macro jmp_if_k8, which jumps if the CPU is K8
   (using an SSE compare)
4. Replace #if CAR_FAM10 sections with runtime checks using jmp_if_k8.
   This is pretty mechanical work. The macro uses local labels
   (1: and 2:) to prevent namespace issues
5. At one time, CPU_ADDR_BITS is used to fill a register. This is
   replaced with hardcoded values for both cases, and switched
   appropriately.
6. Disable SSE

Signed-off-by: Patrick Georgi <patrick.georgi@coresystems.de>
Signed-off-by: Carl-Daniel Hailfinger <c-d.hailfinger.devel.2006@gmx.net>
Acked-by: Stefan Reinauer <stepan@coresystems.de>



git-svn-id: svn://svn.coreboot.org/coreboot/trunk@3951 2b7e53f0-3cfb-0310-b3e9-8179ed1497e1

1 files changed, 56 insertions, 18 deletions

diff --git a/src/cpu/amd/car/cache_as_ram.inc b/src/cpu/amd/car/cache_as_ram.inc
index 8b8f53c515..162680db73 100644
--- a/src/cpu/amd/car/cache_as_ram.inc
+++ b/src/cpu/amd/car/cache_as_ram.inc
@@ -24,12 +24,22 @@
 /* leave some space for global variable to pass to RAM stage */
 #define GlobalVarSize DCACHE_RAM_GLOBAL_VAR_SIZE
 
-#if CAR_FAM10 == 1
+/* for CAR_FAM10 */
 #define CacheSizeAPStack 0x400 /* 1K */
-#endif
+
+#define jmp_if_k8(x)	comisd	%xmm2, %xmm1; jb x
+
+#define CPUID_MASK		0x0ff00f00
+#define CPUID_VAL_FAM10_ROTATED	0x0f000010
 
 #include <cpu/x86/mtrr.h>
 #include <cpu/amd/mtrr.h>
+/*
+ XMM map:
+    xmm1: cpu family
+    xmm2: fam10 comparison value
+    xmm3: backup ebx
+*/
 
 	/* Save the BIST result */
 	movl	%eax, %ebp
@@ -41,6 +51,24 @@ cache_as_ram_setup:
 	movb	$0xA0, %al
 	outb	%al, $0x80
 
+	/* enable SSE */
+	movl %cr4, %eax
+	orl $(3<<9), %eax
+	movl %eax, %cr4
+
+	/* figure out cpu family */
+	cvtsi2sd %ebx, %xmm3
+	movl	$0x01, %eax
+	cpuid
+	/* base family is bits 8..11, extended family is bits 20..27 */
+	andl	$CPUID_MASK, %eax
+	/* reorder bits for easier comparison by value */
+	roll	$0x10, %eax
+	cvtsi2sd %eax, %xmm1
+	movl	$CPUID_VAL_FAM10_ROTATED, %eax
+	cvtsi2sd %eax, %xmm2
+	cvtsd2si %xmm3, %ebx
+
 	/* hope we can skip the double set for normal part */
 #if ((HAVE_FAILOVER_BOOT == 1) && (USE_FAILOVER_IMAGE == 1)) || ((HAVE_FAILOVER_BOOT == 0) && (USE_FALLBACK_IMAGE == 1))
 
@@ -50,7 +78,8 @@ cache_as_ram_setup:
 	andl	$(1 << 11), %eax
 	movl	%eax, %ebx	/* We store the status */
 
-#if CAR_FAM10 == 1
+	jmp_if_k8(CAR_FAM10_out_post_errata)
+
 	/* for GH, CAR need to set DRAM Base/Limit Registers to direct that to node0 */
 
 	/* Only BSP needed, for other nodes set during HT/memory init. */
@@ -86,16 +115,14 @@ cache_as_ram_setup:
 
 CAR_FAM10_out:
 
-#endif
-
-#if CAR_FAM10 == 1
 	/* Errata 193: Disable clean copybacks to L3 cache to allow cached ROM.
 	   Re-enable it in after RAM is initialized and before CAR is disabled */
 	movl	$0xc001102a, %ecx
 	rdmsr
 	bts	$15, %eax
 	wrmsr
-#endif
+
+CAR_FAM10_out_post_errata:
 
 	/* Set MtrrFixDramModEn for clear fixed mtrr */
 enable_fixed_mtrr_dram_modify:
@@ -134,8 +161,10 @@ clear_fixed_var_mtrr_out:
 	 * macro will have a monotonically increasing segs parameter.
 	 */
 	xorl \reg, \reg
-#if CAR_FAM10 == 1
-.elseif \segs == 1
+.else
+	jmp_if_k8(1f)
+
+.if \segs == 1
 	movl $0x1e000000, \reg /* WB MEM type */
 .elseif \segs == 2
 	movl $0x1e1e0000, \reg /* WB MEM type */
@@ -143,8 +172,10 @@ clear_fixed_var_mtrr_out:
 	movl $0x1e1e1e00, \reg /* WB MEM type */
 .elseif \segs >= 4
 	movl $0x1e1e1e1e, \reg /* WB MEM type */
-#else
-.elseif \segs == 1
+.endif
+	jmp 2f
+1:
+.if \segs == 1
 	movl $0x06000000, \reg /* WB IO type */
 .elseif \segs == 2
 	movl $0x06060000, \reg /* WB IO type */
@@ -152,8 +183,9 @@ clear_fixed_var_mtrr_out:
 	movl $0x06060600, \reg /* WB IO type */
 .elseif \segs >= 4
 	movl $0x06060606, \reg /* WB IO type */
-#endif
 .endif
+2:
+.endif /* if \segs <= 0 */
 .endm
 
 /* size is the cache size in bytes we want to use for CAR.
@@ -219,7 +251,10 @@ clear_fixed_var_mtrr_out:
 	wrmsr
 
 	movl	$0x203, %ecx
-	movl	$((1 << (CPU_ADDR_BITS - 32)) - 1), %edx  /* AMD 40 bit for K8, 48 bit for GH */
+	movl	$0xff, %edx /* (1 << (CPU_ADDR_BITS - 32)) - 1 for K8 (CPU_ADDR_BITS = 40) */
+	jmp_if_k8(wbcache_post_fam10_setup)
+	movl	$0xffff, %edx /* (1 << (CPU_ADDR_BITS - 32)) - 1 for FAM10 (CPU_ADDR_BITS = 48) */
+wbcache_post_fam10_setup:
 	movl	$(~(XIP_ROM_SIZE - 1) | 0x800), %eax
 	wrmsr
 #endif /* XIP_ROM_SIZE && XIP_ROM_BASE */
@@ -248,13 +283,14 @@ clear_fixed_var_mtrr_out:
 	movl	%eax, %cr0
 
 
-#if CAR_FAM10 == 1
+	jmp_if_k8(fam10_end_part1)
+
 	/* So we need to check if it is BSP */
 	movl	$0x1b, %ecx
 	rdmsr
 	bt	$8, %eax /*BSC */
 	jnc	CAR_FAM10_ap
-#endif
+fam10_end_part1:
 
 	movb	$0xA2, %al
 	outb	%al, $0x80
@@ -280,8 +316,6 @@ clear_fixed_var_mtrr_out:
 	movb	$0xA3, %al
 	outb	%al, $0x80
 
-#if CAR_FAM10 == 1
-
 	jmp	CAR_FAM10_ap_out
 CAR_FAM10_ap:
 	/* need to set stack pointer for AP */
@@ -326,11 +360,15 @@ roll_cfg:
 	outb	%al, $0x80
 
 CAR_FAM10_ap_out:
-#endif
 
 	movb	$0xA5, %al
 	outb	%al, $0x80
 
+	/* disable SSE */
+	movl %cr4, %eax
+	andl $~(3<<9), %eax
+	movl %eax, %cr4
+
 	/* Restore the BIST result */
 	movl	%ebp, %eax