Fix Clang support

The amalgamated release is now confirmed to be working with Clang,
including its integrated assembler.

Fixes #41

dsp/tty/windex-avx2.S

@@ -24,7 +24,7 @@
 /	@param	rdi points to nonempty array
 /	@param	rsi is item count divisible by 16
 /	@note	needs avx2 (haswell+)
-windex$avx2:
+windex_avx2:
 	push	%rbp
 	mov	%rsp,%rbp
 	.profilable
@@ -64,7 +64,7 @@ windex$avx2:
 5:	vzeroupper
 	leave
 	ret
-	.endfn	windex$avx2,globl
+	.endfn	windex_avx2,globl
 
 	.rodata.cst32
 .Lidx:	.short	0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15

dsp/tty/windex-k8.c

@@ -19,7 +19,7 @@
 #include "dsp/tty/tty.h"
 #include "libc/assert.h"
 
-unsigned windex$k8(short *a, size_t n) {
+unsigned windex_k8(short *a, size_t n) {
   unsigned short min, res, i;
   res = 0;
   min = a[0];

dsp/tty/windex-sse4.S

@@ -24,7 +24,7 @@
 /	@param	rdi points to nonempty array
 /	@param	esi is 16-byte aligned 8+ / 8 multiple array item count
 /	@note	needs sse4 (nehalem+)
-windex$sse4:
+windex_sse4:
 	push	%rbp
 	mov	%rsp,%rbp
 	.profilable
@@ -54,7 +54,7 @@ windex$sse4:
 	movzwl	-32(%rbp,%rax,2),%eax
 	leave
 	ret
-	.endfn	windex$sse4,globl
+	.endfn	windex_sse4,globl
 
 	.rodata.cst16
 .Lidx:	.short	0,1,2,3,4,5,6,7

dsp/tty/windex.S

@@ -27,14 +27,14 @@ windex:	.quad	0
 	.previous
 
 	.init.start 300,_init_windex
-	ezlea	windex$avx2,ax
+	ezlea	windex_avx2,ax
 #if !X86_NEED(AVX2)
-	ezlea	windex$sse4,dx
+	ezlea	windex_sse4,dx
 	testb	X86_HAVE(AVX2)+kCpuids(%rip)
 	cmovz	%rdx,%rax
 #endif /* AVX2 */
 #if !X86_NEED(SSE4_2)
-	ezlea	windex$k8,dx
+	ezlea	windex_k8,dx
 	testb	X86_HAVE(SSE4_2)+kCpuids(%rip)
 	cmovz	%rdx,%rax
 #endif /* SSE4 */