PATCH: Pass -msse2avx to assembler for -mavx

PATCH: Pass -msse2avx to assembler for -mavx

[H.J. Lu]

Linux x86 assembler supports -msse2avx, which encodes SSE instructions
with VEX prefix.  It will turn inline SSE instructions into AVX
instructions.  This patch passes -msse2avx to as for -mavx and adds
-msse2avx to gcc so that it can be turned off with -mno-sse2avx.  OK
for trunk?

Thanks.


H.J.
---
2008-09-29  H.J. Lu  <hongjiu.lu@intel.com>

	* config/i386/i386.opt: Add msse2avx.

	* config/i386/linux.h (ASM_SPEC): New.  Support -msse2avx.
	* config/i386/linux64.h (ASM_SPEC): Likewise.

	* doc/invoke.texi: Document -msse2avx.

--- gcc/config/i386/i386.opt.avx	2008-09-29 09:31:58.000000000 -0700
+++ gcc/config/i386/i386.opt	2008-09-29 10:01:58.000000000 -0700
@@ -342,3 +342,7 @@ Support AES built-in functions and code 
 mpclmul
 Target Report Mask(ISA_PCLMUL) Var(ix86_isa_flags) VarExists Save
 Support PCLMUL built-in functions and code generation
+
+msse2avx
+Target Report Var(ix86_sse2avx)
+Encode SSE instructions with VEX prefix
--- gcc/config/i386/linux.h.avx	2008-07-02 09:06:34.000000000 -0700
+++ gcc/config/i386/linux.h	2008-09-29 10:39:26.000000000 -0700
@@ -102,6 +102,11 @@ along with GCC; see the file COPYING3.  
 #define LINK_EMULATION "elf_i386"
 #define GLIBC_DYNAMIC_LINKER "/lib/ld-linux.so.2"
 
+#undef  ASM_SPEC
+#define ASM_SPEC \
+  "%{v:-V} %{Qy:} %{!Qn:-Qy} %{n} %{T} %{Ym,*} %{Yd,*} %{Wa,*:%*} \
+  %{!mno-sse2avx:%{mavx:-msse2avx}} %{msse2avx:%{!mavx:-msse2avx}}"
+
 #undef  SUBTARGET_EXTRA_SPECS
 #define SUBTARGET_EXTRA_SPECS \
   { "link_emulation", LINK_EMULATION },\
--- gcc/config/i386/linux64.h.avx	2008-07-02 09:06:34.000000000 -0700
+++ gcc/config/i386/linux64.h	2008-09-29 10:35:41.000000000 -0700
@@ -56,6 +56,11 @@ along with GCC; see the file COPYING3.  
 #define GLIBC_DYNAMIC_LINKER32 "/lib/ld-linux.so.2"
 #define GLIBC_DYNAMIC_LINKER64 "/lib64/ld-linux-x86-64.so.2"
 
+#undef ASM_SPEC
+#define ASM_SPEC "%{v:-V} %{Qy:} %{!Qn:-Qy} %{n} %{T} %{Ym,*} %{Yd,*} \
+ %{Wa,*:%*} %{m32:--32} %{m64:--64} \
+ %{!mno-sse2avx:%{mavx:-msse2avx}} %{msse2avx:%{!mavx:-msse2avx}}"
+
 #if TARGET_64BIT_DEFAULT
 #define SPEC_32 "m32"
 #define SPEC_64 "!m32"
--- gcc/doc/invoke.texi.avx	2008-09-25 17:09:34.000000000 -0700
+++ gcc/doc/invoke.texi	2008-09-29 11:06:27.000000000 -0700
@@ -581,7 +581,7 @@ Objective-C and Objective-C++ Dialects}.
 -momit-leaf-frame-pointer  -mno-red-zone -mno-tls-direct-seg-refs @gol
 -mcmodel=@var{code-model} @gol
 -m32  -m64 -mlarge-data-threshold=@var{num} @gol
--mfused-madd -mno-fused-madd}
+-mfused-madd -mno-fused-madd -msse2avx}
 
 @emph{IA-64 Options}
 @gccoptlist{-mbig-endian  -mlittle-endian  -mgnu-as  -mgnu-ld  -mno-pic @gol
@@ -11212,6 +11212,12 @@ Enable automatic generation of fused flo
 if the ISA supports such instructions.  The -mfused-madd option is on by
 default.  The fused multiply-add instructions have a different
 rounding behavior compared to executing a multiply followed by an add.
+
+@item -msse2avx
+@itemx -mno-sse2avx
+@opindex msse2avx
+Specify that the assembler should encode SSE instructions with VEX
+prefix.  The option @option{-mavx} turns this on by default.
 @end table
 
 These @samp{-m} switches are supported in addition to the above

Re: PATCH: Pass -msse2avx to assembler for -mavx

[Richard Henderson]

Ok.


r~