From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([2001:4830:134:3::10]:43813) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1gQCig-0007yY-T5 for qemu-devel@nongnu.org; Fri, 23 Nov 2018 09:46:12 -0500 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1gQCif-0003AS-Oc for qemu-devel@nongnu.org; Fri, 23 Nov 2018 09:46:10 -0500 Received: from mail-wm1-x343.google.com ([2a00:1450:4864:20::343]:38779) by eggs.gnu.org with esmtps (TLS1.0:RSA_AES_128_CBC_SHA1:16) (Exim 4.71) (envelope-from ) id 1gQCif-00038g-HQ for qemu-devel@nongnu.org; Fri, 23 Nov 2018 09:46:09 -0500 Received: by mail-wm1-x343.google.com with SMTP id k198so12261561wmd.3 for ; Fri, 23 Nov 2018 06:46:08 -0800 (PST) From: Richard Henderson Date: Fri, 23 Nov 2018 15:45:28 +0100 Message-Id: <20181123144558.5048-8-richard.henderson@linaro.org> In-Reply-To: <20181123144558.5048-1-richard.henderson@linaro.org> References: <20181123144558.5048-1-richard.henderson@linaro.org> Subject: [Qemu-devel] [PATCH for-4.0 v2 07/37] tcg/i386: Change TCG_REG_L[01] to not overlap function arguments List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: qemu-devel@nongnu.org Cc: Alistair.Francis@wdc.com We will shortly be forcing qemu_ld/st arguments into registers that match the function call abi of the host, which means that the temps must be elsewhere. Signed-off-by: Richard Henderson --- tcg/i386/tcg-target.inc.c | 28 +++++++++++++++++++--------- 1 file changed, 19 insertions(+), 9 deletions(-) diff --git a/tcg/i386/tcg-target.inc.c b/tcg/i386/tcg-target.inc.c index 3234a8d8bf..07df4b2b12 100644 --- a/tcg/i386/tcg-target.inc.c +++ b/tcg/i386/tcg-target.inc.c @@ -121,12 +121,16 @@ static const int tcg_target_call_oarg_regs[] = { #define TCG_CT_CONST_I32 0x400 #define TCG_CT_CONST_WSZ 0x800 -/* Registers used with L constraint, which are the first argument - registers on x86_64, and two random call clobbered registers on - i386. */ +/* Registers used with L constraint, which are two random + * call clobbered registers. These should be free. + */ #if TCG_TARGET_REG_BITS == 64 -# define TCG_REG_L0 tcg_target_call_iarg_regs[0] -# define TCG_REG_L1 tcg_target_call_iarg_regs[1] +# define TCG_REG_L0 TCG_REG_RAX +# ifdef _WIN64 +# define TCG_REG_L1 TCG_REG_R10 +# else +# define TCG_REG_L1 TCG_REG_RDI +# endif #else # define TCG_REG_L0 TCG_REG_EAX # define TCG_REG_L1 TCG_REG_EDX @@ -1628,6 +1632,7 @@ static TCGReg tcg_out_tlb_load(TCGContext *s, TCGReg addrlo, TCGReg addrhi, unsigned a_mask = (1 << a_bits) - 1; unsigned s_mask = (1 << s_bits) - 1; target_ulong tlb_mask; + TCGReg base; if (TCG_TARGET_REG_BITS == 64) { if (TARGET_LONG_BITS == 64) { @@ -1674,7 +1679,12 @@ static TCGReg tcg_out_tlb_load(TCGContext *s, TCGReg addrlo, TCGReg addrhi, before the fastpath ADDQ below. For 64-bit guest and x32 host, MOVQ copies the entire guest address for the slow path, while truncation for the 32-bit host happens with the fastpath ADDL below. */ - tcg_out_mov(s, ttype, r1, addrlo); + if (TCG_TARGET_REG_BITS == 64) { + base = tcg_target_call_iarg_regs[1]; + } else { + base = r1; + } + tcg_out_mov(s, ttype, base, addrlo); /* jne slow_path */ tcg_out_opc(s, OPC_JCC_long + JCC_JNE, 0, 0, 0); @@ -1693,11 +1703,11 @@ static TCGReg tcg_out_tlb_load(TCGContext *s, TCGReg addrlo, TCGReg addrhi, /* TLB Hit. */ - /* add addend(r0), r1 */ - tcg_out_modrm_offset(s, OPC_ADD_GvEv + hrexw, r1, r0, + /* add addend(r0), base */ + tcg_out_modrm_offset(s, OPC_ADD_GvEv + hrexw, base, r0, offsetof(CPUTLBEntry, addend) - which); - return r1; + return base; } /* -- 2.17.2