This is the mail archive of the
binutils@sourceware.org
mailing list for the binutils project.
[PATCH] x86-64: also optimize ANDQ with immediate fitting in 7 bits
- From: "Jan Beulich" <JBeulich at suse dot com>
- To: <binutils at sourceware dot org>
- Cc: "H.J. Lu" <hjl dot tools at gmail dot com>
- Date: Mon, 24 Jun 2019 06:57:12 -0600
- Subject: [PATCH] x86-64: also optimize ANDQ with immediate fitting in 7 bits
The same reasoning applies here as did/does for immediates fitting in
31 bits.
gas/
2019-06-24 Jan Beulich <jbeulich@suse.com>
* config/tc-i386.c (optimize_encoding): Also handle ANDQ with
immediatie fitting in 7 bits.
* testsuite/gas/i386/x86-64-optimize-1.s: Add ANDQ cases with
7- and 8-bit immediates.
* testsuite/gas/i386/x86-64-optimize-1.d: Adjust expectations.
opcodes/
2019-06-24 Jan Beulich <jbeulich@suse.com>
* i386-opc.tbl (and): Mark Imm8S form for optimization.
* i386-tbl.h: Re-generate.
--- a/gas/config/tc-i386.c
+++ b/gas/config/tc-i386.c
@@ -3948,7 +3948,10 @@ optimize_encoding (void)
&& i.tm.extension_opcode == 0x4)
|| ((i.tm.base_opcode == 0xf6
|| i.tm.base_opcode == 0xc6)
- && i.tm.extension_opcode == 0x0)))))
+ && i.tm.extension_opcode == 0x0)))
+ || (fits_in_imm7 (i.op[0].imms->X_add_number)
+ && i.tm.base_opcode == 0x83
+ && i.tm.extension_opcode == 0x4)))
|| (i.types[0].bitfield.qword
&& ((i.reg_operands == 2
&& i.op[0].regs == i.op[1].regs
@@ -3962,6 +3965,7 @@ optimize_encoding (void)
{
/* Optimize: -O:
andq $imm31, %r64 -> andl $imm31, %r32
+ andq $imm7, %r64 -> andl $imm7, %r32
testq $imm31, %r64 -> testl $imm31, %r32
xorq %r64, %r64 -> xorl %r32, %r32
subq %r64, %r64 -> subl %r32, %r32
--- a/gas/testsuite/gas/i386/x86-64-optimize-1.d
+++ b/gas/testsuite/gas/i386/x86-64-optimize-1.d
@@ -15,6 +15,12 @@ Disassembly of section .text:
+[a-f0-9]+: 48 25 00 00 00 80 and \$0xffffffff80000000,%rax
+[a-f0-9]+: 48 81 e3 00 00 00 80 and \$0xffffffff80000000,%rbx
+[a-f0-9]+: 49 81 e6 00 00 00 80 and \$0xffffffff80000000,%r14
+ +[a-f0-9]+: 83 e0 7f and \$0x7f,%eax
+ +[a-f0-9]+: 83 e3 7f and \$0x7f,%ebx
+ +[a-f0-9]+: 41 83 e6 7f and \$0x7f,%r14d
+ +[a-f0-9]+: 48 83 e0 80 and \$0xffffffffffffff80,%rax
+ +[a-f0-9]+: 48 83 e3 80 and \$0xffffffffffffff80,%rbx
+ +[a-f0-9]+: 49 83 e6 80 and \$0xffffffffffffff80,%r14
+[a-f0-9]+: a9 ff ff ff 7f test \$0x7fffffff,%eax
+[a-f0-9]+: f7 c3 ff ff ff 7f test \$0x7fffffff,%ebx
+[a-f0-9]+: 41 f7 c6 ff ff ff 7f test \$0x7fffffff,%r14d
--- a/gas/testsuite/gas/i386/x86-64-optimize-1.s
+++ b/gas/testsuite/gas/i386/x86-64-optimize-1.s
@@ -10,6 +10,12 @@ _start:
andq $-((1<<31)), %rax
andq $-((1<<31)), %rbx
andq $-((1<<31)), %r14
+ andq $((1<<7) - 1), %rax
+ andq $((1<<7) - 1), %rbx
+ andq $((1<<7) - 1), %r14
+ andq $-((1<<7)), %rax
+ andq $-((1<<7)), %rbx
+ andq $-((1<<7)), %r14
testq $((1<<31) - 1), %rax
testq $((1<<31) - 1), %rbx
testq $((1<<31) - 1), %r14
--- a/opcodes/i386-opc.tbl
+++ b/opcodes/i386-opc.tbl
@@ -209,7 +209,7 @@ test, 2, 0xa8, None, 1, 0, W|No_sSuf|No_
test, 2, 0xf6, 0x0, 1, 0, W|Modrm|No_sSuf|No_ldSuf|Optimize, { Imm8|Imm16|Imm32|Imm32S, Reg8|Reg16|Reg32|Reg64|Byte|Word|Dword|Qword|Unspecified|BaseIndex }
and, 2, 0x20, None, 1, 0, D|W|CheckRegSize|Modrm|No_sSuf|No_ldSuf|IsLockable|HLEPrefixOk, { Reg8|Reg16|Reg32|Reg64, Reg8|Reg16|Reg32|Reg64|Byte|Word|Dword|Qword|Unspecified|BaseIndex }
-and, 2, 0x83, 0x4, 1, 0, Modrm|No_bSuf|No_sSuf|No_ldSuf|IsLockable|HLEPrefixOk, { Imm8S, Reg16|Reg32|Reg64|Word|Dword|Qword|Unspecified|BaseIndex }
+and, 2, 0x83, 0x4, 1, 0, Modrm|No_bSuf|No_sSuf|No_ldSuf|IsLockable|HLEPrefixOk|Optimize, { Imm8S, Reg16|Reg32|Reg64|Word|Dword|Qword|Unspecified|BaseIndex }
and, 2, 0x24, None, 1, 0, W|No_sSuf|No_ldSuf|Optimize, { Imm8|Imm16|Imm32|Imm32S, Acc|Byte|Word|Dword|Qword }
and, 2, 0x80, 0x4, 1, 0, W|Modrm|No_sSuf|No_ldSuf|IsLockable|HLEPrefixOk|Optimize, { Imm8|Imm16|Imm32|Imm32S, Reg8|Reg16|Reg32|Reg64|Byte|Word|Dword|Qword|Unspecified|BaseIndex }