Merge branch 'dev'

CMakeLists.txt

@@ -1,6 +1,6 @@
 cmake_minimum_required(VERSION 3.5)
 
-project(xbyak LANGUAGES CXX VERSION 7.02)
+project(xbyak LANGUAGES CXX VERSION 7.03)
 
 file(GLOB headers xbyak/*.h)
 

doc/changelog.md

@@ -1,5 +1,6 @@
 # History
 
+* 2023/Dec/26 ver 7.03 set the default value of dfv to 0
 * 2023/Dec/20 ver 7.02 SHA* support APX
 * 2023/Dec/19 ver 7.01 support AESKLE, WIDE_KL, KEYLOCKER, KEYLOCKER_WIDE, detection of APX10/APX
 * 2023/Dec/01 ver 7.00 support APX

doc/usage.md

@@ -145,6 +145,16 @@ vpdpbusd(xm0, xm1, xm2); // VEX encoding
   - `imul(ax|T_zu, cx, 0x1234);` // Set ND=ZU
   - `imul(ax|T_zu|T_nf, cx, 0x1234);` // Set ND=ZU and EVEX.NF=1
   - `setb(r31b|T_zu);` // same as set(r31b); movzx(r31, r31b);
+  - See [sample/zero_upper.cpp](../sample/zero_upper.cpp)
+
+### ccmpSCC and ctestSCC
+
+- ccmpSCC(op1, op2, dfv = 0); // eflags = eflags == SCC ? cmp(op1, op2) : dfv
+- ctestSCC(op1, op2, dfv = 0); // eflags = eflags == SCC ? test(op1, op2) : dfv
+- SCC means source condition code such as z, a, gt.
+- See [sample/ccmp.cpp](../sample/ccmp.cpp)
+- Specify the union of T_of(=8), T_sf(=4), T_zf(=2), or T_cf(=1) for dfv.
+
 
 ## Label
 Two kinds of Label are supported. (String literal and Label class).

gen/gen_code.cpp

@@ -635,10 +635,10 @@ void put()
 			// ccmpscc
 			// true if SCC = 0b1010, false if SCC = 0b1011 (see APX Architecture Specification p.266)
 			const char *s = p->ext == 10 ? "t" : p->ext == 11 ? "f" : p->name;
-			printf("void ccmp%s(const Operand& op1, const Operand& op2, int dfv) { opCcmp(op1, op2, dfv, 0x38, %d); }\n", s, p->ext);
-			printf("void ccmp%s(const Operand& op, int imm, int dfv) { opCcmpi(op, imm, dfv, %d); }\n", s, p->ext);
-			printf("void ctest%s(const Operand& op, const Reg& r, int dfv) { opCcmp(op, r, dfv, 0x84, %d); }\n", s, p->ext);
-			printf("void ctest%s(const Operand& op, int imm, int dfv) { opTesti(op, imm, dfv, %d); }\n", s, p->ext);
+			printf("void ccmp%s(const Operand& op1, const Operand& op2, int dfv = 0) { opCcmp(op1, op2, dfv, 0x38, %d); }\n", s, p->ext);
+			printf("void ccmp%s(const Operand& op, int imm, int dfv = 0) { opCcmpi(op, imm, dfv, %d); }\n", s, p->ext);
+			printf("void ctest%s(const Operand& op, const Reg& r, int dfv = 0) { opCcmp(op, r, dfv, 0x84, %d); }\n", s, p->ext);
+			printf("void ctest%s(const Operand& op, int imm, int dfv = 0) { opTesti(op, imm, dfv, %d); }\n", s, p->ext);
 		}
 	}
 	{

meson.build

@@ -5,7 +5,7 @@
 project(
 	'xbyak',
 	'cpp',
-	version: '7.02',
+	version: '7.03',
 	license: 'BSD-3-Clause',
 	default_options: 'b_ndebug=if-release'
 )

readme.md

@@ -1,5 +1,5 @@
 
-# Xbyak 7.02 [![Badge Build]][Build Status]
+# Xbyak 7.03 [![Badge Build]][Build Status]
 
 *A C++ JIT assembler for x86 (IA32), x64 (AMD64, x86-64)*
 

readme.txt

@@ -1,5 +1,5 @@
 
-    C++用x86(IA-32), x64(AMD64, x86-64) JITアセンブラ Xbyak 7.02
+    C++用x86(IA-32), x64(AMD64, x86-64) JITアセンブラ Xbyak 7.03
 
 -----------------------------------------------------------------------------
 ◎概要
@@ -404,6 +404,7 @@ sample/{echo,hello}.bfは http://www.kmonos.net/alang/etc/brainfuck.php から
 -----------------------------------------------------------------------------
 ◎履歴
 
+2023/12/26 ver 7.03 dfvのデフォルト値を0に設定
 2023/12/20 ver 7.02 SHA*のAPX対応
 2023/12/19 ver 7.01 AESKLE, WIDE_KL, KEYLOCKER, KEYLOCKER_WIDE対応 APX10/APX判定対応
 2023/12/01 ver 7.00 APX対応

sample/Makefile

@@ -30,7 +30,7 @@ else
 endif
 
 ifeq ($(BIT),64)
-TARGET += test64 bf64 memfunc64 test_util64 jmp_table64
+TARGET += test64 bf64 memfunc64 test_util64 jmp_table64 zero_upper ccmp
 ifeq ($(BOOST_EXIST),1)
 TARGET += calc64 #calc2_64
 endif
@@ -103,6 +103,14 @@ profiler: profiler.cpp ../xbyak/xbyak_util.h
 	$(CXX) $(CFLAGS) profiler.cpp -o $@
 profiler-vtune: profiler.cpp ../xbyak/xbyak_util.h
 	$(CXX) $(CFLAGS) profiler.cpp -o $@ -DXBYAK_USE_VTUNE -I /opt/intel/vtune_amplifier/include/ -L /opt/intel/vtune_amplifier/lib64 -ljitprofiling -ldl
+zero_upper: zero_upper.cpp $(XBYAK_INC)
+	$(CXX) $(CFLAGS) zero_upper.cpp -o $@
+test_zero_upper: zero_upper
+	sde -future -- ./zero_upper
+ccmp: ccmp.cpp $(XBYAK_INC)
+	$(CXX) $(CFLAGS) ccmp.cpp -o $@
+test_ccmp: ccmp
+	sde -future -- ./ccmp
 
 clean:
 	rm -rf $(TARGET) profiler profiler-vtune

sample/ccmp.cpp

@@ -0,0 +1,68 @@
+/*
+	An example of ccmp
+	> g++ ccmp.cpp -I ../xbyak
+	> sde -future -- ./a.out
+*/
+#include <stdio.h>
+#include <xbyak/xbyak.h>
+#include <xbyak/xbyak_util.h>
+
+using namespace Xbyak;
+
+struct Code1 : Xbyak::CodeGenerator {
+	Code1()
+	{
+		Xbyak::util::StackFrame sf(this, 2);
+		const auto& p1 = sf.p[0];
+		const auto& p2 = sf.p[1];
+		int dfv = 0;
+		cmp(p1, 3);
+		ctesta(p2, 1, dfv); // eflags = (p1 > 3) ? (p2 & 1) : dfv;
+		setz(al|T_zu);
+	}
+};
+
+struct Code2 : Xbyak::CodeGenerator {
+	Code2()
+	{
+		Xbyak::util::StackFrame sf(this, 3);
+		const auto& p1 = sf.p[0];
+		const auto& p2 = sf.p[1];
+		const auto& p3 = sf.p[2];
+		int dfv = 0;
+		cmp(p1, 1);
+		ccmpe(p2, 2, dfv); // eflags = p1==1 ? p2==2 : dfv;
+		ccmpe(p3, 3, dfv); // eflags = (p1==1 && p2==2) ? p3==3 : dfv;
+		setz(al|T_zu); // p1==1 && p2==2 && p3==3
+	}
+};
+
+
+int main()
+	try
+{
+	{
+		puts("(p1 > 3) && ((p2 & 1) == 0)");
+		Code1 c;
+		auto f = c.getCode<int (*)(int, int)>();
+		for (int p1 = 2; p1 < 5; p1++) {
+			for (int p2 = 0; p2 < 3; p2++) {
+				printf("p1=%d p2=%d ret=%d (%d)\n", p1, p2, f(p1, p2), p1 > 3 && ((p2&1) == 0));
+			}
+		}
+	}
+	{
+		puts("p1 == 1 && p2 == 2 && p3 == 3");
+		Code2 c;
+		auto f = c.getCode<int (*)(int, int, int)>();
+		for (int p1 = 0; p1 < 3; p1++) {
+			for (int p2 = 1; p2 < 4; p2++) {
+				for (int p3 = 2; p3 < 5; p3++) {
+					printf("p1=%d p2=%d p3=%d ret=%d (%d)\n", p1, p2, p3, f(p1, p2, p3), p1==1 && p2==2 && p3==3);
+				}
+			}
+		}
+	}
+} catch (std::exception& e) {
+	printf("ERR %s\n", e.what());
+}

sample/zero_upper.cpp

@@ -0,0 +1,48 @@
+/*
+	An example of T_zu (zero upper) flag
+	> g++ zero_upper.cpp -I ../xbyak
+	> sde -future -- ./a.out
+*/
+#include <stdio.h>
+#include <xbyak/xbyak.h>
+
+using namespace Xbyak;
+
+struct Code : Xbyak::CodeGenerator {
+	Code(int mode)
+	{
+		mov(eax, 0x12345678);
+		cmp(eax, eax); // ZF=1
+		switch (mode) {
+		case 0: // imul
+			puts("imul");
+			imul(ax,ax, 0x1234);
+			break;
+		case 1: // imul+zu
+			puts("imul+zu");
+			imul(ax|T_zu, ax, 0x1234);
+			break;
+		case 2: // setz
+			puts("setz");
+			setz(al);
+			break;
+		case 3: // setz+zu
+			puts("setz+zu");
+			setz(al|T_zu);
+			break;
+		}
+		ret();
+	}
+};
+
+int main()
+	try
+{
+	for (int mode = 0; mode < 4; mode++) {
+		Code c(mode);
+		auto f = c.getCode<int (*)()>();
+		printf("ret=%08x\n", f());
+	}
+} catch (std::exception& e) {
+	printf("ERR %s\n", e.what());
+}

xbyak/xbyak.h

@@ -155,7 +155,7 @@ namespace Xbyak {
 
 enum {
 	DEFAULT_MAX_CODE_SIZE = 4096,
-	VERSION = 0x7020 /* 0xABCD = A.BC(.D) */
+	VERSION = 0x7030 /* 0xABCD = A.BC(.D) */
 };
 
 #ifndef MIE_INTEGER_TYPE_DEFINED