Use CPUID to get information about CPU capabilities

Using golang.org/x/sys/cpu would be really great, but it makes it
hard to vendor the library with go 1.10.
This patch gets information about CPU capabilities directly from CPUID

Makefile

@@ -4,7 +4,6 @@ PRJ_DIR      = $(abspath $(dir $(MK_FILE_PATH)))
 GOPATH_LOCAL = $(PRJ_DIR)/build
 GOPATH_DIR   = github.com/cloudflare/p751sidh
 CSHAKE_PKG   ?= github.com/henrydcase/nobs/hash/sha3
-CPU_PKG      = golang.org/x/sys/cpu
 TARGETS      = p503 p751 sidh sike
 GO           ?= go
 GOARCH       ?=
@@ -32,7 +31,7 @@ clean:
 	rm -rf coverage*.txt
 
 build_env:
-	GOPATH=$(GOPATH_LOCAL) $(GO) get $(CSHAKE_PKG) $(CPU_PKG)
+	GOPATH=$(GOPATH_LOCAL) $(GO) get $(CSHAKE_PKG)
 	mkdir -p $(GOPATH_LOCAL)/src/$(GOPATH_DIR)
 	cp -rf internal $(GOPATH_LOCAL)/src/$(GOPATH_DIR)
 	cp -rf etc $(GOPATH_LOCAL)/src/$(GOPATH_DIR)

internal/utils/cpuid.go

@@ -0,0 +1,34 @@
+// +build amd64, !noasm
+
+// Sets capabilities flags for x86 according to information received from
+// CPUID. It was written in accordance with
+// "Intel® 64 and IA-32 Architectures Developer's Manual: Vol. 2A".
+// https://www.intel.com/content/www/us/en/architecture-and-technology/64-ia-32-architectures-software-developer-vol-2a-manual.html
+
+package utils
+
+// Signals support for MULX which is in BMI2
+var HasBMI2 bool
+// Signals support for MULX and BMI2
+var HasADX bool
+
+// Performs CPUID and returns values of registers
+// go:nosplit
+func cpuid(eaxArg, ecxArg uint32) (eax, ebx, ecx, edx uint32)
+
+// Returns true in case bit 'n' in 'bits' is set, otherwise false
+func bitn(bits uint32, n uint8) bool {
+    return (bits>>n)&1 == 1
+}
+
+func init() {
+    // CPUID returns max possible input that can be requested
+    max, _, _, _ := cpuid(0,0)
+    if max < 7 {
+        return
+    }
+
+    _, ebx, _, _ := cpuid(7,0)
+    HasBMI2 = bitn(ebx, 19)
+    HasADX = bitn(ebx, 7)
+}

internal/utils/cpuid.s

@@ -0,0 +1,13 @@
+// +build amd64,!noasm
+
+#include "textflag.h"
+
+TEXT ·cpuid(SB), NOSPLIT, $0-4
+    MOVL eaxArg+0(FP), AX
+    MOVL ecxArg+4(FP), CX
+    CPUID
+    MOVL AX, eax+8(FP)
+    MOVL BX, ebx+12(FP)
+    MOVL CX, ecx+16(FP)
+    MOVL DX, edx+20(FP)
+    RET

p503/arith_amd64_test.go

@@ -4,7 +4,7 @@ package p503
 
 import (
 	. "github.com/cloudflare/p751sidh/internal/isogeny"
-	"golang.org/x/sys/cpu"
+	cpu "github.com/cloudflare/p751sidh/internal/utils"
 	"reflect"
 	"testing"
 	"testing/quick"
@@ -43,7 +43,7 @@ func testRedc(t *testing.T, redcFunc func(z *FpElement, x *FpElementX2)) {
 
 // Ensures corretness of implementation of mul operation which uses MULX
 func TestMulWithMULX(t *testing.T) {
-	if !cpu.X86.HasBMI2 {
+	if !cpu.HasBMI2 {
 		t.Skip("MULX not supported by the platform")
 	}
 	testMul(t, mulWithMULX)
@@ -51,7 +51,7 @@ func TestMulWithMULX(t *testing.T) {
 
 // Ensures corretness of implementation of mul operation which uses MULX and ADOX/ADCX
 func TestMulWithMULXADX(t *testing.T) {
-	if !(cpu.X86.HasADX && cpu.X86.HasBMI2) {
+	if !(cpu.HasADX && cpu.HasBMI2) {
 		t.Skip("MULX, ADCX and ADOX not supported by the platform")
 	}
 	testMul(t, mulWithMULXADX)
@@ -59,7 +59,7 @@ func TestMulWithMULXADX(t *testing.T) {
 
 // Ensures corretness of Montgomery reduction implementation which uses MULX
 func TestRedcWithMULX(t *testing.T) {
-	if !cpu.X86.HasBMI2 {
+	if !cpu.HasBMI2 {
 		t.Skip("MULX not supported by the platform")
 	}
 	testRedc(t, redcWithMULX)
@@ -67,7 +67,7 @@ func TestRedcWithMULX(t *testing.T) {
 
 // Ensures corretness of Montgomery reduction implementation which uses MULX
 func TestRedcWithMULXADX(t *testing.T) {
-	if !(cpu.X86.HasADX && cpu.X86.HasBMI2) {
+	if !(cpu.HasADX && cpu.HasBMI2) {
 		t.Skip("MULX, ADCX and ADOX not supported by the platform")
 	}
 	testRedc(t, redcWithMULXADX)

p503/arith_decl.go

@@ -4,7 +4,7 @@ package p503
 
 import (
 	. "github.com/cloudflare/p751sidh/internal/isogeny"
-	"golang.org/x/sys/cpu"
+	cpu "github.com/cloudflare/p751sidh/internal/utils"
 )
 
 // If choice = 0, leave x,y unchanged. If choice = 1, set x,y = y,x.
@@ -74,8 +74,8 @@ func redcWithMULXADX(z *FpElement, x *FpElementX2)
 // On initialization, set the fp503Mul function pointer to the
 // fastest implementation depending on CPU capabilities.
 func init() {
-	if cpu.X86.HasBMI2 {
-		if cpu.X86.HasADX {
+	if cpu.HasBMI2 {
+		if cpu.HasADX {
 			fp503Mul = mulWithMULXADX
 			fp503MontgomeryReduce = redcWithMULXADX
 		} else {

p751/arith_amd64_test.go

@@ -3,10 +3,11 @@
 package p751
 
 import (
-	"golang.org/x/sys/cpu"
+    . "github.com/cloudflare/p751sidh/internal/isogeny"
+	cpu "github.com/cloudflare/p751sidh/internal/utils"
+
 	"testing"
 	"testing/quick"
-    . "github.com/cloudflare/p751sidh/internal/isogeny"
 )
 
 func TestFp751MontgomeryReduce(t *testing.T) {
@@ -15,7 +16,7 @@ func TestFp751MontgomeryReduce(t *testing.T) {
 	fp751MontgomeryReduce = fp751MontgomeryReduceFallback
 	t.Run("PrimeFieldElementToBigInt", TestPrimeFieldElementToBigInt)
 
-	if !cpu.X86.HasBMI2 {
+	if !cpu.HasBMI2 {
 		return
 	}
 
@@ -43,7 +44,7 @@ func TestFp751MontgomeryReduce(t *testing.T) {
 		t.Error(err)
 	}
 
-	if !cpu.X86.HasADX {
+	if !cpu.HasADX {
 		return
 	}
 

p751/arith_decl.go

@@ -3,7 +3,7 @@
 package p751
 
 import (
-	"golang.org/x/sys/cpu"
+	cpu "github.com/cloudflare/p751sidh/internal/utils"
 	. "github.com/cloudflare/p751sidh/internal/isogeny"
 )
 
@@ -59,8 +59,8 @@ func fp751StrongReduce(x *FpElement)
 // On initialization, set the fp751MontgomeryReduce function pointer to the
 // fastest implementation depending on CPU capabilities.
 func init() {
-	if cpu.X86.HasBMI2 {
-		if cpu.X86.HasADX {
+	if cpu.HasBMI2 {
+		if cpu.HasADX {
 			fp751MontgomeryReduce = fp751MontgomeryReduceBMI2ADX
 		} else {
 			fp751MontgomeryReduce = fp751MontgomeryReduceBMI2