X86: Implement the media instructions that convert fp values to ints.

5 files changed, 225 insertions, 29 deletions

diff --git a/src/arch/x86/isa/decoder/two_byte_opcodes.isa b/src/arch/x86/isa/decoder/two_byte_opcodes.isa
index 11fc92abc..c670390ad 100644
--- a/src/arch/x86/isa/decoder/two_byte_opcodes.isa
+++ b/src/arch/x86/isa/decoder/two_byte_opcodes.isa
@@ -377,16 +377,16 @@
                     0x1: Inst::MOVAPS(Wq,Vq);
                     0x2: Inst::CVTPI2PS(Vq,Qq);
                     0x3: movntps_Mo_Vo();
-                    0x4: cvttps2pi_Pq_Wq();
-                    0x5: cvtpS2pi_Pq_Wq();
+                    0x4: Inst::CVTTPS2PI(Pq,Wq);
+                    0x5: Inst::CVTPS2PI(Pq,Wq);
                     0x6: Inst::UCOMISS(Vd,Wd);
                     0x7: Inst::COMISS(Vd,Wd);
                 }
                 // repe (0xF3)
                 0x4: decode OPCODE_OP_BOTTOM3 {
                     0x2: Inst::CVTSI2SS(Vd,Ed);
-                    0x4: cvttss2si_Gd_Wd();
-                    0x5: cvtss2si_Gd_Wd();
+                    0x4: Inst::CVTTSS2SI(Gd,Wd);
+                    0x5: Inst::CVTSS2SI(Gd,Wd);
                     default: Inst::UD2();
                 }
                 // operand size (0x66)
@@ -395,8 +395,8 @@
                     0x1: Inst::MOVAPD(Wo,Vo);
                     0x2: Inst::CVTPI2PD(Vo,Qq);
                     0x3: movntpd_Mo_Vo();
-                    0x4: cvttpd2pi_Pq_Wo();
-                    0x5: cvtpd2pi_Pq_Wo();
+                    0x4: Inst::CVTTPD2PI(Pq,Wo);
+                    0x5: Inst::CVTPD2PI(Pq,Wo);
                     0x6: Inst::UCOMISD(Vq,Wq);
                     0x7: Inst::COMISD(Vq,Wq);
                 }
@@ -406,7 +406,7 @@
                     0x2: Inst::CVTSI2SD(Vdp,Edp);
                     // The size of the W operand should be q, not dp
                     0x4: Inst::CVTTSD2SI(Gdp,Wdp);
-                    0x5: cvtsd2si_Gd_Wq();
+                    0x5: Inst::CVTSD2SI(Gd,Wq);
                     default: Inst::UD2();
                 }
                 default: Inst::UD2();
@@ -511,7 +511,7 @@
                         0x0: ADDSS(Vd,Wd);
                         0x1: MULSS(Vd,Wd);
                         0x2: CVTSS2SD(Vq,Wd);
-                        0x3: WarnUnimpl::cvttps2dq_Vo_Wo();
+                        0x3: CVTTPS2DQ(Vo,Wo);
                         0x4: SUBSS(Vd,Wd);
                         0x5: MINSS(Vd,Wd);
                         0x6: DIVSS(Vd,Wd);
@@ -522,7 +522,7 @@
                         0x0: ADDPD(Vo,Wo);
                         0x1: MULPD(Vo,Wo);
                         0x2: CVTPD2PS(Vo,Wo);
-                        0x3: WarnUnimpl::cvtps2dq_Vo_Wo();
+                        0x3: CVTPS2DQ(Vo,Wo);
                         0x4: SUBPD(Vo,Wo);
                         0x5: MINPD(Vo,Wo);
                         0x6: DIVPD(Vo,Wo);
@@ -1023,12 +1023,12 @@
                         0x3: PAVGW(Vo,Wo);
                         0x4: PMULHUW(Vo,Wo);
                         0x5: PMULHW(Vo,Wo);
-                        0x6: WarnUnimpl::cvttpd2dq_Vo_Wo();
+                        0x6: CVTTPD2DQ(Vo,Wo);
                         0x7: WarnUnimpl::movntdq_Mo_Vo();
                     }
                     // repne (0xF2)
                     0x8: decode OPCODE_OP_BOTTOM3 {
-                        0x6: WarnUnimpl::cvtpd2dq_Vo_Wo();
+                        0x6: CVTPD2DQ(Vo,Wo);
                         default: UD2();
                     }
                     default: UD2();
diff --git a/src/arch/x86/isa/insts/simd128/floating_point/data_conversion/convert_floating_point_to_gpr_integer.py b/src/arch/x86/isa/insts/simd128/floating_point/data_conversion/convert_floating_point_to_gpr_integer.py
index 413026d63..16abd96f4 100644
--- a/src/arch/x86/isa/insts/simd128/floating_point/data_conversion/convert_floating_point_to_gpr_integer.py
+++ b/src/arch/x86/isa/insts/simd128/floating_point/data_conversion/convert_floating_point_to_gpr_integer.py
@@ -54,22 +54,75 @@
 # Authors: Gabe Black
 
 microcode = '''
-# CVTSS2SI
-# CVTSD2SI
-# CVTTSS2SI
+def macroop CVTSS2SI_R_XMM {
+    cvtf2i ufp1, xmmlm, srcSize=4, destSize=dsz, ext=(1 | 4)
+    mov2int reg, ufp1, size=dsz
+};
+
+def macroop CVTSS2SI_R_M {
+    ldfp ufp1, seg, sib, disp, dataSize=8
+    cvtf2i ufp1, ufp1, srcSize=4, destSize=dsz, ext=(1 | 4)
+    mov2int reg, ufp1, size=dsz
+};
+
+def macroop CVTSS2SI_R_P {
+    rdip t7
+    ldfp ufp1, seg, riprel, disp, dataSize=8
+    cvtf2i ufp1, ufp1, srcSize=4, destSize=dsz, ext=(1 | 4)
+    mov2int reg, ufp1, size=dsz
+};
+
+def macroop CVTSD2SI_R_XMM {
+    cvtf2i ufp1, xmmlm, srcSize=8, destSize=dsz, ext=(1 | 4)
+    mov2int reg, ufp1, size=dsz
+};
+
+def macroop CVTSD2SI_R_M {
+    ldfp ufp1, seg, sib, disp, dataSize=8
+    cvtf2i ufp1, ufp1, srcSize=8, destSize=dsz, ext=(1 | 4)
+    mov2int reg, ufp1, size=dsz
+};
+
+def macroop CVTSD2SI_R_P {
+    rdip t7
+    ldfp ufp1, seg, riprel, disp, dataSize=8
+    cvtf2i ufp1, ufp1, srcSize=8, destSize=dsz, ext=(1 | 4)
+    mov2int reg, ufp1, size=dsz
+};
+
+def macroop CVTTSS2SI_R_XMM {
+    cvtf2i ufp1, xmmlm, srcSize=4, destSize=dsz, ext=1
+    mov2int reg, ufp1, size=dsz
+};
+
+def macroop CVTTSS2SI_R_M {
+    ldfp ufp1, seg, sib, disp, dataSize=8
+    cvtf2i ufp1, ufp1, srcSize=4, destSize=dsz, ext=1
+    mov2int reg, ufp1, size=dsz
+};
+
+def macroop CVTTSS2SI_R_P {
+    rdip t7
+    ldfp ufp1, seg, riprel, disp, dataSize=8
+    cvtf2i ufp1, ufp1, srcSize=4, destSize=dsz, ext=1
+    mov2int reg, ufp1, size=dsz
+};
 
 def macroop CVTTSD2SI_R_XMM {
-    cvtf_d2i reg, xmmlm
+    cvtf2i ufp1, xmmlm, srcSize=8, destSize=dsz, ext=1
+    mov2int reg, ufp1, size=dsz
 };
 
 def macroop CVTTSD2SI_R_M {
     ldfp ufp1, seg, sib, disp, dataSize=8
-    cvtf_d2i reg, ufp1
+    cvtf2i ufp1, ufp1, srcSize=8, destSize=dsz, ext=1
+    mov2int reg, ufp1, size=dsz
 };
 
 def macroop CVTTSD2SI_R_P {
     rdip t7
     ldfp ufp1, seg, riprel, disp, dataSize=8
-    cvtf_d2i reg, ufp1
+    cvtf2i ufp1, ufp1, srcSize=8, destSize=dsz, ext=1
+    mov2int reg, ufp1, size=dsz
 };
 '''
diff --git a/src/arch/x86/isa/insts/simd128/floating_point/data_conversion/convert_floating_point_to_mmx_integer.py b/src/arch/x86/isa/insts/simd128/floating_point/data_conversion/convert_floating_point_to_mmx_integer.py
index af579a46f..900e91c99 100644
--- a/src/arch/x86/isa/insts/simd128/floating_point/data_conversion/convert_floating_point_to_mmx_integer.py
+++ b/src/arch/x86/isa/insts/simd128/floating_point/data_conversion/convert_floating_point_to_mmx_integer.py
@@ -54,8 +54,73 @@
 # Authors: Gabe Black
 
 microcode = '''
-# CVTPS2PI
-# CVTPD2PI
-# CVTTPS2PI
-# CVTTPD2PI
+def macroop CVTPS2PI_MMX_XMM {
+    cvtf2i mmx, xmmlm, size=4, ext=4
+};
+
+def macroop CVTPS2PI_MMX_M {
+    ldfp ufp1, seg, sib, disp, dataSize=8
+    cvtf2i mmx, ufp1, size=4, ext=4
+};
+
+def macroop CVTPS2PI_MMX_P {
+    rdip t7
+    ldfp ufp1, seg, riprel, disp, dataSize=8
+    cvtf2i mmx, ufp1, size=4, ext=4
+};
+
+def macroop CVTPD2PI_MMX_XMM {
+    cvtf2i mmx, xmmlm, srcSize=8, destSize=4, ext=4
+    cvtf2i mmx, xmmhm, srcSize=8, destSize=4, ext=(4 | 2)
+};
+
+def macroop CVTPD2PI_MMX_M {
+    ldfp ufp1, seg, sib, "DISPLACEMENT", dataSize=8
+    ldfp ufp2, seg, sib, "DISPLACEMENT + 8", dataSize=8
+    cvtf2i mmx, ufp1, srcSize=8, destSize=4, ext=4
+    cvtf2i mmx, ufp2, srcSize=8, destSize=4, ext=(4 | 2)
+};
+
+def macroop CVTPD2PI_MMX_P {
+    rdip t7
+    ldfp ufp1, seg, riprel, "DISPLACEMENT", dataSize=8
+    ldfp ufp2, seg, riprel, "DISPLACEMENT + 8", dataSize=8
+    cvtf2i mmx, ufp1, srcSize=8, destSize=4, ext=4
+    cvtf2i mmx, ufp2, srcSize=8, destSize=4, ext=(4 | 2)
+};
+
+def macroop CVTTPS2PI_MMX_XMM {
+    cvtf2i mmx, xmmlm, size=4, ext=0
+};
+
+def macroop CVTTPS2PI_MMX_M {
+    ldfp ufp1, seg, sib, disp, dataSize=8
+    cvtf2i mmx, ufp1, size=4, ext=0
+};
+
+def macroop CVTTPS2PI_MMX_P {
+    rdip t7
+    ldfp ufp1, seg, riprel, disp, dataSize=8
+    cvtf2i mmx, ufp1, size=4, ext=0
+};
+
+def macroop CVTTPD2PI_MMX_XMM {
+    cvtf2i mmx, xmmlm, srcSize=8, destSize=4, ext=0
+    cvtf2i mmx, xmmhm, srcSize=8, destSize=4, ext=2
+};
+
+def macroop CVTTPD2PI_MMX_M {
+    ldfp ufp1, seg, sib, "DISPLACEMENT", dataSize=8
+    ldfp ufp2, seg, sib, "DISPLACEMENT + 8", dataSize=8
+    cvtf2i mmx, ufp1, srcSize=8, destSize=4, ext=0
+    cvtf2i mmx, ufp2, srcSize=8, destSize=4, ext=2
+};
+
+def macroop CVTTPD2PI_MMX_P {
+    rdip t7
+    ldfp ufp1, seg, riprel, "DISPLACEMENT", dataSize=8
+    ldfp ufp2, seg, riprel, "DISPLACEMENT + 8", dataSize=8
+    cvtf2i mmx, ufp1, srcSize=8, destSize=4, ext=0
+    cvtf2i mmx, ufp2, srcSize=8, destSize=4, ext=2
+};
 '''
diff --git a/src/arch/x86/isa/insts/simd128/floating_point/data_conversion/convert_floating_point_to_xmm_integer.py b/src/arch/x86/isa/insts/simd128/floating_point/data_conversion/convert_floating_point_to_xmm_integer.py
index 683e2808d..041f891ef 100644
--- a/src/arch/x86/isa/insts/simd128/floating_point/data_conversion/convert_floating_point_to_xmm_integer.py
+++ b/src/arch/x86/isa/insts/simd128/floating_point/data_conversion/convert_floating_point_to_xmm_integer.py
@@ -54,8 +54,89 @@
 # Authors: Gabe Black
 
 microcode = '''
-# CVTPS2DQ
-# CVTPD2DQ
-# CVTTPS2DQ
-# CVTTPD2DQ
+def macroop CVTPS2DQ_XMM_XMM {
+    cvtf2i xmml, xmmlm, size=4, ext=4
+    cvtf2i xmmh, xmmhm, size=4, ext=4
+};
+
+def macroop CVTPS2DQ_XMM_M {
+    ldfp ufp1, seg, sib, "DISPLACEMENT", dataSize=8
+    ldfp ufp2, seg, sib, "DISPLACEMENT + 8", dataSize=8
+    cvtf2i xmml, ufp1, size=4, ext=4
+    cvtf2i xmmh, ufp2, size=4, ext=4
+};
+
+def macroop CVTPS2DQ_XMM_P {
+    rdip t7
+    ldfp ufp1, seg, riprel, "DISPLACEMENT", dataSize=8
+    ldfp ufp2, seg, riprel, "DISPLACEMENT + 8", dataSize=8
+    cvtf2i xmml, ufp1, size=4, ext=4
+    cvtf2i xmmh, ufp2, size=4, ext=4
+};
+
+def macroop CVTPD2DQ_XMM_XMM {
+    cvtf2i xmml, xmmlm, srcSize=8, destSize=4, ext=4
+    cvtf2i xmml, xmmhm, srcSize=8, destSize=4, ext=(4 | 2)
+    lfpimm xmmh, 0
+};
+
+def macroop CVTPD2DQ_XMM_M {
+    ldfp ufp1, seg, sib, "DISPLACEMENT", dataSize=8
+    ldfp ufp2, seg, sib, "DISPLACEMENT + 8", dataSize=8
+    cvtf2i xmml, ufp1, srcSize=8, destSize=4, ext=4
+    cvtf2i xmml, ufp2, srcSize=8, destSize=4, ext=(4 | 2)
+    lfpimm xmmh, 0
+};
+
+def macroop CVTPD2DQ_XMM_P {
+    rdip t7
+    ldfp ufp1, seg, riprel, "DISPLACEMENT", dataSize=8
+    ldfp ufp2, seg, riprel, "DISPLACEMENT + 8", dataSize=8
+    cvtf2i xmml, ufp1, srcSize=8, destSize=4, ext=4
+    cvtf2i xmml, ufp2, srcSize=8, destSize=4, ext=(4 | 2)
+    lfpimm xmmh, 0
+};
+
+def macroop CVTTPS2DQ_XMM_XMM {
+    cvtf2i xmml, xmmlm, size=4, ext=0
+    cvtf2i xmmh, xmmhm, size=4, ext=0
+};
+
+def macroop CVTTPS2DQ_XMM_M {
+    ldfp ufp1, seg, sib, "DISPLACEMENT", dataSize=8
+    ldfp ufp2, seg, sib, "DISPLACEMENT + 8", dataSize=8
+    cvtf2i xmml, ufp1, size=4, ext=0
+    cvtf2i xmmh, ufp2, size=4, ext=0
+};
+
+def macroop CVTTPS2DQ_XMM_P {
+    rdip t7
+    ldfp ufp1, seg, riprel, "DISPLACEMENT", dataSize=8
+    ldfp ufp2, seg, riprel, "DISPLACEMENT + 8", dataSize=8
+    cvtf2i xmml, ufp1, size=4, ext=0
+    cvtf2i xmmh, ufp2, size=4, ext=0
+};
+
+def macroop CVTTPD2DQ_XMM_XMM {
+    cvtf2i xmml, xmmlm, srcSize=8, destSize=4, ext=0
+    cvtf2i xmml, xmmhm, srcSize=8, destSize=4, ext=2
+    lfpimm xmmh, 0
+};
+
+def macroop CVTTPD2DQ_XMM_M {
+    ldfp ufp1, seg, sib, "DISPLACEMENT", dataSize=8
+    ldfp ufp2, seg, sib, "DISPLACEMENT + 8", dataSize=8
+    cvtf2i xmml, ufp1, srcSize=8, destSize=4, ext=0
+    cvtf2i xmml, ufp2, srcSize=8, destSize=4, ext=2
+    lfpimm xmmh, 0
+};
+
+def macroop CVTTPD2DQ_XMM_P {
+    rdip t7
+    ldfp ufp1, seg, riprel, "DISPLACEMENT", dataSize=8
+    ldfp ufp2, seg, riprel, "DISPLACEMENT + 8", dataSize=8
+    cvtf2i xmml, ufp1, srcSize=8, destSize=4, ext=0
+    cvtf2i xmml, ufp2, srcSize=8, destSize=4, ext=2
+    lfpimm xmmh, 0
+};
 '''
diff --git a/src/arch/x86/isa/insts/simd64/floating_point/data_conversion.py b/src/arch/x86/isa/insts/simd64/floating_point/data_conversion.py
index 6c942065d..1ade48152 100644
--- a/src/arch/x86/isa/insts/simd64/floating_point/data_conversion.py
+++ b/src/arch/x86/isa/insts/simd64/floating_point/data_conversion.py
@@ -54,10 +54,7 @@
 # Authors: Gabe Black
 
 microcode = '''
-# CVTPS2PI
-# CVTTPS2PI
-# CVTPD2PI
-# CVTTPD2PI
+# CVTPS2PI, CVTTPS2PI, CVTPD2PI, and CVTTPD2PI are implemented in simd128
 # PF2IW
 # PF2ID
 '''