diff options
author | Andrew Waterman <andrew@sifive.com> | 2021-12-17 11:27:19 -0800 |
---|---|---|
committer | GitHub <noreply@github.com> | 2021-12-17 11:27:19 -0800 |
commit | 6751b49ac173676c927143d6822910ac2d06a548 (patch) | |
tree | 83ce944a0166b1a823efcdfcefc345e62125c8b9 | |
parent | f1a6fb9695cffcdea9d8219c14b288c33e046716 (diff) | |
parent | f25933a48392d2141bf557fe16b3e3c56957b2d2 (diff) | |
download | spike-6751b49ac173676c927143d6822910ac2d06a548.zip spike-6751b49ac173676c927143d6822910ac2d06a548.tar.gz spike-6751b49ac173676c927143d6822910ac2d06a548.tar.bz2 |
Merge pull request #881 from eopXD/simplify-float-convert
Simplify float convert instructions
-rw-r--r-- | riscv/decode.h | 175 | ||||
-rw-r--r-- | riscv/insns/vfclass_v.h | 6 | ||||
-rw-r--r-- | riscv/insns/vfcvt_f_x_v.h | 19 | ||||
-rw-r--r-- | riscv/insns/vfcvt_f_xu_v.h | 19 | ||||
-rw-r--r-- | riscv/insns/vfcvt_rtz_x_f_v.h | 16 | ||||
-rw-r--r-- | riscv/insns/vfcvt_rtz_xu_f_v.h | 16 | ||||
-rw-r--r-- | riscv/insns/vfcvt_x_f_v.h | 16 | ||||
-rw-r--r-- | riscv/insns/vfcvt_xu_f_v.h | 16 | ||||
-rw-r--r-- | riscv/insns/vfncvt_f_f_w.h | 30 | ||||
-rw-r--r-- | riscv/insns/vfncvt_f_x_w.h | 31 | ||||
-rw-r--r-- | riscv/insns/vfncvt_f_xu_w.h | 31 | ||||
-rw-r--r-- | riscv/insns/vfncvt_rod_f_f_w.h | 38 | ||||
-rw-r--r-- | riscv/insns/vfncvt_rtz_x_f_w.h | 32 | ||||
-rw-r--r-- | riscv/insns/vfncvt_rtz_xu_f_w.h | 32 | ||||
-rw-r--r-- | riscv/insns/vfncvt_x_f_w.h | 32 | ||||
-rw-r--r-- | riscv/insns/vfncvt_xu_f_w.h | 32 | ||||
-rw-r--r-- | riscv/insns/vfwcvt_f_f_v.h | 30 | ||||
-rw-r--r-- | riscv/insns/vfwcvt_f_x_v.h | 32 | ||||
-rw-r--r-- | riscv/insns/vfwcvt_f_xu_v.h | 32 | ||||
-rw-r--r-- | riscv/insns/vfwcvt_rtz_x_f_v.h | 31 | ||||
-rw-r--r-- | riscv/insns/vfwcvt_rtz_xu_f_v.h | 31 | ||||
-rw-r--r-- | riscv/insns/vfwcvt_x_f_v.h | 31 | ||||
-rw-r--r-- | riscv/insns/vfwcvt_xu_f_v.h | 31 |
23 files changed, 322 insertions, 437 deletions
diff --git a/riscv/decode.h b/riscv/decode.h index 90025f3..02d0569 100644 --- a/riscv/decode.h +++ b/riscv/decode.h @@ -852,6 +852,18 @@ static inline bool is_aligned(const unsigned val, const unsigned pos) float##width##_t rs1 = f##width(READ_FREG(rs1_num)); \ float##width##_t vs2 = P.VU.elt<float##width##_t>(rs2_num, i); +#define CVT_FP_TO_FP_PARAMS(from_width, to_width) \ + auto vs2 = P.VU.elt<float##from_width##_t>(rs2_num, i); \ + auto &vd = P.VU.elt<float##to_width##_t>(rd_num, i, true); + +#define CVT_INT_TO_FP_PARAMS(from_width, to_width, sign) \ + auto vs2 = P.VU.elt<sign##from_width##_t>(rs2_num, i); \ + auto &vd = P.VU.elt<float##to_width##_t>(rd_num, i, true); + +#define CVT_FP_TO_INT_PARAMS(from_width, to_width, sign) \ + auto vs2 = P.VU.elt<float##from_width##_t>(rs2_num, i); \ + auto &vd = P.VU.elt<sign##to_width##_t>(rd_num, i, true); + // // vector: integer and masking operation loop // @@ -2369,39 +2381,142 @@ reg_t index[P.VU.vlmax]; \ for (reg_t i=P.VU.vstart->read(); i<vl; ++i){ \ VI_LOOP_ELEMENT_SKIP(); -#define VI_VFP_CVT_SCALE(BODY8, BODY16, BODY32, \ - CHECK8, CHECK16, CHECK32, \ - is_widen, eew_check) \ - if (is_widen) { \ - VI_CHECK_DSS(false);\ - } else { \ - VI_CHECK_SDS(false); \ - } \ - require(eew_check); \ +#define VI_VFP_CVT_LOOP(CVT_PARAMS, CHECK, BODY) \ + CHECK \ + VI_VFP_LOOP_SCALE_BASE \ + CVT_PARAMS \ + BODY \ + set_fp_exceptions; \ + VI_VFP_LOOP_END + +#define VI_VFP_CVT_INT_TO_FP(BODY16, BODY32, BODY64, sign) \ + VI_CHECK_SDS(false); \ switch(P.VU.vsew) { \ - case e8: {\ - CHECK8 \ - VI_VFP_LOOP_SCALE_BASE \ - BODY8 \ - set_fp_exceptions; \ - VI_VFP_LOOP_END \ - } \ + case e16: \ + { VI_VFP_CVT_LOOP(CVT_INT_TO_FP_PARAMS(16, 16, sign), {}, BODY16); } \ break; \ - case e16: {\ - CHECK16 \ - VI_VFP_LOOP_SCALE_BASE \ - BODY16 \ - set_fp_exceptions; \ - VI_VFP_LOOP_END \ - } \ + case e32: \ + { VI_VFP_CVT_LOOP(CVT_INT_TO_FP_PARAMS(32, 32, sign), {}, BODY32); } \ break; \ - case e32: {\ - CHECK32 \ - VI_VFP_LOOP_SCALE_BASE \ - BODY32 \ - set_fp_exceptions; \ - VI_VFP_LOOP_END \ - } \ + case e64: \ + { VI_VFP_CVT_LOOP(CVT_INT_TO_FP_PARAMS(64, 64, sign), {}, BODY64); } \ + break; \ + default: \ + require(0); \ + break; \ + } + +#define VI_VFP_CVT_FP_TO_INT(BODY16, BODY32, BODY64, sign) \ + VI_CHECK_SDS(false); \ + switch(P.VU.vsew) { \ + case e16: \ + { VI_VFP_CVT_LOOP(CVT_FP_TO_INT_PARAMS(16, 16, sign), {}, BODY16); } \ + break; \ + case e32: \ + { VI_VFP_CVT_LOOP(CVT_FP_TO_INT_PARAMS(32, 32, sign), {}, BODY32); } \ + break; \ + case e64: \ + { VI_VFP_CVT_LOOP(CVT_FP_TO_INT_PARAMS(64, 64, sign), {}, BODY64); } \ + break; \ + default: \ + require(0); \ + break; \ + } + +#define VI_VFP_WCVT_FP_TO_FP(BODY8, BODY16, BODY32, \ + CHECK8, CHECK16, CHECK32) \ + VI_CHECK_DSS(false); \ + switch(P.VU.vsew) { \ + case e16: \ + { VI_VFP_CVT_LOOP(CVT_FP_TO_FP_PARAMS(16, 32), CHECK16, BODY16); } \ + break; \ + case e32: \ + { VI_VFP_CVT_LOOP(CVT_FP_TO_FP_PARAMS(32, 64), CHECK32, BODY32); } \ + break; \ + default: \ + require(0); \ + break; \ + } + +#define VI_VFP_WCVT_INT_TO_FP(BODY8, BODY16, BODY32, \ + CHECK8, CHECK16, CHECK32, \ + sign) \ + VI_CHECK_DSS(false); \ + switch(P.VU.vsew) { \ + case e8: \ + { VI_VFP_CVT_LOOP(CVT_INT_TO_FP_PARAMS(8, 16, sign), CHECK8, BODY8); } \ + break; \ + case e16: \ + { VI_VFP_CVT_LOOP(CVT_INT_TO_FP_PARAMS(16, 32, sign), CHECK16, BODY16); } \ + break; \ + case e32: \ + { VI_VFP_CVT_LOOP(CVT_INT_TO_FP_PARAMS(32, 64, sign), CHECK32, BODY32); } \ + break; \ + default: \ + require(0); \ + break; \ + } + +#define VI_VFP_WCVT_FP_TO_INT(BODY8, BODY16, BODY32, \ + CHECK8, CHECK16, CHECK32, \ + sign) \ + VI_CHECK_DSS(false); \ + switch(P.VU.vsew) { \ + case e16: \ + { VI_VFP_CVT_LOOP(CVT_FP_TO_INT_PARAMS(16, 32, sign), CHECK16, BODY16); } \ + break; \ + case e32: \ + { VI_VFP_CVT_LOOP(CVT_FP_TO_INT_PARAMS(32, 64, sign), CHECK32, BODY32); } \ + break; \ + default: \ + require(0); \ + break; \ + } + +#define VI_VFP_NCVT_FP_TO_FP(BODY8, BODY16, BODY32, \ + CHECK8, CHECK16, CHECK32) \ + VI_CHECK_SDS(false); \ + switch(P.VU.vsew) { \ + case e16: \ + { VI_VFP_CVT_LOOP(CVT_FP_TO_FP_PARAMS(32, 16), CHECK16, BODY16); } \ + break; \ + case e32: \ + { VI_VFP_CVT_LOOP(CVT_FP_TO_FP_PARAMS(64, 32), CHECK32, BODY32); } \ + break; \ + default: \ + require(0); \ + break; \ + } + +#define VI_VFP_NCVT_INT_TO_FP(BODY8, BODY16, BODY32, \ + CHECK8, CHECK16, CHECK32, \ + sign) \ + VI_CHECK_SDS(false); \ + switch(P.VU.vsew) { \ + case e16: \ + { VI_VFP_CVT_LOOP(CVT_INT_TO_FP_PARAMS(32, 16, sign), CHECK16, BODY16); } \ + break; \ + case e32: \ + { VI_VFP_CVT_LOOP(CVT_INT_TO_FP_PARAMS(64, 32, sign), CHECK32, BODY32); } \ + break; \ + default: \ + require(0); \ + break; \ + } + +#define VI_VFP_NCVT_FP_TO_INT(BODY8, BODY16, BODY32, \ + CHECK8, CHECK16, CHECK32, \ + sign) \ + VI_CHECK_SDS(false); \ + switch(P.VU.vsew) { \ + case e8: \ + { VI_VFP_CVT_LOOP(CVT_FP_TO_INT_PARAMS(16, 8, sign), CHECK8, BODY8); } \ + break; \ + case e16: \ + { VI_VFP_CVT_LOOP(CVT_FP_TO_INT_PARAMS(32, 16, sign), CHECK16, BODY16); } \ + break; \ + case e32: \ + { VI_VFP_CVT_LOOP(CVT_FP_TO_INT_PARAMS(64, 32, sign), CHECK32, BODY32); } \ break; \ default: \ require(0); \ diff --git a/riscv/insns/vfclass_v.h b/riscv/insns/vfclass_v.h index 658f28a..a307d2d 100644 --- a/riscv/insns/vfclass_v.h +++ b/riscv/insns/vfclass_v.h @@ -1,11 +1,11 @@ // vfclass.v vd, vs2, vm VI_VFP_V_LOOP ({ - vd.v = f16_classify(vs2); + vd = f16(f16_classify(vs2)); }, { - vd.v = f32_classify(vs2); + vd = f32(f32_classify(vs2)); }, { - vd.v = f64_classify(vs2); + vd = f64(f64_classify(vs2)); }) diff --git a/riscv/insns/vfcvt_f_x_v.h b/riscv/insns/vfcvt_f_x_v.h index c53b0e1..d094c14 100644 --- a/riscv/insns/vfcvt_f_x_v.h +++ b/riscv/insns/vfcvt_f_x_v.h @@ -1,14 +1,7 @@ // vfcvt.f.x.v vd, vd2, vm -VI_VFP_VF_LOOP -({ - auto vs2_i = P.VU.elt<int16_t>(rs2_num, i); - vd = i32_to_f16(vs2_i); -}, -{ - auto vs2_i = P.VU.elt<int32_t>(rs2_num, i); - vd = i32_to_f32(vs2_i); -}, -{ - auto vs2_i = P.VU.elt<int64_t>(rs2_num, i); - vd = i64_to_f64(vs2_i); -}) +VI_VFP_CVT_INT_TO_FP( + { vd = i32_to_f16(vs2); }, // BODY16 + { vd = i32_to_f32(vs2); }, // BODY32 + { vd = i64_to_f64(vs2); }, // BODY64 + int // sign +) diff --git a/riscv/insns/vfcvt_f_xu_v.h b/riscv/insns/vfcvt_f_xu_v.h index bd03768..64dbb1c 100644 --- a/riscv/insns/vfcvt_f_xu_v.h +++ b/riscv/insns/vfcvt_f_xu_v.h @@ -1,14 +1,7 @@ // vfcvt.f.xu.v vd, vd2, vm -VI_VFP_VF_LOOP -({ - auto vs2_u = P.VU.elt<uint16_t>(rs2_num, i); - vd = ui32_to_f16(vs2_u); -}, -{ - auto vs2_u = P.VU.elt<uint32_t>(rs2_num, i); - vd = ui32_to_f32(vs2_u); -}, -{ - auto vs2_u = P.VU.elt<uint64_t>(rs2_num, i); - vd = ui64_to_f64(vs2_u); -}) +VI_VFP_CVT_INT_TO_FP( + { vd = ui32_to_f16(vs2); }, // BODY16 + { vd = ui32_to_f32(vs2); }, // BODY32 + { vd = ui64_to_f64(vs2); }, // BODY64 + uint // sign +) diff --git a/riscv/insns/vfcvt_rtz_x_f_v.h b/riscv/insns/vfcvt_rtz_x_f_v.h index e7241bd..ecdfa22 100644 --- a/riscv/insns/vfcvt_rtz_x_f_v.h +++ b/riscv/insns/vfcvt_rtz_x_f_v.h @@ -1,11 +1,7 @@ // vfcvt.rtz.x.f.v vd, vd2, vm -VI_VFP_VF_LOOP -({ - P.VU.elt<int16_t>(rd_num, i) = f16_to_i16(vs2, softfloat_round_minMag, true); -}, -{ - P.VU.elt<int32_t>(rd_num, i) = f32_to_i32(vs2, softfloat_round_minMag, true); -}, -{ - P.VU.elt<int64_t>(rd_num, i) = f64_to_i64(vs2, softfloat_round_minMag, true); -}) +VI_VFP_CVT_FP_TO_INT( + { vd = f16_to_i16(vs2, softfloat_round_minMag, true); }, // BODY16 + { vd = f32_to_i32(vs2, softfloat_round_minMag, true); }, // BODY32 + { vd = f64_to_i64(vs2, softfloat_round_minMag, true); }, // BODY64 + int // sign +) diff --git a/riscv/insns/vfcvt_rtz_xu_f_v.h b/riscv/insns/vfcvt_rtz_xu_f_v.h index d3d266d..87585d2 100644 --- a/riscv/insns/vfcvt_rtz_xu_f_v.h +++ b/riscv/insns/vfcvt_rtz_xu_f_v.h @@ -1,11 +1,7 @@ // vfcvt.rtz.xu.f.v vd, vd2, vm -VI_VFP_VF_LOOP -({ - P.VU.elt<uint16_t>(rd_num, i) = f16_to_ui16(vs2, softfloat_round_minMag, true); -}, -{ - P.VU.elt<uint32_t>(rd_num, i) = f32_to_ui32(vs2, softfloat_round_minMag, true); -}, -{ - P.VU.elt<uint64_t>(rd_num, i) = f64_to_ui64(vs2, softfloat_round_minMag, true); -}) +VI_VFP_CVT_FP_TO_INT( + { vd = f16_to_ui16(vs2, softfloat_round_minMag, true); }, // BODY16 + { vd = f32_to_ui32(vs2, softfloat_round_minMag, true); }, // BODY32 + { vd = f64_to_ui64(vs2, softfloat_round_minMag, true); }, // BODY64 + uint // sign +) diff --git a/riscv/insns/vfcvt_x_f_v.h b/riscv/insns/vfcvt_x_f_v.h index 7e507a3..4f21b52 100644 --- a/riscv/insns/vfcvt_x_f_v.h +++ b/riscv/insns/vfcvt_x_f_v.h @@ -1,11 +1,7 @@ // vfcvt.x.f.v vd, vd2, vm -VI_VFP_VF_LOOP -({ - P.VU.elt<int16_t>(rd_num, i) = f16_to_i16(vs2, STATE.frm->read(), true); -}, -{ - P.VU.elt<int32_t>(rd_num, i) = f32_to_i32(vs2, STATE.frm->read(), true); -}, -{ - P.VU.elt<int64_t>(rd_num, i) = f64_to_i64(vs2, STATE.frm->read(), true); -}) +VI_VFP_CVT_FP_TO_INT( + { vd = f16_to_i16(vs2, softfloat_roundingMode, true); }, // BODY16 + { vd = f32_to_i32(vs2, softfloat_roundingMode, true); }, // BODY32 + { vd = f64_to_i64(vs2, softfloat_roundingMode, true); }, // BODY64 + int // sign +) diff --git a/riscv/insns/vfcvt_xu_f_v.h b/riscv/insns/vfcvt_xu_f_v.h index 51c00ca..ba50fff 100644 --- a/riscv/insns/vfcvt_xu_f_v.h +++ b/riscv/insns/vfcvt_xu_f_v.h @@ -1,11 +1,7 @@ // vfcvt.xu.f.v vd, vd2, vm -VI_VFP_VV_LOOP -({ - P.VU.elt<uint16_t>(rd_num, i) = f16_to_ui16(vs2, STATE.frm->read(), true); -}, -{ - P.VU.elt<uint32_t>(rd_num, i) = f32_to_ui32(vs2, STATE.frm->read(), true); -}, -{ - P.VU.elt<uint64_t>(rd_num, i) = f64_to_ui64(vs2, STATE.frm->read(), true); -}) +VI_VFP_CVT_FP_TO_INT( + { vd = f16_to_ui16(vs2, softfloat_roundingMode, true); }, // BODY16 + { vd = f32_to_ui32(vs2, softfloat_roundingMode, true); }, // BODY32 + { vd = f64_to_ui64(vs2, softfloat_roundingMode, true); }, // BODY64 + uint // sign +) diff --git a/riscv/insns/vfncvt_f_f_w.h b/riscv/insns/vfncvt_f_f_w.h index 5448eb5..f4996f5 100644 --- a/riscv/insns/vfncvt_f_f_w.h +++ b/riscv/insns/vfncvt_f_f_w.h @@ -1,23 +1,9 @@ // vfncvt.f.f.v vd, vs2, vm -VI_VFP_CVT_SCALE -({ - ; -}, -{ - auto vs2 = P.VU.elt<float32_t>(rs2_num, i); - P.VU.elt<float16_t>(rd_num, i, true) = f32_to_f16(vs2); -}, -{ - auto vs2 = P.VU.elt<float64_t>(rs2_num, i); - P.VU.elt<float32_t>(rd_num, i, true) = f64_to_f32(vs2); -}, -{ - ; -}, -{ - require(p->extension_enabled(EXT_ZFH)); -}, -{ - require(p->extension_enabled('D')); -}, -false, (P.VU.vsew >= 16)) +VI_VFP_NCVT_FP_TO_FP( + {;}, // BODY16 + { vd = f32_to_f16(vs2); }, // BODY32 + { vd = f64_to_f32(vs2); }, // BODY64 + {;}, // CHECK16 + { require_extension(EXT_ZFH); }, // CHECK32 + { require_extension('D'); } // CHECK64 +) diff --git a/riscv/insns/vfncvt_f_x_w.h b/riscv/insns/vfncvt_f_x_w.h index 10a6f7b..d587be2 100644 --- a/riscv/insns/vfncvt_f_x_w.h +++ b/riscv/insns/vfncvt_f_x_w.h @@ -1,23 +1,10 @@ // vfncvt.f.x.v vd, vs2, vm -VI_VFP_CVT_SCALE -({ - ; -}, -{ - auto vs2 = P.VU.elt<int32_t>(rs2_num, i); - P.VU.elt<float16_t>(rd_num, i, true) = i32_to_f16(vs2); -}, -{ - auto vs2 = P.VU.elt<int64_t>(rs2_num, i); - P.VU.elt<float32_t>(rd_num, i, true) = i64_to_f32(vs2); -}, -{ - ; -}, -{ - require(p->extension_enabled(EXT_ZFH)); -}, -{ - require(p->extension_enabled('F')); -}, -false, (P.VU.vsew >= 16)) +VI_VFP_NCVT_INT_TO_FP( + {;}, // BODY16 + { vd = i32_to_f16(vs2); }, // BODY32 + { vd = i64_to_f32(vs2); }, // BODY64 + {;}, // CHECK16 + { require_extension(EXT_ZFH); }, // CHECK32 + { require_extension('F'); }, // CHECK64 + int // sign +) diff --git a/riscv/insns/vfncvt_f_xu_w.h b/riscv/insns/vfncvt_f_xu_w.h index 32b4b02..5e0e34f 100644 --- a/riscv/insns/vfncvt_f_xu_w.h +++ b/riscv/insns/vfncvt_f_xu_w.h @@ -1,23 +1,10 @@ // vfncvt.f.xu.v vd, vs2, vm -VI_VFP_CVT_SCALE -({ - ; -}, -{ - auto vs2 = P.VU.elt<uint32_t>(rs2_num, i); - P.VU.elt<float16_t>(rd_num, i, true) = ui32_to_f16(vs2); -}, -{ - auto vs2 = P.VU.elt<uint64_t>(rs2_num, i); - P.VU.elt<float32_t>(rd_num, i, true) = ui64_to_f32(vs2); -}, -{ - ; -}, -{ - require(p->extension_enabled(EXT_ZFH)); -}, -{ - require(p->extension_enabled('F')); -}, -false, (P.VU.vsew >= 16)) +VI_VFP_NCVT_INT_TO_FP( + {;}, // BODY16 + { vd = ui32_to_f16(vs2); }, // BODY32 + { vd = ui64_to_f32(vs2); }, // BODY64 + {;}, // CHECK16 + { require_extension(EXT_ZFH); }, // CHECK32 + { require_extension('F'); }, // CHECK64 + uint // sign +) diff --git a/riscv/insns/vfncvt_rod_f_f_w.h b/riscv/insns/vfncvt_rod_f_f_w.h index 20a14c9..89bdc05 100644 --- a/riscv/insns/vfncvt_rod_f_f_w.h +++ b/riscv/insns/vfncvt_rod_f_f_w.h @@ -1,25 +1,15 @@ // vfncvt.rod.f.f.v vd, vs2, vm -VI_VFP_CVT_SCALE -({ - ; -}, -{ - softfloat_roundingMode = softfloat_round_odd; - auto vs2 = P.VU.elt<float32_t>(rs2_num, i); - P.VU.elt<float16_t>(rd_num, i, true) = f32_to_f16(vs2); -}, -{ - softfloat_roundingMode = softfloat_round_odd; - auto vs2 = P.VU.elt<float64_t>(rs2_num, i); - P.VU.elt<float32_t>(rd_num, i, true) = f64_to_f32(vs2); -}, -{ - ; -}, -{ - require(p->extension_enabled(EXT_ZFH)); -}, -{ - require(p->extension_enabled('F')); -}, -false, (P.VU.vsew >= 16)) +VI_VFP_NCVT_FP_TO_FP( + {;}, // BODY16 + { // BODY32 + softfloat_roundingMode = softfloat_round_odd; + vd = f32_to_f16(vs2); + }, + { // BODY64 + softfloat_roundingMode = softfloat_round_odd; + vd = f64_to_f32(vs2); + }, + {;}, // CHECK16 + { require_extension(EXT_ZFH); }, // CHECK32 + { require_extension('F'); } // CHECK64 +) diff --git a/riscv/insns/vfncvt_rtz_x_f_w.h b/riscv/insns/vfncvt_rtz_x_f_w.h index 0629b8d..23b4d5e 100644 --- a/riscv/insns/vfncvt_rtz_x_f_w.h +++ b/riscv/insns/vfncvt_rtz_x_f_w.h @@ -1,24 +1,10 @@ // vfncvt.rtz.x.f.w vd, vs2, vm -VI_VFP_CVT_SCALE -({ - auto vs2 = P.VU.elt<float16_t>(rs2_num, i); - P.VU.elt<int8_t>(rd_num, i, true) = f16_to_i8(vs2, softfloat_round_minMag, true); -}, -{ - auto vs2 = P.VU.elt<float32_t>(rs2_num, i); - P.VU.elt<int16_t>(rd_num, i, true) = f32_to_i16(vs2, softfloat_round_minMag, true); -}, -{ - auto vs2 = P.VU.elt<float64_t>(rs2_num, i); - P.VU.elt<int32_t>(rd_num, i, true) = f64_to_i32(vs2, softfloat_round_minMag, true); -}, -{ - require(p->extension_enabled(EXT_ZFH)); -}, -{ - require(p->extension_enabled('F')); -}, -{ - require(p->extension_enabled('D')); -}, -false, (P.VU.vsew <= 32)) +VI_VFP_NCVT_FP_TO_INT( + { vd = f16_to_i8(vs2, softfloat_round_minMag, true); }, // BODY16 + { vd = f32_to_i16(vs2, softfloat_round_minMag, true); }, // BODY32 + { vd = f64_to_i32(vs2, softfloat_round_minMag, true); }, // BODY64 + { require_extension(EXT_ZFH); }, // CHECK16 + { require(p->extension_enabled('F')); }, // CHECK32 + { require(p->extension_enabled('D')); }, // CHECK64 + int // sign +) diff --git a/riscv/insns/vfncvt_rtz_xu_f_w.h b/riscv/insns/vfncvt_rtz_xu_f_w.h index 82aa63e..f55c680 100644 --- a/riscv/insns/vfncvt_rtz_xu_f_w.h +++ b/riscv/insns/vfncvt_rtz_xu_f_w.h @@ -1,24 +1,10 @@ // vfncvt.rtz.xu.f.w vd, vs2, vm -VI_VFP_CVT_SCALE -({ - auto vs2 = P.VU.elt<float16_t>(rs2_num, i); - P.VU.elt<uint8_t>(rd_num, i, true) = f16_to_ui8(vs2, softfloat_round_minMag, true); -}, -{ - auto vs2 = P.VU.elt<float32_t>(rs2_num, i); - P.VU.elt<uint16_t>(rd_num, i, true) = f32_to_ui16(vs2, softfloat_round_minMag, true); -}, -{ - auto vs2 = P.VU.elt<float64_t>(rs2_num, i); - P.VU.elt<uint32_t>(rd_num, i, true) = f64_to_ui32(vs2, softfloat_round_minMag, true); -}, -{ - require(p->extension_enabled(EXT_ZFH)); -}, -{ - require(p->extension_enabled('F')); -}, -{ - require(p->extension_enabled('D')); -}, -false, (P.VU.vsew <= 32)) +VI_VFP_NCVT_FP_TO_INT( + { vd = f16_to_ui8(vs2, softfloat_round_minMag, true); }, // BODY16 + { vd = f32_to_ui16(vs2, softfloat_round_minMag, true); }, // BODY32 + { vd = f64_to_ui32(vs2, softfloat_round_minMag, true); }, // BODY64 + { require_extension(EXT_ZFH); }, // CHECK16 + { require(p->extension_enabled('F')); }, // CHECK32 + { require(p->extension_enabled('D')); }, // CHECK64 + uint // sign +) diff --git a/riscv/insns/vfncvt_x_f_w.h b/riscv/insns/vfncvt_x_f_w.h index a8a6dfb..a7f3c33 100644 --- a/riscv/insns/vfncvt_x_f_w.h +++ b/riscv/insns/vfncvt_x_f_w.h @@ -1,24 +1,10 @@ // vfncvt.x.f.w vd, vs2, vm -VI_VFP_CVT_SCALE -({ - auto vs2 = P.VU.elt<float16_t>(rs2_num, i); - P.VU.elt<int8_t>(rd_num, i, true) = f16_to_i8(vs2, STATE.frm->read(), true); -}, -{ - auto vs2 = P.VU.elt<float32_t>(rs2_num, i); - P.VU.elt<int16_t>(rd_num, i, true) = f32_to_i16(vs2, STATE.frm->read(), true); -}, -{ - auto vs2 = P.VU.elt<float64_t>(rs2_num, i); - P.VU.elt<int32_t>(rd_num, i, true) = f64_to_i32(vs2, STATE.frm->read(), true); -}, -{ - require(p->extension_enabled(EXT_ZFH)); -}, -{ - require(p->extension_enabled('F')); -}, -{ - require(p->extension_enabled('D')); -}, -false, (P.VU.vsew <= 32)) +VI_VFP_NCVT_FP_TO_INT( + { vd = f16_to_i8(vs2, softfloat_roundingMode, true); }, // BODY16 + { vd = f32_to_i16(vs2, softfloat_roundingMode, true); }, // BODY32 + { vd = f64_to_i32(vs2, softfloat_roundingMode, true); }, // BODY64 + { require_extension(EXT_ZFH); }, // CHECK16 + { require(p->extension_enabled('F')); }, // CHECK32 + { require(p->extension_enabled('D')); }, // CHECK64 + int // sign +) diff --git a/riscv/insns/vfncvt_xu_f_w.h b/riscv/insns/vfncvt_xu_f_w.h index bff733e..02046e8 100644 --- a/riscv/insns/vfncvt_xu_f_w.h +++ b/riscv/insns/vfncvt_xu_f_w.h @@ -1,24 +1,10 @@ // vfncvt.xu.f.w vd, vs2, vm -VI_VFP_CVT_SCALE -({ - auto vs2 = P.VU.elt<float16_t>(rs2_num, i); - P.VU.elt<uint8_t>(rd_num, i, true) = f16_to_ui8(vs2, STATE.frm->read(), true); -}, -{ - auto vs2 = P.VU.elt<float32_t>(rs2_num, i); - P.VU.elt<uint16_t>(rd_num, i, true) = f32_to_ui16(vs2, STATE.frm->read(), true); -}, -{ - auto vs2 = P.VU.elt<float64_t>(rs2_num, i); - P.VU.elt<uint32_t>(rd_num, i, true) = f64_to_ui32(vs2, STATE.frm->read(), true); -}, -{ - require(p->extension_enabled(EXT_ZFH)); -}, -{ - require(p->extension_enabled('F')); -}, -{ - require(p->extension_enabled('D')); -}, -false, (P.VU.vsew <= 32)) +VI_VFP_NCVT_FP_TO_INT( + { vd = f16_to_ui8(vs2, softfloat_roundingMode, true); }, // BODY16 + { vd = f32_to_ui16(vs2, softfloat_roundingMode, true); }, // BODY32 + { vd = f64_to_ui32(vs2, softfloat_roundingMode, true); }, // BODY64 + { require_extension(EXT_ZFH); }, // CHECK16 + { require(p->extension_enabled('F')); }, // CHECK32 + { require(p->extension_enabled('D')); }, // CHECK64 + uint // sign +) diff --git a/riscv/insns/vfwcvt_f_f_v.h b/riscv/insns/vfwcvt_f_f_v.h index fcaf65c..0700070 100644 --- a/riscv/insns/vfwcvt_f_f_v.h +++ b/riscv/insns/vfwcvt_f_f_v.h @@ -1,23 +1,9 @@ // vfwcvt.f.f.v vd, vs2, vm -VI_VFP_CVT_SCALE -({ - ; -}, -{ - auto vs2 = P.VU.elt<float16_t>(rs2_num, i); - P.VU.elt<float32_t>(rd_num, i, true) = f16_to_f32(vs2); -}, -{ - auto vs2 = P.VU.elt<float32_t>(rs2_num, i); - P.VU.elt<float64_t>(rd_num, i, true) = f32_to_f64(vs2); -}, -{ - ; -}, -{ - require(p->extension_enabled(EXT_ZFH)); -}, -{ - require(p->extension_enabled('D')); -}, -true, (P.VU.vsew >= 16)) +VI_VFP_WCVT_FP_TO_FP( + {;}, // BODY8 + { vd = f16_to_f32(vs2); }, // BODY16 + { vd = f32_to_f64(vs2); }, // BODY32 + {;}, // CHECK8 + { require_extension(EXT_ZFH); }, // CHECK16 + { require_extension('D'); } // CHECK32 +) diff --git a/riscv/insns/vfwcvt_f_x_v.h b/riscv/insns/vfwcvt_f_x_v.h index 8d8283c..f51e8e3 100644 --- a/riscv/insns/vfwcvt_f_x_v.h +++ b/riscv/insns/vfwcvt_f_x_v.h @@ -1,24 +1,10 @@ // vfwcvt.f.x.v vd, vs2, vm -VI_VFP_CVT_SCALE -({ - auto vs2 = P.VU.elt<int8_t>(rs2_num, i); - P.VU.elt<float16_t>(rd_num, i, true) = i32_to_f16(vs2); -}, -{ - auto vs2 = P.VU.elt<int16_t>(rs2_num, i); - P.VU.elt<float32_t>(rd_num, i, true) = i32_to_f32(vs2); -}, -{ - auto vs2 = P.VU.elt<int32_t>(rs2_num, i); - P.VU.elt<float64_t>(rd_num, i, true) = i32_to_f64(vs2); -}, -{ - require(p->extension_enabled(EXT_ZFH)); -}, -{ - require(p->extension_enabled('F')); -}, -{ - require(p->extension_enabled('D')); -}, -true, (P.VU.vsew >= 8)) +VI_VFP_WCVT_INT_TO_FP( + { vd = i32_to_f16(vs2); }, // BODY8 + { vd = i32_to_f32(vs2); }, // BODY16 + { vd = i32_to_f64(vs2); }, // BODY32 + { require(p->extension_enabled(EXT_ZFH)); }, // CHECK8 + { require_extension('F'); }, // CHECK16 + { require_extension('D'); }, // CHECK32 + int // sign +) diff --git a/riscv/insns/vfwcvt_f_xu_v.h b/riscv/insns/vfwcvt_f_xu_v.h index e8036ce..7dd4972 100644 --- a/riscv/insns/vfwcvt_f_xu_v.h +++ b/riscv/insns/vfwcvt_f_xu_v.h @@ -1,24 +1,10 @@ // vfwcvt.f.xu.v vd, vs2, vm -VI_VFP_CVT_SCALE -({ - auto vs2 = P.VU.elt<uint8_t>(rs2_num, i); - P.VU.elt<float16_t>(rd_num, i, true) = ui32_to_f16(vs2); -}, -{ - auto vs2 = P.VU.elt<uint16_t>(rs2_num, i); - P.VU.elt<float32_t>(rd_num, i, true) = ui32_to_f32(vs2); -}, -{ - auto vs2 = P.VU.elt<uint32_t>(rs2_num, i); - P.VU.elt<float64_t>(rd_num, i, true) = ui32_to_f64(vs2); -}, -{ - require(p->extension_enabled(EXT_ZFH)); -}, -{ - require(p->extension_enabled('F')); -}, -{ - require(p->extension_enabled('D')); -}, -true, (P.VU.vsew >= 8)) +VI_VFP_WCVT_INT_TO_FP( + { vd = ui32_to_f16(vs2); }, // BODY8 + { vd = ui32_to_f32(vs2); }, // BODY16 + { vd = ui32_to_f64(vs2); }, // BODY32 + { require(p->extension_enabled(EXT_ZFH)); }, // CHECK8 + { require_extension('F'); }, // CHECK16 + { require_extension('D'); }, // CHECK32 + uint // sign +) diff --git a/riscv/insns/vfwcvt_rtz_x_f_v.h b/riscv/insns/vfwcvt_rtz_x_f_v.h index 83fa764..74e5b9a 100644 --- a/riscv/insns/vfwcvt_rtz_x_f_v.h +++ b/riscv/insns/vfwcvt_rtz_x_f_v.h @@ -1,23 +1,10 @@ // vfwcvt.rtz.x.f.v vd, vs2, vm -VI_VFP_CVT_SCALE -({ - ; -}, -{ - auto vs2 = P.VU.elt<float16_t>(rs2_num, i); - P.VU.elt<int32_t>(rd_num, i, true) = f16_to_i32(vs2, softfloat_round_minMag, true); -}, -{ - auto vs2 = P.VU.elt<float32_t>(rs2_num, i); - P.VU.elt<int64_t>(rd_num, i, true) = f32_to_i64(vs2, softfloat_round_minMag, true); -}, -{ - ; -}, -{ - require(p->extension_enabled(EXT_ZFH)); -}, -{ - require(p->extension_enabled('F')); -}, -true, (P.VU.vsew >= 16)) +VI_VFP_WCVT_FP_TO_INT( + {;}, // BODY8 + { vd = f16_to_i32(vs2, softfloat_round_minMag, true); }, // BODY16 + { vd = f32_to_i64(vs2, softfloat_round_minMag, true); }, // BODY32 + {;}, // CHECK8 + { require_extension(EXT_ZFH); }, // CHECK16 + { require_extension('F'); }, // CHECK32 + int // sign +) diff --git a/riscv/insns/vfwcvt_rtz_xu_f_v.h b/riscv/insns/vfwcvt_rtz_xu_f_v.h index 43d1979..72b8c6e 100644 --- a/riscv/insns/vfwcvt_rtz_xu_f_v.h +++ b/riscv/insns/vfwcvt_rtz_xu_f_v.h @@ -1,23 +1,10 @@ // vfwcvt.rtz,xu.f.v vd, vs2, vm -VI_VFP_CVT_SCALE -({ - ; -}, -{ - auto vs2 = P.VU.elt<float16_t>(rs2_num, i); - P.VU.elt<uint32_t>(rd_num, i, true) = f16_to_ui32(vs2, softfloat_round_minMag, true); -}, -{ - auto vs2 = P.VU.elt<float32_t>(rs2_num, i); - P.VU.elt<uint64_t>(rd_num, i, true) = f32_to_ui64(vs2, softfloat_round_minMag, true); -}, -{ - ; -}, -{ - require(p->extension_enabled(EXT_ZFH)); -}, -{ - require(p->extension_enabled('F')); -}, -true, (P.VU.vsew >= 16)) +VI_VFP_WCVT_FP_TO_INT( + {;}, // BODY8 + { vd = f16_to_ui32(vs2, softfloat_round_minMag, true); }, // BODY16 + { vd = f32_to_ui64(vs2, softfloat_round_minMag, true); }, // BODY32 + {;}, // CHECK8 + { require_extension(EXT_ZFH); }, // CHECK16 + { require_extension('F'); }, // CHECK32 + uint // sign +) diff --git a/riscv/insns/vfwcvt_x_f_v.h b/riscv/insns/vfwcvt_x_f_v.h index 5e0c064..74497f4 100644 --- a/riscv/insns/vfwcvt_x_f_v.h +++ b/riscv/insns/vfwcvt_x_f_v.h @@ -1,23 +1,10 @@ // vfwcvt.x.f.v vd, vs2, vm -VI_VFP_CVT_SCALE -({ - ; -}, -{ - auto vs2 = P.VU.elt<float16_t>(rs2_num, i); - P.VU.elt<int32_t>(rd_num, i, true) = f16_to_i32(vs2, STATE.frm->read(), true); -}, -{ - auto vs2 = P.VU.elt<float32_t>(rs2_num, i); - P.VU.elt<int64_t>(rd_num, i, true) = f32_to_i64(vs2, STATE.frm->read(), true); -}, -{ - ; -}, -{ - require(p->extension_enabled(EXT_ZFH)); -}, -{ - require(p->extension_enabled('F')); -}, -true, (P.VU.vsew >= 16)) +VI_VFP_WCVT_FP_TO_INT( + {;}, // BODY8 + { vd = f16_to_i32(vs2, softfloat_roundingMode, true); }, // BODY16 + { vd = f32_to_i64(vs2, softfloat_roundingMode, true); }, // BODY32 + {;}, // CHECK8 + { require_extension(EXT_ZFH); }, // CHECK16 + { require_extension('F'); }, // CHECK32 + int // sign +) diff --git a/riscv/insns/vfwcvt_xu_f_v.h b/riscv/insns/vfwcvt_xu_f_v.h index f3243c8..ad96c9c 100644 --- a/riscv/insns/vfwcvt_xu_f_v.h +++ b/riscv/insns/vfwcvt_xu_f_v.h @@ -1,23 +1,10 @@ // vfwcvt.xu.f.v vd, vs2, vm -VI_VFP_CVT_SCALE -({ - ; -}, -{ - auto vs2 = P.VU.elt<float16_t>(rs2_num, i); - P.VU.elt<uint32_t>(rd_num, i, true) = f16_to_ui32(vs2, STATE.frm->read(), true); -}, -{ - auto vs2 = P.VU.elt<float32_t>(rs2_num, i); - P.VU.elt<uint64_t>(rd_num, i, true) = f32_to_ui64(vs2, STATE.frm->read(), true); -}, -{ - ; -}, -{ - require(p->extension_enabled(EXT_ZFH)); -}, -{ - require(p->extension_enabled('F')); -}, -true, (P.VU.vsew >= 16)) +VI_VFP_WCVT_FP_TO_INT( + {;}, // BODY8 + { vd = f16_to_ui32(vs2, softfloat_roundingMode, true); }, // BODY16 + { vd = f32_to_ui64(vs2, softfloat_roundingMode, true); }, // BODY32 + {;}, // CHECK8 + { require_extension(EXT_ZFH); }, // CHECK16 + { require_extension('F'); }, // CHECK32 + uint // sign +) |