[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[Qemu-devel] [PATCH 2/8] target-arm: Create and use neon_unarrow_sat* he
From: |
christophe.lyon |
Subject: |
[Qemu-devel] [PATCH 2/8] target-arm: Create and use neon_unarrow_sat* helpers |
Date: |
Mon, 31 Jan 2011 19:06:40 +0100 |
From: Christophe Lyon <address@hidden>
Fix VQMOVUN, improve VQSHRUN and VQRSHRUN.
Signed-off-by: Christophe Lyon <address@hidden>
---
target-arm/helpers.h | 3 ++
target-arm/neon_helper.c | 63 ++++++++++++++++++++++++++++++++++++++++++++++
target-arm/translate.c | 43 ++++++++++++++++++++++---------
3 files changed, 96 insertions(+), 13 deletions(-)
diff --git a/target-arm/helpers.h b/target-arm/helpers.h
index b88ebae..8cc6a44 100644
--- a/target-arm/helpers.h
+++ b/target-arm/helpers.h
@@ -295,10 +295,13 @@ DEF_HELPER_3(neon_qrdmulh_s32, i32, env, i32, i32)
DEF_HELPER_1(neon_narrow_u8, i32, i64)
DEF_HELPER_1(neon_narrow_u16, i32, i64)
+DEF_HELPER_2(neon_unarrow_sat8, i32, env, i64)
DEF_HELPER_2(neon_narrow_sat_u8, i32, env, i64)
DEF_HELPER_2(neon_narrow_sat_s8, i32, env, i64)
+DEF_HELPER_2(neon_unarrow_sat16, i32, env, i64)
DEF_HELPER_2(neon_narrow_sat_u16, i32, env, i64)
DEF_HELPER_2(neon_narrow_sat_s16, i32, env, i64)
+DEF_HELPER_2(neon_unarrow_sat32, i32, env, i64)
DEF_HELPER_2(neon_narrow_sat_u32, i32, env, i64)
DEF_HELPER_2(neon_narrow_sat_s32, i32, env, i64)
DEF_HELPER_1(neon_narrow_high_u8, i32, i64)
diff --git a/target-arm/neon_helper.c b/target-arm/neon_helper.c
index 6c832b4..891b812 100644
--- a/target-arm/neon_helper.c
+++ b/target-arm/neon_helper.c
@@ -1005,6 +1005,33 @@ uint32_t HELPER(neon_narrow_round_high_u16)(uint64_t x)
return ((x >> 16) & 0xffff) | ((x >> 32) & 0xffff0000);
}
+uint32_t HELPER(neon_unarrow_sat8)(CPUState *env, uint64_t x)
+{
+ uint16_t s;
+ uint8_t d;
+ uint32_t res = 0;
+#define SAT8(n) \
+ s = x >> n; \
+ if (s & 0x8000) { \
+ SET_QC(); \
+ } else { \
+ if (s > 0xff) { \
+ d = 0xff; \
+ SET_QC(); \
+ } else { \
+ d = s; \
+ } \
+ res |= (uint32_t)d << (n / 2); \
+ }
+
+ SAT8(0);
+ SAT8(16);
+ SAT8(32);
+ SAT8(48);
+#undef SAT8
+ return res;
+}
+
uint32_t HELPER(neon_narrow_sat_u8)(CPUState *env, uint64_t x)
{
uint16_t s;
@@ -1051,6 +1078,29 @@ uint32_t HELPER(neon_narrow_sat_s8)(CPUState *env,
uint64_t x)
return res;
}
+uint32_t HELPER(neon_unarrow_sat16)(CPUState *env, uint64_t x)
+{
+ uint32_t high;
+ uint32_t low;
+ low = x;
+ if (low & 0x80000000) {
+ low = 0;
+ SET_QC();
+ } else if (low > 0xffff) {
+ low = 0xffff;
+ SET_QC();
+ }
+ high = x >> 32;
+ if (high & 0x80000000) {
+ high = 0;
+ SET_QC();
+ } else if (high > 0xffff) {
+ high = 0xffff;
+ SET_QC();
+ }
+ return low | (high << 16);
+}
+
uint32_t HELPER(neon_narrow_sat_u16)(CPUState *env, uint64_t x)
{
uint32_t high;
@@ -1085,6 +1135,19 @@ uint32_t HELPER(neon_narrow_sat_s16)(CPUState *env,
uint64_t x)
return (uint16_t)low | (high << 16);
}
+uint32_t HELPER(neon_unarrow_sat32)(CPUState *env, uint64_t x)
+{
+ if (x & 0x8000000000000000ull) {
+ SET_QC();
+ return 0;
+ }
+ if (x > 0xffffffffu) {
+ SET_QC();
+ return 0xffffffffu;
+ }
+ return x;
+}
+
uint32_t HELPER(neon_narrow_sat_u32)(CPUState *env, uint64_t x)
{
if (x > 0xffffffffu) {
diff --git a/target-arm/translate.c b/target-arm/translate.c
index b44f7a1..6dd024d 100644
--- a/target-arm/translate.c
+++ b/target-arm/translate.c
@@ -4071,6 +4071,16 @@ static inline void gen_neon_narrow_satu(int size, TCGv
dest, TCGv_i64 src)
}
}
+static inline void gen_neon_unarrow_sats(int size, TCGv dest, TCGv_i64 src)
+{
+ switch(size) {
+ case 0: gen_helper_neon_unarrow_sat8(dest, cpu_env, src); break;
+ case 1: gen_helper_neon_unarrow_sat16(dest, cpu_env, src); break;
+ case 2: gen_helper_neon_unarrow_sat32(dest, cpu_env, src); break;
+ default: abort();
+ }
+}
+
static inline void gen_neon_shift_narrow(int size, TCGv var, TCGv shift,
int q, int u)
{
@@ -4841,13 +4851,14 @@ static int disas_neon_data_insn(CPUState * env,
DisasContext *s, uint32_t insn)
dead_tmp(tmp3);
}
tmp = new_tmp();
- if (op == 8 && !u) {
- gen_neon_narrow(size - 1, tmp, cpu_V0);
- } else {
- if (op == 8)
- gen_neon_narrow_sats(size - 1, tmp, cpu_V0);
- else
- gen_neon_narrow_satu(size - 1, tmp, cpu_V0);
+ if (op == 8) {
+ if (u) { /* VQSHRUN / VQRSHRUN */
+ gen_neon_unarrow_sats(size - 1, tmp, cpu_V0);
+ } else { /* VSHRN / VRSHRN */
+ gen_neon_narrow(size - 1, tmp, cpu_V0);
+ }
+ } else { /* VQSHRN / VQRSHRN */
+ gen_neon_narrow_satu(size - 1, tmp, cpu_V0);
}
neon_store_reg(rd, pass, tmp);
} /* for pass */
@@ -5471,12 +5482,18 @@ static int disas_neon_data_insn(CPUState * env,
DisasContext *s, uint32_t insn)
for (pass = 0; pass < 2; pass++) {
neon_load_reg64(cpu_V0, rm + pass);
tmp = new_tmp();
- if (op == 36 && q == 0) {
- gen_neon_narrow(size, tmp, cpu_V0);
- } else if (q) {
- gen_neon_narrow_satu(size, tmp, cpu_V0);
- } else {
- gen_neon_narrow_sats(size, tmp, cpu_V0);
+ if (op == 36) {
+ if (q) { /* VQMOVUN */
+ gen_neon_unarrow_sats(size, tmp, cpu_V0);
+ } else { /* VMOVN */
+ gen_neon_narrow(size, tmp, cpu_V0);
+ }
+ } else { /* VQMOVN */
+ if (q) {
+ gen_neon_narrow_satu(size, tmp, cpu_V0);
+ } else {
+ gen_neon_narrow_sats(size, tmp, cpu_V0);
+ }
}
if (pass == 0) {
tmp2 = tmp;
--
1.7.2.3
- [Qemu-devel] [PATCH v2 0/8] target-arm: Fix Neon instructions VQMOVUN VQRSHL VQRSHRN VQRSHRUN VQSHRN VQSHRUN VSLI VSRI, christophe.lyon, 2011/01/31
- [Qemu-devel] [PATCH 1/8] target-arm: Fixes for several shift instructions: VRSHL, VRSHR, VRSHRN, VSHLL, VRSRA., christophe.lyon, 2011/01/31
- [Qemu-devel] [PATCH 4/8] target-arm: fiddle decoding of 64 bit shift by imm and narrow, christophe.lyon, 2011/01/31
- [Qemu-devel] [PATCH 2/8] target-arm: Create and use neon_unarrow_sat* helpers,
christophe.lyon <=
- [Qemu-devel] [PATCH 8/8] target-arm: Fix VQRSHL Neon instructions (signed/unsigned 64 bits and signed 32 bits variants)., christophe.lyon, 2011/01/31
- [Qemu-devel] [PATCH 6/8] target-arm: Fix Neon VQ(R)SHRN instructions., christophe.lyon, 2011/01/31
- [Qemu-devel] [PATCH 5/8] target-arm: fix neon vqrshl instruction, christophe.lyon, 2011/01/31
- [Qemu-devel] [PATCH 3/8] target-arm: VQRSHRN related changes, christophe.lyon, 2011/01/31
- [Qemu-devel] [PATCH 7/8] target-arm: implement vsli.64, vsri.64, christophe.lyon, 2011/01/31