From: "Alex Bennée" <alex.bennee@linaro.org>
To: Richard Henderson <richard.henderson@linaro.org>
Cc: qemu-devel@nongnu.org
Subject: Re: [PATCH v2 13/36] tcg: Use tcg_constant_{i32,i64} with tcg int expanders
Date: Wed, 22 Apr 2020 21:04:51 +0100 [thread overview]
Message-ID: <87ftcvb82k.fsf@linaro.org> (raw)
In-Reply-To: <20200422011722.13287-14-richard.henderson@linaro.org>
Richard Henderson <richard.henderson@linaro.org> writes:
> Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
We have a regression. Setting up a build dir with:
../../configure --disable-tools --disable-docs --target-list=sparc-softmmu,sparc64-softmmu
make -j30 && make check-acceptance
And then running a bisect between HEAD and master:
git bisect run /bin/sh -c "cd builds/bisect && make -j30 && ./tests/venv/bin/avocado run ./tests/acceptance/boot_linux_console.py:BootLinuxConsole.test_sparc_ss20"
Fingers:
a4d42b76dd29818e4f393c4c3eb59601b0015b2f is the first bad commit
commit a4d42b76dd29818e4f393c4c3eb59601b0015b2f
Author: Richard Henderson <richard.henderson@linaro.org>
Date: Tue Apr 21 18:16:59 2020 -0700
tcg: Use tcg_constant_{i32,i64} with tcg int expanders
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20200422011722.13287-14-richard.henderson@linaro.org>
:040000 040000 45283ae0961f2794f5f15e09c29f160372fb5fae 92939e91645a5cf4fc36d475ff5dddd0839a7314 M include
:040000 040000 1083f94f8f045924fbf1e1f9c116f05827c25345 31a5dfc97636fcd0a114b910095b11cb767a22db M tcg
bisect run success
> ---
> include/tcg/tcg-op.h | 13 +--
> tcg/tcg-op.c | 216 ++++++++++++++++++++-----------------------
> 2 files changed, 100 insertions(+), 129 deletions(-)
>
> diff --git a/include/tcg/tcg-op.h b/include/tcg/tcg-op.h
> index 230db6e022..11ed9192f7 100644
> --- a/include/tcg/tcg-op.h
> +++ b/include/tcg/tcg-op.h
> @@ -271,6 +271,7 @@ void tcg_gen_mb(TCGBar);
>
> /* 32 bit ops */
>
> +void tcg_gen_movi_i32(TCGv_i32 ret, int32_t arg);
> void tcg_gen_addi_i32(TCGv_i32 ret, TCGv_i32 arg1, int32_t arg2);
> void tcg_gen_subfi_i32(TCGv_i32 ret, int32_t arg1, TCGv_i32 arg2);
> void tcg_gen_subi_i32(TCGv_i32 ret, TCGv_i32 arg1, int32_t arg2);
> @@ -349,11 +350,6 @@ static inline void tcg_gen_mov_i32(TCGv_i32 ret, TCGv_i32 arg)
> }
> }
>
> -static inline void tcg_gen_movi_i32(TCGv_i32 ret, int32_t arg)
> -{
> - tcg_gen_op2i_i32(INDEX_op_movi_i32, ret, arg);
> -}
> -
> static inline void tcg_gen_ld8u_i32(TCGv_i32 ret, TCGv_ptr arg2,
> tcg_target_long offset)
> {
> @@ -467,6 +463,7 @@ static inline void tcg_gen_not_i32(TCGv_i32 ret, TCGv_i32 arg)
>
> /* 64 bit ops */
>
> +void tcg_gen_movi_i64(TCGv_i64 ret, int64_t arg);
> void tcg_gen_addi_i64(TCGv_i64 ret, TCGv_i64 arg1, int64_t arg2);
> void tcg_gen_subfi_i64(TCGv_i64 ret, int64_t arg1, TCGv_i64 arg2);
> void tcg_gen_subi_i64(TCGv_i64 ret, TCGv_i64 arg1, int64_t arg2);
> @@ -550,11 +547,6 @@ static inline void tcg_gen_mov_i64(TCGv_i64 ret, TCGv_i64 arg)
> }
> }
>
> -static inline void tcg_gen_movi_i64(TCGv_i64 ret, int64_t arg)
> -{
> - tcg_gen_op2i_i64(INDEX_op_movi_i64, ret, arg);
> -}
> -
> static inline void tcg_gen_ld8u_i64(TCGv_i64 ret, TCGv_ptr arg2,
> tcg_target_long offset)
> {
> @@ -698,7 +690,6 @@ static inline void tcg_gen_sub_i64(TCGv_i64 ret, TCGv_i64 arg1, TCGv_i64 arg2)
>
> void tcg_gen_discard_i64(TCGv_i64 arg);
> void tcg_gen_mov_i64(TCGv_i64 ret, TCGv_i64 arg);
> -void tcg_gen_movi_i64(TCGv_i64 ret, int64_t arg);
> void tcg_gen_ld8u_i64(TCGv_i64 ret, TCGv_ptr arg2, tcg_target_long offset);
> void tcg_gen_ld8s_i64(TCGv_i64 ret, TCGv_ptr arg2, tcg_target_long offset);
> void tcg_gen_ld16u_i64(TCGv_i64 ret, TCGv_ptr arg2, tcg_target_long offset);
> diff --git a/tcg/tcg-op.c b/tcg/tcg-op.c
> index e2e25ebf7d..07eb661a07 100644
> --- a/tcg/tcg-op.c
> +++ b/tcg/tcg-op.c
> @@ -104,15 +104,18 @@ void tcg_gen_mb(TCGBar mb_type)
>
> /* 32 bit ops */
>
> +void tcg_gen_movi_i32(TCGv_i32 ret, int32_t arg)
> +{
> + tcg_gen_mov_i32(ret, tcg_constant_i32(arg));
> +}
> +
> void tcg_gen_addi_i32(TCGv_i32 ret, TCGv_i32 arg1, int32_t arg2)
> {
> /* some cases can be optimized here */
> if (arg2 == 0) {
> tcg_gen_mov_i32(ret, arg1);
> } else {
> - TCGv_i32 t0 = tcg_const_i32(arg2);
> - tcg_gen_add_i32(ret, arg1, t0);
> - tcg_temp_free_i32(t0);
> + tcg_gen_add_i32(ret, arg1, tcg_constant_i32(arg2));
> }
> }
>
> @@ -122,9 +125,7 @@ void tcg_gen_subfi_i32(TCGv_i32 ret, int32_t arg1, TCGv_i32 arg2)
> /* Don't recurse with tcg_gen_neg_i32. */
> tcg_gen_op2_i32(INDEX_op_neg_i32, ret, arg2);
> } else {
> - TCGv_i32 t0 = tcg_const_i32(arg1);
> - tcg_gen_sub_i32(ret, t0, arg2);
> - tcg_temp_free_i32(t0);
> + tcg_gen_sub_i32(ret, tcg_constant_i32(arg1), arg2);
> }
> }
>
> @@ -134,15 +135,12 @@ void tcg_gen_subi_i32(TCGv_i32 ret, TCGv_i32 arg1, int32_t arg2)
> if (arg2 == 0) {
> tcg_gen_mov_i32(ret, arg1);
> } else {
> - TCGv_i32 t0 = tcg_const_i32(arg2);
> - tcg_gen_sub_i32(ret, arg1, t0);
> - tcg_temp_free_i32(t0);
> + tcg_gen_sub_i32(ret, arg1, tcg_constant_i32(arg2));
> }
> }
>
> void tcg_gen_andi_i32(TCGv_i32 ret, TCGv_i32 arg1, int32_t arg2)
> {
> - TCGv_i32 t0;
> /* Some cases can be optimized here. */
> switch (arg2) {
> case 0:
> @@ -165,9 +163,8 @@ void tcg_gen_andi_i32(TCGv_i32 ret, TCGv_i32 arg1, int32_t arg2)
> }
> break;
> }
> - t0 = tcg_const_i32(arg2);
> - tcg_gen_and_i32(ret, arg1, t0);
> - tcg_temp_free_i32(t0);
> +
> + tcg_gen_and_i32(ret, arg1, tcg_constant_i32(arg2));
> }
>
> void tcg_gen_ori_i32(TCGv_i32 ret, TCGv_i32 arg1, int32_t arg2)
> @@ -178,9 +175,7 @@ void tcg_gen_ori_i32(TCGv_i32 ret, TCGv_i32 arg1, int32_t arg2)
> } else if (arg2 == 0) {
> tcg_gen_mov_i32(ret, arg1);
> } else {
> - TCGv_i32 t0 = tcg_const_i32(arg2);
> - tcg_gen_or_i32(ret, arg1, t0);
> - tcg_temp_free_i32(t0);
> + tcg_gen_or_i32(ret, arg1, tcg_constant_i32(arg2));
> }
> }
>
> @@ -193,9 +188,7 @@ void tcg_gen_xori_i32(TCGv_i32 ret, TCGv_i32 arg1, int32_t arg2)
> /* Don't recurse with tcg_gen_not_i32. */
> tcg_gen_op2_i32(INDEX_op_not_i32, ret, arg1);
> } else {
> - TCGv_i32 t0 = tcg_const_i32(arg2);
> - tcg_gen_xor_i32(ret, arg1, t0);
> - tcg_temp_free_i32(t0);
> + tcg_gen_xor_i32(ret, arg1, tcg_constant_i32(arg2));
> }
> }
>
> @@ -205,9 +198,7 @@ void tcg_gen_shli_i32(TCGv_i32 ret, TCGv_i32 arg1, int32_t arg2)
> if (arg2 == 0) {
> tcg_gen_mov_i32(ret, arg1);
> } else {
> - TCGv_i32 t0 = tcg_const_i32(arg2);
> - tcg_gen_shl_i32(ret, arg1, t0);
> - tcg_temp_free_i32(t0);
> + tcg_gen_shl_i32(ret, arg1, tcg_constant_i32(arg2));
> }
> }
>
> @@ -217,9 +208,7 @@ void tcg_gen_shri_i32(TCGv_i32 ret, TCGv_i32 arg1, int32_t arg2)
> if (arg2 == 0) {
> tcg_gen_mov_i32(ret, arg1);
> } else {
> - TCGv_i32 t0 = tcg_const_i32(arg2);
> - tcg_gen_shr_i32(ret, arg1, t0);
> - tcg_temp_free_i32(t0);
> + tcg_gen_shr_i32(ret, arg1, tcg_constant_i32(arg2));
> }
> }
>
> @@ -229,9 +218,7 @@ void tcg_gen_sari_i32(TCGv_i32 ret, TCGv_i32 arg1, int32_t arg2)
> if (arg2 == 0) {
> tcg_gen_mov_i32(ret, arg1);
> } else {
> - TCGv_i32 t0 = tcg_const_i32(arg2);
> - tcg_gen_sar_i32(ret, arg1, t0);
> - tcg_temp_free_i32(t0);
> + tcg_gen_sar_i32(ret, arg1, tcg_constant_i32(arg2));
> }
> }
>
> @@ -250,9 +237,7 @@ void tcg_gen_brcondi_i32(TCGCond cond, TCGv_i32 arg1, int32_t arg2, TCGLabel *l)
> if (cond == TCG_COND_ALWAYS) {
> tcg_gen_br(l);
> } else if (cond != TCG_COND_NEVER) {
> - TCGv_i32 t0 = tcg_const_i32(arg2);
> - tcg_gen_brcond_i32(cond, arg1, t0, l);
> - tcg_temp_free_i32(t0);
> + tcg_gen_brcond_i32(cond, arg1, tcg_constant_i32(arg2), l);
> }
> }
>
> @@ -271,9 +256,7 @@ void tcg_gen_setcond_i32(TCGCond cond, TCGv_i32 ret,
> void tcg_gen_setcondi_i32(TCGCond cond, TCGv_i32 ret,
> TCGv_i32 arg1, int32_t arg2)
> {
> - TCGv_i32 t0 = tcg_const_i32(arg2);
> - tcg_gen_setcond_i32(cond, ret, arg1, t0);
> - tcg_temp_free_i32(t0);
> + tcg_gen_setcond_i32(cond, ret, arg1, tcg_constant_i32(arg2));
> }
>
> void tcg_gen_muli_i32(TCGv_i32 ret, TCGv_i32 arg1, int32_t arg2)
> @@ -283,9 +266,7 @@ void tcg_gen_muli_i32(TCGv_i32 ret, TCGv_i32 arg1, int32_t arg2)
> } else if (is_power_of_2(arg2)) {
> tcg_gen_shli_i32(ret, arg1, ctz32(arg2));
> } else {
> - TCGv_i32 t0 = tcg_const_i32(arg2);
> - tcg_gen_mul_i32(ret, arg1, t0);
> - tcg_temp_free_i32(t0);
> + tcg_gen_mul_i32(ret, arg1, tcg_constant_i32(arg2));
> }
> }
>
> @@ -433,9 +414,7 @@ void tcg_gen_clz_i32(TCGv_i32 ret, TCGv_i32 arg1, TCGv_i32 arg2)
>
> void tcg_gen_clzi_i32(TCGv_i32 ret, TCGv_i32 arg1, uint32_t arg2)
> {
> - TCGv_i32 t = tcg_const_i32(arg2);
> - tcg_gen_clz_i32(ret, arg1, t);
> - tcg_temp_free_i32(t);
> + tcg_gen_clz_i32(ret, arg1, tcg_constant_i32(arg2));
> }
>
> void tcg_gen_ctz_i32(TCGv_i32 ret, TCGv_i32 arg1, TCGv_i32 arg2)
> @@ -468,10 +447,9 @@ void tcg_gen_ctz_i32(TCGv_i32 ret, TCGv_i32 arg1, TCGv_i32 arg2)
> tcg_gen_clzi_i32(t, t, 32);
> tcg_gen_xori_i32(t, t, 31);
> }
> - z = tcg_const_i32(0);
> + z = tcg_constant_i32(0);
> tcg_gen_movcond_i32(TCG_COND_EQ, ret, arg1, z, arg2, t);
> tcg_temp_free_i32(t);
> - tcg_temp_free_i32(z);
> } else {
> gen_helper_ctz_i32(ret, arg1, arg2);
> }
> @@ -487,9 +465,7 @@ void tcg_gen_ctzi_i32(TCGv_i32 ret, TCGv_i32 arg1, uint32_t arg2)
> tcg_gen_ctpop_i32(ret, t);
> tcg_temp_free_i32(t);
> } else {
> - TCGv_i32 t = tcg_const_i32(arg2);
> - tcg_gen_ctz_i32(ret, arg1, t);
> - tcg_temp_free_i32(t);
> + tcg_gen_ctz_i32(ret, arg1, tcg_constant_i32(arg2));
> }
> }
>
> @@ -547,9 +523,7 @@ void tcg_gen_rotli_i32(TCGv_i32 ret, TCGv_i32 arg1, unsigned arg2)
> if (arg2 == 0) {
> tcg_gen_mov_i32(ret, arg1);
> } else if (TCG_TARGET_HAS_rot_i32) {
> - TCGv_i32 t0 = tcg_const_i32(arg2);
> - tcg_gen_rotl_i32(ret, arg1, t0);
> - tcg_temp_free_i32(t0);
> + tcg_gen_rotl_i32(ret, arg1, tcg_constant_i32(arg2));
> } else {
> TCGv_i32 t0, t1;
> t0 = tcg_temp_new_i32();
> @@ -653,9 +627,8 @@ void tcg_gen_deposit_z_i32(TCGv_i32 ret, TCGv_i32 arg,
> tcg_gen_andi_i32(ret, arg, (1u << len) - 1);
> } else if (TCG_TARGET_HAS_deposit_i32
> && TCG_TARGET_deposit_i32_valid(ofs, len)) {
> - TCGv_i32 zero = tcg_const_i32(0);
> + TCGv_i32 zero = tcg_constant_i32(0);
> tcg_gen_op5ii_i32(INDEX_op_deposit_i32, ret, zero, arg, ofs, len);
> - tcg_temp_free_i32(zero);
> } else {
> /* To help two-operand hosts we prefer to zero-extend first,
> which allows ARG to stay live. */
> @@ -1052,7 +1025,7 @@ void tcg_gen_bswap32_i32(TCGv_i32 ret, TCGv_i32 arg)
> } else {
> TCGv_i32 t0 = tcg_temp_new_i32();
> TCGv_i32 t1 = tcg_temp_new_i32();
> - TCGv_i32 t2 = tcg_const_i32(0x00ff00ff);
> + TCGv_i32 t2 = tcg_constant_i32(0x00ff00ff);
>
> /* arg = abcd */
> tcg_gen_shri_i32(t0, arg, 8); /* t0 = .abc */
> @@ -1067,7 +1040,6 @@ void tcg_gen_bswap32_i32(TCGv_i32 ret, TCGv_i32 arg)
>
> tcg_temp_free_i32(t0);
> tcg_temp_free_i32(t1);
> - tcg_temp_free_i32(t2);
> }
> }
>
> @@ -1237,6 +1209,14 @@ void tcg_gen_mul_i64(TCGv_i64 ret, TCGv_i64 arg1, TCGv_i64 arg2)
> tcg_temp_free_i64(t0);
> tcg_temp_free_i32(t1);
> }
> +
> +#else
> +
> +void tcg_gen_movi_i64(TCGv_i64 ret, int64_t arg)
> +{
> + tcg_gen_mov_i64(ret, tcg_constant_i64(arg));
> +}
> +
> #endif /* TCG_TARGET_REG_SIZE == 32 */
>
> void tcg_gen_addi_i64(TCGv_i64 ret, TCGv_i64 arg1, int64_t arg2)
> @@ -1244,10 +1224,12 @@ void tcg_gen_addi_i64(TCGv_i64 ret, TCGv_i64 arg1, int64_t arg2)
> /* some cases can be optimized here */
> if (arg2 == 0) {
> tcg_gen_mov_i64(ret, arg1);
> + } else if (TCG_TARGET_REG_BITS == 64) {
> + tcg_gen_add_i64(ret, arg1, tcg_constant_i64(arg2));
> } else {
> - TCGv_i64 t0 = tcg_const_i64(arg2);
> - tcg_gen_add_i64(ret, arg1, t0);
> - tcg_temp_free_i64(t0);
> + tcg_gen_add2_i32(TCGV_LOW(ret), TCGV_HIGH(ret),
> + TCGV_LOW(arg1), TCGV_HIGH(arg1),
> + tcg_constant_i32(arg2), tcg_constant_i32(arg2 >> 32));
> }
> }
>
> @@ -1256,10 +1238,12 @@ void tcg_gen_subfi_i64(TCGv_i64 ret, int64_t arg1, TCGv_i64 arg2)
> if (arg1 == 0 && TCG_TARGET_HAS_neg_i64) {
> /* Don't recurse with tcg_gen_neg_i64. */
> tcg_gen_op2_i64(INDEX_op_neg_i64, ret, arg2);
> + } else if (TCG_TARGET_REG_BITS == 64) {
> + tcg_gen_sub_i64(ret, tcg_constant_i64(arg1), arg2);
> } else {
> - TCGv_i64 t0 = tcg_const_i64(arg1);
> - tcg_gen_sub_i64(ret, t0, arg2);
> - tcg_temp_free_i64(t0);
> + tcg_gen_sub2_i32(TCGV_LOW(ret), TCGV_HIGH(ret),
> + tcg_constant_i32(arg1), tcg_constant_i32(arg1 >> 32),
> + TCGV_LOW(arg2), TCGV_HIGH(arg2));
> }
> }
>
> @@ -1268,17 +1252,17 @@ void tcg_gen_subi_i64(TCGv_i64 ret, TCGv_i64 arg1, int64_t arg2)
> /* some cases can be optimized here */
> if (arg2 == 0) {
> tcg_gen_mov_i64(ret, arg1);
> + } else if (TCG_TARGET_REG_BITS == 64) {
> + tcg_gen_sub_i64(ret, arg1, tcg_constant_i64(arg2));
> } else {
> - TCGv_i64 t0 = tcg_const_i64(arg2);
> - tcg_gen_sub_i64(ret, arg1, t0);
> - tcg_temp_free_i64(t0);
> + tcg_gen_sub2_i32(TCGV_LOW(ret), TCGV_HIGH(ret),
> + TCGV_LOW(arg1), TCGV_HIGH(arg1),
> + tcg_constant_i32(arg2), tcg_constant_i32(arg2 >> 32));
> }
> }
>
> void tcg_gen_andi_i64(TCGv_i64 ret, TCGv_i64 arg1, int64_t arg2)
> {
> - TCGv_i64 t0;
> -
> if (TCG_TARGET_REG_BITS == 32) {
> tcg_gen_andi_i32(TCGV_LOW(ret), TCGV_LOW(arg1), arg2);
> tcg_gen_andi_i32(TCGV_HIGH(ret), TCGV_HIGH(arg1), arg2 >> 32);
> @@ -1313,9 +1297,8 @@ void tcg_gen_andi_i64(TCGv_i64 ret, TCGv_i64 arg1, int64_t arg2)
> }
> break;
> }
> - t0 = tcg_const_i64(arg2);
> - tcg_gen_and_i64(ret, arg1, t0);
> - tcg_temp_free_i64(t0);
> +
> + tcg_gen_and_i64(ret, arg1, tcg_constant_i64(arg2));
> }
>
> void tcg_gen_ori_i64(TCGv_i64 ret, TCGv_i64 arg1, int64_t arg2)
> @@ -1331,9 +1314,7 @@ void tcg_gen_ori_i64(TCGv_i64 ret, TCGv_i64 arg1, int64_t arg2)
> } else if (arg2 == 0) {
> tcg_gen_mov_i64(ret, arg1);
> } else {
> - TCGv_i64 t0 = tcg_const_i64(arg2);
> - tcg_gen_or_i64(ret, arg1, t0);
> - tcg_temp_free_i64(t0);
> + tcg_gen_or_i64(ret, arg1, tcg_constant_i64(arg2));
> }
> }
>
> @@ -1351,9 +1332,7 @@ void tcg_gen_xori_i64(TCGv_i64 ret, TCGv_i64 arg1, int64_t arg2)
> /* Don't recurse with tcg_gen_not_i64. */
> tcg_gen_op2_i64(INDEX_op_not_i64, ret, arg1);
> } else {
> - TCGv_i64 t0 = tcg_const_i64(arg2);
> - tcg_gen_xor_i64(ret, arg1, t0);
> - tcg_temp_free_i64(t0);
> + tcg_gen_xor_i64(ret, arg1, tcg_constant_i64(arg2));
> }
> }
>
> @@ -1415,9 +1394,7 @@ void tcg_gen_shli_i64(TCGv_i64 ret, TCGv_i64 arg1, int64_t arg2)
> } else if (arg2 == 0) {
> tcg_gen_mov_i64(ret, arg1);
> } else {
> - TCGv_i64 t0 = tcg_const_i64(arg2);
> - tcg_gen_shl_i64(ret, arg1, t0);
> - tcg_temp_free_i64(t0);
> + tcg_gen_shl_i64(ret, arg1, tcg_constant_i64(arg2));
> }
> }
>
> @@ -1429,9 +1406,7 @@ void tcg_gen_shri_i64(TCGv_i64 ret, TCGv_i64 arg1, int64_t arg2)
> } else if (arg2 == 0) {
> tcg_gen_mov_i64(ret, arg1);
> } else {
> - TCGv_i64 t0 = tcg_const_i64(arg2);
> - tcg_gen_shr_i64(ret, arg1, t0);
> - tcg_temp_free_i64(t0);
> + tcg_gen_shr_i64(ret, arg1, tcg_constant_i64(arg2));
> }
> }
>
> @@ -1443,9 +1418,7 @@ void tcg_gen_sari_i64(TCGv_i64 ret, TCGv_i64 arg1, int64_t arg2)
> } else if (arg2 == 0) {
> tcg_gen_mov_i64(ret, arg1);
> } else {
> - TCGv_i64 t0 = tcg_const_i64(arg2);
> - tcg_gen_sar_i64(ret, arg1, t0);
> - tcg_temp_free_i64(t0);
> + tcg_gen_sar_i64(ret, arg1, tcg_constant_i64(arg2));
> }
> }
>
> @@ -1468,12 +1441,17 @@ void tcg_gen_brcond_i64(TCGCond cond, TCGv_i64 arg1, TCGv_i64 arg2, TCGLabel *l)
>
> void tcg_gen_brcondi_i64(TCGCond cond, TCGv_i64 arg1, int64_t arg2, TCGLabel *l)
> {
> - if (cond == TCG_COND_ALWAYS) {
> + if (TCG_TARGET_REG_BITS == 64) {
> + tcg_gen_brcond_i64(cond, arg1, tcg_constant_i64(arg2), l);
> + } else if (cond == TCG_COND_ALWAYS) {
> tcg_gen_br(l);
> } else if (cond != TCG_COND_NEVER) {
> - TCGv_i64 t0 = tcg_const_i64(arg2);
> - tcg_gen_brcond_i64(cond, arg1, t0, l);
> - tcg_temp_free_i64(t0);
> + l->refs++;
> + tcg_gen_op6ii_i32(INDEX_op_brcond2_i32,
> + TCGV_LOW(arg1), TCGV_HIGH(arg1),
> + tcg_constant_i32(arg2),
> + tcg_constant_i32(arg2 >> 32),
> + cond, label_arg(l));
> }
> }
>
> @@ -1499,9 +1477,19 @@ void tcg_gen_setcond_i64(TCGCond cond, TCGv_i64 ret,
> void tcg_gen_setcondi_i64(TCGCond cond, TCGv_i64 ret,
> TCGv_i64 arg1, int64_t arg2)
> {
> - TCGv_i64 t0 = tcg_const_i64(arg2);
> - tcg_gen_setcond_i64(cond, ret, arg1, t0);
> - tcg_temp_free_i64(t0);
> + if (TCG_TARGET_REG_BITS == 64) {
> + tcg_gen_setcond_i64(cond, ret, arg1, tcg_constant_i64(arg2));
> + } else if (cond == TCG_COND_ALWAYS) {
> + tcg_gen_movi_i64(ret, 1);
> + } else if (cond == TCG_COND_NEVER) {
> + tcg_gen_movi_i64(ret, 0);
> + } else {
> + tcg_gen_op6i_i32(INDEX_op_setcond2_i32, TCGV_LOW(ret),
> + TCGV_LOW(arg1), TCGV_HIGH(arg1),
> + tcg_constant_i32(arg2),
> + tcg_constant_i32(arg2 >> 32), cond);
> + tcg_gen_movi_i32(TCGV_HIGH(ret), 0);
> + }
> }
>
> void tcg_gen_muli_i64(TCGv_i64 ret, TCGv_i64 arg1, int64_t arg2)
> @@ -1690,7 +1678,7 @@ void tcg_gen_bswap32_i64(TCGv_i64 ret, TCGv_i64 arg)
> } else {
> TCGv_i64 t0 = tcg_temp_new_i64();
> TCGv_i64 t1 = tcg_temp_new_i64();
> - TCGv_i64 t2 = tcg_const_i64(0x00ff00ff);
> + TCGv_i64 t2 = tcg_constant_i64(0x00ff00ff);
>
> /* arg = ....abcd */
> tcg_gen_shri_i64(t0, arg, 8); /* t0 = .....abc */
> @@ -1706,7 +1694,6 @@ void tcg_gen_bswap32_i64(TCGv_i64 ret, TCGv_i64 arg)
>
> tcg_temp_free_i64(t0);
> tcg_temp_free_i64(t1);
> - tcg_temp_free_i64(t2);
> }
> }
>
> @@ -1850,16 +1837,16 @@ void tcg_gen_clzi_i64(TCGv_i64 ret, TCGv_i64 arg1, uint64_t arg2)
> if (TCG_TARGET_REG_BITS == 32
> && TCG_TARGET_HAS_clz_i32
> && arg2 <= 0xffffffffu) {
> - TCGv_i32 t = tcg_const_i32((uint32_t)arg2 - 32);
> - tcg_gen_clz_i32(t, TCGV_LOW(arg1), t);
> + TCGv_i32 t = tcg_temp_new_i32();
> + tcg_gen_clzi_i32(t, TCGV_LOW(arg1), arg2 - 32);
> tcg_gen_addi_i32(t, t, 32);
> tcg_gen_clz_i32(TCGV_LOW(ret), TCGV_HIGH(arg1), t);
> tcg_gen_movi_i32(TCGV_HIGH(ret), 0);
> tcg_temp_free_i32(t);
> } else {
> - TCGv_i64 t = tcg_const_i64(arg2);
> - tcg_gen_clz_i64(ret, arg1, t);
> - tcg_temp_free_i64(t);
> + TCGv_i64 t0 = tcg_const_i64(arg2);
> + tcg_gen_clz_i64(ret, arg1, t0);
> + tcg_temp_free_i64(t0);
> }
> }
>
> @@ -1881,7 +1868,7 @@ void tcg_gen_ctz_i64(TCGv_i64 ret, TCGv_i64 arg1, TCGv_i64 arg2)
> tcg_gen_clzi_i64(t, t, 64);
> tcg_gen_xori_i64(t, t, 63);
> }
> - z = tcg_const_i64(0);
> + z = tcg_constant_i64(0);
> tcg_gen_movcond_i64(TCG_COND_EQ, ret, arg1, z, arg2, t);
> tcg_temp_free_i64(t);
> tcg_temp_free_i64(z);
> @@ -1895,8 +1882,8 @@ void tcg_gen_ctzi_i64(TCGv_i64 ret, TCGv_i64 arg1, uint64_t arg2)
> if (TCG_TARGET_REG_BITS == 32
> && TCG_TARGET_HAS_ctz_i32
> && arg2 <= 0xffffffffu) {
> - TCGv_i32 t32 = tcg_const_i32((uint32_t)arg2 - 32);
> - tcg_gen_ctz_i32(t32, TCGV_HIGH(arg1), t32);
> + TCGv_i32 t32 = tcg_temp_new_i32();
> + tcg_gen_ctzi_i32(t32, TCGV_HIGH(arg1), arg2 - 32);
> tcg_gen_addi_i32(t32, t32, 32);
> tcg_gen_ctz_i32(TCGV_LOW(ret), TCGV_LOW(arg1), t32);
> tcg_gen_movi_i32(TCGV_HIGH(ret), 0);
> @@ -1911,9 +1898,9 @@ void tcg_gen_ctzi_i64(TCGv_i64 ret, TCGv_i64 arg1, uint64_t arg2)
> tcg_gen_ctpop_i64(ret, t);
> tcg_temp_free_i64(t);
> } else {
> - TCGv_i64 t64 = tcg_const_i64(arg2);
> - tcg_gen_ctz_i64(ret, arg1, t64);
> - tcg_temp_free_i64(t64);
> + TCGv_i64 t0 = tcg_const_i64(arg2);
> + tcg_gen_ctz_i64(ret, arg1, t0);
> + tcg_temp_free_i64(t0);
> }
> }
>
> @@ -1969,9 +1956,7 @@ void tcg_gen_rotli_i64(TCGv_i64 ret, TCGv_i64 arg1, unsigned arg2)
> if (arg2 == 0) {
> tcg_gen_mov_i64(ret, arg1);
> } else if (TCG_TARGET_HAS_rot_i64) {
> - TCGv_i64 t0 = tcg_const_i64(arg2);
> - tcg_gen_rotl_i64(ret, arg1, t0);
> - tcg_temp_free_i64(t0);
> + tcg_gen_rotl_i64(ret, arg1, tcg_constant_i64(arg2));
> } else {
> TCGv_i64 t0, t1;
> t0 = tcg_temp_new_i64();
> @@ -2089,9 +2074,8 @@ void tcg_gen_deposit_z_i64(TCGv_i64 ret, TCGv_i64 arg,
> tcg_gen_andi_i64(ret, arg, (1ull << len) - 1);
> } else if (TCG_TARGET_HAS_deposit_i64
> && TCG_TARGET_deposit_i64_valid(ofs, len)) {
> - TCGv_i64 zero = tcg_const_i64(0);
> + TCGv_i64 zero = tcg_constant_i64(0);
> tcg_gen_op5ii_i64(INDEX_op_deposit_i64, ret, zero, arg, ofs, len);
> - tcg_temp_free_i64(zero);
> } else {
> if (TCG_TARGET_REG_BITS == 32) {
> if (ofs >= 32) {
> @@ -3102,9 +3086,8 @@ void tcg_gen_atomic_cmpxchg_i32(TCGv_i32 retv, TCGv addr, TCGv_i32 cmpv,
>
> #ifdef CONFIG_SOFTMMU
> {
> - TCGv_i32 oi = tcg_const_i32(make_memop_idx(memop & ~MO_SIGN, idx));
> - gen(retv, cpu_env, addr, cmpv, newv, oi);
> - tcg_temp_free_i32(oi);
> + TCGMemOpIdx oi = make_memop_idx(memop & ~MO_SIGN, idx);
> + gen(retv, cpu_env, addr, cmpv, newv, tcg_constant_i32(oi));
> }
> #else
> gen(retv, cpu_env, addr, cmpv, newv);
> @@ -3147,9 +3130,8 @@ void tcg_gen_atomic_cmpxchg_i64(TCGv_i64 retv, TCGv addr, TCGv_i64 cmpv,
>
> #ifdef CONFIG_SOFTMMU
> {
> - TCGv_i32 oi = tcg_const_i32(make_memop_idx(memop, idx));
> - gen(retv, cpu_env, addr, cmpv, newv, oi);
> - tcg_temp_free_i32(oi);
> + TCGMemOpIdx oi = make_memop_idx(memop, idx);
> + gen(retv, cpu_env, addr, cmpv, newv, tcg_constant_i32(oi));
> }
> #else
> gen(retv, cpu_env, addr, cmpv, newv);
> @@ -3210,9 +3192,8 @@ static void do_atomic_op_i32(TCGv_i32 ret, TCGv addr, TCGv_i32 val,
>
> #ifdef CONFIG_SOFTMMU
> {
> - TCGv_i32 oi = tcg_const_i32(make_memop_idx(memop & ~MO_SIGN, idx));
> - gen(ret, cpu_env, addr, val, oi);
> - tcg_temp_free_i32(oi);
> + TCGMemOpIdx oi = make_memop_idx(memop & ~MO_SIGN, idx);
> + gen(ret, cpu_env, addr, val, tcg_constant_i32(oi));
> }
> #else
> gen(ret, cpu_env, addr, val);
> @@ -3255,9 +3236,8 @@ static void do_atomic_op_i64(TCGv_i64 ret, TCGv addr, TCGv_i64 val,
>
> #ifdef CONFIG_SOFTMMU
> {
> - TCGv_i32 oi = tcg_const_i32(make_memop_idx(memop & ~MO_SIGN, idx));
> - gen(ret, cpu_env, addr, val, oi);
> - tcg_temp_free_i32(oi);
> + TCGMemOpIdx oi = make_memop_idx(memop & ~MO_SIGN, idx);
> + gen(ret, cpu_env, addr, val, tcg_constant_i32(oi));
> }
> #else
> gen(ret, cpu_env, addr, val);
--
Alex Bennée
next prev parent reply other threads:[~2020-04-22 20:06 UTC|newest]
Thread overview: 75+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-04-22 1:16 [PATCH v2 00/36] tcg 5.1 omnibus patch set Richard Henderson
2020-04-22 1:16 ` [PATCH v2 01/36] tcg: Add tcg_gen_gvec_dup_imm Richard Henderson
2020-04-22 1:16 ` [PATCH v2 02/36] target/s390x: Use tcg_gen_gvec_dup_imm Richard Henderson
2020-04-22 1:16 ` [PATCH v2 03/36] target/ppc: " Richard Henderson
2020-04-22 1:16 ` [PATCH v2 04/36] target/arm: " Richard Henderson
2020-04-22 1:16 ` [PATCH v2 05/36] tcg: Use tcg_gen_gvec_dup_imm in logical simplifications Richard Henderson
2020-04-22 1:16 ` [PATCH v2 06/36] tcg: Remove tcg_gen_gvec_dup{8,16,32,64}i Richard Henderson
2020-04-22 1:16 ` [PATCH v2 07/36] tcg: Add tcg_gen_gvec_dup_tl Richard Henderson
2020-04-22 1:16 ` [PATCH v2 08/36] tcg: Improve vector tail clearing Richard Henderson
2020-04-22 1:16 ` [PATCH v2 09/36] tcg: Consolidate 3 bits into enum TCGTempKind Richard Henderson
2020-04-22 11:25 ` Alex Bennée
2020-04-22 19:58 ` Aleksandar Markovic
2020-04-23 9:00 ` Philippe Mathieu-Daudé
2020-04-23 15:40 ` Richard Henderson
2020-04-23 17:24 ` Daniel P. Berrangé
2020-04-23 23:11 ` Richard Henderson
2020-04-24 9:08 ` Daniel P. Berrangé
2020-04-22 1:16 ` [PATCH v2 10/36] tcg: Add temp_readonly Richard Henderson
2020-04-22 11:26 ` Alex Bennée
2020-04-22 1:16 ` [PATCH v2 11/36] tcg: Introduce TYPE_CONST temporaries Richard Henderson
2020-04-22 15:17 ` Alex Bennée
2020-04-22 16:55 ` Richard Henderson
2020-04-22 1:16 ` [PATCH v2 12/36] tcg: Use tcg_constant_i32 with icount expander Richard Henderson
2020-04-22 15:40 ` Alex Bennée
2020-04-22 1:16 ` [PATCH v2 13/36] tcg: Use tcg_constant_{i32, i64} with tcg int expanders Richard Henderson
2020-04-22 16:18 ` [PATCH v2 13/36] tcg: Use tcg_constant_{i32,i64} " Alex Bennée
2020-04-22 17:02 ` Richard Henderson
2020-04-22 17:57 ` Alex Bennée
2020-04-22 20:04 ` Alex Bennée [this message]
2020-04-23 23:13 ` Richard Henderson
2020-04-24 13:23 ` Alex Bennée
2020-04-22 1:17 ` [PATCH v2 14/36] tcg: Use tcg_constant_{i32, vec} with tcg vec expanders Richard Henderson
2020-04-22 17:00 ` [PATCH v2 14/36] tcg: Use tcg_constant_{i32,vec} " Alex Bennée
2020-04-22 1:17 ` [PATCH v2 15/36] tcg: Use tcg_constant_{i32,i64} with tcg plugins Richard Henderson
2020-04-22 17:18 ` [PATCH v2 15/36] tcg: Use tcg_constant_{i32, i64} " Alex Bennée
2020-04-22 1:17 ` [PATCH v2 16/36] tcg: Rename struct tcg_temp_info to TempOptInfo Richard Henderson
2020-04-22 17:19 ` Alex Bennée
2020-04-22 1:17 ` [PATCH v2 17/36] tcg/optimize: Adjust TempOptInfo allocation Richard Henderson
2020-04-22 17:53 ` Alex Bennée
2020-04-22 18:28 ` Alex Bennée
2020-04-22 1:17 ` [PATCH v2 18/36] tcg/optimize: Use tcg_constant_internal with constant folding Richard Henderson
2020-04-22 18:28 ` Alex Bennée
2020-04-22 1:17 ` [PATCH v2 19/36] tcg/tci: Add special tci_movi_{i32,i64} opcodes Richard Henderson
2020-04-22 19:02 ` Alex Bennée
2020-04-22 1:17 ` [PATCH v2 20/36] tcg: Remove movi and dupi opcodes Richard Henderson
2020-04-22 9:12 ` Aleksandar Markovic
2020-04-22 19:03 ` Alex Bennée
2020-04-22 1:17 ` [PATCH v2 21/36] tcg: Use tcg_out_dupi_vec from temp_load Richard Henderson
2020-04-22 19:28 ` Alex Bennée
2020-04-22 1:17 ` [PATCH v2 22/36] tcg: Increase tcg_out_dupi_vec immediate to int64_t Richard Henderson
2020-04-22 19:33 ` Alex Bennée
2020-04-22 1:17 ` [PATCH v2 23/36] tcg: Add tcg_reg_alloc_dup2 Richard Henderson
2020-04-22 19:40 ` Alex Bennée
2020-04-22 1:17 ` [PATCH v2 24/36] tcg/i386: Use tcg_constant_vec with tcg vec expanders Richard Henderson
2020-04-22 19:43 ` Alex Bennée
2020-04-22 1:17 ` [PATCH v2 25/36] tcg: Remove tcg_gen_dup{8,16,32,64}i_vec Richard Henderson
2020-04-23 9:11 ` Alex Bennée
2020-04-22 1:17 ` [PATCH v2 26/36] tcg: Add load_dest parameter to GVecGen2 Richard Henderson
2020-04-23 9:37 ` Alex Bennée
2020-04-22 1:17 ` [PATCH v2 27/36] tcg: Fix integral argument type to tcg_gen_rot[rl]i_i{32, 64} Richard Henderson
2020-04-22 10:19 ` Philippe Mathieu-Daudé
2020-04-23 9:38 ` [PATCH v2 27/36] tcg: Fix integral argument type to tcg_gen_rot[rl]i_i{32,64} Alex Bennée
2020-04-22 1:17 ` [PATCH v2 28/36] tcg: Implement gvec support for rotate by immediate Richard Henderson
2020-04-23 13:28 ` Alex Bennée
2020-04-22 1:17 ` [PATCH v2 29/36] tcg: Implement gvec support for rotate by vector Richard Henderson
2020-04-23 13:41 ` Alex Bennée
2020-04-22 1:17 ` [PATCH v2 30/36] tcg: Remove expansion to shift by vector from do_shifts Richard Henderson
2020-04-22 1:17 ` [PATCH v2 31/36] tcg: Implement gvec support for rotate by scalar Richard Henderson
2020-04-23 13:46 ` Alex Bennée
2020-04-22 1:17 ` [PATCH v2 32/36] tcg/i386: Implement INDEX_op_rotl[is]_vec Richard Henderson
2020-04-22 1:17 ` [PATCH v2 33/36] tcg/aarch64: Implement INDEX_op_rotli_vec Richard Henderson
2020-04-22 1:17 ` [PATCH v2 34/36] tcg/ppc: Implement INDEX_op_rot[lr]v_vec Richard Henderson
2020-04-22 1:17 ` [PATCH v2 35/36] target/ppc: Use tcg_gen_gvec_rotlv Richard Henderson
2020-04-22 1:17 ` [PATCH v2 36/36] target/s390x: Use tcg_gen_gvec_rotl{i,s,v} Richard Henderson
2020-04-23 13:50 ` [PATCH v2 00/36] tcg 5.1 omnibus patch set Alex Bennée
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=87ftcvb82k.fsf@linaro.org \
--to=alex.bennee@linaro.org \
--cc=qemu-devel@nongnu.org \
--cc=richard.henderson@linaro.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).